Professional Documents
Culture Documents
ZHANG, YUAN. Cooperative Distributed Energy Management for Smart Grid. (Under the
direction of Dr. Mo-Yuen Chow.)
gathering information from all the installed devices within the system for energy management
applications. However, this centralized management structure makes the system vulnerable to
single point of failure and communication failures, and raises privacy concerns. To
accommodate for the increasing penetration of distributed energy resources and increasing
requirement of energy security and resiliency, the energy management for smart grid is
witnessing a paradigm shift from centralized structure to a distributed structure. The objective
and focus of this dissertation is to develop fully distributed algorithms for smart grid
cooperative energy management for different applications and scenarios around an energy
management framework.
Within this distributed energy management framework, each bus of the system is modeled
exchanging information with its neighbor agents, the optimal dispatch decisions and schedules
of the dispatchable devices connected at each bus can be achieved in an iterative manner. Four
algorithms are developed throughout this dissertation for different energy management
scenarios. The Robust distributed system Incremental Cost Estimation (RICE) algorithm is
developed to solve the economic dispatch problem for a smart grid environment in a
Optimal Dispatch (CDOD) algorithm is developed to solve the DC optimal power flow
problem with line loss consideration in a distributed manner. The Cooperative Distributed
Energy Scheduling (CoDES) algorithm is developed for a smart grid to solve the optimal
dispatch schedule problem with renewable and energy storage integration. The CoDES
algorithm considers modified DC power flow constraints, branch energy losses, and energy
storage charging and discharging efficiencies. Finally, based on the CoDES algorithm, the
developed for active energy management for smart grid with the capability of adjusting the
scheduled set points to compensate for the prediction errors arising from the intermittencies of
by
Yuan Zhang
Electrical Engineering
2016
APPROVED BY:
_______________________________ _______________________________
Dr. Mo-Yuen Chow Dr. Subhashish Bhattacharya
Chair of Advisory Committee
_______________________________ _______________________________
Dr. David Lubkeman Dr. Alyson Wilson
DEDICATION
To My Parents
ii
BIOGRAPHY
Yuan Zhang was born in Hangzhou, the beautiful capital city of Zhejiang Province in
China. He received the B.Eng degree in Electrical Engineering from Huazhong University
of Science and Technology, Wuhan, Hubei, China, in 2011, and the MS degree in
Electrical Engineering from North Carolina State University, Raleigh, NC, USA, in 2013.
He is currently pursuing the Ph.D. degree from the Department of Electrical and Computer
Engineering, North Carolina State University. His research interest includes microgrid
control, cooperative distributed control, smart grid energy management, and smart grid
iii
ACKNOWLEDGMENTS
I would not be able to finish this dissertation without the invaluable guidance and
support from many people around me. First, I would like to express my deepest sincere
gratitude to my advisor, Dr. Mo-Yuen Chow, who has always been patient, supportive and
encouraging during my study in ADAC lab. Without his support, guidance, and
knowledge, this dissertation would not have been possible. The methodology and
perspective of analytical thinking and problem solving that I have learned from him will
Second, I would like to thank my committee members, Dr. Alyson Wilson, Dr. David
Lubkeman, and Dr. Subhashish Bhattacharya. Their insightful comments and feedbacks
Third, I would also like to thank my dear friends and colleagues in ADAC lab for the
numerous fruitful discussions we have had. Specifically, I would like to thank Dr. Ziang
Zhang, Dr. Wente Zeng, Dr. Navid Rahbari-Asr, and Jie Duan. Without their co-
privilege to work with you. In addition, I also need to thank Dr. Habiballah Rahimi Eichi,
Dr. Unnati Ojha, Dr. Wengcong Su, Xichun Ying, Cong-Sheng Huang, Bharat Balagopal,
Alberto Castelo and all the others who had joined ADAC lab, for the fruitful discussions
I would also like to thank my former roommates Dr. Yi Xu, and Rong Jing, and current
roommates Rui Gao, and Suxuan Guo, you made me feel like at home.
iv
Finally, there will never be enough words to express my gratitude to my parents, Ying
Zhu and Jian-Qiang Zhang. Without your love, support, and faith, none of these would be
possible.
v
TABLE OF CONTENTS
vi
3.3 CONVERGENCE FACTOR DESIGN ......................................................................... 43
3.4 CASE STUDIES ..................................................................................................... 45
3.4.1 Case Study 1: Convergence Factor Range ................................................. 46
3.4.2 Case Study 2: Non-negative Function Trajectory with Stable and Unstable
Convergence Factor .................................................................................................. 48
3.4.3 Case Study 3: Decaying Factor and Trajectory Bound .............................. 50
3.5 CONCLUSION AND REMARKS............................................................................... 51
Chapter 6. Cooperative Distributed Energy Scheduling for Microgrid with Real-
time Re-Dispatch ............................................................................................................. 97
6.1 INTRODUCTION .................................................................................................... 97
6.2 PROBLEM FORMULATION .................................................................................. 100
6.2.1 Day-ahead Energy Scheduling ................................................................. 101
6.2.2 Real-Time Energy Re-dispatch ................................................................. 104
6.3 COOPERATIVE DISTRIBUTED ENERGY SCHEDULING WITH RE-DISPATCH .......... 106
6.3.1 Cooperative Distributed Energy Scheduling ............................................ 106
6.3.2 Localized Convergence Assessment for CoDES Algorithm ...................... 117
6.3.3 Cooperative Distributed Energy Re-dispatch ........................................... 126
6.4 NUMERICAL SIMULATION ................................................................................. 132
6.4.1 Test System Description ............................................................................ 132
6.4.2 Cooperative Distributed Energy Scheduling Result ................................. 134
6.4.3 Cooperative Distributed Re-Dispatch Result............................................ 140
6.4.4 Cost Comparison ...................................................................................... 142
6.5 CONCLUSION ..................................................................................................... 144
viii
LIST OF TABLES
ix
LIST OF FIGURES
x
Figure 5-5 Generator schedule in per unit: (a) Generator 1 schedule; (b) Generator 2
schedule; (c) Generator 3 schedule ................................................................................... 87
Figure 5-6 Locational marginal price ($/MWh) at bus 4 .................................................. 87
Figure 5-7: (a) ~ (e) Battery 1 ~ 5 charging/discharging schedule in per unit ................. 88
Figure 5-8 Convergence of hour 1 generator dispatching ................................................ 89
Figure 5-9 (a) ~ (e): Actual SoC of battery 1 ~5 considering charging/discharging
efficiency........................................................................................................................... 90
Figure 5-10 (a) ~ (e): Actual SoC of battery 1 ~5 without considering
charging/discharging efficiency ........................................................................................ 91
Figure 5-11 Algorithm objective value with one generator removal at 4000th iteration. 93
Figure 5-12 Histogram of percentage difference in objective values between the
centralized fmincon function and the proposed CoDES algorithm. ................................. 95
Figure 6-1 CoDES-ReD structure block diagram. ............................................................ 99
Figure 6-2 4-bus system example. .................................................................................. 100
Figure 6-3. CoDES algorithm localized convergence assessment flow chart ................ 126
Figure 6-4 34-bus microgrid one-line diagram. .............................................................. 132
Figure 6-5 Randomly created system communications topology. .................................. 133
Figure 6-6 Demand and renewable generation profiles example. .................................. 134
Figure 6-7 Convergence of cooperative distributed energy scheduling. ........................ 135
Figure 6-8 Day ahead dispatch schedules. ...................................................................... 136
Figure 6-9. Local convergence flag average and the local convergence measure .......... 137
Figure 6-10. Local convergence measure component at different iterations .................. 138
Figure 6-11. External component example for the first scheduling time step ................ 138
Figure 6-12. Local and global convergence measures .................................................... 139
Figure 6-13 Convergence of cooperative distributed energy re-dispatch command for the
first 5-minute interval. .................................................................................................... 141
Figure 6-14 Re-dispatch power mismatch histogram for all 288 executions. ................ 141
Figure 6-15 Real-time dispatch commands. ................................................................... 142
Figure 6-16 Grid power drawn under different cases. .................................................... 143
Figure 6-17 Total electricity bill for three cases. ............................................................ 143
xi
Chapter 1. INTRODUCTION
being integrated into the current power systems [1]. As a result, the legacy power system
is gradually evolving into its 2.0 version: the smart grid [2], as shown in Figure 1-1.
One of the most essential aspects of a power system is energy management, which
greatly affects the efficiency, and reliability of the system operations. The Energy
Independence and Security Act of 2007 (EISA) has defined a list of distinguishing
characteristics of a smart grid [3]. Some of the energy management related characteristics
renewable resources;
1
3. Deployment and integration of advanced electricity storage and peak-shaving
technologies, including plug-in electric and hybrid electric vehicles, and thermal-
applications in the power system, such as volt/var control, economic dispatch, unit
the control center. In order to do this, each of the controllable devices in the system will
need to send information to the control center through the Supervisory, Control, and Data
Acquisition (SCADA) systems. However, as the numbers of distributed resources that are
deployed and integrated into the system keep increasing, the traditional centralized energy
management scheme may hit its limitations if all of them will need to communication with
the control center, and it will be more vulnerable to errors and faults.
Distributed algorithms have already been studied in many areas such as animal science
[4], [5], computer science [6], and control engineering [7], [8]. In a distributed system,
each individual or agent only needs to take local measurements, exchange information
with neighbors, and do local calculations. By doing this kind of coordination and local
control concepts on the smart grid to solve different energy management problems. This
dissertation includes multiple distributed algorithms the author has developed during his
2
Chapter 2 presents a Robust distributed system Incremental Cost Estimation (RICE)
algorithm for solving power system classic economic dispatch problems (EDP).
Communications information losses due to communications time delays and packet losses
are considered in the algorithm, and robust correction scheme is designed to compensate
for the information losses. The convergence and optimality of the RICE algorithm is also
proved provided the convergence factor value of the algorithm is small enough.
tuning method for the RICE algorithm is developed that can automatically tune the
convergence factor value of the RICE algorithm to guarantee its convergence and
optimality.
The EDP that the RICE algorithm solves does not consider system line capacity and
is presented to solve the DC optimal power flow problem in a distributed way while
that further extends the CDOD algorithm by considering renewable energy and energy
storage integration into the smart grid. Different from the previous chapters where single-
step energy-scheduling problems are being solved, the CoDES algorithm solves a multi-
step energy-scheduling problem to find the optimal dispatch schedule for a smart grid.
System topology constraints and energy storage efficiencies are also considered in the
3
Chapter 5, but in a distribution system scenario. The CoDES-ReD algorithm has two
major functionalities: 1) determines the optimal energy storage devices dispatch schedules
based on the day ahead predicted profiles; and 2) adjusts the scheduled set points to
compensate for the prediction errors arising from the intermittencies of the renewables and
real-time load patterns. In this algorithm, each bus of the system is modeled as an agent.
By solely exchanging information with neighbors, the dispatch command of the energy
Chapter 7 concludes the dissertation and proposes several directions for future work.
4
Chapter 2. A ROBUST DISTRIBUTED SYSTEM INCREMENTAL COST ESTIMATION
INFORMATION LOSSES
Abstract:
integrated into the power system, how to manage these distributed resources economically
will become a challenge for the future smart grid. To solve the issue, consensus based
algorithms degrades when information losses occur. In this chapter, a Robust distributed
algorithms that solve EDP, the RICE algorithm has two updating layers running in parallel
in each distributed controller: one layer uses the gossip-based updating rule to estimate the
system average power mismatch, while the other layer uses the consensus-based updating
rule to update the system Incremental Cost (IC) estimation. In this approach, the
eliminated. The convergence and optimality of the algorithm is guaranteed as long as the
studies are presented to illustrate the performance of the proposed algorithm and show the
5
robustness under different information loss scenarios with different communications
topologies.
2.1 INTRODUCTION
The power system today is facing many challenges that it was not originally designed
to handle. Smart grid is believed to be the future for the power system. One key feature of
the smart gird is the deployment and integration of controllable distributed energy
resources. However, economically manage these energy resources will be a challenge for
The Economic Dispatch Problem (EDP) is one of the fundamental energy management
minimize the total generation cost by determining the required output power of each
generation unit under constraints. Various algorithms have been proposed by researchers
to solve the EDP under different conditions [9]–[11]. However, all of these algorithms
need to have a control center that has access to the states of the entire system. With a
variety of controllable distributed devices installed in the future smart grid, the central
controller will have a high computational and communications burden, and the system will
be vulnerable to central node failures. Besides, even a small change in the smart grid such
makes the centralized methods highly inefficient and unscalable. Thus, a new paradigm is
6
includes the physical system dynamics, a cyber layer that includes the communications
system, and a control middleware that runs the distributed algorithm that links the physical
Cyber Layer
Distributed
Control and
Management
Induction Middleware
Generator motor load
been studied by researchers [13]–[17]. The power system dispatch problem is one of their
applications and is discussed in many references [17]–[19]. In [17], the Incremental Cost
Consensus (ICC) algorithm is proposed. The incremental cost (IC) of each generation unit
is chosen as the information state for the consensus network. However, in order to satisfy
the power balance constraint, one of the generation units needs to be chosen as the leader
agent to have access to the total system demand information. In [18], a consensus +
innovation approach is proposed to solve the EDP. This method eliminates the
requirement of the leader agent by using a consensus term to achieve the convergence of a
common IC value and an innovation term to ensure the power balance constraint is
satisfied. In [19], instead of using a constant local reference, each generator can estimate
7
the mismatch between generation and total demand. Then, this local estimation is used to
communications network. However, in real world applications, packet loss may occur due
more likely to be used, and is more prone to transmission errors [21]. In this chapter, we
do not specify the communications protocol that is used by the distributed algorithm. The
packet losses due to congestion and transmission error are treated to have similar effect on
the distributed algorithm in terms of information loss, and we use information loss rate to
denote this effect in a general structure. It has been shown in [22] that the performance of
the average consensus algorithm degrades when there are information losses in the
communications network. A case study will be shown in Section 2.4 that under the
information loss environment, the algorithm proposed in [19] for EDP no longer
Within unreliable communications network conditions, the gossip algorithm has drawn
a lot of attention [23], [24]. Compared with the consensus algorithm, the gossip algorithm
the packet routing and just exchanges information with at most one of its direct neighbors
during a given time slot. Thus, it takes much less communications bandwidth and has
much fewer constraint than the consensus case. If an information loss event occurs, both
neighbors will not update their states. In [25], an asynchronous distributed algorithm is
8
proposed using pairwise gossip protocol to solve the EDP; however, a leader agent is still
information losses. The RICE algorithm has two updating layers running in parallel: one
layer uses the gossip-based updating rule to estimate the average power mismatch of the
system, while the other layer uses the consensus-based updating rule to update the system
IC estimation of each generation unit. The convergence and optimality of the algorithm
are guaranteed even under communications information losses as long as the undirected
The rest of the chapter is organized as follows. In Section 2.2, the preliminaries are
introduced. The formulation and proof of the proposed RICE algorithm are provided in
Section 2.3. The performance of the RICE algorithm is studied via several representative
case studies in Section 2.4. Finally, the concluding remarks are brought in Section 2.5.
2.2 PRELIMINARY
column vector with all the elements equal to 1. The product of two stochastic matrices is
stochastic matrix if 1TQ = 1T. A nonnegative square matrix D is called a doubly stochastic
Moreover, assume a column vector ψ has the same length as the row stochastic matrix
P, and the elements in this vector satisfy a probability distribution function, that is, ψT1 =
9
1. Based on the Perron-Frobenius theorem [26], we can always find such a vector which
satisfies:
ψT P ψT . (2-1)
Similarly, for a column stochastic matrix Q, there exists a probability distribution φ such
that:
Qφ φ. (2-2)
In a multi-agent consensus network, each agent agrees on the value of a shared variable
among all the agents by exchanging information with its neighbors. This shared value is
called information state xi. The consensus algorithm for single-integrator dynamics can be
xi k 1 xi k aij x j k xi k , (2-3)
j Ni
where k denotes the number of iteration, η is the step size, aij is the entry (i, j) of the
adjacency matrix A associated with the communications topology, and Ni is the neighbor
set of agent i. The system can also be written in matrix form as:
x k 1 Px k , (2-4)
where P = I − ηL is called the Perron matrix of the graph with parameter η, and L is the
In the gossip algorithm, each agent exchanges information with at most one of its
neighbors during each iteration [16]. Denote S to be the set of node pairs that are
exchanging information during the kth iteration, the following updating rule will be used:
10
x k 1 W' k x k , (2-5)
Remark 1: The consensus and gossip algorithms are distributed updating rules for
guidelines on how to design the system communications topology and determine how
each local controller should update its local state based on the received information from
its neighbors. Different communications protocols such as TCP/IP, UDP, etc. can be
implemented to handle the packet exchange among local controllers. In this chapter, we do
Assume that each generation unit has the following quadratic cost function:
2
Cgen.i Pgen.i ai bi Pgen.i ci Pgen .i , (2-7)
where ai, bi, and ci are known coefficients, and Pgen.i is the output power of the ith
generation unit, respectively. The objective of solving the EDP is to find the optimal
generation schedule Pgen.i for each generation unit i, such that the total generation cost of
n
min Ci Pgen.i . (2-8)
i 1
The problem is subjected to the following two constraints: 1) the power balance constraint:
11
n
Pgen.i Pd , (2-9)
i 1
where Pd is the fixed system load; and 2) the generation limit constraints:
where Pgen.i.min and Pgen.i.max are the lower and upper limits of the power that can be
Assume a smart grid has n buses. Each bus is equipped with a distributed controller that
has two-way communications capability. The distributed controller also has access to the
information of its local devices such as generation cost function and local demand.
Without loss of generality, assume each bus has one local generation unit and one local
load connected to it. With these assumptions, the optimization problem (2-8) ~ (2-10) can
be written as:
n
min Ci Pgen.i
i 1
s.t.
n n , (2-11)
Pgen.i Pd .i
i 1 i 1
n
where Pd .i is the fixed load connected to bus i, and Pd i 1 d .i
P .
Without considering the generation limits, the Lagrangian function of (2-11) can be
written as:
12
n n
2
J ai bi Pgen.i ci Pgen .i Pd .i Pgen.i , (2-12)
i 1 i 1
where λ > 0 is the Lagrangian multiplier, and here it is also called the system IC. The first
J
bi 2ci Pgen.i 0, (2-13)
Pgen.i
J n n
Pd .i Pgen.i P 0. (2-14)
i 1 i 1
bi
Pgen.i . (2-15)
2ci
Therefore, the updating rule for each variable can be written as:
k 1 k P k , (2-16)
n n
where P k Pd .i k Pgen.i k , ε is the convergence factor affecting the
i 1 i 1
k bi
Pgen.i k 1 Pgen.i.min , if Pgen.i.min
2ci
k bi
Pgen.i k 1 . (2-17)
2ci
k bi
Pgen.i k 1 Pgen.i.max , if Pgen.i.max
2ci
However, the algorithm described by (2-16) and (2-17) requires each device to have
access to the global information ΔP(k) and λ(k). Thus, each device will need to have
access to the decisions made by all the other devices in the system as well as the demand
13
of all the loads for each iteration. This may not be feasible nor desirable in a large
To make the iterative algorithm fully distributed such that no device is required to have
access to the global information, the global variable can be estimated by using consensus
and gossip networks within two updating layers. Together, the two updating layers serve
The system IC can be iteratively estimated by each controller as the information state
n
ˆ k 1 Wij ˆ j k P k , (2-18)
i
j 1
where Wij is the entry (i, j) of a row stochastic nonnegative consensus update weighting
matrix W, ΔP(k) is the system power mismatch, and ˆi k is the estimated system IC of
the ith bus at iteration k. Note that the update weighting matrix W here is not necessary to
be the Perron matrix shown in (2-4). Later it will be shown that as long as W is row
Let ΔPi(k) represent the power mismatch between the local demand and generation
Pi k Pd .i k Pgen.i k , (2-19)
where Pd.i(k) and Pgen.i(k) are the demand and generation, respectively, of bus i at iteration
k. Define Pˆi k as the system power mismatch estimation state of bus i at iteration k.
Then, the average power mismatch can be estimated by each local controller using a
14
n
Pˆi k 1 Wij ' Pˆ j k Pi k 1 Pi k , (2-20)
j 1
where Wij′ is the entry (i, j) of a random gossip updating matrix W′. If write (2-20) into
matrix form,
Pˆ k 1 W ' Pˆ k P k 1 P k , (2-21)
1T Pˆ k 1 P k 1 1T W ' Pˆ k P k . (2-22)
1T Pˆ k 1 P k 1 1T Pˆ k P k . (2-23)
Using (2-23),
1T Pˆ k 1 P k 1 1T Pˆ k P k
1T Pˆ k 1 P k 1
(2-24)
1T Pˆ 0 P 0 .
n n
Pˆi k Pi k , (2-25)
i 1 i 1
Pˆi k is the actual system power mismatch between the demand and
n
which means i 1
the generation. Now, each distributed controller will have partial information of the
system power mismatch. Pˆi can be used in the system IC estimation rule to represent the
15
n
ˆ k 1 Wij ˆ j k Pˆi k . (2-26)
i
j 1
Equations (2-26), (2-19) and (2-20) represent the proposed RICE algorithm. From these
three equations, it is clear that the only information each local controller needs to send to
its neighbors are its local system IC estimation and its local system power mismatch
estimation. The flowchart of the RICE algorithm for each agent is shown in Figure 2-2.
Start
1 ˆ bi
Calculate Pgen.i k 1 i k 1
2ci 2ci
Converge? No
Yes
End
Remark 2: The definition of system IC is required for this algorithm. Therefore, if for a
certain bus i, there is no generation unit connected, then the generation cost function
associated with that particular bus is set to a default value while both of the generation
16
limits are set to 0. In this chapter the default values are chosen as a = 0, b = 8, and c =
0.001. If multiple generation units are connected to bus i, then they can be aggregated into
one single local generation unit. If multiple loads are connected to one single bus i, then
information losses. As shown in the previous section, the communications involved in the
RICE algorithm are mainly represented in equation (2-20) using gossip algorithm for
system power mismatch estimation, and (2-26) using consensus algorithm for system IC
estimation.
n n n n n n
ˆ k 1 Wij ˆ j k Pˆi k ˆ k Pˆi k . (2-27)
i i
i 1 i 1 j 1 i 1 i 1 i 1
As k ,
n
ˆ k 1 n
ˆ k will hold if the algorithm converges. Thus,
i 1 i i 1 i
n
lim Pˆi k 0. (2-28)
k
i 1
If the system power mismatch estimation information Pˆm sent from agent m to agent l is
lost during iteration k, and no actions are taken, agent l will need to use some estimation of
this value Pm which may not be equal to the actual value. Based on (2-22) ~ (2-25) we
will get:
17
1T Pˆ k 1 P k 1
n n n
Wij' Pˆj k Wlm' Pˆm k Wlm' Pm k Pi k
i 1 j 1 i 1
n n
, (2-29)
Pi k W '
lm Pˆm k W '
lm Pm k Pi k
i 1 i 1
Wlm' Pm k Pˆm k
n n
Pˆi k Pi k Wlm' Pm k Pˆm k . (2-30)
i 1 i 1
n
lim Pi k Wlm' Pm k Pˆm k 0. (2-31)
k
i 1
n n
lim Pd .i k Pgen.i k 0, (2-32)
k
i 1 i 1
which means that any information loss scenario, will eventually result in power mismatch
between the generation and load, and thus the power balance equality constraint will not
be satisfied. As a result, the algorithm can no longer converge to a feasible solution. Thus,
the effect of the information loss is essentially embedded in the optimization problem.
The mitigation of information loss in the RICE algorithm is achieved by using gossip
updating rule in the power mismatch layer and a correction mechanism in the system IC
estimation layer. In the power mismatch layer, each distributed controller only needs to
exchange information with at most one of its neighbors at any iteration. If any of the
information being exchanged are lost, none of the controllers will update the power
mismatch estimation. Therefore, the power mismatch information that is stored in the
18
summation of all local mismatch estimations (as shown in (2-25)) will remain intact. Thus,
the system power mismatch estimation layer would be robust to information loss.
In the system IC estimation layer, as shown in (2-26), the RICE algorithm only requires
a row stochastic updating matrix W. One simple correction mechanism can be applied to
mitigate the effect of information loss. For instance, assume a four-bus system with the
specific communications topology running the RICE algorithm with the W matrix given
below:
13 13 13 0
13 13 13 0
W . (2-33)
14 14 14 14
0 0 12 12
If the IC information being sent from bus 2 to bus 1 is lost during the kth iteration, the first
row of the W matrix for that iteration can be modified by bus 1 to keep the W matrix row
2
stochastic. The update weighting of node 1 can be increased temporarily. That is, W11
3
and W12 0 while all the other elements remain the same. This ensures the correctness of
the updating, and the row stochastic structure of the updating matrix is preserved. This
method can also be applied to the scenarios when multiple information losses happen at
general structure. The packet loss due to congestion and transmission error are treated as
having similar effect on the distributed algorithm in terms of information loss, and we use
information loss rate to denote this effect. Each data packet is transmitted only once
acknowledgment will be sent to the sender regarding the correctness of the received
19
information. If the packet were not received, there would be no acknowledgement. It
might be argued that the acknowledgement can also be lost and another acknowledgement
will be needed if the communications network has a communication packet loss rate. Then,
this becomes the Byzantine’s general problem [27] which is beyond the scope of this
Remark 4: Since the RICE algorithm runs in an iterative manner, there will be a
deadline for each updating. For example, assume a system is configured in such a way that
the updating should be executed every 100 ms. If a communication link in the system is
highly unreliable, using ARQ may lead to multiple packet retransmission requests during
iteration k, which may violate the 100 ms requirement and result in a time delay scenario.
In this section, we show that the RICE algorithm converges by choosing a sufficiently
small value of ε. The RICE algorithm can be written in the matrix form as:
λˆ k 1 Wλˆ k Pˆ k , (2-34)
P k 1 Pd k 1 Pgen k 1 , (2-35)
Pˆ k 1 W ' Pˆ k P k 1 P k , (2-36)
where W is a row stochastic matrix based on consensus updating rule and W′ is a random
20
Pˆ k 1 W' G Pˆ k G W I λˆ k Pd k 1 Pd k . (2-38)
λˆ k 1 W I λˆ k 0
Pd , (2-39)
Pˆ k 1 G W I W' G Pˆ k I
where ΔPd = Pd(k+1) − Pd(k) is the step load change and diag … .
λˆ k 1 W I λˆ k 0
Pd , (2-40)
E Pˆ k 1 G W I W G E Pˆ k I
W I
M ,
G W I W G
W 0
A ,
G W I W
0
and ∆ ; thus,
0
M A . (2-41)
Matrix A is a triangular matrix so its eigenvalues are the eigenvalues of W and . Since
W is a row stochastic matrix and is a doubly stochastic matrix, for both of them, all the
eigenvalues are located within the unit circle except one, which is located at one. Thus, A
has two eigenvalues located at θ1 = θ2 = 1, and all the other eigenvalues are within the unit
circle.
Based on the eigenvalue derivation theory [28], the derivative of the eigenvalue θi with
respect to ε is:
21
M
i
y Ti | 0 xi , (2-42)
where yi and xi are the left and right eigenvectors of matrix M associated with eigenvalue
θi satisfying:
that y1T = [1TG 1T] and y2T = [ψT 0] are the left eigenvectors of A, and 0
θ1 = θ2 = 1. They also satisfy (2-43). Here, ψ and φ are the probability distributions
0
1 T 0 I T T 0 I
y 1 x1 1 G 1 1 0, (2-44)
0 G 0 G 1
n
0 I 0 I 1
2
y T2 x2 ψT 0 ψT φ 0. (2-45)
0 G 0 G φ
Based on (2-44) and (2-45), it is clear that θ1 is not perturbed by , while θ2 will move
toward the center of the unit circle as ε increases. Therefore, there exists a small enough
positive constant σ1 such that when 0 < ε < σ1, θ2 is located within the unit circle.
have
min X p
X 1
p
, (2-46)
p
22
where ||.||p is any matrix p-norm and κp = ||X||p||X−1||p is called the condition number of the
eigenvalue problem for matrix A. Substituting for Δ and using the infinity norm yields:
Inequality (2-47) states that the perturbation of the eigenvalues of M from the eigenvalues
small value σ2 such that for all 0 < ε < σ2, those eigenvalues of A, which are inside the unit
circle for ε = 0, remain inside. Therefore, a small number σ = min{σ1, σ2} can be found
such that when 0 < ε < σ, the eigenvalues of M are all located inside the unity circle except
one of them which is always equal to one. Thus, for 0 < ε < σ, the system described by
(2-40) is stable and converges. The eigenvalue located at one leads the evolution of the
Remark 5: In order to converge to the global optimal, the RICE algorithm requires the
system will form several subsystems and each subsystem will converge to its own optimal
operation point.
In this section, we show that the convergence point of the RICE algorithm is optimal.
Considering the generation limits, the Lagrangian function for this optimization problem
is:
23
n n n
J C Pgen.i Pgen.i Pd .i
i 1 i 1 i 1
n n
(2-48)
i Pgen.i Pgen.i.min i Pgen.i Pgen.i. max .
i 1 i 1
In order to verify the convergence value of (2-40) is the optimal solution, it suffices to
show that the convergence point satisfies the Karush-Kuhn-Tucker (KKT) conditions [30].
J C Pgen.i
i 1,2,..., N : i i 0, (2-49)
Pgen.i Pgen.i *
Pgen .i Pgen .i
n n
*
Pgen .i Pd .i 0, (2-50)
i 1 i 1
*
i 1, 2,..., N : i Pgen .i Pgen.i. min 0, (2-51)
*
i 1,2,..., N : i Pgen .i Pgen.i. max 0, (2-52)
i 1,2,..., N : i , i 0, (2-53)
∗
where . is the optimal solution to the optimization problem. Assume the local demand
is kept constant before the system converges. That is ΔPd = 0. Then, (2-40) becomes
λˆ k 1 W I λˆ k
. (2-54)
E Pˆ k 1 G W I W G E Pˆ k
It has been proven in the previous section that as long as 0 < ε < σ, all the eigenvalues of
M are located inside the unity circle except one, which is always equal to one. And it is
easy to verify that [1 0]T is the eigenvector of M associated with the eigenvalue θ1 = 1. It
T
is also clear that M has 2n independent eigenvectors and λ 0 E Pˆ 0 can be
24
λˆ 0
x1 v 1 x2 v 2 ... x2 n v 2 n , (2-55)
E Pˆ 0
where {x1, …, x2n} are constants and {v1, …, v2n} are the eigenvectors of M. According to
(2-54),
λˆ 1 λˆ 0
M x1 1 v 1 x2 2 v 2 ... x2 n v 2n . (2-56)
Pˆ 1 Pˆ 0
2n
E E
Thus,
λˆ k k k k
x1 v1 x2 v2 ... x2 n v 2n . (2-57)
Pˆ k
1 2 2n
E
Since there is only one eigenvalue (θ1) that is equal to 1 and all of the others are less than
1:
λˆ k 1
lim x1 . (2-58)
k E Pˆ k 0
Pˆi k
n
That is, E will converge to 0. According to (2-25), i 1
Pi k 0 ; thus, (2-50)
is satisfied. This result also indicates that all the ˆi will converge to the same value, which
is the system IC. However, the actual IC for each generation unit depends on the amount
C Pgen.i
First, define ˆ* , the initial value of P*gen.i is the generation
Pgen.i *
Pgen . i Pgen .i
command calculated based on the convergence value of the system IC ˆ * . Then, we will
25
1. Pgen.i.min *
Pgen Pgen.i.max : In this case, by choosing βi = 0 and γi = 0 and ˆ* ,
.i
* * *
2. Pgen .i Pgen.i.min : In this case, Pgen .i is taking as Pgen.i Pgen.i. min . Here,
C Pgen.i
ˆ* .
Pgen.i
Pgen.i Pgen.i .min
By taking 0, ˆ* , and
i
C Pgen.i
ˆ* 0,
i
Pgen.i
Pgen .i Pgen .i .min
* * *
3. Pgen .i Pgen.i.max : In this case, Pgen .i is taking as Pgen.i Pgen.i.max . Here,
C Pgen.i
ˆ* .
Pgen.i
Pgen.i Pgen.i .max
By taking 0, ˆ* , and
i
C Pgen.i
ˆ* 0,
i
Pgen.i
Pgen . i Pgen . i .max
Therefore, for the convergence point of the RICE we can always find a set of α, βi, and
γi for each bus such that the KKT condition can be satisfied. As (2-11) is a convex
problem with affine constraints, satisfaction of KKT conditions translates into the global
optimality [31].
26
2.4 CASE STUDIES
In this section, three case studies are presented to demonstrate the performance of the
proposed RICE algorithm, and its robustness under different information loss rates
The IEEE 9 bus system is chosen as the base case in this chapter. The one-line diagram
of the system is shown in Figure 2-3. It has three generation units and three loads. The
cost functions of the generation units are obtained from [32] and are shown in Table 2-1.
The total system power demand is 850 megawatts (MW). The goal is to determine the
power generation of each generator to minimize the total system generation cost and meet
the demand. By using the centralized quadratic programming algorithm to solve this EDP,
Load C
100 MW
G2 T T G3
8
2 3
7 9
5 6
Load A Load B
200 MW 550 MW
T
1
G1
27
Table 2-1 Generation Unit Parameters
Minimum Maximum
Unit a b c
generation (MW) generation (MW)
1 561 7.92 0.001562 150 600
Now, the RICE algorithm is applied to this system to solve the EDP in a distributed
dashed lines. As proven in Section 2.3, the algorithm will converge to global optimal if the
convergence factor ε is small enough. In this chapter, ε = 0.001 is used. Also, each
controller assigns equal weighting to the local information coming from its neighbors.
Therefore, the non-zero elements in each row of W matrix as described in (2-26) are all
the same and their sum equals to 1. The system simulation result is shown in Figure 2-4. It
can be seen that the system IC estimation of each distributed controller converges to the
same value as the quadratic programming result, and the power balance constraint is
satisfied.
28
(a) RICE System IC convergence
Incremental Cost
9
X: 134
Y: 9.148
8.5
8
50 100 150 200
Iterations
(b) Demand vs Generation
Active Power (MW)
800
600
Demand
400
Generation
200
50 100 150 200
Iterations
Figure 2-4 RICE algorithm simulation: (a) the system IC estimation of each bus and (b)
communications information losses in actual applications. In this part, one case study will
As described in Section 2.3, each pair of the agents will need to exchange system IC
estimation and system power mismatch estimation with each other during each iteration if
there is a communications link between them. In this and the following case studies, we
assume that there is an information loss rate p in all the communication links. The
information loss of packets between agents i and j during the kth iteration of the algorithm
29
P Lij k 0 p (The information is lost),
and
for an information loss scenario. In its problem formulation, a row stochastic matrix P is
used as the updating weighting matrix for the IC of each generation unit and a column
stochastic matrix Q is used as the updating weighting matrix to estimate the power
mismatch:
i k 1 pij j k yi k ,
j Ni
xi k 1 i i k 1 i , (2-59)
yi k 1 qij y j k xi k 1 xi k ,
j Ni
where λi(k) is the estimation of optimal IC of generator i, xi(k) is the corresponding power
generation, yi(k) is the local power mismatch, pij and qij are related elements in P and Q
matrices. If information loss happens to generation unit i, in order to minimize the effect
of information loss, this agent will need to use the information received in the previous
time step. Because of the outdated information, the convergence value of the algorithm
Remark 6: The correction technique described in Section 2.3 may not be able to be
applied here. According to the problem formulation in [19], this technique can only be
applied if each local controller knows the updating weighting assigned by its neighbors.
However, this may be the information the local controllers do not want to disclose to
others.
30
The same test system as described in case study 1 is used here. The simulation result of
Figure 2-5. This result confirms that the consensus-based algorithm is sensitive to
communications information loss since there is a steady state error between total demand
and total generation based on the convergence value of the system IC.
9
X: 146
8.5 Y: 8.906
8
20 40 60 80 100 120 140 160 180 200
Iterations
(b) Demand vs Generation
Active Power (MW)
800
600
Demand
400
Generation
200
20 40 60 80 100 120 140 160 180 200
Iterations
Figure 2-5 Consensus based algorithm simulation with 5% information loss rate; (a) IC of
In the RICE algorithm, when a local controller detects the communications information
loss, the correction technique described in Section 2.3 is applied. The simulation results of
the RICE algorithm under a 5% information loss rate scenario using the same test system
is shown in Figure 2-6. It is clear that the system IC converges to the same global optimal
value as the quadratic programming result, and the power balance constraint is satisfied
31
(a) RICE System IC convergence with 5% information loss
Incremental Cost
9
X: 158
Y: 9.148
8.5
8
20 40 60 80 100 120 140 160 180 200
Iterations
(b) Demand vs Generation
Active Power (MW)
800
600 Demand
Generation
400
200
20 40 60 80 100 120 140 160 180 200
Iterations
Figure 2-6 RICE algorithm simulation with 5% information loss rate: (a) IC of each bus,
Communication Topologies
By comparing Figure 2-4 and Figure 2-6, we can see that the RICE algorithm takes
slightly more iterations to converge when there is information loss. In this case study, the
convergence rate of RICE algorithm under different information loss rates and
(0%, 5%, …, 50%) and three communications topologies (high sparsity, medium sparsity
and low sparsity) are considered. Each case runs 3000 iterations for 20 times. Each
controller assigns equal weighting to the local information received from each of its
neighbors, and ε = 0.001 is chosen. To compare the convergence rates, in each case the
number of iterations required for the system IC estimation to reach a certain neighborhood
of the optimal system IC is used as the convergence index. That is, in each case we record
k such that:
32
n
*
error k i k 1 10 3 , (2-60)
i 1
where n is the number of the buses, λi(k) is the estimated system IC by the ith bus at
The RICE algorithm convergence rate with three different topologies under different
communications information loss rates are shown in Figure 2-7. For each case, the
convergence index, k, as defined by (2-60) is recorded and plotted in a box plot. For each
box, the central mark is the median, the edges of the box are the 25th and 75th percentiles,
and the whiskers extend to the most extreme data points not considered outliers, and the
outliers are plotted individually. The corresponding statistics are also recorded in Table
9 9 9
1
1 1
5
5 5
3
3 3
(a) High sparsity topology (c) Medium sparsity topology (e) Low sparsity topology
900
3000 3000
800
2500 2500
700
500
1500 1500
400
1000
1000 300
500 200
500
100
0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%
Information loss rate Information loss rate Information loss rate
(b) Convergence rate under (d) Convergence rate under (f) Convergence rate under
different information loss rate different information loss rate different information loss rate
Figure 2-7 RICE algorithm convergence rate under different communication information
33
Table 2-2 Convergence Rate Statistics of High Sparsity Topology (Iterations)
Information
loss rate 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%
Statistics
Max 301 379 449 582 694 1008 1253 1675 2525 3000 3000
75% quartile 288 351 418 511 661 890 1192 1581 2284 2979 3000
Median 276 328 399 492 614 845 1099 1473 2100 2798 3000
25% quartile 250 301 375 473 596 794 1034 1416 1971 2633 3000
Min 229 258 330 435 565 737 934 1140 1822 2328 3000
Information
loss rate 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%
Statistics
Max 191 223 303 379 516 609 826 1211 1657 2273 3000
75% quartile 172 205 263 334 445 554 773 1122 1499 2032 2976
Median 162 196 251 303 403 531 706 997 1389 1805 2682
25% quartile 152 184 244 282 377 485 620 916 1096 1533 2473
Min 147 175 208 246 321 409 487 721 763 1266 862
34
Table 2-4 Convergence Rate Statistics of Low Sparsity Topology (Iterations)
Information
loss rate 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%
Statistics
Max 119 116 119 119 145 170 252 294 459 606 875
75% quartile 117 113 108 111 128 148 207 273 391 544 762
Median 116 111 103 101 121 138 177 249 334 497 724
25% quartile 115 106 99 96 113 130 158 220 300 449 677
Min 113 103 95 83 100 110 135 187 248 371 508
Also, for a certain information loss rate, more iterations are required for convergence of
the algorithm if the communication topology connectivity is low. This is because with
higher connectivity, the information propagates faster throughout the network. The ideal
case will be a fully connected communications topology, in which each agent will have a
direct link to every local agent in the system, which will lead to the highest convergence
rate.
2.5 CONCLUSION
In this chapter, the RICE algorithm for the cooperative distributed energy management
of smart grids is proposed to solve the EDP. It eliminates the vulnerability of the
a fully distributed algorithm which works solely based on local information. The
convergence and optimality proofs of the algorithm are provided to guarantee its
convergence to the optimal operation point under the connectivity of the communications
35
topology. The case studies also demonstrate the robustness and performance of the RICE
algorithm under different information loss rates scenarios with different communications
topologies.
36
Chapter 3. ONLINE CONVERGENCE FACTOR TUNING FOR ROBUST COOPERATIVE
Abstract:
Solving the economic dispatch problem (EDP) in a distributed way has attracted lots of
attention in recent years due to its scalability and robustness to single points of failure. The
information losses. However, assuring the stability of the algorithm without knowing the
global information of the system is a challenging issue. This chapter provides a distributed
online approach to tune a certain parameter of the algorithm called the “convergence
factor” using only local information to assure the algorithm is stable. To do this, a local
energy function is defined for each agent. As the algorithm proceeds, each agent uses a
decaying mechanism to tune its convergence factor to ensure that its local energy function
is within a certain bound. The summation of local energy functions represents an energy
function for the entire network. Therefore, if each agent uses the tuning mechanism, the
energy of the system would be forced to be constrained and the system will become stable.
The effectiveness of the proposed approach is verified through several case studies.
3.1 INTRODUCTION
With the increasing number of controllable distributed energy resources deployed and
integrated into the smart grid, how to economically control and manage these resources
will become a challenge for the future smart grid. In this paradigm, the future smart grid
will become a multi-agent system. If we still use the conventional centralized approaches,
the central controller will face high computational and communications burdens.
37
Moreover, the system will be vulnerable to central node failures and have a poor
investigated in the literature as robust and computationally scalable alternatives [12], [16]–
[19], [33].
local agent needs to be purely distributed, which means all the algorithm parameters
should be calculated or tuned only based on its local states or the information received
The classic Economic Dispatch Problem (EDP) is a very basic and essential energy
management problem for the power system. Many consensus and gossip based approaches
have been proposed by researchers in literature to solve this problem in a distributed way.
In all of these algorithms, there is a certain gain that needs to be carefully designed to
ensure system stability and convergence. This nonnegative gain is called by different
names such as convergence factor [17], learning gain [19], and agreement term [18]. In
[19], a systematic method is used to design this gain to assure stability. However, the
method they have proposed requires centralized information and the gain needs to be
redesigned once there is a system topology change. In [18], the authors decay this gain (αt)
at every iteration so that αt→0 as t→∞. The problem of this approach is that, if the system
load changes after a sufficient long time, the algorithm will fail to respond to this change
In this chapter, using the concept of energy functions, we propose a distributed online
approach to tune the convergence factor for distributed economic dispatch algorithms. In
the proposed approach, a local energy function is defined for each agent. As the algorithm
38
proceeds, each agent tunes its local convergence factor to keep its energy function within a
certain bound. By choosing this bound small enough, the summation of the energy
functions behaves as a Lyapunov like function [34] for the entire system and therefore the
stability of the algorithm is guaranteed. We apply the proposed method on the Robust
concept. The effectiveness of the proposed approach is verified through several case
studies. The main contributions of the proposed gain adapting method are:
The rest of the chapter is organized as follows. In Section 3.2, the preliminaries are
introduced. The online convergence factor tuning method is formulated in Section 3.3.
Some representative case studies are presented in Section 3.4. Finally, the concluding
3.2 PRELIMINARY
Assume that each generation unit has the following quadratic cost function:
2
Cgen.i Pgen.i ai bi Pgen.i ci Pgen .i , (3-1)
where ai, bi, and ci are known coefficients and Pgen.i is the output power of the ith
generation unit, respectively. The objective of solving the EDP is to minimize the total
n
min Ci Pgen.i , (3-2)
i 1
39
under the power balance constraint:
n n
Pgen.i Pload .i , (3-3)
i 1 i 1
The incremental cost (IC), λi, of each generation unit is defined as:
C Pgen.i
i . (3-5)
Pgen.i
i bi
Pgen.i . (3-6)
2ci
In a multi-agent consensus network, each agent agrees on the value of a shared variable
among all the agents by exchanging information with its neighbors. This shared value is
called information state xi. The consensus algorithm for single-integrator dynamics can be
xi k 1 xi k aij x j k xi k , (3-7)
j Ni
where k denotes a communications event, η is the step size, aij is the entry (i, j) of the
adjacency matrix A associated with the communications topology, and Ni is the neighbor
set of agent i.
40
xk 1 Wx k , (3-8)
In the gossip algorithm, each agent i exchanges information with at most one of its
xk 1 W' k x k , (3-9)
where W′ is a random updating matrix. Assume the set of the node pairs is denoted by S,
Assume a smart grid has n buses and each bus is equipped with a distributed controller
with two-way communications capability. The distributed controller also has access to the
information of its local devices such as generation cost function and local demand size.
Under this assumption, the RICE algorithm has been proposed in [33] to solve the EDP as
information losses.
The RICE algorithm has two updating layers running in parallel in each distributed
controller: one layer uses the gossip protocol to estimate the system’s average power
mismatch, while the other layer uses the consensus protocol to update the system IC
41
Pi k Pload.i k Pgen.i k , (3-12)
n
Pˆi k 1 Wij ' Pˆj k Pi k 1 Pi k , (3-13)
j 1
where for each agent i, λi is the system IC estimation, ε is the convergence factor, ΔPi is
the local power mismatch, Pˆi is the system power mismatch estimation, Wij is the entry
(i, j) of a row stochastic updating matrix based on consensus protocol, and Wij′ is the entry
Since the RICE algorithm uses random double stochastic matrix for system power
mismatch estimation, we can write equations (3-11) ~ (3-13) into matrix form and take the
expectation of it:
λ k 1 W I λ k
. (3-14)
E Pˆ k 1 G W I W G E Pˆ k
1 1 1
where G diag and W is the expectation of the random matrix
2c1 2c2 2cn
W′. Here, we assume the local demand is kept constant before the system converges.
By using eigenvalue derivation theory [28] and Bauer-Fike theorem [29], it can be
proven that with a small enough ε, the RICE algorithm is stable and converges. It can be
further shown that the convergence state also satisfies the Karush-Kuhn-Tucker (KKT)
conditions [35] of the EDP problem formulated by equations (3-2) ~ (3-4) for optimality.
Thus, with a small enough ε, the RICE algorithm will always converge to the global
optimal as long as the communication topology is connected. For more details about the
42
Remark 1: The definition of IC is required for the RICE algorithm. Therefore, if for a
certain bus i there is no generation unit connected, then the generation cost function
associated with that particular bus is set to a default value while both of the generation
limits are set to 0. In this chapter, the default values are chosen as a = 0, b = 8, and c =
0.001. If multiple generation units are connected to bus i, then they can be aggregated into
one single local generation unit. If multiple loads are connected to one single bus i, then
an actual system, instead of using the same convergence factor, each agent can have its
own ε to keep the system stable and to help it converge optimally. In this section, we will
describe how each agent can tune its convergence factor value online such that the system
will remain stable and converge to the optimal point even with an initially unstable
convergence factor.
Remark 2: As can be seen in (3-14), the dynamic of the RICE algorithm is affected by
the updating weighting matrix and generation cost function parameters as well. However,
the parameters of generation cost function for each agent are not tunable. Moreover, the
updating weightings are chosen according to the communications topology of the system
and changing them online may not be feasible nor desirable. Therefore, the convergence
factor is the most feasible tunable parameter to ensure system stability and convergence.
2
2 Pˆi (k )
Vi k i k j k , (3-15)
j Ni S B.i
43
where Ni is the set of neighbors of agent i, and SB is the base power of the ith bus. In this
chapter, since the generation of the bus without generation unit is zero, the base power for
bus i is defined as max{100, Pgen.i.max}. If the system is stable, for all i, j, lim
→
and lim Δ 0. Therefore, Vi will have a decreasing trend and will converge to
→
zero. Thus, this function can be used as a local indicator of stability by each agent to tune
the convergence factor value online. We propose the following online tuning process:
i k if Vi k
i k 1 i k , (3-16)
if Vi k
i
where ζi > 1 is a decay factor set by each agent to tune the convergence factor value.
Based on (3-16), the convergence factor is decayed whenever Vi(k) violates a certain band
δ. The summation of these local energy functions can be seen as an energy function for the
entire system:
2
N
2
N
Pˆi (k )
V k i k j k . (3-17)
i 1 j Ni i 1 S B.i
So the decaying mechanism in (3-16) is forcing the energy of the system to decrease to
remain within a certain bound. A case study will be shown in Section 3.4 that by using this
method, only based on local information; the system will be kept stable even with an
The advantage of this stabilizing mechanism is that, each local agent is able to tune its
convergence factor value online during the convergence of the algorithm only based on its
local states and the information received from its direct neighbors.
44
3.4 CASE STUDIES
The IEEE 9 bus system is chosen as the base case in this chapter. The one-line diagram
of the system is shown in Figure 3-1. It has three generation units and three loads with a
total system power demand of 850 MW. The corresponding two-way communications
topology is also shown in Figure 3-1 by dashed lines. The cost functions of the generation
units are obtained from [32] and are shown in Table 3-1. The goal is to determine the
power generation of each generator to minimize the total system generation cost and meet
the demand. By using quadratic programming algorithm to solve this EDP, the optimal
Load C
100 MW
G2 T T G3
8
2 3
7 9
5 6
Load A Load B
200 MW 550 MW
T
1
G1
45
Table 3-1 Generation Unit Parameters
Minimum Maximum
Unit a b c
generation (MW) generation (MW)
In this chapter, we assume that each controller assigns equal weighting to the local
information received from each of its neighbors. Therefore, the non-zero elements in each
row of W matrix as described in (3-14) are all the same and their sum equals to one. To
determine the convergence rates, the number of iterations required for the system IC
estimation in each case to reach a certain neighborhood of the optimal system IC is used as
the convergence index. That is, in each case we record k such that:
n
*
error k i k 1 10 3 , (3-18)
i 1
where n is the number of the buses, λi(k) is the system IC by the ith bus at iteration k, and
$/MWh.
In this case study, we will show how the convergence factor value will affect the
topology as shown in Figure 3-1 and gradually increase the value of ε from 5×10-5 to
3×10-3 with an interval of 5×10-5. We run each case up to 5000 iterations for 20 times, and
the convergence index k of each run is recorded and plotted in a compact box plot. For
46
each box, the central black circle is the median, the edges of the box are the 25th and 75th
percentiles, the whiskers extend to the most extreme data points not considered outliers,
The simulation result is shown in Figure 3-2. As we can see from the figure, when the
convergence factor value is too large and beyond a certain region, the system cannot
converge within 5000 iterations. In fact, in these cases the system is unstable and cannot
converge to the global optimal. In our base system, this threshold of the ε value is 0.0027.
When the convergence factor value is too small, it will also increase the number of
ε = 0.0027
Figure 3-2 Convergence rate with different convergence factor ranging from 5×10-5 to
3×10-3
As mentioned before, the change in system updating matrix and generation cost
function of each agent will also affect the performance of the RICE algorithm. Thus, the
convergence factor will need to be tuned online during the convergence process such that
47
even with an unstable initial convergence factor or changes in the system, the algorithm
3.4.2 Case Study 2: Non-negative Function Trajectory with Stable and Unstable
Convergence Factor
In this case study, we will show the trajectory of the local energy function defined in
Section 3.3 with both stable and unstable convergence factors. Based on the results in case
study 1, two different ε values are chosen here for each agent: 1) ε = 0.001, which
The system simulation results with two ε values are shown in Figure 3-3 and Figure 3-4.
In each figure, four plots with respect to iterations are provided: a) System IC estimation
Local energy function trajectory of each agent; d) System power mismatch. As can be
seen from the simulation results, when the system is stable, the system IC estimation of
each agent converges to the same system global optimal value which is 9.14823 $/MWh,
and the local energy function trajectory of each agent is also decreasing and converges to
zero. However, for the unstable case, the system IC estimation of each agent oscillates and
cannot converge to a common value, and the local energy function trajectory of each agent
is also oscillating with a large amplitude compared to the stable case. This confirms that
the local energy function defined in (3-15) can be used by each local agent to tune the
48
(a) Incremental Cost vs. Iteration (b) Generation vs. Iteration
9.5 500
400
9 X: 317
Y: 9.148 300
8.5
200
8
100
7.5 0
0 100200 300 400 100 200 300 400
Iterations Iterations
(c) Local Energy function vs. Iteration (d) Pgen mismatch vs. iteration
40 1000
30
500
20
0
10 X: 356
Y: -7.643e-06
0 -500
0 100 200 300 400 0 100 200 300 400
Iterations Iterations
400
12
300
10
200
8
100
6 0
0 100 200 300 400 100 200 300 400
Iterations Iterations
(c) Local Energy function vs. Iteration (d) Pgen mismatch vs. iteration
40 1000
30
500
20
0
10
0 -500
0 100 200 300 400 0 100 200 300 400
Iterations Iterations
49
3.4.3 Case Study 3: Decaying Factor and Trajectory Bound
As shown in case study 2, when the convergence factor value is large and the system is
unstable, the trajectory of Vi(k) of each agent is oscillating, while in the stable case, the
trajectory is decreasing. Thus, a decaying factor as described in Section 3.3 can be applied
To further analyze the system, the zoomed-in plot of the non-negative function
trajectory as shown in case study 1 with ε = 0.001 is shown in Figure 3-5. As can be seen
from the figure, with a stable convergence factor value, the trajectory decreasing speed of
each agent is different. As a matter of fact, if the bound δ is not chosen properly, we may
not be able to get a desired system performance. If the δ is too small, it may require more
iterations for the convergence factor to stop decreasing, and will lead to a very small
convergence factor value, which will further slowdown the system convergence rate. On
the other hand, if δ is too large, the convergence factor might stop decreasing very quickly
before it falls into the stability region, which will still lead to an unstable system.
Local Energy function vs. Iteration
0.15
0.1
0.05
0
0 20 40 60 80 100
Iterations
proposed in Section 3.3, one simulation case of the same system is presented here with a
50
trajectory bound of δ = 0.05, initial convergence factor of each agent set to be 0.005, and ζ
= 1.2. The simulation results are shown in Figure 3-6. By comparing Figure 3-6 and
Figure 3-4, it is clear that the algorithm converges to the global optimal with an unstable
X: 299 400
10
Y: 9.148
300
9
200
8
100
7 0
0 100 200 300 400 100 200 300 400
Iterations Iterations
(c) Local Energy function vs. Iteration (d) Pgen mismatch vs. iteration
40 1000
30
500
20
0
10 X: 365
Y: -1.879e-05
0 -500
0 100 200 300 400 0 100 200 300 400
Iterations Iterations
In this chapter, an online convergence factor tuning method for cooperative distributed
local energy function is defined for each agent. Each agent tunes the value of ε online to
enforce the local energy function value remain in a certain range to maintain the system’s
stability and convergence to global optimal. Representative case studies are presented to
51
Chapter 4. DISTRIBUTED OPTIMAL GENERATION DISPATCH CONSIDERING
TRANSMISSION LOSSES
Abstract:
problem, the convex DC optimal power flow (DCOPF) problem is used in many studies.
In this chapter, the DCOPF with transmission line losses (DCOPFL) is formulated to
The convergence and correctness of the CDOD algorithm are verified through two
representative case studies. The DCOPFL is also verified to have the smallest
Pij: Active power flow in the branch connecting buses i and j in p.u.
Pijmax: Active power capacity of the branch connecting buses i and j in p.u.
52
QG.imin: Minimum output reactive power of generator at bus i in p.u.
rij, xij: Resistance and reactance of the branch connecting buses i and j
gij, bij: Conductance and susceptance of the branch connecting buses i and j
n: Number of bus
4.1 INTRODUCTION
The optimal power flow (OPF) problem is one of the most important energy
to minimize the total generation cost of a power system subject to power balance
constraints and various transmission system physical constraints, such as bus voltage
limits and line capacities [32]. Because of the nonconvex AC power flow constraints,
solving the AC optimal power flow (ACOPF) problem is a very challenging task. In the
The formulation of the AC power flow can be approximated as a DC power flow under
several assumptions, such as neglecting transmission line resistance and assuming same
voltage magnitude at each bus [32]. As a result, the ACOPF problem can be approximated
as the DC optimal power flow (DCOPF) problem. If the transmission line capacity
constraint is also neglected, the DCOPF can be further simplified as the classic economic
53
dispatch (ED) problem. Compared with ACOPF, DCOPF is a convex optimization
problem that is much faster and easier to set up and to solve; therefore, it is used in many
With the deployment of distributed generation (DG) units into the power system,
distributed approaches have been proposed in the literatures to solve various energy
[19], [33], [37]–[39]. In a distributed energy management paradigm, since the local
information with its neighbors, the system will also be scalable and robust to single points
of failure. As customers sometimes privately own DGs, the distributed approach ensures
customers’ privacy as well. In [37], [38], a distributed approach is proposed to solve the
DCOPF by solving the first order optimality condition in a distributed manner. In [17]–
[19], distributed consensus based approaches are presented to solve the classic ED
algorithm is proposed to solve the ED problem while considering demand side response.
ED only considers power balance and generation limit constraints. In addition to these two
transmission line losses play an important role in the economic operation of the power
system. Neglecting it will result in a less optimal generation dispatch and will pose extra
54
burden to the real-time operation as well as system frequency regulation [40]. The EDL
problem in [39] considers transmission line losses but the transmission line capacity is not
considered.
In order to better approximate the ACOPF, in this chapter, DCOPF with transmission
in a distributed manner. In the CDOD algorithm, each bus of the system is assumed to be
equipped with a local controller that has communications capability to exchange limited
information with its neighbors through a communications network. As a result, the CDOD
algorithm is scalable and also ensures participants’ privacy. Two case studies will be
shown in Section 4.4 to demonstrate the effectiveness of the CDOD algorithm. By taking
also compared with both DCOPF as well as the EDL in terms of the aggregated
approximation error between the result dispatch of each generation unit and the generation
The remaining of the chapter is organized as follows. In Section 4.2, the formulation of
standard ACOPF and DCOPFL problems, as well as the definition of the approximation
error index are presented. The detailed derivation of the CDOD algorithm is described in
Section 4.3. In Section 4.4, two representative case studies are presented. Finally, the
55
4.2 PRELIMINARY
Consider an n-bus power system with several generators and fixed loads connected to it,
and each generator has its operation cost function in quadratic form:
where ai, bi and ci are defined in Table 4-1. Assume each bus possesses at most one
generation unit and one load, the standard ACOPF problem can be formulated as an
optimization problem. Its objective is to find the optimal generation dispatch for each
generator in the system with fixed loads such that the summation of the operation cost of
each generation unit is minimized while satisfying a number of constraints. Usually, these
generation capacity constraint, and 4) bus voltage quality constraint. Mathematically, the
n
min ai bi PG .i ci PG2.i
i 1
s.t. i 1, ..., n
PG .i Pload .i Pij 0, QG .i Qload .i Qij 0, (4-2)
j Ni j Ni
Its nonconvexity mainly comes from the power flow equations that represent the
relationship between the voltage and the complex power drew or injected at each bus. In
56
many types of studies, the AC power flow equations can be approximated by using the DC
power flow equations. The DC power flow equation can be derived under the following
three assumptions:
The resistance rij for any branch ij is negligible compared to the reactance xij,
therefore, set to 0.
The voltage magnitude at each bus is equal to the base voltage, therefore, set as
1.0 p.u.
The active and reactive power flow in the branch connecting buses i and j can be
Pij Vi 2 g ij VV
i j g ij cos i j bij sin i j , (4-3)
Qij Vi 2bij VV
i i g ij sin i j bij cos i j , (4-4)
where gij and bij can be calculated based on rij and xij using equation (4-5) and (4-6)
rij
gij 2
, (4-5)
r
ij xij2
xij
bij 2
. (4-6)
r
ij xij2
Based on the above assumptions, the power flow equations (4-3) and (4-4) can be
approximated as:
i j
Pij , (4-7)
xij
57
Qij 0. (4-8)
In this chapter, we are interested in the DCOPF problem that considers transmission
line losses as well. A well-known method to include transmission line losses into problem
formulation is to make each bus consider half of the losses associated with the branches
connected to it [39], [43]. As a result, the ACOPF represented by equation (4-2) can be
n
min ai bi PG .i ci PG2.i
i 1
s.t. i 1, ..., n
2
i j rij i j
PG .i Pload .i , j Ni . (4-9)
j Ni xij j Ni 2 xij
i j
Pijmax Pijmax , j Ni , i j
xij
PGmin
.i PG.i PGmax
.i
In this optimization problem, the unknowns are PG.i, and δi at each bus i. Note, the
standard DCOPF formulation is similar to the DCOPFL problem, except the first
constraint becomes:
i j
PG .i Pload .i , j Ni . (4-10)
j Ni xij
As EDL, DCOPF and DCOPFL problems are all approximations to ACOPF problem,
an evaluation index is used to quantify the approximation error between the optimal
solution of these three formulations with respect to the optimal solution of the ACOPF.
Denote PG.i* as the optimal solution of the ACOPF in p.u., the aggregated percentage
58
approximation error between the result generation dispatch and the dispatch result given
n
PG*.i PG .i
i 1
n
100% . (4-11)
*
P G .i
i 1
This approximation error index will be used later to evaluate the solution obtained by the
CDOD algorithm.
The cost function and constraints of the optimization problem represented by equation
(4-9) are all convex. Define λi, as the KKT multiplier associated with each equality
constraint at bus i, μij and μji as the KKT multipliers associated with inequality constraint
at each branch connecting bus i and j, and ρ as the penalty factor, the augmented
n n n
L ai bi PG .i ci PG2.i i Pi ij Pij
i 1 i 1 i 1 j Ni , i j
n n n 2
n 2
Pji
2 i 1 j Ni , i j
0,
xmin x xmin
x xmin , xmax
x xmin x xmax . (4-13)
xmax x xmax
59
2
i j rij i j
Pi PG .i PD .i , (4-14)
j Ni xij j Ni 2 xij
Pijmax ,
i j
Pij (4-15)
xij
Pijmax .
j i
Pji (4-16)
xij
The primal variables of the Lagrangian function are PG.i and δi for each bus i, and the dual
variables are λi for each bus i and μij for each branch connecting bus i and j.
Assume each bus is equipped with a local controller with the capability to
communications system topology is the same as the physical system topology. Based on
the primal-dual subgradient method [41], the primal variables (i.e., PG.i and δi ) and dual
variables (i.e., λi and μij) associated to bus i can be updated in an iterative manner using the
L
PG .i k 1 PG .i k 1 k , (4-17)
PG .i PGmin max
. i , PG . i
L
i k 1 i k 2 k , (4-18)
i
L
i k 1 i k 3 k , (4-19)
i
L
ij k 1 ij k 4 k , (4-20)
ij 0,
60
L
Pi , (4-21)
i
L
bi 2ci PG.i i Pi , (4-22)
PG.i
L 1 rij i j 1 rij j i
i 2 j 2
]
i j Ni xij x ij j Ni xij x ij
1
ij ji
j Ni xij
i j
, (4-23)
1 rij i j 1 rij j i
Pi Pj
j Ni xij xij2 j Ni xij xij2
1 1
Pij Pji
j Ni xij j Ni xij
i j 0, i j 0,
L
Pij . (4-24)
ij
The equation (4-17) ~ (4-20) consists of the proposed CDOD algorithm. If each bus i
also has the direct access to the KKT multiplier μij as well as the line resistance and
reactance associated with the transmission line connected to it, the CDOD algorithm only
requires each bus i to have its local and the its direct neighbors’ information. More
specifically, each bus i will only need to have δj, λj, ΔPj and μji from its direct neighbor j,
and all the other information are locally available. As a result, the proposed CDOD
parameters, local generation amount, and local demands at each bus are disclosed, the
61
4.4 CASE STUDY
In this section, a 6-bus system from [32] and a 30-bus system from the MATPOWER
library are used to demonstrate the effectiveness of the proposed CDOD algorithm by
solving the DCOPFL problem. The DCOPF as well as the EDL problems are also solved
for these two test systems using the MATLAB optimization toolbox. Based on the error
index defined in Section 4.2.3, the DCOPF and EDL results are then compared with the
CDOD algorithm’s result using ACOPF result solved by MATPOWER as the reference.
The one-line diagram of the 6-bus system is shown in Figure 4-1. It contains three
generators and three loads. The generator cost function parameters, generator generation
limits, system loading conditions, system branch parameters and branch transmission
4 1
5 1
6 1
62
Table 4-4 Branch Data
1 2 0.10 0.20 1
1 4 0.05 0.20 1
1 5 0.08 0.30 1
3
G3
2
G2
6
1
5
G1
63
Assume the communications topology is the same as the physical system topology. The
CDOD algorithm is applied to solve the DCOPFL for this system with the algorithm
parameters given in Table 4-5. The convergence of Pgen.i, δi and λi of each bus i are shown
in Figure 4-2. As can be seen from Figure 4-2, the algorithm converges in around 1500
iterations. In order to verify the correctness of the converged solution, the MATLAB
fmincon function is used to solve the same DCOPFL problem as a centralized approach.
Both the CDOD solution and the fmincon solution are shown in Table 4-6 for comparison.
As can be seen from Table 4-6, the proposed method converges to the same result as the
centralized method.
ρ η1 η2 η3 η4
1 1.0866 0 1.0868 0
4 0 -0.096299 0 -0.0963
5 0 -0.12284 0 -0.1228
6 0 -0.12008 0 -0.1201
64
Generation (p.u.)
1.5
0.5
Gen1 Gen2 Gen3
0
0 500 1000 1500 2000 2500 3000
Iteration
(rads/sec)
0.2
1 2 3 4 5 6
0.1
-0.1
-0.2
0 500 1000 1500 2000 2500 3000
Iteration
Lambda
1500
1000
500
1 2 3 4 5 6
0
0 500 1000 1500 2000 2500 3000
Iteration
The DCOPFL problem not only can be solved in a distributed manner using the CDOD
algorithm, but also is a better approximation to the ACOPF problem in terms of the
aggregated error between its optimal solution and the ACOPF solution.
The EDL as well as DCOPF problem are solved for this 6-bus system by using
MATLAB optimization toolbox. The ACOPF solution is obtained by using the runopf
function from MATPOWER toolbox. The optimal solution to all of the four problems
(EDL, DCOPF, DCOPFL and ACOPF) are shown in Table 4-7. Using the aggregated
65
percentage error index defined in Section 4.2.3, the percentage aggregated approximation
error of EDL, DCOPF and DCOPFL comparing with the ACOPF results are shown in
Table 4-8. As can be seen from Table 4-8, the DCOPFL gives the best approximation to
In this Section, a 30-bus system is used to demonstrate the CDOD algorithm. The one-
line diagram of the 30-bus system is shown in Figure 4-3. It contains six generation units
and twenty loads. Due to the page limitation, the generator cost function parameters,
generator generation limits, system loading conditions, system branch parameters and
branch transmission capacities are not shown in this chapter, readers are referred to
1
http://www.pserc.cornell.edu/matpower/
66
G G
1 2
15 18
28 3
4 14 19
G
8 12 13
6 7 5
9 11
16 17
20
10
26 G
23
25
G
22 21 24
G
27 29 30
Similar to the previous case study, the DCOPFL problem for the 30-bus system is
solved using CDOD algorithm with the parameters given in Table 4-9. The convergence
of Pgen.i, δi and λi of each bus i are shown in Figure 4-4. As can be seen from Figure 4-4,
the algorithm converges in around 1500 iterations, which is similar to the 6-bus system
ρ η1 η2 η3 η4
20 20 0.0015 0.00001 1
67
Generation (p.u.)
0.8
G1 G2 G13 G22 G23 G27
0.6
0.4
0.2
0
0 500 1000 1500 2000 2500 3000
Iteration
(rads/sec)
0.2
0.1
-0.1
-0.2
0 500 1000 1500 2000 2500 3000
Iteration
Lambda
500
400
300
200
100
0
0 500 1000 1500 2000 2500 3000
Iteration
Similar to the 6-bus case, EDL, DCOPF and ACOPF are also solved and their results
are recorded in Table 4-10. Using the approximation error index defined in Section 4.2.3,
the percentage aggregated error of EDL, DCOPF and DCOPFL comparing with the
ACOPF results are shown in Table 4-10. As shown in Table 4-10, the DCOPFL still gives
the best approximation to ACOPF. In this case, because none of the transmission branches
have reached their capacity, the optimal results of EDL and DCOPFL are the same.
68
Table 4-10 Different optimization problem solution for 6-bus system
4.5 CONCLUSION
In this chapter, the DCOPFL problem is formulated by adding the transmission line
loss effect into the DCOPF formulation. A CDOD algorithm is proposed to solve the
demonstrated and its result is verified with MATLAB fmincon function results. An
approximation error index is defined and the DCOPFL formulation is verified to have the
69
Chapter 5. DAY-AHEAD SMART GRID COOPERATIVE DISTRIBUTED ENERGY
Abstract:
Day-ahead scheduling of generation units and storage devices is essential for the
calculates the dispatch schedule by gathering information from all of the devices.
However, this centralized control structure makes the system vulnerable to single point of
failure and communication failures, and raises privacy concerns. In this chapter, a fully
distributed algorithm is proposed to find the optimal dispatch schedule for a smart grid
with renewable and energy storage integration. The algorithm considers modified DC
power flow constraints, branch energy losses, and energy storage charging and
discharging efficiencies. In this algorithm, each bus of the system is modeled as an agent.
By solely exchanging information with its neighbors, the optimal dispatch schedule of the
conventional generators and energy storage can be achieved in an iterative manner. The
5.1 INTRODUCTION
with the objective being to find the optimal schedule for power generation of generators
and charging/discharging of storage devices, which will minimize the total system
operating cost for the next operation day. This optimization problem is subject to various
constraints related to the system’s physical limitations, such as branch capacity constraints,
70
Conventionally this kind of optimization problem can be solved by utilizing centralized
and particle swarm algorithm [46]. However, as more distributed generators and energy
storage devices are integrated into the system, a centralized energy management scheme
may face certain limitations such as vulnerability to single point of failure and
individual device to disclose certain information (e.g., generation cost function, demand
profile, etc.) to the control center, the privacy issue will concern the system participants as
well.
distributed solutions for energy management. For example, distributed load shedding [47],
distributed economic dispatch [17], [39], [48], and distributed optimal power flow [37],
[49], [50] solutions have been proposed. However, these are all single-step optimization
solutions for the optimal smart grid operation. With energy storage integration, the State
of Charge (SoC) of the energy storage devices will come into play. As a result, these
renewable and energy storage integration. The algorithm is able to find the optimal
schedule of the charging and discharging of energy storage devices, but the branch
capacity, branch energy losses, and energy storage efficiencies of the system are not
approach is proposed to solve the problem in a distributed way. Similar to paper [51], this
71
paper also does not consider branch capacity, branch energy losses and energy storage
efficiencies. Furthermore, it also requires defining quadratic cost functions for all the
controllable devices, which is not a trivial process especially for storage devices. In paper
losses and energy storage efficiencies, but the branch capacities are not considered.
Besides, in the energy management system framework presented in paper [53], each local
controller solves a sub-problem. The solutions of the sub-problems are then sent to the
will also send its result to each of the local controllers. Because the optimal dispatch
schedule is jointly computed by the local controllers and microgrid control center, the
algorithm is not a fully distributed solution. Papers [54]–[57] also work on similar energy
scheduling applications. In paper [54], the energy storage efficiency and system line
capacity are not considered. In paper [55], [56], the system line capacity and energy losses
are not considered, and in paper [57], the energy losses on the transmission lines are
neglected.
In fact, the topology and component characteristics of the system such as branch energy
losses, branch capacity, and energy storage efficiencies play important roles in getting the
shows that by considering branch capacity and losses in the DC optimal power flow
formulation, the optimization will yield to a closer solution to the AC optimal power flow.
for a smart grid with renewable and energy storage integration is proposed. In this
algorithm, each bus of the system is modeled as an agent with communications capability.
72
By only exchanging information with its direct neighbors, the optimal energy schedule can
1. Different from the energy management solution presented in paper [53] that
requires both local and central controllers, our algorithm is a fully distributed
the system participants’ privacy can also be ensured by only allowing limited
reconfiguration;
2. Different from the algorithm presented in paper [52], CoDES algorithm does
not require defining quadratic cost functions for energy storage and PV systems
3. The CoDES algorithm considers branch capacity, branch energy losses and
energy storage efficiencies at the same time, while the problem formulated in
The following of this chapter will be organized as follows. In Section 5.2, a detailed
optimization problem formulation is presented. Section 5.3 explains the derivation process
Section 5.4. Finally, the conclusion of the chapter is discussed in Section 5.5.
For an n-bus power system topology, assume each bus has at most one generator, one
energy storage device, one renewable resource and multiple loads connected to it. The
objective of the problem is to find the optimal dispatch schedule of generators, and the
73
charging/discharging schedule of storage devices to minimize the total generation cost of
T n
min ai bi Pg .i t ci Pg2.i t t, (5-1)
t 1 i 1
where ai, bi and ci are the coefficients of the cost function of the generator connected to the
ith bus, Δt is the time duration, and Pg.i(t) is the power dispatch command of the generator
Remark 1: The communication cost is not considered in this chapter as this will involve
the optimal design of the communications network which relates to the capital cost of the
system, which is currently an active research area in academia [59]. However, the scope of
this chapter is the system operating cost, therefore, the optimal design and the cost of the
At all times, the energy stored in the energy storage devices should be within the
EBmin
.i EB.i t EBmax
.i , (5-2)
where EB.i(t) is the amount of energy stored in the storage device at the beginning of time t,
EB.imin and EB.imax are the minimum and maximum amount of energy that are required for
In an actual system, the charging and discharging commands are sent to the power
conversion devices that are interfacing the storage devices with the grid. The efficiencies
of the energy storage device and the power conversion device as a module makes a big
74
difference when considering the actual power that flows into and out of the storage device.
scenario), respectively, and refer to them as the charging and discharging efficiencies of
Generally, assume the energy stored in a storage device is E(t) at the beginning of time
period t. Then, the energy stored in the storage device at the beginning of the time period
PB.i t
EB.i t t, PB.i t 0
EB.i t 1 i.d , (5-3)
EB.i t P
i .c B .i t t, PB.i t 0
where PB.i(t) is the charging/discharging command of the storage device that is connected
to the ith bus during the time period t. A positive PB.i indicates the storage device is in
t PB.i s 0,
t
EBmin
.i EB.i.0 t i .c PB.i s ,0
t EBmax
.i , (5-4)
s 1 i .d s 1
a x a
x a ,b
x a x b. (5-5)
b x b
It is clear that equation (5-4) is a nonlinear constraint due to the charging and discharging
efficiencies of the energy storage devices. This nonlinear constraint can be linearized
using the linearization method that is widely used in linear programming [60]. Denote two
75
PB.i t if PB.i t 0
PB(.i ) t , (5-6)
0 otherwise
0 if PB.i t 0
PB(.i ) t . (5-7)
PB.i t otherwise
Then the actual command of the storage device will be PB.i(t) = PB.i+(t) - PB.i- (t). As a
t
PB .i s t
E Bmin
.i E B .i .0 t i PB .i s t E Bmax
.i . (5-8)
s 1 i s 1
In this chapter, we name PB.i(+)(t) and PB.i(-)(t) as the virtual discharging and charging
power, respectively.
Theorem 1: Assume PB.i(+)*(t) and PB.i(-)*(t) are the solutions to the optimization
Case 1: PB.i*(t) = PB.i(+)*(t) - PB.i(-)*(t) ≥ 0. In this case, the energy in the storage device
t
PB(.i )* s t
( )*
EB.i t 1 EB.i.0 t i .c B.i P s t. (5-9)
s 1 i .d s 1
PB*.i 1
EB.i t 1 EB.i t t t i .c PB(.i )* t t. (5-10)
i .d i .d
1
Based on (5-10), as i .c 0 , it can be proved by mathematical induction that
i .d
76
energy would be lost which would have negative effect on the cost. Therefore,
Case 2: PB.i*(t) = PB.i(+)*(t) – PB.i(-)*(t) ≤ 0. Using the technique similar to case 1, we can
considered as part of the power balance constraints. Denote the resistance and reactance of
the branch connecting buses i and j with rij and xij in per unit, and the voltage amplitude
and phase angle of the ith bus with Vi and δi, the per unit active power flow on the branch
Pij Vi 2 gij VV
i j g ij cos i j bij sin i j , (5-11)
where
rij
g ij 2
, (5-12)
r ij xij2
xij
bij 2
. (5-13)
r ij xij2
Due to the nonconvex nature of equation (5-11), solving the optimization problem will
be NP hard under the worst case scenarios [36]. As a result, in many of power system
studies the AC power flow equation are usually approximated with DC power flow
equation to convexify the optimization problem [32]. In this chapter, the nodal power
balance constraint is also derived based on the DC power flow equations that can be
77
1. The resistance rij for any branch ij is negligible compared to the reactance xij;
therefore, set it to 0,
2. The voltage magnitude at each bus is equal to the base voltage; therefore, set it
to 1.0 p.u.,
library [42], we have found that the system branch r/x ratio may not be small enough to be
approximation. Therefore, in this chapter, only the second and third assumptions are
considered for the problem formulation. Similar approximation results are also reported in
[62], [63] indicating that the DC power flow formulation is an appropriate approximation
for the type of energy management application that is considered in this chapter. For real-
time power system operation, AC power flow equations are more appropriate, but the real-
xij
Pij 2 i j . (5-14)
r
ij xij2
Thus, the nodal power balance constraint considering branch energy losses is formulated
as:
xij i t j t
2 2
Pg .i t PR.i t PB(.i ) t PB(.i ) t
j Ni r ij x
ij
2 (5-15)
rij xij i t j t
PD.i t 2 2
,
j Ni 2 r ij x ij
78
where PR.i(t) and PD.i(t) are the forecast renewable generation and demand during time
step t at the ith bus in per unit and Ni is the set of buses that have a branch connected to bus
i. Each bus also considers half of the losses associated with the branches connected to it
[39], [43], [58], such that the system branch energy losses can be included in the
formulation.
xij i t j t
Pijmax 2 2
Pijmax , j Ni , i j (5-16)
r x
ij ij
,
where Pijmax is the power limit that can flow on the branch connecting bus i and j in per
unit.
Pgmin
.i Pg .i t Pgmax
.i , (5-17)
Pgmin
.i Pg .i t Pgmax
.i , (5-18)
0 PB.i t PBmax
.i , (5-19)
where Pg.imin and Pg.imax are the lower and upper generation limit of the generator
connected to the ith bus. PB.imax is the maximum charging/discharging rate of the energy
i t . (5-20)
1 t 0. (5-21)
79
5.3 COOPERATIVE DISTRIBUTED ENERGY SCHEDULING ALGORITHM
The previous section formulates an energy scheduling problem for a smart grid. In this
optimization problem, the control variables are Pg.i(t), PB.i(+)(t), PB.i(-)(t), and i(t) for 1 ≤ ∀t
≤ T. To solve the optimization problem formulated in the previous section, first the
T n T n
L ai bi Pg .i t ci Pg2.i t t i t Pi t
t 1 i 1 t 1 i 1
T n T n
1i t Pi1 t 2i t Pi 2 t
t 1 i 1 t 1 i 1
T n
ij t Pij t
t 1 i 1 j Ni , i j
T n T n (5-22)
2
ji t Pji t Pi t
t 1 i 1 j Ni , i j 2 t 1 i 1
T n 2 T n 2
Pi1 t Pi 2 t
2 t 1 i 1
0,
2 t 1 i 1
0,
T n 2 T n 2
Pij t Pji t ,
2 t 1 i 1 j Ni , i j
0, 2 t 1 i 1 j Ni , i j
0,
where λi, ξ1i, ξ2i, μij and μji are the KKT multipliers, ρ is the penalty factor, and:
xij i t j t
Pij t 2 2
Pijmax , (5-23)
r ij x ij
xij j t i t
Pji t 2 2
Pijmax , (5-24)
r ij x ij
2
xij i t j t rij xij i t j t
Pi t
j Ni rij2 xij2 j Ni 2 rij2 xij2 , (5-25)
PD.i t Pg .i t PR.i t PB(.i ) t PB(.i ) t
t
PB(.i ) s t
( )
Pi1 t EB.i.0 EB.i.max t P
i .c B.i s t , (5-26)
s 1 i .d s 1
80
t
PB(.i ) s t
( )
Pi 2 t t P
i .c B .i s t E B .i.0 EB .i. min . (5-27)
s 1 i .d s 1
By taking the derivative of L with respect to Pg.i, PB.i, i and the KKT multipliers for all
time step t, the distributed updating rules of the CoDES algorithm can be derived by
k 1 k L
i t i t 1 k
t , (5-28)
i
L
Pgk.i 1 t Pgk.i t 2 t , (5-29)
Pgk.i
Pgmin
.i , Pgmax
.i
k 1 k L
PB.i t PB.i t 3 k
t , (5-30)
PB.i PBmax , PBmax
.i .i
k 1 k L
PB.i t PB.i t 3 k
t , (5-31)
PB.i PBmax , PBmax
.i .i
k 1 k L
i t i t 4 k
t , (5-32)
i ,
k 1 k L
ij t ij t 5 k
t , (5-33)
ij 0,
k 1 k L
2i t 2i t 6 k
t , (5-34)
2i 0,
k 1 k L
1i t 1i t 6 k
t , (5-35)
1i 0,
where η1 to η6 are the updating gains of the algorithm that can be adjusted by the user
based on different system topologies. The subscript i in the equations (5-28) ~ (5-35)
81
represents that they are the updating equations that the ith agent will execute. The
The equations (5-28) ~ (5-35) form the proposed CoDES algorithm. In this algorithm,
each bus of the system is modeled as one agent. Each agent has an embedded controller
installed that has local computation power as well as the communications capability to
exchange information with its neighbor agents. As a result, the proposed CoDES
Assume the system communications topology is the same as the system physical
topology. Figure 5-1 shows the flowchart that describes the execution of the CoDES
Start
Wait
No
Received all messages?
Yes
Local update based on equations (28) – (35)
No
Converge?
Yes
End
Remark 2: In this chapter, the cost function parameters of generators (ai, bi, ci, Pg.imin
and Pg.imax), local generation forecast (PR.i), local demand forecast (PD.i), and local energy
82
storage parameters (EB.imin, EB.imax, κc, κd, and PB.imax) are considered as private information
of the system components as they describe their key characteristics. Based on equations
(5-28) ~ (5-35), each bus i only needs to have δj, λj, ΔPj and μji from its direct neighbor j
when the algorithm is executing, and all the other information is locally available. Thus, in
contrast to the centralized management schemes in which the central controller will need
all the information from each of the agents, the CoDES algorithm also ensures the privacy
Remark 3: In this chapter, each bus of the system is modeled as an agent to show the
generality of the algorithm. Problems on how to partition the system in several groups and
execute the distributed algorithm among the groups are also being studied in literature [64].
However, since privacy is also one of the concerns in this algorithm, more cost benefit
studies should be made on how to partition the system without jeopardizing the privacy of
To illustrate the performance of the algorithm, the CoDES algorithm is applied to a 30-
bus system, and its results are benchmarked against the centralized method results. Figure
5-2 shows the one-line diagram of the 30-bus system topology considered in this case
study. It has six generators, five battery storage systems, three PV systems and twenty
loads. The system parameters are obtained from MATPOWER [42] library case 30. The
battery parameters with initial states are shown in Table 5-1. In this table, the parameters
are shown in per unit with a base power of 100MW. For example, 0.3 per unit battery
capacity equals to 30MWh. The generator parameters, the branch parameters and branch
83
energy-transfer capacities are not included in this chapter. Please refer to MATPOWER
case 30 for detailed system parameters. In this chapter, both the charging and discharging
G1 G2
1 2 PV2
15 18
28 3
4 14
B1 19
G3
8 12 13
6 7 5
PV1
9 11
16 17
10 B2 20
26 PV3
G4
23
25 B4
22 G5
B3
21 24
G6
27 29 30 B5
The other inputs to the algorithm are the forecast 24-hour demand, and PV generation
profiles at each corresponding bus with 1-hour resolution, meaning that in this case study
Δt is set to 1 hour. Figure 5-3 shows the aggregated demand, aggregated PV generation,
84
and net aggregated demand (which is the difference between aggregated demand and
aggregated PV generation) profiles of the system in per unit. In this chapter, the actual
demand and PV generation profiles at each corresponding bus have the same shape but
(a)
4
0
0 5 10 15 20 25
(b)
0.5
0
0 5 10 15 20 25
(c)
4
0
0 5 10 15 20 25
Time (hour)
Figure 5-3 Aggregated forecasted profiles in per unit: (a) aggregated demand profile; (b)
Assume the system communications topology is the same as the physical system
topology. The proposed CoDES algorithm can be applied to solve the scheduling problem
with the algorithm parameters given in Table 5-2. While the algorithm is running, each
bus only needs to exchange information with its direct neighbors. For example, bus 1 only
needs to exchange information with buses 2 and 3 during the algorithm execution.
ρ η1 η2 η3 η4 η5 η6 Δt
85
In order to verify the optimality of the convergence states of the algorithm, the
optimization problem as described in Section 5.2 is also implemented and solved by using
MATLAB fmincon function. The interior-point algorithm is used for the fmincon function
and its result is used as the centralized reference for evaluation. The simulation in this
chapter is done on a laptop computer with a 2.5GHz quad-core Intel i7 processor with
6MB shared L3 cache. It takes 2.5 minutes to finish the 8000-iteration execution. Figure
5-4 shows the objective function values of both fmincon result (green line) and the CoDES
algorithm result (blue line). It is clear from the figure that the proposed distributed
algorithm converges to the same optimality as the centralized algorithm. Based on this
figure, we can also see that the algorithm converges within around 2000 iterations. For
detailed description about how to use the fmincon function, please refer to the MATLAB
help document.
4
x 10
3
2.5
2 Algorithm Objective
Global Optimum
1.5
0.5
0
0 1000 2000 3000 4000 5000 6000 7000 8000
Iterations
Figure 5-4 Convergence of distributed energy scheduling objective value vs. global
optimum
Figure 5-5 shows the 24-hour dispatching schedule of the six generators in the system
in per unit. Based on the generator cost function parameters, the generation cost of
generators 1, 2 and 4 are relatively cheaper than that of generator 3, 5, and 6 when the
86
generation amount is low. As a result, generators 1, 2 and 4 are the first to be dispatched
during the first few hours when the demand is low. As the demand increases, the generator
3, 5, and 6 starts to inject power into the system to support the load.
(a) (b)
0.5 1
0.5
0 0
0 5 10 15 20 25 0 5 10 15 20 25
(c) (d)
0.2 0.4
0.1 0.2
0 0
0 5 10 15 20 25 0 5 10 15 20 25
(e) (f)
0.2 0.4
0.1 0.2
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time (hour) Time (hour)
Figure 5-5 Generator schedule in per unit: (a) Generator 1 schedule; (b) Generator 2
Figure 5-6 shows the 24-hour Locational Marginal Price (LMP), which is also the KKT
variable λ, at bus 4 as an example. The LMPs at the other buses have similar patterns. By
comparing Figure 5-3(c) and Figure 5-6, it can be seen that the LMP has a positive
correlation with the amount of the total demand at each hour. A lower demand
0
0 5 10 15 20 25
Time (hour)
87
Figure 5-7 (a)-(e) shows the 24-hour schedule in per unit of the five battery storage
devices that are installed in the system. As can be seen from Figure 5-7, during the time
when the net demand is low such as hours 2 to hour 5 (which also corresponds to a low
LMP) the energy storage devices are being charged to store energy for later use. As the net
demand increases, the battery storage devices are being discharged to help the generator
(a) (b)
0.2 0.2
0 0
-0.2 -0.2
0 5 10 15 20 25 0 5 10 15 20 25
(c) (d)
0.1 0.2
0 0
-0.1 -0.2
0 5 10 15 20 25 0 5 10 15 20 25
Time (hour)
(e)
0.2
-0.2
0 5 10 15 20 25
Time (hour)
Figure 5-8 shows the convergence of the dispatch decisions for generators for the first
responses.
88
Stage 1 Generation (p.u.)
0.8
Gen1
0.7
Gen2
0.6 Gen3
Gen4
0.5 Gen5
Gen6
0.4
0.3
0.2
0.1
0
0 1000 2000 3000 4000 5000 6000 7000 8000
Iteration
board controllers to test the performance. Beaglebone Black board is chosen to build the
platform. The Beaglebone Black board is a single board computer that running a Debian
Linux distribution. It equips a 1GHz ARM Cortex processor and 512 MB RAM. TCP/IP is
used to implement the communications among the boards through a LAN. For a case with
three renewables, three loads and three batteries, the implemented algorithm takes about 3
minutes to calculate the optimal dispatch for the next 24 hours with a 1-hour resolution.
Remark 5: The communications latency between the agents are also considered in the
hardware implementation. Each agent keeps a data log that records the time stamp
(iteration) of the message it receives from neighbors. The agent will not execute the local
updating function until all the messages from its neighbors for that iteration are received.
Another scenario is communication packet loss. In one of our previous papers [48], we
have used the gossip algorithm to handle the packet loss among the agents.
89
5.4.2 Energy Storage Efficiency Effect
In this section, we will discuss what might happen if the energy storage charging and
One serious consequence for not considering energy storage charging and discharging
efficiencies is the possibility of over discharge of the battery during actual operations.
Based on the battery charging/discharging schedules that are calculated in Section 5.4.1
and the battery initial states from Table 5-1, the actual SoC of the batteries as a function of
time are plotted in Figure 5-9 (a)-(e). In each subplot of Figure 5-9, the two dotted red
lines represent the upper and lower bounds of the desired SoC as specified in Table 5-1.
As can be seen from the figure, the actual SoC of each battery always falls within the
desired boundary.
(a) (b)
1.5 1.5
1 1
0.5 0.5
0 0
0 5 10 15 20 25 0 5 10 15 20 25
(c) (d)
1.5 1.5
1 1
0.5 0.5
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time (hour)
(e)
1.5
1
0.5
0
0 5 10 15 20 25
Time (hour)
efficiency
90
In this case, the same test case as described in Section 5.4.1 is simulated, except that
the battery charging/discharging efficiencies are all set to be equal to one. However, as a
physical constraint, the charging/discharging efficiencies do exist for each battery module.
If we still assume the actual efficiencies of the battery are equal to 0.9, the actual SoC of
each battery is numerically calculated and plotted in Figure 5-10. As can be seen from the
figure, at some points the SoC falls below the minimum SoC limit and even becomes
negative. In actual operations, this means that without considering the energy storage
discharge. Moreover, once the battery is not able to follow the scheduled dispatch
(a) (b)
1.5
1 1
0.5
0
0
-1 -0.5
0 5 10 15 20 25 0 5 10 15 20 25
(c) (d)
1.5 1.5
1 1
0.5 0.5
0 0
-0.5 -0.5
0 5 10 15 20 25 0 5 10 15 20 25
Time (hour)
(e)
1.5
1
0.5
0
-0.5
0 5 10 15 20 25
Time (hour)
charging/discharging efficiency
91
5.4.3 System Configuration Change
reconfigurations. Let us take the 30-bus system as shown in Figure 5-2 as an example. If
another generator is going to be installed at bus 8, only the program installed at bus 8
needs to be updated to reflect the change, while the programs in all the other agents will
remain intact. The scenario of removing one generator is similar. Similarly, if one more
branch is installed to connect bus 8 and 9, only the programs installed at buses 8 and 9
need to be updated while the programs in all the other agents would not change. In
addition, these changes can also be done online automatically as the algorithm is running.
In this section, we are showing a case that demonstrates the capability of the CoDES
algorithm to adjust its operation automatically when the system configuration changes. In
this case, the generator installed at bus 22 is removed from the system at the 4000th
iteration as the algorithm is running. Instead of stopping the system and reconfiguring the
algorithm, the CoDES is able to automatically adapt this system configuration change and
The algorithm objective value for this case is shown in Figure 5-11. As can be seen
from the figure, before the system configuration change (at the 4000th iteration) happens,
the system has already converged to the optimal state and the algorithm objective value is
$4499, which is also the same as Figure 5-4. After the generator at bus 22 is removed, the
program installed at the 22nd bus automatically adjusts this configuration change and the
CoDES converges to the new system optimal states, which is $4898. Similar to the results
shown in Section 5.4.1, the new optimization problem with one generator removed is also
92
solved by using the fmincon function. The fmincon function result is also used as the
4
x 10 Objective value ($)
3
Algorithm Objective
Global Optimum
2
X: 2958 X: 6596
1
Y: 4499 Y: 4898
0
0 1000 2000 3000 4000 5000 6000 7000 8000
Iterations
Figure 5-11 Algorithm objective value with one generator removal at 4000th iteration.
Remark 6: Besides the physical system reconfiguration, another aspect of the system
configuration change is the communications system topology change. Once this scenario
happens, the communications system and the physical system no longer have the same
topology. That means extra measures need to be taken to ensure the information sent by
one agent is received by the other agent through another route. This will require each
agent to maintain a “routing table” to direct the correct information flow. The design of
this mechanism is beyond the scope of this chapter. In this chapter we only consider that
there will be a communication network to deliver the required information from agent i to
agent j during each iteration base on the iterative updating equations derived in (5-28)-
(5-35). For the CoDES algorithm, as long as the required information can be delivered
during each iteration, the algorithm will operate as it is supposed and converge to the
5.4.4 Validate the Convergence of the Algorithm using Monte Carlo Simulation
random scenarios are generated and the algorithm objective values for each case are
93
compared with the objective value calculated by using the centralized fmincon function.
The CoDES algorithm considers the physical system topology and line parameters. In
order to make the random scenarios make sense to actual physical system, the system
topology is not changed and the same 30-bus system is used in this simulation. For each
1. Randomly select parameters for each of the batteries in the ranges shown in Table
5-3.
2. Use the PV profiles in Section 5.4.1 and scale them for each PV system based on
3. Randomly generate demand profiles based on the Ppeak randomly selected in the
4. Apply the CoDES algorithm as well as the fmincon function on the system and
record the objective values. Stop the CoDES algorithm when the summation of L1
Table 5-3 Range of random parameters for Monte Carlo simulation (p.u.).
Figure 5-12 shows the histogram of the percentage difference of the objective values
between the centralized fmincon function and the proposed CoDES algorithm in 50
random scenarios. The maximum percentage difference is less than 0.01%. This result
94
indicates that there is no significant difference between the proposed CoDES algorithm
Frequency
5.5 CONCLUSION
In this chapter, a novel CoDES algorithm is proposed that can optimally schedule the
distributed way to minimize the total system day-ahead operating cost. Comparing to
existing literatures that solve similar problems, the CoDES is a fully distributed solution
and considers system branch power-transfer capacity, branch losses and energy storage
efficiencies in the formulation. By modeling each bus of the system as one agent, each
agent only needs to exchange limited information with its direct neighbors, which also
state of the algorithm is benchmarked against the centralized MATLAB fmincon function
to verify the optimality. The convergence of the CoDES algorithm is also verified through
a Monte-Carlo simulation.
95
The test case considered in this chapter is a transmission system, however, the
microgrid with renewable and energy storage integration [65]. The application of the
96
Chapter 6. COOPERATIVE DISTRIBUTED ENERGY SCHEDULING FOR MICROGRID
Abstract:
management for microgrids. The CoDES-ReD algorithm has two major functionalities: 1)
determine the optimal energy storage dispatch schedules based on the day ahead predicted
profiles; and 2) adjust the scheduled set points to compensate for the prediction errors
arising from the intermittencies of the renewables and real-time load patterns. In this
information with neighbors, the dispatch command of the energy storage device installed
at each bus can be calculated in an iterative manner. The performance of the algorithm is
6.1 INTRODUCTION
Energy management is one of the most important applications for a power system to
maintain its economic and efficient operations. Energy management algorithms solve
microgrid active energy management algorithm which does the followings: 1) distributed
97
With the development of smart grid technologies and the integration of distributed
generation and energy storage devices, there is an increasing interest in the literature on
the distributed energy management approaches for microgrids [17], [37], [39], [47]–[57],
have advantages such as robustness to single point of failure and communication failure.
In references [17], [37], [39], [47]–[50], different single time-step distributed energy
[47]. Several different distributed economic dispatch algorithms are proposed in [17], [39],
[48]. Distributed optimal power flow algorithms are proposed in [37], [49], [50]. One
major drawback of these single time-step algorithms is their inability to consider energy
storage into problem formulation. With energy storage integration, the State of Charge
(SoC) of the energy storage devices will come into play, and decisions for different time
steps would be tied together. As a result, these single time-step optimization solutions may
algorithms are proposed. All of these energy scheduling algorithms have the objective of
minimizing the system operating cost. However, one of the important issues of the energy
scheduling in the above mentioned algorithms is that the predicted/forecasted demand and
generation profiles are used for the planning process. In fact, due to the intermittency
nature of renewable resources and the actual patterns of the demands, the forecasts may
not be 100% accurate. If the scheduled dispatch commands are not adjusted based on the
actual system operating conditions, the system may not operate economically. Indeed,
most of the above mentioned distributed scheduling algorithms do not have a mechanism
98
to re-dispatch the scheduled dispatch commands based on the actual system status, except
[52]. In [52], a receding horizon approach is used to continuously update the system
prediction. But [52] does not consider the efficiencies of the energy storages. Moreover, it
also requires defining quadratic cost functions for all the controllable devices, which is not
(CoDES-ReD) algorithm for microgrid active energy management is presented. Figure 6-1
Real-time Re-dispatch 1
9
21
23 24
32
5 33
system status 22
14 15 16 17
Operation schedule
System
Real-time
performance
system Status
measures Cooperative
Distributed Energy
System
Scheduling
constraints
Data Analytics and External
Forecasted demand Management information
& generation
commands of the energy storage devices in the system based on the forecast
demand and renewable generation profiles 24 hours before the operating day
99
ReD algorithm that acts as a tuning loop and adjusts the dispatch commands by
The remainder of the chapter is organized as follows: In Section 6.2, the detailed
optimization problem formulation is presented. Section 6.3 presents the proposed CoDES-
demonstrate the performance of the algorithm. Finally, the conclusion of the chapter is
general example. Except for the grid interface bus (substation), we consider each of the
other n−1 buses be connected with an energy storage device, a renewable energy
generation unit, and a demand. One example of this is to consider each bus as a residential
house which installs a battery storage, a rooftop PV panel, and household appliances. If
one of the n-1 buses does not have one or more of these three components, the
corresponding parameters (defined later in this section) in the algorithm will be set to zero.
Figure 6-2 shows an example of a 4-bus system. In this chapter, we use the terms node,
Bus 1 Bus 2
Storage 1
Grid
PV 1
Load 1
PV 2
Storage 3
Load 2
Load 3
Bus 4 Bus 3
100
6.2.1 Day-ahead Energy Scheduling
the total electricity bill of the microgrid for the next operating day. The major inputs to
this optimization problem are: 1) forecasted 24-hour renewable generation and demand
profiles with 1-hour resolution; 2) utility electricity price tariff; and 3) energy storage
parameters (e.g., capacity, power rating, efficiency, etc.). The decision variables of the
optimization problem are the grid power drawn/injection for the grid interface
and the charging/discharging command for each of the energy storage devices . for
each hour for the next operating day. The remaining of this subsection formulates this
optimization problem.
The objective of the energy scheduling is to minimize the total electricity cost during a
time period T (24 hours) for the n-bus grid connected microgrid. The objective function
T
min C t , (6-1)
t 1
where
and are the electricity price and customer surplus power sell back price
during time step t, and is the grid power during time step t. Δt is the period
101
As it can be seen from (6-2), the objective function is nonlinear based on different signs
Pgrid t , if Pgrid t 0
Pgrid t , (6-3)
0, otherwise
0, if Pgrid t 0
Pgrid t , (6-4)
Pgrid t , otherwise
T
min pbuy t Pgrid t t psell t Pgrid t t . (6-6)
t 1
Similar to the proof included in Section 5.2.1, it can be proved that at the optimal
6.2.1.2 Constraints
1. Battery capacity
Eimin
.B Ei.B t Eimax
.B , (6-7)
where . and . are the minimum and maximum limits of the energy that are
allowed to be scheduled during the scheduling, and . is the energy stored in battery
During the charging and discharging, the energy stored in the battery has the following
dynamics:
102
Pi.B t
Ei.B t t, Pi. B t 0
Ei.B t 1 i .d , (6-8)
Ei.B t P
i .c i . B t t , Pi. B t 0
where . and . are the charging and discharging efficiency of the battery connected to
the ith bus, respectively. . is the charging/discharging power of the battery, and B
As it can be seen from (6-8), the battery energy dynamics is also nonlinear. Similar
linearization approach used in (6-3) and (6-4) can be used here by denoting two new
variables . and . , and the actual command of the energy storage device will be:
t
Pi. B s t
Eimin
.B Ei. B.0 t P
i .c i . B s t Eimax
.B . (6-10)
s 1 i .d s 1
Similar to the proof included in Section 5.2.1, it can also be proved that at the optimal
n n n
Pi.D t Pi.B t Pi.B t Pi.R t Pgrid t Pgrid t , (6-11)
i 1 i 1 i 1
where . and . are the power generation of the renewable unit and the power
consumption of the demand connected to the ith bus during time step t respectively. R
denotes the set of renewable generation units, and D denotes the set of demand units. Note
that if we specify the 1st bus as the grid interface, then . , . , . , and
. will all be 0.
3. Power rating
103
0 Pi.B t , Pi.B t Pi.max
B , (6-12)
max
0 Pgrid t , Pgrid t Pgrid , (6-13)
where . and are the maximum battery charging and discharging power, and
maximum grid power, respectively. Thus, the day-ahead energy scheduling problem is
programming problem. This optimization problem will be solved every Δτ minutes (e.g., 5
three different objectives: 1) reducing the electricity cost; 2) tracking the grid schedule
calculated by day-ahead energy scheduling; and 3) tracking the battery dispatch schedule
calculated by the day-ahead energy scheduling. The decision variables of this optimization
problem are the grid power draw/injection for the grid interface and the
charging/discharging command . for each of the energy storage device for each time
step with a Δτ-minute interval. The remaining of this subsection will formulate this
optimization problem.
by day-ahead energy scheduling for time step τ, the objective function of the real-time
104
* 2
min Pgrid p Pgrid Pgrid
n
2
, (6-14)
Pi.B ( ) Pi.*B ,
i 1
∗
where is the utility electricity price. If 0, then is set to be equal to
6.2.2.2 Constraints
1. Power balance
n n n
Pi.B Pi.R Pgrid Pi.D , (6-15)
i 1 i 1 i 1
where . and . are the renewable generation power and demand sampled at the
2. Power rating
max max
Pgrid Pgrid Pgrid , (6-16)
Pi.min'
B Pi.B Pi.max'
B , (6-17)
where
Ei.B t Eimax
.B
Pi.min'
B max Pi.max
B , . (6-19)
i .c
105
6.3 COOPERATIVE DISTRIBUTED ENERGY SCHEDULING WITH RE-DISPATCH
In this section, the proposed CoDES-ReD algorithm is presented. The two problems
T
L pbuy t Pgrid t t psell t Pgrid t t
t 1
T n T n T
t P t 1i t P1i t 2i t P2i t (6-20)
t 1 i 1 t 1 i 1 t 1
T T T
2 2 2
P t P1i t P2i t ,
2 t 1 2 t 1 2 t 1
where λ, μ1i and μ2i are the KKT multipliers, ρ is the penalty factor, and:
n n n
P t Pi.D t Pi.B t Pi.B t Pi.R t Pgrid t Pgrid t , (6-21)
i 1 i 1 i 1
t
Pi. B s t
P1i t Ei. B .0 t P
i .c i . B s t Eimax
.B , (6-22)
s 1 i .d s 1
t
Pi . B s t
P2 i t Eimin
.B Ei . B .0 t P
i .c i . B s t. (6-23)
s 1 i .d s 1
Note that the local power rating constraints are not included in the Lagrangian as they
can be treated as the domain of each decision variables. As the objective function is
convex and the constraints are affine, the optimal solution is the saddle point of the
. and the KKT multipliers for all time step t, the following iterative equations can be
derived to find the saddle point by utilizing the primal-dual gradient descent method [41].
106
k 1 k L
Pgrid t Pgrid t 1 k
P grid t max
0, Pgrid , (6-24)
k k k
P
grid t 1 pbuy t t t P t max
0, Pgrid
k 1 k L
Pgrid t Pgrid t 1 k
P grid t max
0, Pgrid , (6-25)
k k k
Pgrid t 1 psell t t t P t max
0, Pgrid
L
Pi.B k 1
t Pi.B k t 2 k
Pi.B t 0, Pi max
.B
k k
t P t
1 T
k 1 T
k
1i l t 2i l t , (6-26)
i .d l t i .d l t
Pi.B k t 2 1 T
k
t P1i l
0,
i .d l t
1 T
k
t P2i l
0,
i .d l t
0, Pi max
.B
L
Pi.B k 1
t Pi.B k t 2 k
Pi.B t 0, Pimax
.B
k
t Pk t
T T
k k
i .c 1i l t i .c 2i l t , (6-27)
l t l t
Pi.B k t 2 T
i .c t P1i k l
0,
l t
T
i .c t P2ik l
0,
l t 0, Pimax
.B
k 1 k L k k
t t 3 k
t 3 P t , (6-28)
t
107
k 1 k L k k
1i t 1i t 4 k 1i t 4 P1i t , (6-29)
1i t 0,
0,
k 1 k L k k
2i t 2i t 4 k 2i t 4 P2i t , (6-30)
2i t 0,
0,
a, if x a
x [ a ,b ]
x, otherwise , (6-31)
b, if x b
ξ1 to ξ4 are the updating gains of the algorithm that can be adjusted by the user based on
different system configurations. The subscript i in the equations (6-24) ~ (6-30) represents
that they are the updating equations that the ith agent will execute. The superscript k
denotes the index of iteration. Please refer to [41] for more details on the convergence of
Based on (6-20), we can see that the partial derivative terms in (6-24) ~ (6-30) involves
the global variables and Δ . That means in order for each agent to update
to have access to these global variables. However, in a distributed system, it is not likely
that each of the agents will have access to them. So, a consensus algorithm [67] is used for
ˆk 1
t ˆk t ˆk t ˆk t Pˆi k t ,
i i ij j i 3 (6-32)
j Ni
Pˆi k 1
t Pˆi k t ij Pˆj k t Pˆi k t Pi k 1
t Pi k t , (6-33)
j Ni
108
k k k k k
Pi t Pi. D t Pi. B t Pi.B t Pi. R t Pi. grid t Pi. grid t . (6-34)
the day-ahead scheduling problem can be solved in a fully distributed way. We call this
will all be zero. Similarly, if the agent i is a household node, . and . will
be zero. For more analysis on CoDES algorithm performance and optimality, readers are
referred to our previous paper [66] for details. Readers are also referred to paper [67] for
In this subsection we will show that the converged states of the CoDES algorithm
∗ ∗ ∗ ∗
, , . . for each time step are the optimal solution of the
First, form the Lagrangian of the optimization problem by considering all the
constraints:
109
T
L pbuy t Pgrid t t psell t Pgrid t t
t 1
T n T n T
t P t 1i t P1i t 2i t P2i t
t 1 i 1 t 1 i 1 t 1
n T n T
The KKT conditions of the optimality imply that at the optimal state, for ∀1 ,
* * * * * * * *
there exist real values t , 1i t , 2i t , g1 t , g2 t , g1 t , g2 t , 1i t ,
* * *
2i t , 1i t , and 2i t for all i such that the following conditions hold:
1. Derivatives of the Lagrangian with respect to the primal variables shall be zero:
* * *
pbuy t t t g1 t g2 t 0, (6-36)
* * *
psell t t t g1 t g2 t 0, (6-37)
T T
* 1 * 1 * * *
t 1i l t 2i l t 1i t 2i t 0, (6-38)
i .d l t i .d l t
T T
* * * * *
t i .c 1i l t i .c 2i l t 1i t 2i t 0. (6-39)
l t l t
n n n
Pi.D t Pi.B* t Pi.B* t Pi.R t Pgrid* t Pgrid* t 0, (6-40)
i 1 i 1 i 1
t
Pi.B* s t
*
Ei. B.0 t P
i .c i . B s t Eimax
.B 0, (6-41)
s 1 i.d s 1
110
min
t
Pi. B* s t
*
E i.B Ei.B .0 t P
i .c i . B s t 0, (6-42)
s 1 i .d s 1
max
0 Pgrid t , Pgrid t Pgrid . (6-44)
3. Complementary slackness shall hold for dual variables for the inequality
constraints:
*
t
Pi.B* s t
*
1i t Ei.B.0 t P
i .c i . B s t Eimax
.B 0, (6-45)
s 1 i .d s 1
* min
t
Pi.B* s t
*
2i t E i. B Ei.B.0 t P
i .c i . B s t 0, (6-46)
s 1 i.d s 1
*
g1 t 0 Pgrid* t 0, (6-47)
*
g2 t Pgrid* t max
Pgrid 0, (6-48)
*
g1 t 0 Pgrid* t 0, (6-49)
*
g2 t Pgrid* t max
Pgrid 0, (6-50)
*
1i t 0 Pi.B* t 0, (6-51)
*
2i t Pi.B* t Pi.max
B 0, (6-52)
*
1i t 0 Pi.B* t 0, (6-53)
*
2i t Pi.B* t Pi.max
B 0. (6-54)
As can be seen based on the updating equations (6-24) ~ (6-27), the conditions (6-43)
and (6-44) are automatically satisfied. Here, we will show that the converged states of the
111
1. Satisfaction of global equality constraint
n n n n
ˆk 1
t ˆk t ˆk t ˆk t Pˆi
* k
t . (6-55)
i i ij j i 3
i 1 i 1 i 1 j Ni i 1
ˆk 1
t ˆk t ˆk 1
t ˆ k
t ˆ* t . (6-56)
i i j j
n n n
ˆ* t ˆ* t Pˆi
* k
t . (6-57)
ij 3
i 1 i 1 i 1
n n
ˆ* t ˆ* t . (6-58)
ij
i 1 i 1
n
Pˆi
* k
t 0. (6-59)
i 1
n n
Pˆi Pˆi t
k 1 k k 1 k
t Pi t Pi t . (6-60)
i 1 i 1
n n
Pˆi t
k k
Pi t . (6-61)
i 1 i 1
Based on (6-59) we can conclude that global equality constraint (6-40) is satisfied at
112
2. Satisfaction of Pgrid t and Pgrid t related conditions.
Based on the property of average consensus, the consensus variables converge to their
average, therefore:
n
Pˆi t
k
(6-62)
lim Pˆi t
k i 1
.
k n
Pˆi * t 0, (6-63)
As we have already proved that Pgrid t Pgrid t 0 , there will be mainly four cases:
1) 0 Pgrid* t max
Pgrid and Pgrid* t 0
*
pbuy t t t 0, (6-66)
*
psell t t t 0, (6-67)
* *
by choosing g1 t 0 and g2 t 0 , conditions (6-36), (6-47) and (6-48) are satisfied.
* * *
By taking g1 t psell t t t and g2 t 0 , conditions (6-37), (6-49), and
2) Pgrid* t max
Pgrid and Pgrid* t 0
113
*
pbuy t t t 0, (6-68)
*
psell t t t 0, (6-69)
* * *
by choosing g1 t 0 and g2 t pbuy t t t , conditions (6-36), (6-47) and
* * *
(6-48) are satisfied. By taking g1 t psell t t t and g2 t 0 , conditions
*
pbuy t t t 0, (6-70)
*
psell t t t 0, (6-71)
* * *
by choosing g1 t pbuy t t t and g2 t 0 , conditions (6-36), (6-47) and
* * *
(6-48) are satisfied. By taking g1 t 0 and g2 t psell t t t , conditions
*
pbuy t t t 0, (6-72)
*
psell t t t 0, (6-73)
* * *
by choosing g1 t pbuy t t t and g2 t 0 , conditions (6-36), (6-47) and
* *
(6-48) are satisfied. By taking g1 t 0 and g2 t 0 , conditions (6-37), (6-49), and
114
At the convergence point, we can re-write the updating equation (6-29) as
* *
1i t 1i t 4 P1*i t , (6-74)
0,
*
1) If 1i t 0
* *
1i t 1i t 4 P1*i t P1*i t 0. (6-75)
*
2) If 1i t 0
P1*i t 0. (6-76)
Thus, conditions (6-41) and (6-45) are satisfied. Similar steps can be taken to show (6-42)
At the convergence point, we can re-write the updating equations (6-26) and (6-27) as:
k 1 T
k
t 1i l t
k 1 k i .d l t
Pi.B t Pi.B t 2 , (6-77)
1 T
k
2i l t
i .d l t 0, Pimax
.B
T
k k
t i .c 1i l t
k 1 k l t
Pi. B t Pi.B t 2 . (6-78)
T
k
i .c 2i l t
l t 0, Pi max
.B
As we have already proved that Pgrid t Pgrid t 0 , there will be mainly four cases:
1) 0 Pi.B* t Pi.max
B and Pi.B* t 0
115
T T
* 1 * 1 *
t 1i l t 2i l t 0, (6-79)
i.d l t i .d l t
T T
* * *
t i .c 1i l t i .c 2i l t 0, (6-80)
l t l t
* *
by choosing 1i t 0 and 2i t 0 , conditions (6-38), (6-51) and (6-52) are satisfied.
T T
* * * * *
By taking 1i t t i .c 1i l t i .c 2i l t and 2i t 0 ,
l t l t
2) Pi.B* t Pi.max
B and Pi.B* t 0
* 1 T
* 1 T
*
t 1i l t 2i l t 0, (6-81)
i.d l t i .d l t
T T
* * *
t i .c 1i l t i .c 2i l t 0, (6-82)
l t l t
T T
* 1 1
by choosing 1i t 0 and 2i
*
t *
t *
1i l t *
2i l t ,
i .d l t i .d l t
T T
* * * * *
1i t t i .c 1i l t i .c 2i l t and 2i t 0 , conditions (6-39),
l t l t
T T
* 1 * 1 *
t 1i l t 2i l t 0, (6-83)
i.d l t i .d l t
116
T T
* * *
t i .c 1i l t i .c 2i l t 0, (6-84)
l t l t
T T
1 1 *
by choosing 1i
*
t *
t *
1i l t *
2i l t and 2i t 0 ,
i .d l t i .d l t
*
conditions (6-38), (6-51) and (6-52) are satisfied. By taking 1i t 0 and
T T
* * * *
2i t t i .c 1i l t i .c 2i l t , conditions (6-39), (6-53) and
l t l t
* 1 T
* 1 T
*
t 1i l t 2i l t 0, (6-85)
i.d l t i .d l t
T T
* * *
t i .c 1i l t i .c 2i l t 0, (6-86)
l t l t
T T
1 1 *
by choosing 1i
*
t *
t *
1i l t *
2i l t and 2i t 0 ,
i .d l t i .d l t
* *
conditions (6-38), (6-51) and (6-52) are satisfied. By taking 1i t 0 and 2i t 0,
By using the primal-dual method along with the consensus network, the CoDES
algorithm will asymptotically converge to the optimal solution of the scheduling problem.
It is also necessary to have a method that can determine the convergence status of the
algorithm such that the algorithm result can be send to the physical devices for execution
117
assessment strategy is presented that enables the determination of convergence status of
CoDES algorithm in a distributed way that only utilizes local available information. It
should be noted that the presented method is also applicable to other distributed energy
From the iterative updating equations (6-24) ~ (6-30), and the consensus updating
equations (6-32) and (6-33) presented in Section 6.3.1, it can be observed that the updating
equations for each agent i during each time step are essentially functions of ,
1. at each agent i for each schedule time step t will converge to a common
value. This is due to the property of consensus algorithm.
2. Δ at each agent i for each schedule time step t will converge to 0. This is
because Δ is the estimation of system power mismatch Δ . It can be
shown that if configure Δ Δ during the initialization,
Thus, based on the abovementioned three conditions and the identified variables, the
three components:
118
24 n
1 ˆk t ˆk t
2
cgk1 i j , (6-87)
n t 1 i 1 j Ni
which quantifies in average how far away the of each agent is from its neighbors;
1 24 n 2
cgk2 Pˆi k t , (6-88)
n t 1 i 1
which quantifies in average how far away the Δ of each agent is from 0;
1 24 n 2 2
cgk3 Pi1k t Pi 2k t , (6-89)
n t 1 i 1
0, 0,
which quantifies in average how much the battery capacity inequality constraint of each
As have been analyzed, as the algorithm converges, (6-90) will converge to zeros. It
can thus be used to assess the convergence status of the algorithm. However, this strategy
will require each agent to have the knowledge of the information from all the other agents,
or a central entity that has the knowledge of the status of each agent within the system. In
fact, in an actual distributed application, requiring each agent to know the information
from all the other agents may not be a practical assumption. Using a single entity will also
Like the global convergence measure, each agent can also define local measures to
assess the convergence among the agent itself and its direct neighbors. The local
119
24 2
ci1k ˆk t ˆk t , (6-91)
i j
t 1 j Ni
which quantifies in how far away the of agent i is from its neighbors’;
24 2
ci 2k Pˆi k t , (6-92)
t 1
24 2 2
ci 3k Pi1k t Pi 2k t , (6-93)
0, 0,
t 1
which quantifies how much the battery capacity inequality constraint at agent i is violated.
It should be noted here that as there is no battery installed at the grid interface, node 1
does not have this component in its convergence measure. Thus, the local convergence
ci k ci1k ci 2k ci 3k . (6-94)
By observing the global and local convergence measures, it is obvious that the following
equation stands.
1 n
cgk ci k . (6-95)
n i 1
Thus, if is satisfied for all the agent i, it is guaranteed that will also be
satisfied. So, the problem becomes to develop a strategy to determine how many agents in
the system have reached local convergence such that the system convergence status can be
assessed.
Before the discussion of the strategy to determine the convergence status of the
algorithm, the local convergence criteria at each node need to be defined. The definition of
120
local convergence measures in Section 6.3.2.1 enables each local agent to independently
select its own convergence criteria based on the available information and local
component parameters.
∗
By analyzing equations (6-24) and (6-25), and denote to be the converged KKT
multiplier value at the ith bus during time step t, it can be observed that as the algorithm
converges the following condition will hold for a grid connected system:
Assuming each node has the knowledge of utility tariff, by solving the following equation
for as an example, each node can obtain a threshold . Its physical meaning states
that, the approximate deviation of between agent i and its neighbors at each time
ci1k
24 100% 1% . (6-97)
min psell t , t 1, 24
Assume each agent has the knowledge of the system size in terms of number of nodes,
each agent can estimate the system peak load by using its local peak load multiply by the
number of nodes (except the grid interface) in the system. Then by solving the following
equation for as an example, each agent can obtain a threshold . Its physical
meaning states that, the approximate system power mismatch at each time interval is less
ci 2k n
24 100% 1% . (6-98)
max Pi. D t , t 1, , 24 n 1
121
As for the grid interface node, it will be replacing the denominator in (6-98) with historical
system loading condition recorded at the gird interface to have an estimated system peak
Similarly, since each node has the knowledge of the local installed battery, by solving
the following equation for as an example, each node can obtain a threshold . Its
physical meaning states that, the capacity violation of the battery at agent i is less than 1%
k
ci 3
24 100% 1% . (6-99)
max
E i.B
and in dependently at each node based on their physical meaning, the local
ci k i min i1 , i2 , i3 , (6-100)
such that each node can ensure that the three local convergence criteria components are all
satisfied. Furthermore, based on the analysis in Section 6.3.2.1, the following condition
will always hold, meaning that the global convergence measure is also smaller than a
small constant.
1 n
1 n
cgk ci k i . (6-101)
n i 1 n i 1
Note the calculation example shown in (6-98) only requires each agent has an
estimation of the scale of system loading (e.g., 10kW range or 100kW range, etc.) such
that reasonable small local thresholds can be determined. It is not necessary for each node
122
6.3.2.3 Localized Convergence Flags and Localized Algorithm Convergence
Assessment
The analysis in Sections 6.3.2.1 and 6.3.2.2 indicates that the convergence status of the
agent. Once all the local agents have met their local convergence criteria (6-100) then the
global convergence measure of the algorithm is also less than a small constant (6-101),
assessment strategy is presented that enables each agent to assess the algorithm
Define a local convergence flag . for each of the agent with an initial value set to be
0, once the local convergence criteria for agent i has been met (i.e., is satisfied),
the . will then be set to 1, indicating the agent considers it is locally converged. Then a
minimum time consensus [14] is utilized for each agent to calculate the average of . of
k 1
cˆi. f ij cˆ jk. f . (6-102)
j Ni
During the operation, the consensus network will sample the local convergence flag of
each agent every m iterations, and equation (6-102) will be executed every iteration only
with the exception that the following updating equation being executed once every m
iterations.
k 1
cˆi. f ij c jk. f (6-103)
j Ni
123
1. To assess the convergence status of the algorithm, it is necessary for each
agent to know the true average value of all the . once sampled. By
using the minimum time consensus, this average value can be obtained in
finite time steps.
2. As discussed before, this strategy requires each node to sample the local
convergence flag every m iteration. The property of minimum time
consensus enables the determination of m that guarantees the true average
can be obtained.
Assume each agent has the capability to store the value of consensus variable ̂ . for
each iteration k, a Hankel Matrix can be constructed every 2k iterations (even iterations).
Hi (6-104)
k k 1 2k
cˆi. f cˆi. f cˆi. f
constructed. Once it loses full rank (becomes singular), its normalized kernel will be
rank. The exact average of the local convergence flag can then be calculated as:
k
cˆ M i β i
lim cˆi. f i
(6-105)
k 1T β
where ̂ ̂ . ⋯ ̂ . ̂. .
It is also stated in [14] that the minimum number of iterations required for node i to
compute the final consensus average value is bounded by twice the longest distance from
node i to any other node in the communication topology (lower bound) and twice the
124
number of cells in the minimum external equitable partition of the communication
topology with respect to node i (upper bound). In fact, for the worst case, the number of
the cells in the minimum external equitable partition of the communication topology with
respect to node i is equal to the number of nodes within the graph. So, by assuming each
agent of the system knows the system size, the local convergence flag sampling interval m
can be set to a value that is larger than two times the number of agents in the system.
and detailed process is derived by using CoDES algorithm as example. As can be seen
from the derivation that all the calculations are locally done with all the required
information locally available through the CoDES iterative updating process There is only
one signal ̂ . (a floating type number) need to be packed into the message for peer-to-
usage of minimum time consensus also enables the exact assessment of the algorithm
convergence status. Figure 6-3 shows the general flow chart of the localized convergence
125
k
cˆi. f
As shown in Figure 6-3 the assessment will continue running until the average of local
convergence flag equal to 1. In fact, this assessment method can also be used to evaluate
the algorithm convergence status in terms of percentage of the agents in the system that
To solve the real-time energy re-dispatch problem, first write down the Lagrangian
function:
* 2
L Pgrid p Pgrid Pgrid
n 2 (6-106)
Pi.B Pi.*B P ,
i 1
n n n
P Pi.D Pi.B Pi.R Pgrid . (6-107)
i 1 i 1 i 1
126
Note that the local power rating constraints are not included in the Lagrangian as they
can be treated as the domain of each decision variables. The first order optimality
condition dictates:
L *
p 2 Pgrid Pgrid 0, (6-108)
Pgrid
L
2 Pi.B Pi.*B 0, (6-109)
Pi.B
L
P 0. (6-110)
k
k 1 * p
P
grid P
grid
, (6-111)
2r max max
Pgrid , Pgrid
k
k 1 *
Pi.B Pi.B . (6-112)
2
Pi min' max'
. B , Pi . B
Similar to the situation that we are facing in Section 6.3.1, the global variables and
ˆk 1 ˆk ˆk ˆk Pˆi k ,
i i ij j i (6-113)
j Ni
Pˆi k 1
Pˆi k ij Pˆj k Pˆi k Pi k 1
Pi k . (6-114)
j Ni
ˆ k
p
k 1 * i
P
grid P
grid
, (6-115)
2 max max
Pgrid , Pgrid
127
λˆ
k
k 1 *
P
B P B , (6-116)
2
PBmin' , PBmax'
λˆ Wλˆ Pˆ
k 1 k k
, (6-117)
Pˆ W Pˆ
k 1 k k 1 k
P P , (6-118)
k k k
Pi Pi. D Pi.B Pi. R Pi. grid . (6-119)
approach used in our previous paper [48] can be used to prove that as long as long as the
communication topology is connected, there always exist a small enough ε such that the
cooperative distributed energy re-dispatch algorithm will converge. We call this “the ReD
algorithm”.
In this subsection we will show that the converged states of the ReD algorithm ,
. for each time step are the optimal solution of the optimization problem as shown
in (6-14) ~ (6-19).
First, form the Lagrangian of the optimization problem by considering all the
constraints:
128
* 2
L Pgrid p Pgrid Pgrid
n
2
Pi. B Pi.*B P
i 1
(6-120)
max max
g1 Pgrid Pgrid g2 Pgrid Pgrid
n n
1i Pi.min'
B Pi. B 2i Pi. B Pi.max'
B .
i 1 i 1
The KKT conditions of the optimality imply that at the optimal state, for each time step
, there exist real values , , , and and for all agent i such
1. Derivatives of the Lagrangian with respect to the primal variables shall be zero:
p '
2 Pgrid *
Pgrid ˆ' ' '
0, (6-121)
i g1 g2
Note here that the is replaced by . That is because as the consensus algorithm
n n n
Pi.D Pi.' B Pi.R '
Pgrid 0, (6-123)
i 1 i 1 i 1
Pi.min'
B Pi.' B Pi.max'
B . (6-125)
3. Complementary slackness shall hold for dual variables for the inequality
constraints:
129
'
1i Pi.min'
B Pi.' B 0, (6-128)
'
2i Pi.' B Pi.max'
B 0. (6-129)
As can be seen based on the updating equations(6-111), and (6-112), the conditions
(6-124) and (6-125) are automatically satisfied. Here, we will show that the converged
For each time step , if we sum (6-117) over all the nodes, we will get:
n n n
ˆ' ˆ' Pˆi ' . (6-130)
i ij j
i 1 i 1 i 1
For the consensus network, the is a doubly stochastic matrix. Also since
n
Pˆi ' 0. (6-131)
i 1
If we pre-multiply (6-118) with (1 is a column vector with all the elements equal to
1T Pˆ 1T W Pˆ
k 1 k k 1 k
1T P P . (6-132)
n n
Pˆi Pˆi
k 1 k k 1 k
Pi Pi . (6-133)
i 1 i 1
n n
Pˆi
k k
Pi . (6-134)
i 1 i 1
130
Based on (6-131) we can conclude that global equality constraint (6-123) is satisfied at
1)
'
g1 0,
'
g2 0,
2)
In this case
ˆ' p '
2 Pgrid *
Pgrid 0. (6-135)
i
By choosing
'
g1 0,
3)
In this case
p '
2 Pgrid *
Pgrid ˆ' 0. (6-136)
i
By choosing
'
p '
2 Pgrid *
Pgrid ˆ' ,
g1 i
131
'
g2 0,
Similar technique can be applied to show that conditions (6-122), (6-128) and (6-129)
are satisfied.
Thus, we have proved that the converged states of the ReD algorithm are the optimal
In this chapter, a 34-bus microgrid will be used as a test system to demonstrate the
CoDES-ReD algorithm. The one-line diagram of the microgrid is shown in Figure 6-4.
Without the loss of generality, we assume that each bus of the microgrid, except the
substation (bus 1), will be equipped with one battery storage device, one roof-top PV
panel, and a demand. One example of this microgrid is a residential community. In this
system, each bus of the microgrid will also have a distributed controller embedded that has
the communications capability for sending and receiving information to and from its
29
12
28
11
20 27
10
26
2 3 4 6 7 8 13 18 25 30 31
19 34
9
21 32
1 23 24
5 33
22
14 15 16 17
132
Figure 6-5 Randomly created system communications topology.
In all the case studies, we assume that the battery energy storage device installed at
each bus is Tesla Powerwall, which is a 7kWh Li-ion battery pack with peak power of
3.3kW and 92% efficiency2. The demand and renewable generation profiles at each bus
are obtained from Pecan Street database3. The examples of these profiles are shown in
Figure 6-6, where the forecast profiles are the actual profiles from one household 24 hours
before the operating day with 1-hour resolution, and the real-time profiles are the actual
profiles of the same household during the operating day with 5-minute resolution. Detailed
profiles that used in this chapter can also be found in Appendix-A. For all the case studies
in this chapter, a Duke Energy Progress North Carolina residential service time of use
(TOU) tariff is used4. In this chapter, the surplus power sell-back price is assumed as half
2
https://www.teslamotors.com/powerwall
3
https://dataport.pecanstreet.org/
4
https://www.duke-energy.com/pdfs/R3-NC-Schedule-R-TOU-dep.pdf
133
Figure 6-6 Demand and renewable generation profiles example.
In this subsection, the results of CoDES algorithm are presented. 33 household demand
and renewable generation profiles are randomly selected from Pecan Street database with
1-hour resolution as forecasted profiles for the scheduling. The battery parameters for the
By assuming the Duke Energy TOU tariff, the day-ahead energy scheduling problem
can be solved by the CoDES algorithm with the parameters shown in Table 6-2.
ξ1 ξ2 ξ3 ξ4 ρ
134
Figure 6-7 shows the convergence of the CoDES algorithm. As can be seen from the
figure, the algorithm converges within around 2000 iterations. The subplot at the top
shows the convergence of the algorithm objective function value as the blue line. The
same scheduling problem is also solved by MATLAB linprog function using the interior-
point algorithm as benchmark, and the result is shown as the red star markers. As can be
seen from the subplot, the distributed algorithm converges to the same optimal value as
the centralized linear programming result. The subplot at the bottom shows the
Figure 6-8 shows the day-ahead dispatch schedules of grid power drawn, and the
battery dispatch schedule at bus 9 as examples. The corresponding state of charge (SoC)
of the battery is also shown in the figure indicating the SoC of the battery is bounded
135
Power (kW)
SoC Power (kW)
In this chapter, for the ease of analysis, the local convergence flag of each agent is set
as 0.01 for most of the case studies. For the test case in this chapter, 0.01 has
1. The approximate deviation of between each agent and its neighbors at each
time interval is less than 0.6% of the minimum possible converged .
2. The approximate system power mismatch at each time interval is less than 0.2%
of the estimated system peak load.
3. The capacity violation of the battery at each agent is less than 0.4% of its
maximum scheduling capacity.
Based on the localized convergence assessment strategy presented in section 6.3.2. The
minimum time consensus is used to calculated the average of the sampled local
convergence flag from all the agents. As have analyzed, the worst-case scenario will
require the minimum time step to be no less than twice the size of the system to allow each
agent in the system to obtain the exact average value of the converged consensus state. In
this case, assume each agent has the knowledge of the size of the system, the sampling
136
interval m is set to be 100. The reason of choosing 100 as the sampling interval is for the
ease of plotting the result, an interval of 68 will be enough for this example.
Figure 6-9 shows both the average of local convergence flags (left axis), and the local
convergence measures (right axis) of each agent at a zoomed-in scale. As can be seen
from the figure, the local convergence criteria at each agent is less than 0.01 prior to the
1600th iteration. Thus, when the assessment strategy samples the local convergence flag at
the 1600th iteration, all the agents will have a convergence flag equals to 1. Thus, based on
the convergence assessment strategy shown in Figure 6-3, each of the agent will declare
the algorithm has been converged at 1700th iteration and stop the execution.
Figure 6-9. Local convergence flag average and the local convergence measure
It can be observed from Figure 6-9 that the convergence measures at each agent are not
decreasing at the same rate. There is one node that decreases slower than the others. To
identify this slow node, Figure 6-10 shows the bar plots of for each agent at both the
1000th iteration and the 1500th iteration as example. It can be observed from the bar plots
that at node 1 (the grid interface) is the factor that leads to the slow decreasing rate in
137
10-3
0.04 1
0.02 0.5
0 0
0 20 40 0 20 40
0.5 0.02
0.01
0 0
0 20 40 0 20 40
-4
10
0.02 4
0.01 2
0 0
0 20 40 0 20 40
To explain this observation, we need to go back to the CoDES updating equation (6-33).
Based on the equation, there is an external component being added to the consensus
agents during the first scheduling time step at the 1000th iteration. As can be seen from the
plot, the value of this external component at the node 1 is much higher than the ones at the
other nodes. This explains why at the grid interface node has a slower decreasing rate.
0.06
0.04
0.02
-0.02
0 5 10 15 20 25 30 35
Figure 6-11. External component example for the first scheduling time step
From here, a natural question will rise: what if the grid interface node chooses an that
is too small and thus slows down the declaration of the algorithm convergence? In fact, if
138
we look at the two sampling interval between 1300th iteration and 1600th iteration, each
agent calculates an exact average local convergence flag as 0.9706, which indicates there
is only one agent (grid interface node in this case) has not declared local converge yet.
Figure 6-12 plots both local and global convergence measures. From the figure, it can be
observed that the global convergence measure of the algorithm has already dropped below
0.01 at around 1400th iteration. That means at 1400th iteration, the algorithm state is
already close enough to the optimal solution. As a result, we can essentially declare the
Due to the utilization of minimum time consensus, each agent can obtain the exact
average of the local convergence flags during the sampling interval. Thus, the above
analysis also provides an insight that it is not necessary to wait until all the agents declare
local converge before declare the convergence of the algorithm. If each of the agent
identifies that there is only one agent has not declared convergence yet (in this case an
average local convergence flag equals to 0.9706), it is also safe to use the current states as
139
6.4.3 Cooperative Distributed Re-Dispatch Result
In this subsection, the results of the ReD algorithm are presented. The ReD algorithm is
executed every 5 minutes by sampling the system real-time operating status. The battery
We will mainly show one execution as an example. The remaining 287 executions
throughout the operating day will have similar dynamics. Figure 6-13 shows the
convergence of the re-dispatch command for the first interval (12:00AM – 12:05AM) as
an example. The subplot at the top shows the convergence of for each agent. As can be
seen from this subplot, the re-dispatch command can be calculated within only 100
that the ReD algorithm is suitable for real-time operation with short execution intervals.
The subplot in the middle and bottom shows the convergence of the corresponding grid
power drawn, and the re-dispatch command of the battery at bus 9 as examples. The star
∗ ∗
markers in these two subplots are the and . respectively. Based on Figure 6-6, the
forecast demand at bus 9 at 00:00AM is 2.5506kW, while the real-time measured demand
at 00:00AM is 1.7985kW. As a result, more energy is being stored in the battery compared
140
Power (kW)
Power (kW)
Figure 6-13 Convergence of cooperative distributed energy re-dispatch command for the
Figure 6-14 is the histogram of the violation of equality constraint (6-15) for all the 288
executions. As can be seen from the figure, the maximum power mismatch after 150-
certain that all the 288 executions converge to the optimal solution of the real-time energy
200
100
0
-2 -1.5 -1 -0.5 0 0.5 1 1.5
Power mismatch for each 5-min step 10 -4
Figure 6-14 Re-dispatch power mismatch histogram for all 288 executions.
Figure 6-15 shows the actual grid power drawn, and the battery dispatch command at
bus 9 as examples throughout the operating day. For comparison, both the day-ahead
141
schedule and actual commands are shown in the top and middle subplots. The
corresponding state of charge (SoC) of the battery is shown in the subplot at the bottom. It
can be seen that the battery SoC is well bounded within its minimum and maximum limits.
Power (kW)
SoC Power (kW)
This Section compares the actual system total electricity bill of two different cases as
shown in Table 6-4 with the benchmark case. The benchmark case is determined by
assuming the actual 5-minute resolution profiles are perfectly predicted by each of the
buses, and the optimal dispatch commands are calculated by using MATLAB linprog
function.
Case 1 2
Re-dispatch No Yes
142
Figure 6-16 shows the actual power drawn from the grid under the two cases as well as
the benchmark case. The blue line in each of the subplots is the power drawn schedule
calculated by the day-ahead energy scheduling as comparison. The red lines are the actual
power drawn from the grid when the real-time profiles are applied. Based on the actual
power drawn, the total electricity cost of the three cases are calculated and plotted in
Figure 6-17. As can be seen from the figure, there is a 19.28% total cost reduction when
comparing case 2 with case 1 results for this test system. Also, case 2 result is only 4.2%
higher than the optimal case, which demonstrates the effectiveness and the necessity of the
40
20
0
0 5 10 15 20
Time (hour)
143
6.5 CONCLUSION
In this chapter, the novel CoDES-ReD algorithm is proposed that has two
based on forecasted demand and renewable generation profiles, and 2) ReD algorithm to
re-dispatch the energy storage devices based on the scheduled command and actual system
status. By modeling each bus of the system as one agent, each agent only needs to
exchange limited information with its direct neighbors and the dispatch command of
By using a 34-bus microgrid as the test system, the effectiveness of the CoDES-ReD
algorithm is demonstrated through several representative case studies. For this test system,
the CoDES algorithm converged in 1300 iterations and the ReD algorithm converges in
100 iterations. That shows the ReD algorithm is suitable for real-time operation with short
execution intervals. The simulation results indicate that for this test system, there is a
19.28% cost reduction when the re-dispatch tuning loop is included in addition to the day-
ahead scheduling. Furthermore, the case study result also indicates that the total system
operating cost is only 4.2% higher than the ideal case for the test system.
144
Chapter 7. CONCLUSION AND FUTURE WORK
This dissertation demonstrates our understanding of and vision for the cooperative
distributed energy management framework for the future smart grid. Four algorithms are
developed throughout this dissertation for different energy management scenarios. The
solve the economic dispatch problem for a smart grid environment in a distributed way
Dispatch (CDOD) algorithm is developed to solve the DC optimal power flow problem
with line loss consideration in a distributed manner. The Cooperative Distributed Energy
Scheduling (CoDES) algorithm is developed for a smart grid to solve the optimal dispatch
schedule problem with renewable and energy storage integration while considering
modified DC power flow constraints, branch energy losses, and energy storage charging
and discharging efficiencies. Finally, based on the CoDES algorithm, the CoDES with Re-
Dispatch (CoDES-ReD) algorithm is developed for active energy management for smart
grid, with the capability of adjusting the scheduled set points to compensate for the
prediction errors arising from the intermittencies of the renewables and real-time load
patterns.
145
These four algorithms not only present how we solve different energy management
problems, but also present our understandings and suggestions of improvements for the
cooperative distributed energy management framework for smart grid. Therefore, based
on the latest framework presented in Chapter 6, the following are several directions for
further development:
structure and fully distributed structure, and analysis the cost effect of one
146
PUBLICATION LIST
Journal Papers
[J1] Yuan Zhang, Navid Rahbari-Asr, Jie Duan, and Mo-Yuen Chow, “Day-ahead
Smart Grid Cooperative Distributed Energy Scheduling with Renewable and Storage
Integration”, in IEEE Transactions on Sustainable Energy, vol.7, no.4, pp.1739-1748,
Oct. 2016
[J3] Wente Zeng, Yuan Zhang, and Mo-Yuen. Chow, "Resilient Distributed Energy
Management Subject to Unexpected Misbehaving Generation Units," in IEEE
Transactions on Industrial Informatics, vol.PP, no.99, pp.1-1. (In press)
[J4] Yuan Zhang, Navid Rahbari-Asr, and Mo-Yuen Chow, “A Robust Distributed
System Incremental Cost Estimation Algorithm for Smart Grid Economic Dispatch with
Communications Information Losses”, in Journal of Network and Control Applications,
vol. 59, pp. 315-324, 2015.
[C1] Yuan Zhang, and Mo-Yuen Chow, “Microgrid Cooperative Distributed Energy
Scheduling (CoDES) Considering Battery Degradation Cost”, 2016 IEEE International
Symposium on Industrial Electronics (ISIE), Santa Clara, CA, USA, 2016, pp. 1-6.
[C2] Yuan Zhang, and Mo-Yuen Chow, “Distributed Optimal Generation Dispatch
Considering Transmission Losses”, North American Power Symposium (NAPS), 2015,
Charlotte, NC, USA, 2015, pp. 1-6.
[C3] Wente Zeng, Yuan Zhang, and Mo-Yuen Chow, “A resilient distributed energy
management algorithm for economic dispatch in the presence of misbehaving generation
units”, Resilience Week (RWS), 2015, Philadelphia, PA, 2015, pp. 1-5.
[C4] Navid Rahbari-Asr, Yuan Zhang, and Mo-Yuen Chow, “Cooperative Distributed
Scheduling for Storage Devices in Microgrids using Dynamic KKT Multipliers and
Consensus Networks”, 2015 IEEE Power and Energy Society General Meeting, Denver,
CO, USA, 2015, pp. 1-5.
[C5] Yuan Zhang, Navid Rahbari-Asr, and Mo-Yuen Chow, “Online Convergence
Factor Tuning for Robust Cooperative Distributed Economic Dispatch”, 2015 IEEE
147
Power and Energy Society General Meeting, Denver, CO, USA, 2015, pp. 1-5 (Best
paper)
[C6] Navid Rahbari-Asr, Yuan Zhang, and Mo-Yuen Chow, “Stability Analysis for
Cooperative Distributed Generation Dispatch in a Cyber-Physical Environment”, IECON
2014 - 40th Annual Conference on IEEE Industrial Electronics Society, Dallas, TX, USA,
2014, pp. 5002-5008.
[C7] Yuan Zhang, and Mo-Yuen Chow, “Hybrid Incremental Cost Consensus
Algorithm for Smart Grid Distributed Energy Management under Packet Loss
Environment”, IECON 2014 - 40th Annual Conference on IEEE Industrial Electronics
Society, Dallas, TX, USA, 2014, pp. 3529-3535.
[C8] Ziang Zhang, Yuan Zhang, and Mo-Yuen. Chow, “Distributed Energy
Management under Smart Grid Plug-and-Play operation”, 2013 IEEE Power and Energy
Society General Meeting, Vancouver, British Columbia, Canada, 2013, pp. 1-5.
148
REFERENCE
[2] H. Farhangi, “The path of the smart grid,” IEEE Power and Energy Magazine, vol.
8, no. 1, pp. 18–28, 2010.
[3] “NIST Framework and Roadmap for Smart Grid Interoperability Standards,
Release 2.0,” 2010.
[10] J.-B. Park, Y.-W. Jeong, J.-R. Shin, and K. Y. Lee, “An Improved Particle Swarm
Optimization for Nonconvex Economic Dispatch Problems,” IEEE Transactions
on Power Systems, vol. 25, no. 1. pp. 156–166, 2010.
[11] A. Bakirtzis, “Genetic algorithm solution to the economic dispatch problem,” IEE
Proceedings - Generation, Transmission and Distribution, vol. 141, no. 4. p. 377,
1994.
[12] Z. Zhang, Y. Zhang, and M.-Y. Chow, “Distributed energy management under
smart grid plug-and-play operations,” in 2013 IEEE Power and Energy Society
General Meeting (PES), 2013, pp. 1–5.
149
Tradeoff on D-NCS Using the Coevolutionary Paradigm,” IEEE Transactions on
Industrial Informatics, vol. 9, no. 1, pp. 394–402, 2013.
[17] Z. Zhang and M.-Y. Chow, “Convergence Analysis of the Incremental Cost
Consensus Algorithm Under Different Communication Network Topologies in a
Smart Grid,” IEEE Transactions on Power Systems, vol. 27, no. 4, pp. 1761–1768,
2012.
[18] S. Kar and G. Hug, “Distributed robust economic dispatch in power systems: A
consensus innovations approach,” in IEEE Power and Energy Society General
Meeting, 2012, pp. 1–8.
[19] S. Yang, S. Tan, and J.-X. Xu, “Consensus Based Approach for Economic
Dispatch Problem in a Smart Grid,” IEEE Transactions on Power Systems, vol. 28,
no. 4, pp. 4416–4426, Nov. 2013.
[20] Y. Bai and M. R. Ito, “A new technique for minimizing network loss from users’
perspective,” Journal of Network and Computer Applications, vol. 30, no. 2, pp.
637–649, Apr. 2007.
[21] C. Lee, L. Shiu, F. Lin, and C. Yang, “Distributed topology control algorithm on
broadcasting in wireless sensor network,” Journal of Network and Computer
Applications, vol. 36, no. 4, pp. 1–10, Jul. 2013.
[22] F. Fagnani and S. Zampieri, “Average consensus with packet drop communication,”
Proceedings of the 45th IEEE Conference on Decision and Control, vol. 48, no. 1,
pp. 102–133, 2006.
150
[26] G.-C. Rota, Non-negative matrices in the mathematical sciences, vol. 47, no. 1.
Society for Industrial and Applied Mathematics, 1983.
[27] L. Lamport, R. Shostak, and M. Pease, “The Byzantine Generals Problem,” ACM
Transactions on Programming Languages and Systems, vol. 4, no. 3, pp. 382–401,
1982.
[28] N. Van Der AA, H. Ter Morsche, and R. Mattheij, “Computation of eigenvalue
and eigenvector derivatives for a general complex-valued eigensystem,” Electronic
Journal of Linear Algebra, pp. 300–314, 2007.
[29] F. L. Bauer and C. T. Fike, Norms and exclusion theorems, vol. 2, no. 1.
Numerische Mathematik, 1960.
[32] A. J. Wood and B. F. Wollenberg, Power generation, operation, and control, 3rd
ed. John Wiley & Sons, 2012.
[33] Y. Zhang and M.-Y. Chow, “Hybrid Incremental Cost Consensus Algorithm for
Smart Grid Distributed Energy Management under Packet Loss Environment,” in
Industrial Electronics Society, IECON 2014 - 40th Annual Conference of the IEEE,
2014, pp. 3529–3535.
[36] J. Lavaei and S. H. Low, “Zero duality gap in optimal power flow problem,” IEEE
Transactions on Power Systems, vol. 27, no. 1, pp. 92–107, 2012.
[37] S. Kar, G. Hug, J. Mohammadi, and S. Member, “Distributed State Estimation and
Energy Management in Smart Grids : A Consensus + Innovations Approach,”
IEEE Journal of Selected Topics in Signal Processing, vol. 8, no. 6, pp. 1022–
1038, 2014.
[38] J. Mohammadi, S. Kar, and G. Hug, “Distributed Approach for DC Optimal Power
Flow Calculations,” Arxiv.org [Online]. Available: http:// arxiv.org/abs/1410.4236.
2014.
151
Consensus-Based Economic Dispatch With Transmission Losses,” IEEE
Transactions on Power Systems, vol. 29, no. 4, pp. 1711–1720, 2014.
[41] M. Zhu and S. Martinez, “On distributed convex optimization under inequality and
equality constraints,” IEEE Transactions on Automatic Control, vol. 57, no. 1, pp.
151–164, 2012.
[43] T. N. dos Santos and A. L. Diniz, “A Dynamic Piecewise Linear Model for DC
Transmission Losses in Optimal Scheduling Problems,” IEEE Transactions on
Power Systems, vol. PP, no. 99, p. 1, 2010.
[45] C. Chen, S. Duan, T. Cai, B. Liu, and G. Hu, “Optimal Allocation and Economic
Analysis of Energy Storage System in Microgrids,” IEEE Transactions on Power
Electronics, vol. 26, no. 10, pp. 2762–2773, Jun. 2011.
[46] J. Soares, M. Silva, T. Sousa, Z. Vale, and H. Morais, “Distributed energy resource
short-term scheduling using Signaled Particle Swarm Optimization,” Energy, vol.
42, no. 1, pp. 466–476, Jun. 2012.
[47] Y. Xu, W. Liu, and J. Gong, “Stable Multi-Agent-Based Load Shedding Algorithm
for Power Systems,” IEEE Transactions on Power Systems, vol. 26, no. 4. pp.
2006–2014, 2011.
[49] T. Erseghe, “Distributed Optimal Power Flow Using ADMM,” IEEE Transactions
on Power Systems, vol. 29, no. 5, pp. 2370–2380, 2014.
[50] E. Dall’Anese, H. Zhu, and G. B. Giannakis, “Distributed optimal power flow for
smart microgrids,” IEEE Transactions on Smart Grid, vol. 4, no. 3, pp. 1464–1475,
2013.
152
[51] N. Rahbari-Asr, Y. Zhang, and M.-Y. Chow, “Cooperative Distributed Scheduling
for Storage Devices in Microgrids using Dynamic KKT Multipliers and Consensus
Networks,” in 2015 IEEE Power and Energy Society General Meeting, 2015, pp.
1–5.
[52] G. Hug, S. Kar, and C. Wu, “Consensus + Innovations Approach for Distributed
Multiagent Coordination in a Microgrid,” IEEE Transactions on Smart Grid, vol. 6,
no. 4, pp. 1893–1903, 2015.
[53] W. Shi, X. Xie, C. Chu, and R. Gadh, “Distributed Optimal Energy Management
in Microgrids,” IEEE Transactions on Smart Grid, vol. 6, no. 3, pp. 1137–1146,
2015.
[55] Y. Xiang, J. Liu, and Y. Liu, “Robust Energy Management of Microgrid With
Uncertain Renewable Generation and Load,” IEEE Transactions on Smart Grid,
vol. 7, no. 2, pp. 1034–1043, 2016.
[57] K. Baker, J. Guo, G. Hug, and X. Li, “Distributed MPC for Efficient Coordination
of Storage and Renewable Energy Sources Across Control Areas,” IEEE
Transactions on Smart Grid, vol. PP, no. 99, pp. 1–10, 2016.
[60] S.-C. Fang and S. Puthenpura, Linear Optimization and Extensions: Theory and
Algorithms, First. Prentice-Hall, 1993.
[61] T. Gonen, Modern Power System Analysis, 2nd ed. CRC Press, 2013.
153
“Exploiting the use of DC SCOPF approximation to improve iterative AC SCOPF
algorithms,” IEEE Transactions on Power Systems, vol. 27, no. 3, pp. 1459–1466,
2012.
154
APPENDIX
155
Appendix A- SYSTEM PARAMETERS FOR CHAPTER 6 CASE
Bus 2 3 4 5 6 7 8 9 10
Time (h)
156
Table A.1 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (h)
157
Table A.1 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (h)
158
Table A.1 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (h)
159
Table A.1 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (h)
160
Table A.1 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (h)
161
Table A.1 Continued
Bus 29 30 31 32 33 34
Time (h)
162
Table A.1 Continued
Bus 29 30 31 32 33 34
Time (h)
163
Table A.2 34-node System 1-hour Resolution Renewable Profile (kW)
Bus 2 3 4 5 6 7 8 9 10
Time (h)
1 0 0 0 0 0 0 0 0 0
2 0 0 0 0 0 0 0 0 0
3 0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0 0 0
5 0 0 0 0 0 0 0 0 0
6 0 0 0 0 0 0 0 0 0
7 0 0 0 0 0 0 0 0 0
8 0 0 0 0 0 0 0.013967 0 0.016033
164
Table A.2 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (h)
22 0 0 0 0 0 0 0 0 0
23 0 0.001317 0 0 0 0 0 0 0
24 0 0 0 0 0 0 0 0 0
165
Table A.2 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (h)
1 0 0 0 0 0 0 0 0 0
2 0 0 0 0 0 0 0 0 0
3 0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0 0 0
5 0 0 0 0 0 0 0 0 0
6 0 0 0 0 0 0 0 0 0
7 0 0 0 0 0 0 0 0 0
166
Table A.2 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (h)
22 0 0 0 0 0 0 0 0 0.006133
23 0 0 0 0 0 0 0 0 0
24 0 0 0 0 0 0 0 0 0
167
Table A.2 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (h)
1 0 0 0 0 0 0.002133 0 0 0
2 0 0 0 0 0 0.002133 0 0 0
3 0 0 0 0 0 0.002083 0 0 0
4 0 0 0 0 0 0.002067 0 0 0
5 0 0 0 0 0 0.00305 0 0 0
6 0 0 0 0 0 0.003483 0 0 0
7 0 0 0 0 0 0.003667 0 0 0
8 0 0 0 0 0 0.009767 0 0 0
168
Table A.2 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (h)
22 0 0 0 0 0 0.018817 0.000317 0 0
23 0 0 0 0 0 0.003067 0 0 0
24 0 0 0 0 0 0.003217 0 0 0
169
Table A.2 Continued
Bus 29 30 31 32 33 34
Time (h)
1 0 0 0 0 0 0
2 0 0 0 0 0 0
3 0 0 0 0 0 0
4 0 0 0 0 0 0
5 0 0 0 0 0 0
6 0 0 0 0 0 0
7 0 0 0 0 0 0
8 0 0 0 0.003967 0 0
170
Table A.2 Continued
Bus 29 30 31 32 33 34
Time (h)
22 0.001017 0.002233 0 0 0 0
23 0 0 0 0 0 0
24 0 0 0 0 0 0
171
Table A.3 34-node System 5-min Resolution Demand Profile (kW)
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
172
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
173
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
174
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
175
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
176
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
177
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
178
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
179
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
100 1.6572 1.1394 0.4674 1.0626 0.3384 0.8205 0.4986 0.942 3.1092
101 1.0872 1.0788 0.6894 1.0017 0.3054 0.786 0.4839 0.9219 3.2616
102 0.858 0.765 0.6924 0.6534 0.24 0.8214 0.3942 0.8802 3.2205
103 0.882 0.783 0.69 0.8514 0.2685 1.1955 0.3273 0.7863 3.3564
104 0.8964 0.9264 0.6849 0.8709 0.2757 0.7056 0.3201 0.7857 3.1617
105 0.7188 0.7695 0.6729 1.2792 0.3531 0.7611 0.3183 0.7839 2.7132
106 0.8433 0.6495 2.7678 3.1704 0.3909 1.5243 0.3288 0.7815 2.9613
107 0.6561 0.5943 2.7174 1.5162 0.3813 0.8748 0.4683 0.7932 2.4378
108 0.7011 0.5298 0.5148 1.1142 0.3774 0.7869 0.5808 0.7998 2.4093
180
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
109 1.485 0.5514 0.5151 1.1157 0.4071 1.1355 0.5172 0.8007 2.6106
110 2.5926 0.6795 0.5118 1.0902 0.7287 0.6213 0.5085 0.8001 2.4282
111 2.028 1.6095 0.51 1.305 0.7899 0.5235 0.5049 0.7998 2.1891
112 0.8715 3.3024 0.5157 1.4169 0.6723 0.5208 0.4965 0.8823 2.0985
113 0.8826 0.5712 0.5115 1.3086 0.4284 0.7539 0.5085 1.2327 2.3631
114 0.8784 0.6681 0.5175 0.993 1.2621 0.7719 0.4623 1.2834 2.6373
115 0.8742 0.6546 0.5727 0.8109 0.4599 0.7701 0.3369 1.3158 2.4102
116 0.7818 0.6546 0.7008 0.8724 1.8399 0.7677 0.3288 1.356 2.0454
117 0.5637 0.8088 0.69 1.131 0.4467 0.7539 0.3279 1.344 1.9353
118 0.5139 4.1805 0.6993 1.3557 0.4341 0.636 0.3282 1.3269 1.9173
119 0.6654 0.8871 1.6191 1.2576 0.357 0.6363 0.3414 1.2636 2.1402
120 0.7533 0.6759 3.7461 0.7914 0.3528 0.624 0.4332 1.2669 2.3898
181
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
121 0.7299 0.6657 0.7947 0.6261 0.3939 0.5889 0.5451 1.2753 2.3706
122 0.7137 0.705 0.531 0.5814 0.8508 0.6306 0.5172 1.2711 2.5233
123 0.6975 0.8616 0.5265 0.6888 0.9318 0.6237 0.5154 1.212 2.8971
124 0.6912 1.1472 0.5145 1.1319 0.9411 0.6132 0.51 1.0863 2.817
125 0.6927 3.7089 0.5277 1.0773 0.9867 0.5313 0.489 1.167 2.7147
126 0.6432 0.8034 0.549 0.5925 1.0485 0.4998 0.405 1.1757 2.0454
127 0.6477 0.8097 0.5439 0.5865 0.888 0.4992 0.3357 1.0719 2.1291
128 0.6756 0.7845 0.6405 0.5805 0.8883 0.5268 0.3258 1.1973 2.0964
129 0.5169 0.837 0.7086 0.663 0.9495 0.6351 0.324 0.8178 2.2086
130 0.6126 0.9267 0.7104 1.1478 0.8772 0.6876 0.3282 0.7998 1.8672
131 0.7137 2.1924 1.2651 1.3419 0.9885 0.78 0.3888 0.7986 1.9041
132 0.7032 3.2277 3.9204 0.792 0.9696 0.7512 0.5601 0.7854 1.8294
182
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
133 0.702 0.9237 1.2363 0.795 0.8904 0.6393 0.5283 0.7878 1.7064
134 0.7251 0.9999 0.5598 0.8061 0.9591 0.6495 0.5178 0.7986 1.7379
135 0.7302 1.0041 0.564 1.0641 0.9456 0.6486 0.5169 0.8145 1.8561
136 0.7095 0.948 0.5586 1.1784 0.7866 0.7377 0.4935 0.9969 1.7922
137 0.7017 4.4046 0.5667 1.155 0.8115 0.7833 0.4335 1.2027 1.7523
138 0.7821 1.1052 0.5622 0.6168 1.1508 0.7347 0.3402 1.2018 1.8108
139 1.3167 1.065 0.5592 0.5904 0.8217 0.6303 0.4581 0.8373 7.1814
140 0.7071 1.0554 0.6312 0.5967 0.8805 0.5571 0.4896 0.7035 6.8178
141 0.6057 1.3479 0.6123 0.8277 2.3484 0.5064 0.4944 0.849 7.7886
142 0.8919 4.6422 3.9285 1.1463 2.403 0.504 0.498 0.7833 7.8852
143 1.0791 1.0155 2.3202 1.2 2.3259 0.5394 0.7623 0.8421 6.6507
144 1.3173 0.8499 0.4443 0.6501 2.2524 0.6378 0.9777 0.8148 7.7805
183
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
145 1.3461 0.7236 0.4137 0.5886 2.1339 0.6303 4.53 0.8031 6.7992
146 0.9705 4.3398 0.4239 0.8154 0.6009 0.6321 4.2747 0.81 6.0714
147 0.9486 1.2954 0.3699 1.1208 0.6348 0.6132 1.035 0.8187 2.61
148 0.9432 0.8253 0.3186 1.4178 0.5676 0.5946 1.1442 0.8361 2.3349
149 0.9342 0.8856 0.2757 1.3539 0.5397 0.6429 0.9873 0.8049 2.2518
150 7.4742 4.0947 0.2931 0.8997 0.6669 0.6525 4.9545 0.81 2.5974
151 7.8288 5.2776 0.7695 0.5958 0.4587 0.741 5.6286 0.7989 2.6286
152 10.3152 1.2156 5.9526 0.5793 0.4977 0.7731 2.1495 0.792 2.9112
153 10.3323 0.9657 1.1823 0.8496 0.4869 0.7746 2.2401 0.8007 5.9463
154 7.0362 2.0808 0.7164 1.1424 0.6081 0.7728 2.2425 0.7974 8.4045
155 0.9006 6.507 0.8097 1.1925 0.6456 0.6882 5.8425 0.8121 7.7433
156 0.9192 3.2622 0.8025 0.7503 0.6711 0.5829 4.8585 0.807 8.286
184
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
157 1.8555 1.1229 2.3703 0.6438 0.9813 0.5247 1.0461 0.8139 6.7749
158 9.2073 1.1166 3.7605 0.6258 0.444 0.5622 0.9678 0.7998 7.8915
159 9.624 4.8705 0.8358 0.7848 0.3906 0.8274 2.1273 0.7929 7.5243
160 9.7635 5.7969 0.7416 1.2564 0.402 0.8721 6.6672 0.8052 6.8346
161 7.8465 1.2801 0.7041 1.3578 0.384 0.858 3.7086 0.8103 6.4674
162 4.5651 0.9267 2.0646 0.9621 0.4398 0.8565 1.4955 0.8079 3.7992
163 1.3461 2.0733 3.3276 0.7998 0.4101 0.7698 4.0506 0.8112 6.57
164 0.8547 7.1556 0.6219 0.7998 0.4401 0.7353 5.3967 0.8154 6.15
165 4.227 9.9147 0.5568 0.8562 0.4356 0.822 2.5803 0.8181 5.1582
166 3.951 6.3027 0.5547 1.1463 0.4014 0.9459 1.4235 0.81 5.1507
167 0.8025 6.0945 0.5619 1.1505 0.3387 0.972 3.0273 0.8049 4.5954
168 5.9277 9.3888 3.2421 0.7998 0.294 0.9672 5.8296 0.801 6.6423
185
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
169 6.1983 10.71 2.352 0.5952 0.3501 0.9195 1.6005 0.7965 7.7085
170 8.4375 6.321 0.6093 0.6132 0.4983 0.8388 1.1289 0.8112 6.6435
171 8.0337 6.1278 0.7182 0.7491 0.5169 0.8328 0.645 0.8073 7.8255
172 6.4731 9.2859 0.8037 1.1955 0.5313 0.8541 4.7004 0.8382 6.7077
173 4.9005 10.7337 2.7606 1.2159 0.513 5.7849 5.2224 0.8469 7.4943
174 3.7989 8.0325 3.2403 0.8973 0.4701 6.087 4.776 0.8493 6.4959
175 0.7962 5.94 0.7752 0.7887 0.4059 6.0012 0.6789 0.8265 6.6837
176 0.5745 6.6408 0.8202 0.7986 0.3693 1.5405 0.4812 0.8004 4.8777
177 0.573 10.8333 0.8157 1.1478 0.36 0.7032 2.0346 0.8055 1.842
178 4.5789 10.7574 3.1263 3.4692 0.4467 0.678 5.0436 0.7989 2.298
179 11.6523 10.3077 2.5761 1.341 0.4497 0.6726 5.1951 0.7845 2.2449
180 9.1725 6.1854 0.5808 0.9345 0.4785 0.7398 2.9481 0.7851 2.1072
186
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
181 8.1123 5.8038 0.5475 0.5793 0.4728 0.8133 0.7827 0.7821 1.2324
182 7.8552 10.0281 0.534 2.6013 0.3948 0.8037 0.6735 0.7749 1.2867
183 7.9329 10.3968 2.388 1.6509 0.3477 0.8214 4.2936 0.7788 1.2822
184 4.3539 9.8916 3.732 0.5688 0.3021 0.8877 4.9392 0.7902 3.0984
185 5.4006 2.1372 0.8529 1.0506 0.3453 0.807 2.8152 0.7953 6.2019
186 4.3215 0.9933 0.6918 0.882 0.441 0.8082 0.6519 0.7971 1.4277
187 4.2981 5.3436 0.6888 2.4627 0.4182 0.8217 0.5136 0.8991 1.1529
188 8.697 3.6642 0.8292 2.0655 0.4104 0.9369 3.9423 1.1124 1.1535
189 6.8304 0.9006 3.8469 0.7005 0.3759 0.9432 4.7469 1.0995 2.4738
190 5.271 0.9087 2.5446 0.7872 0.3099 0.9402 1.9476 1.0734 6.6219
191 4.4301 5.3229 0.69 1.3737 0.2538 0.9351 1.1175 1.1472 5.6373
192 1.4766 3.8751 0.5649 3.4362 0.2574 0.693 1.7907 0.8805 1.2282
187
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
193 1.371 1.1301 0.5487 0.861 0.2661 0.6813 5.1918 0.7872 1.1481
194 0.9402 1.1586 1.6011 0.6891 0.3738 0.6888 1.9515 0.7971 2.9619
195 0.6519 4.623 3.7683 0.5985 0.3966 0.78 0.5031 0.8127 6.8073
196 0.6483 3.108 0.8133 3.6063 0.3963 0.8349 0.7494 0.8241 2.0739
197 4.6815 1.0386 0.6564 0.8355 0.4644 0.8532 3.9225 0.8469 1.3176
198 4.2195 1.0338 0.66 0.9534 0.477 0.846 4.0839 0.8412 1.3263
199 7.0044 5.4213 0.6789 0.6015 0.4083 0.7338 0.7266 0.8199 6.7197
200 4.6476 2.7252 1.701 0.5985 0.567 0.6924 0.6903 0.8094 6.384
201 3.6603 1.0176 4.1091 3.5964 0.8895 0.7668 1.575 0.8199 1.5285
202 3.5508 1.0734 1.4223 0.7686 1.077 0.9039 5.3232 0.8253 1.3662
203 3.3528 5.2128 0.7974 0.8673 2.5272 0.9945 2.1474 0.8169 1.4187
204 2.1078 2.7819 0.7581 1.2594 2.5794 0.981 0.6894 0.8052 1.4397
188
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
205 0.6984 0.8475 0.7983 2.1537 2.6415 0.9867 0.6885 0.8154 7.149
206 0.4908 0.8448 3.4383 3.1473 2.5563 0.9399 3.6057 0.9141 7.269
207 0.4806 5.5926 2.5899 0.8271 1.5552 0.8391 4.1742 0.9048 7.1508
208 0.48 5.6229 0.6147 0.813 0.5853 0.8472 0.6684 1.0008 7.0581
209 0.3522 1.8144 0.612 1.5384 0.5175 0.8472 0.6699 0.8709 7.0446
210 0.2955 1.0755 0.5799 3.5052 0.5577 0.8154 2.3835 0.837 7.08
211 3.1338 3.5103 1.5453 0.6735 0.5202 0.8307 5.4231 0.8355 7.1433
212 0.426 5.8995 4.0353 0.6039 0.579 0.8277 3.2148 0.8322 7.146
213 0.4557 4.8384 1.4322 2.6895 0.5223 0.8295 0.7377 0.8385 7.1367
214 0.4767 1.0053 0.7581 4.4661 0.5664 0.702 0.7062 0.8223 7.1706
215 0.4758 0.9423 0.7785 4.2801 0.4848 0.7008 0.9498 0.8361 7.1493
216 8.1009 5.5047 1.8966 0.837 0.4377 0.7005 5.2176 0.8355 7.1736
189
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
217 3.3597 5.8398 4.2096 0.5982 0.6687 0.7632 5.3295 0.81 7.2033
218 2.6646 3.7128 1.0161 0.7806 0.5847 0.8406 5.3007 1.5093 7.3128
219 5.0043 1.0494 0.7836 3.8787 0.4086 1.0131 1.4145 4.9776 7.3854
220 5.3106 2.7138 1.5408 4.764 0.4533 7.0398 0.5487 4.9956 2.7681
221 7.0344 5.8323 3.8082 4.2663 0.4341 6.7683 0.7107 4.9743 4.3224
222 4.6803 5.5998 1.2132 0.9738 0.3534 6.816 2.2587 1.1994 7.3113
223 4.1391 5.5389 0.3414 0.6555 0.3405 6.8658 5.5212 0.8331 7.2213
224 3.4485 1.458 0.2991 3.1347 0.285 6.7272 5.4654 0.8313 4.1121
225 0.5913 0.7416 1.9983 4.4754 0.2883 6.6336 10.7613 3.6423 2.1912
226 0.7002 5.1138 3.8448 4.4847 0.4041 6.7065 10.8462 4.8354 3.576
227 0.678 5.7687 1.2702 1.563 0.4125 6.732 10.5885 0.9966 8.0532
228 2.7804 5.703 1.1265 0.5994 0.4383 6.9192 10.6257 0.8022 8.2104
190
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
229 7.9059 1.521 4.0356 2.0151 0.4308 7.2129 10.4916 0.8001 2.4969
230 9.1647 0.8802 4.2864 4.5135 0.3552 7.0881 10.2636 2.1003 2.0106
231 6.7083 4.626 1.1553 4.4775 0.4668 6.9168 10.0986 5.3478 2.7588
232 8.6391 5.721 0.8022 1.4574 0.465 6.8478 10.4001 1.9437 8.0583
233 8.9178 5.6607 3.1647 0.8112 0.3441 1.845 12.6861 0.708 7.974
234 9.3261 3.2769 4.2366 2.5467 0.4482 1.1838 8.5254 0.9165 5.0397
235 10.0125 1.0191 0.9672 4.6959 4.9317 1.1724 7.3458 3.0522 2.0964
236 6.8505 1.5783 0.7104 3.0435 4.9869 1.2081 7.0749 5.5227 7.6419
237 7.1592 5.8455 3.0213 0.9192 4.9704 1.3617 5.6427 1.8759 7.9611
238 10.2594 5.6625 3.7248 2.304 4.9113 1.3527 7.0905 1.4502 7.7808
239 10.6896 5.6358 0.7146 4.947 4.8618 1.1598 7.2642 1.6485 7.8246
240 10.5738 3.1974 0.5556 3.8925 4.8525 1.1595 7.047 5.1147 8.1258
191
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
241 6.4086 0.741 1.665 1.0953 4.8759 1.1715 6.5604 5.394 7.9059
242 7.5453 0.8517 3.9873 1.5138 4.9674 4.7877 6.5856 1.7298 7.869
243 9.6282 3.3972 1.3269 4.824 4.9407 6.6717 5.715 1.536 7.7496
244 9.5922 5.6745 0.5715 1.6413 5.133 6.6711 6.4671 3.2754 7.7475
245 6.8268 5.6211 0.7452 1.0449 5.5416 6.6846 7.0536 5.6778 7.7583
246 6.0624 5.4507 3.7497 1.5792 5.4192 6.6264 6.513 4.1664 7.7196
247 9.3036 2.5221 2.8512 4.968 5.4216 6.4653 6.1959 2.3829 7.6989
248 7.3416 0.7143 0.711 2.7819 5.4267 6.5373 5.7132 1.9059 7.9638
249 5.0988 3.6687 0.6969 1.452 5.4402 6.5181 5.8143 5.2536 7.7205
250 5.6811 5.6643 3.369 1.9095 5.5353 3.9936 5.7735 4.287 7.9863
251 4.386 5.7075 3.1761 5.0502 5.5989 1.4673 5.4678 1.2261 7.9401
252 1.1166 3.1167 0.5334 2.5665 5.6115 1.638 5.46 1.263 7.9473
192
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
253 1.8708 1.0086 0.5199 1.5015 5.7081 6.7857 5.4249 1.6062 7.9338
254 5.1168 2.9223 0.5178 2.3673 5.4846 6.9348 5.421 4.3527 7.7826
255 5.5599 5.5473 3.4632 6.3294 5.2608 6.9342 5.2569 4.929 7.4499
256 5.1786 5.5284 1.965 6.303 5.0409 6.9426 5.2272 1.4091 7.3869
257 2.4225 1.4946 0.5706 3.1395 5.3169 6.7566 5.403 1.4631 7.3845
258 1.2276 0.9984 0.5697 2.2611 5.3073 6.4911 5.4867 2.5452 7.3857
259 4.5099 2.6073 1.2189 4.9413 5.2938 6.9204 5.5143 5.4033 7.3182
260 4.6758 5.5854 4.0356 5.5335 5.4465 6.5775 5.4603 6.4992 4.9713
261 4.3536 1.7877 1.7058 3.1425 5.3256 1.8141 5.4438 6.2325 1.6713
262 1.6179 1.0389 0.6993 1.7856 5.1174 1.3839 1.5441 9.5394 1.5051
263 2.8428 2.7732 0.6912 2.946 4.9368 1.3851 0.9708 10.2609 6.2256
264 5.364 5.8023 2.1282 5.3793 4.938 1.3785 4.9395 10.2969 4.6845
193
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
265 5.2023 1.9545 3.7071 5.2104 5.0325 1.3575 4.4745 9.0816 1.5396
266 2.2068 1.2597 0.723 1.8228 5.0664 1.2312 0.9903 3.4521 1.7793
267 1.2567 2.1732 0.5307 1.5807 5.2077 1.2987 0.9381 2.9091 4.9074
268 2.6748 6.0147 0.531 4.7475 7.0056 1.3287 3.4158 6.8721 5.0268
269 4.785 6.0021 0.5721 5.2302 5.2218 1.3773 5.1801 3.2859 1.2702
270 4.1916 2.1975 2.7162 2.1636 5.205 1.1712 1.6494 1.3767 1.5834
271 1.3992 1.4136 2.7864 1.7358 6.0999 1.2195 0.7782 3.2505 3.9294
272 1.2588 5.2689 0.4188 4.4718 5.0049 1.4433 0.7506 5.2905 6.7314
273 1.9302 5.7168 0.3561 5.616 4.9176 1.3986 4.5258 3.2055 2.0061
274 4.7133 2.1465 0.5607 5.1258 4.9974 4.6545 5.2836 1.6482 1.8051
275 4.0026 1.2306 0.9255 4.9677 5.0646 6.2997 1.6992 3.8592 2.3274
276 1.1937 5.2611 3.7755 3.4308 5.0625 6.2133 0.9027 5.5698 7.0218
194
Table A.3 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
277 1.2711 5.2959 1.2993 1.3668 5.0037 6.1818 1.8879 2.1948 2.1111
278 7.4055 1.3341 0.5415 2.0058 4.8963 6.2355 5.2614 1.6746 1.8126
279 10.4256 0.6666 0.456 4.8219 1.5921 5.2629 2.8995 1.4991 1.7997
280 8.9925 2.0688 0.2967 4.7805 3.0546 1.4934 0.8925 5.3769 6.6693
281 7.9044 5.0478 0.21 4.7817 5.0361 1.3794 0.8505 5.3649 1.9653
282 7.8198 3.7359 0.5814 4.7796 5.0583 1.3851 3.5832 5.4222 1.3746
283 10.2768 0.519 3.4521 2.4909 5.1474 1.3578 0.8535 3.7269 1.3767
284 7.8099 0.6321 0.5868 3.6891 5.1105 1.2408 0.8568 1.4922 5.6025
285 5.5653 3.9063 0.2106 4.7793 4.9872 1.158 0.8631 1.8498 2.301
286 5.8329 3.0927 0.2103 4.7808 4.689 1.1922 1.0755 5.3538 1.3641
287 8.436 0.5133 0.21 2.0076 4.5999 1.2888 0.9555 5.301 1.3794
288 6.8577 0.471 0.2109 1.2012 4.563 1.3584 0.9111 1.827 3.7755
195
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
196
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
197
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
198
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
199
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
200
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
201
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
202
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
203
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
100 1.6389 0.7626 1.3008 5.8641 4.2057 1.2189 1.0527 0.3939 0.7389
101 1.9917 0.7476 1.7613 1.8429 4.4781 2.0217 1.047 0.3486 0.8355
102 3.0735 0.7224 3.4632 1.8873 1.7058 1.9881 1.3863 0.3147 0.6966
103 1.2834 0.8742 1.7769 1.614 1.6797 1.1739 1.1613 2.8113 0.6132
104 1.1862 0.9081 1.4166 1.3707 1.6344 1.2714 1.1388 1.5381 0.7431
105 1.0218 1.2585 0.8097 1.5195 1.5138 0.9588 0.6063 1.0833 0.7533
106 1.0239 1.2144 0.642 2.1675 1.524 0.9699 0.5835 0.6321 0.6726
107 1.0248 0.9309 0.6063 3.6132 3.687 1.818 0.549 0.6072 0.621
108 2.2455 0.6519 0.6231 5.5056 5.1321 1.7535 0.5349 0.5886 0.5655
204
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
109 2.0871 0.6594 0.7353 2.6034 1.6851 0.9708 0.4119 1.1397 0.4134
110 1.0374 0.6513 0.8199 2.1114 1.6884 0.6561 0.3558 1.0317 0.4149
111 1.1127 0.7806 0.795 1.992 1.8438 0.66 0.3552 0.7998 0.4131
112 1.1451 0.8004 0.6627 2.5323 1.5885 0.9036 0.3534 0.4227 0.4098
113 1.1511 0.5853 0.5988 1.8249 2.5098 1.6806 0.369 0.4155 0.5142
114 1.8396 0.4527 0.6117 1.9407 5.5311 1.6866 0.4269 0.507 0.6216
115 2.9745 0.8526 0.7401 7.3503 1.9878 1.0353 0.5877 0.6654 0.6189
116 1.2021 1.3098 0.8163 2.8275 1.1301 0.66 0.5598 0.8754 0.6111
117 0.8988 1.8168 0.8016 2.4567 1.1658 0.6792 0.5901 0.8829 0.5895
118 0.9918 1.2405 0.6186 2.4141 1.3098 0.8463 0.6657 0.5031 0.3897
119 1.0077 0.864 0.6255 2.595 1.3062 1.6884 0.534 0.4521 0.3906
120 1.8264 0.7899 0.6468 3.4734 4.269 1.6827 0.4074 0.3786 0.3912
205
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
121 2.6241 0.7734 0.9402 2.2842 3.7599 1.098 0.36 0.6537 0.3936
122 1.2687 0.7596 1.3686 1.7799 1.2585 0.6552 0.3531 0.9057 0.4692
123 1.4085 0.7404 1.0212 1.6017 1.1022 0.6423 0.3501 1.0155 0.615
124 1.398 0.7245 0.8295 2.5917 1.0743 0.6405 0.4422 0.3651 0.6153
125 1.6755 0.7443 2.2434 3.4923 1.1493 1.5708 0.5865 0.4329 0.6066
126 3.3852 0.6381 1.0101 2.1435 1.5321 1.6623 0.5757 0.4047 0.5622
127 1.425 0.5553 1.323 1.9221 1.3635 1.1493 0.5364 0.8811 0.3831
128 1.0245 0.5817 1.3044 9.3621 1.3569 0.6411 0.5295 0.8613 0.3972
129 1.0179 0.9522 1.1106 11.8716 1.3554 0.6423 0.4323 0.918 0.3882
130 0.9453 1.5243 0.9072 10.5246 1.3626 0.6465 0.3582 0.4128 0.3747
131 1.8054 1.5093 0.9 9.5112 1.2066 1.5036 0.3447 0.9963 0.4356
132 3.1416 1.3557 1.1658 9.096 1.1877 1.6572 0.339 0.594 0.6111
206
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
133 1.2351 0.9201 0.9603 8.8155 1.2204 1.2171 0.3462 0.9489 0.5982
134 1.1397 0.7278 1.1169 5.1273 1.2435 0.6492 0.4134 1.1394 0.5664
135 1.1172 0.6639 1.1118 1.7643 1.3476 0.6489 0.5745 0.996 0.5301
136 2.1624 0.6621 1.0113 1.3956 1.4442 0.6465 0.5523 0.4074 0.3639
137 2.4825 0.6705 1.0464 1.7727 1.6263 1.4337 0.5367 0.4314 0.3501
138 0.8541 0.6684 0.966 3.4068 1.4007 1.6593 0.5382 0.4137 0.3561
139 0.8463 0.6468 0.9762 2.5605 1.3977 1.2822 0.4023 0.7089 0.3783
140 0.8373 0.609 1.1016 3.3624 1.2018 0.6489 0.3537 0.9429 0.4518
141 1.9734 0.5388 1.194 6.5397 1.2117 0.6528 0.3429 0.8475 0.5454
142 2.7942 0.6243 1.173 5.481 1.1997 0.6483 0.3504 0.4008 0.5718
143 1.2978 0.6933 1.008 5.7363 1.2258 1.3839 0.3522 0.339 0.5052
144 1.2468 0.6144 0.9735 5.6058 1.329 1.6503 0.471 0.2832 0.5151
207
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
145 1.2621 0.549 0.9792 3.4773 1.3512 1.3389 0.5592 0.3474 0.3915
146 2.7963 0.6813 1.0059 2.0934 1.3605 0.6039 0.543 0.9174 0.3741
147 3.4776 0.7089 1.194 1.8618 1.4196 0.561 0.5349 0.8007 0.345
148 1.2345 0.7683 1.1955 3.3105 1.4601 0.5736 0.5325 0.4572 0.3105
149 0.9507 0.6843 1.0614 3.8385 1.2786 1.5402 0.3639 0.2811 0.4893
150 1.0686 0.7134 0.9771 2.4597 1.2915 1.5711 0.3459 0.3027 0.5466
151 3.0261 0.6657 0.9735 2.0664 1.3995 1.3239 0.3417 0.3243 0.5853
152 2.0376 0.6453 0.9708 1.8558 1.5285 0.5538 0.3378 0.7716 0.5964
153 1.0941 0.6627 1.1292 3.2769 1.4946 0.6087 0.3579 0.7767 0.4866
154 1.3596 0.7464 1.2963 3.2343 1.5546 0.6561 0.6156 0.4449 0.3693
155 2.1246 0.7482 1.3212 1.8354 1.6572 1.5957 0.6387 0.3318 0.2847
156 3.5127 0.6948 1.149 1.0122 1.635 1.68 0.5319 0.3282 0.2904
208
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
157 1.8144 0.7344 1.1646 0.6924 1.6101 1.4787 0.5418 0.4557 0.2388
158 1.3809 0.5559 1.1967 2.5935 1.3485 0.6093 0.5226 0.888 0.4749
159 1.2024 0.6669 1.3101 3.2274 1.0647 0.5967 0.4359 0.7872 0.5664
160 2.763 0.6984 1.5123 1.971 1.1091 0.5997 0.3516 0.4914 0.5595
161 2.8185 0.6774 1.467 1.1001 1.0686 1.515 0.3474 0.2838 0.5784
162 0.9414 0.6642 1.5405 1.7529 1.878 1.6029 0.3417 0.2943 0.4611
163 0.9984 0.6636 1.6419 2.6433 5.3898 1.4988 0.3555 0.4725 0.3804
164 1.5291 0.669 1.8444 1.7247 1.4388 0.6033 0.5124 0.7728 0.3741
165 3.2778 0.6681 1.4853 0.9858 1.2678 0.6006 0.5754 0.768 0.3717
166 1.8489 0.6636 1.4802 1.2996 1.2831 0.5967 0.5532 0.5061 0.3717
167 1.161 0.7914 1.6602 2.7561 1.1529 1.4436 0.5418 0.3006 0.5628
168 1.2477 0.6459 1.6743 2.7813 1.1085 1.5996 0.5241 0.3072 0.5727
209
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
169 6.7575 0.6177 1.5702 1.4658 1.1148 1.5588 0.3987 0.5946 0.5757
170 2.6469 0.6714 1.26 1.2741 1.1886 0.6033 0.5406 0.8085 0.5193
171 1.146 0.723 4.6878 1.4643 1.3992 0.597 0.5598 0.8349 0.3804
172 1.2321 0.7776 2.8878 4.3602 1.3068 0.6078 0.5532 0.6441 0.285
173 1.6272 0.798 1.5591 4.5444 1.2957 0.9987 0.5016 0.6396 0.2523
174 3.2415 0.8106 1.7604 4.389 1.2963 1.6428 0.3423 0.684 0.2424
175 1.7136 0.7092 1.8078 3.4644 1.29 1.6065 0.3417 0.7062 0.3729
176 1.5057 0.6441 1.5531 4.4169 2.7576 0.6228 0.5727 1.1007 0.6294
177 2.5758 0.6567 1.3341 4.8147 4.9869 0.5901 0.564 1.3668 0.7485
178 5.0529 0.6594 1.8429 5.1471 1.1847 0.5916 0.5592 1.3356 0.7797
179 3.2244 0.6426 1.5981 3.7155 1.0032 0.9006 0.5493 0.969 0.7743
180 1.2123 0.7893 1.4283 3.4014 1.0482 1.5978 0.5244 0.9681 0.6081
210
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
181 1.0554 0.813 2.4921 1.9584 1.1733 1.5948 0.5259 0.9684 0.6156
182 0.9993 0.7755 5.7213 1.8852 1.1922 0.6849 0.4194 0.7902 0.6267
183 2.433 0.6528 3.6171 3.4845 1.3986 0.5835 0.3405 0.8553 0.9183
184 2.4282 0.4002 1.6848 3.3165 2.0043 0.5874 0.3366 0.8418 1.7391
185 0.9405 0.3945 1.8054 2.1429 2.0775 0.9015 0.3306 0.2964 1.7277
186 4.497 0.3969 1.9032 1.8627 1.7631 1.6038 0.3444 0.255 1.6728
187 1.0131 0.4026 3.0813 2.118 1.7976 1.6026 0.3576 0.3192 3.981
188 2.6664 0.6018 5.6985 2.4885 5.2947 0.7536 0.5739 0.8007 5.6553
189 2.4165 0.6291 2.1051 2.1024 3.0375 0.5841 0.5526 0.7929 5.6469
190 1.1157 0.6345 1.749 0.9756 2.1204 0.5061 0.552 0.6633 5.6472
191 1.4268 0.6441 1.8378 0.6909 1.6596 0.7452 0.6606 0.3192 2.511
192 1.4214 0.6579 2.3733 2.2599 1.3164 1.5162 0.5574 0.3213 0.7416
211
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
193 3.489 0.8196 3.9252 2.4162 1.3239 1.5105 0.4974 0.3267 0.9333
194 2.709 0.8271 1.3185 0.8721 1.4616 0.7434 0.3648 0.7926 0.816
195 1.419 0.828 1.3347 1.5747 7.4691 0.5286 0.3486 0.8163 0.8172
196 4.488 0.8154 1.4037 3.078 7.5033 0.5496 0.3504 0.7284 0.9342
197 4.464 0.7428 5.607 3.2169 11.1978 1.3098 0.3438 0.3915 0.7455
198 2.6919 0.5802 2.2998 1.5063 6.9564 1.6611 0.522 0.3837 0.375
199 0.8571 0.4764 1.3995 0.6378 6.5859 1.6152 0.5679 0.3447 0.4737
200 0.8466 0.471 1.3161 2.3604 5.0319 0.9072 0.5529 0.7566 0.4827
201 2.8521 0.7356 4.6806 2.4066 2.8887 0.6054 0.5421 0.8418 0.5823
202 4.9788 0.8247 2.0196 0.9219 4.1127 0.6135 0.5319 0.798 0.5715
203 3.5331 0.7851 1.545 1.0335 6.0771 3.9348 0.3747 0.372 0.6102
204 1.1475 0.7383 1.4286 2.6826 1.7082 4.4358 0.3459 0.3558 0.6507
212
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
205 2.9904 0.7938 2.8875 2.2191 1.5354 2.4408 1.5288 0.3753 0.4803
206 3.3444 0.9345 4.1931 2.733 1.4907 1.8492 1.5165 0.6024 0.4254
207 1.1391 0.8946 1.4466 6.3288 4.023 1.0776 1.5093 0.8319 0.4587
208 5.1105 0.9342 1.9863 4.2432 5.7747 1.1916 0.6252 0.8145 0.4143
209 2.6076 0.864 6.2757 2.5203 1.7652 3.9582 3.2505 0.3567 0.4503
210 3.2319 0.7095 4.4022 1.1247 5.3706 6.5448 6.6435 0.3414 0.6945
211 1.8897 0.6963 1.8741 1.2732 4.7454 3.3219 6.6342 0.3369 0.6948
212 1.7994 0.7185 2.019 2.5236 2.5998 0.9723 6.6339 0.5424 0.6825
213 7.1145 0.7428 4.1616 2.1558 6.0528 0.9117 6.6666 0.8322 0.6555
214 3.5301 0.7053 6.3534 0.9171 5.9361 1.8723 6.7866 0.8067 0.501
215 2.2425 0.5637 4.1367 1.3293 5.973 6.9351 7.0026 0.5121 0.4518
216 1.1694 0.5532 1.9818 3.5667 3.1503 6.8316 7.0059 0.3576 0.4503
213
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
217 3.9093 0.4857 2.1984 7.6467 1.9359 1.2414 6.9519 0.3321 0.4842
218 7.0011 0.5718 5.0322 2.1273 6.138 0.9063 6.9588 0.5262 0.5139
219 2.0895 0.7908 6.732 1.1916 3.7941 3.0858 6.8484 0.8313 4.0473
220 0.9336 0.9591 6.5556 2.3991 1.26 6.6843 6.8493 0.8376 5.238
221 2.2572 0.8871 2.1651 1.7049 5.1846 6.5859 6.8139 0.3633 5.2413
222 8.0343 0.8139 1.9533 0.735 6.1218 6.6309 6.9549 0.3273 5.1693
223 3.8475 0.7386 4.5624 1.485 2.493 6.5844 6.9915 0.3384 2.9568
224 1.2456 0.7236 6.6669 2.6394 1.2564 6.5988 6.9963 0.4617 0.4815
225 2.4219 0.6867 4.3695 2.5314 3.5076 6.5334 6.9843 0.8004 0.4893
226 6.1206 0.7551 2.04 6.3645 6.1155 6.6423 7.0119 0.8118 0.4419
227 5.3445 0.7404 3.5163 4.0071 4.7418 6.7185 6.8937 0.3942 0.5874
228 1.3845 0.7224 6.4536 2.244 1.2645 6.6153 6.9375 0.3324 0.6693
214
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
229 2.4903 0.6732 5.8089 1.3278 3.1779 6.5922 6.7782 0.4824 0.672
230 3.6006 0.5133 2.2293 0.4803 6.5526 6.5976 6.9345 0.4209 0.6729
231 8.5683 0.5574 3.699 1.3761 5.0817 6.5988 7.0374 0.7962 0.5697
232 2.7798 0.6675 9.9906 2.2368 2.2185 6.6105 7.0035 0.7938 0.4767
233 1.7592 0.8901 10.1871 1.4784 1.8252 6.5604 6.9702 0.4092 0.4779
234 3.4455 0.8694 6.5373 4.7235 6.4245 6.5472 6.9735 0.3159 0.4839
235 5.6802 0.7707 6.5262 7.7436 6.6999 6.5391 2.6082 0.3063 0.4941
236 5.6277 0.6612 11.0361 5.337 3.4401 6.6594 0.9885 0.6672 0.7047
237 1.4193 0.6738 10.758 5.0139 2.103 6.6717 4.0359 0.7839 0.6969
238 3.3111 0.6768 6.4335 4.0344 2.8185 6.5856 6.8412 0.7794 0.6921
239 4.2738 0.6837 5.8287 3.4263 6.9372 6.6096 6.8469 0.4221 0.7044
240 7.6131 0.7827 8.8146 4.9224 6.0981 6.5514 6.9978 0.3063 0.5607
215
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
241 2.7789 1.6611 10.3008 6.6135 2.1534 6.5637 6.1752 0.3054 0.5154
242 1.2153 1.7271 6.672 10.2477 2.0355 6.5253 1.4823 0.7788 0.5148
243 3.357 1.9956 6.0486 10.4013 6.501 6.5391 2.0892 0.7683 3.2811
244 6.4059 2.0439 9.0516 9.8607 6.6606 6.5907 7.212 0.7722 5.0289
245 5.5686 1.9977 10.3338 8.79 2.1897 6.5727 7.0551 0.4626 5.1969
246 1.5129 3.7314 6.6081 7.1943 2.0916 6.5385 6.9117 0.3099 5.1921
247 1.8636 6.2616 5.8395 8.2671 5.9157 6.4116 6.909 0.3087 5.1912
248 3.5958 6.7881 8.4606 9.096 6.8124 6.4146 6.8997 0.6243 1.0785
249 3.4824 6.9678 10.2975 9.1125 2.6592 6.5529 7.1271 0.7617 0.6102
250 5.7027 6.3885 7.0398 9.0786 1.9557 6.5028 7.2054 0.7566 0.567
251 2.5836 5.9637 5.9955 6.0477 5.7594 6.4491 3.2583 0.4689 0.5838
252 2.7264 5.1282 5.8113 9.5091 6.627 6.3513 1.2258 0.2931 0.6609
216
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
253 3.4179 1.8915 9.2082 8.0469 2.1906 6.2679 1.2297 0.2922 0.9102
254 3.0585 1.9227 10.104 6.7101 2.0295 6.2634 1.1106 0.5793 1.2573
255 1.5057 2.1006 6.6369 9.6522 6.1896 6.2511 1.131 0.9093 1.4175
256 1.6092 2.2851 6.4272 4.6758 6.7953 6.2775 1.0626 0.7689 1.9266
257 3.5472 6.2457 7.3377 9.0756 2.4525 6.2418 1.2093 0.51 1.6089
258 3.5214 6.4338 10.6359 5.8137 2.1405 2.169 1.3065 0.3006 1.8387
259 4.9278 6.4314 8.9721 7.5615 5.4462 1.7496 1.2819 0.3012 1.6047
260 2.5437 5.9946 6.5619 6.8499 6.7002 6.1848 2.1024 0.3738 1.6086
261 2.6946 5.9604 6.4914 5.4432 2.6115 1.758 2.4924 0.7683 1.3266
262 3.2763 5.9475 8.4636 8.2749 2.0904 0.6711 2.3694 0.765 1.5669
263 1.7904 5.9169 10.5351 3.8517 2.8173 0.6684 1.6374 0.537 2.7534
264 1.1829 5.9856 6.9459 10.0293 6.6858 5.5938 3.2997 0.2835 6.0456
217
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
265 2.373 5.9553 6.5148 10.5615 4.4934 5.0772 7.0491 0.282 5.9274
266 3.5439 5.9361 5.5077 11.8896 2.0931 0.6702 7.0758 0.324 5.0991
267 2.3379 5.9436 7.0542 12.7263 1.9635 0.6657 7.0578 0.7773 5.2782
268 1.6581 5.9244 5.0811 9.1302 3.3633 3.8625 7.0383 0.7767 5.3736
269 1.8426 5.9286 2.4603 14.4507 6.8241 4.9614 7.0023 0.5703 2.466
270 3.5904 6.0666 2.5848 11.3358 4.2966 0.6618 2.3991 0.2835 1.3197
271 2.6073 6.1191 3.246 14.9463 2.2839 0.6597 1.0956 0.2823 1.3128
272 1.2888 6.072 5.4795 14.7951 2.2725 2.3118 1.0866 0.2904 1.3557
273 1.185 6.0723 2.1996 14.6688 3.8436 6.0351 1.0794 0.7608 1.8204
274 1.6962 6.0642 1.3077 13.6758 6.8571 1.0911 1.2165 0.7596 1.1826
275 3.354 5.9514 1.2672 11.9703 3.3396 0.6594 1.3668 0.5979 0.6996
276 1.3434 5.8491 2.4081 11.3787 2.1579 0.7788 1.3494 0.2805 0.699
218
Table A.3 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
277 1.2018 5.844 5.6355 11.4999 3.9585 6.0279 1.3236 0.2781 0.7008
278 1.1925 5.8302 5.5308 9.6873 6.6831 2.6094 1.3071 0.2775 0.7014
279 1.1379 5.8398 5.3451 9.3456 6.5391 0.6651 2.2116 0.7389 0.7752
280 0.945 5.8533 5.2191 8.8311 5.6433 0.6669 2.3739 0.7599 0.9261
281 0.8892 5.8413 5.3103 8.6712 2.0928 4.7085 2.3643 0.627 1.0245
282 1.3176 5.8287 5.1231 8.292 3.1905 4.1193 1.5144 0.2772 5.2884
283 2.8503 5.9292 5.0535 6.5349 6.36 0.6732 1.2483 0.2772 5.517
284 1.0269 5.7759 4.9494 7.1334 3.4842 0.6582 1.3581 0.4173 6.4692
285 1.0818 5.772 4.9287 8.1105 1.7088 3.1887 1.3392 0.7602 5.517
286 1.2252 5.7504 5.0907 8.1138 2.0715 6.0207 1.3239 0.7593 5.061
287 1.2975 5.0844 5.0811 7.8609 6.1995 6.0042 1.3233 0.6609 1.6233
288 1.3047 4.7304 3.987 1.2378 5.1447 6.0483 1.1601 0.2772 1.2528
219
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
220
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
221
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
222
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
223
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
224
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
225
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
226
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
227
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
100 1.1421 1.9566 0.438 1.203 0.5223 0.858 5.3505 1.3035 0.4644
101 0.9636 2.0313 0.4377 2.2659 0.7398 2.8839 4.1898 1.3749 0.582
102 0.981 1.9026 0.4395 0.8628 0.7032 1.4283 4.1784 1.3029 0.5802
103 1.4379 1.8348 0.5958 0.828 2.4645 0.7365 4.2165 1.3914 0.5784
104 1.4136 1.5936 1.251 0.7074 3.1683 0.7386 4.0362 1.6269 0.5724
105 1.0287 1.5633 1.2525 0.6474 0.8646 0.7449 3.2469 1.1355 2.2884
106 0.7908 3.1404 1.2498 0.735 0.96 0.7497 8.0658 1.1001 2.6169
107 0.7911 3.4878 1.2498 0.9354 0.8718 1.3188 8.3556 1.2105 0.3786
108 0.8058 1.341 0.9492 1.5534 0.3597 2.9169 7.8849 1.0944 0.3798
228
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
109 1.2765 1.4199 0.5955 2.2287 0.3759 1.0659 7.8546 1.0671 0.3798
110 1.2429 1.5378 0.5865 1.0341 0.4263 0.8808 7.8465 1.0533 0.3768
111 0.9573 1.5369 0.4425 0.813 0.5208 0.8673 7.8507 1.0485 0.5748
112 0.6888 1.5366 0.438 0.7431 0.4875 0.8709 4.6935 1.0119 0.5877
113 0.8904 1.5003 0.4377 0.6456 0.6822 0.7467 2.8563 1.0776 0.5838
114 0.903 1.3476 0.4413 0.5367 3.8805 1.9929 2.5497 1.0524 0.5829
115 1.269 3.1542 0.4404 0.5475 1.0854 2.2737 2.547 1.0497 0.5985
116 1.3764 3.5988 1.0536 0.5337 0.5181 0.8007 2.5902 1.0527 4.2552
117 1.2393 1.3527 1.0875 0.6651 0.4722 0.7845 2.5968 1.0473 5.9445
118 1.1739 1.3479 0.924 0.6993 0.4215 0.9396 2.5629 1.0539 4.2045
119 0.8319 1.3005 0.4422 0.6384 0.4215 0.9714 2.5203 1.0605 3.5763
120 0.9765 1.266 0.6096 0.6279 0.4203 0.9591 2.2488 0.9876 4.9278
229
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
121 1.0098 1.4316 0.6051 0.5124 0.5508 2.736 2.208 0.7035 4.758
122 0.9951 1.4646 0.6057 0.4053 0.4509 1.869 2.733 0.6906 2.9793
123 0.84 1.4589 0.6054 0.4065 0.4131 0.8379 2.0961 0.6864 3.0819
124 0.5019 2.7027 0.603 0.4125 0.408 0.7638 1.9932 0.6768 3.1281
125 0.507 4.2969 0.6012 0.4689 0.3894 0.7887 2.2032 0.6753 1.4802
126 0.5169 1.2969 1.0692 0.6504 0.3588 0.8601 2.2155 0.6822 1.0419
127 0.852 1.2615 4.1178 0.6348 0.2823 1.2336 2.2125 0.6954 7.3863
128 0.9816 1.2594 4.4988 0.6228 1.9077 2.9625 2.2425 0.6864 3.2361
129 0.9717 1.2591 3.4587 0.5118 2.8029 1.1988 2.2062 0.6891 2.6952
130 0.7134 1.2612 0.4938 0.4086 0.486 0.9831 2.1498 0.6885 2.5773
131 0.7158 1.2624 0.4404 0.4113 0.3486 1.0518 2.1348 0.6897 5.3427
132 0.6957 1.2588 0.4389 0.4134 0.3396 0.9138 2.1246 0.6795 8.4648
230
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
133 1.0077 2.7879 0.4386 0.447 0.408 0.9324 2.1234 0.6321 6.2826
134 1.1784 4.2534 0.4407 0.6552 0.4224 1.893 2.1255 0.6174 6.3396
135 1.0977 1.4601 0.4965 0.6471 0.4185 2.7309 2.1666 0.6138 5.9823
136 0.6936 1.2771 1.0839 0.6339 0.507 0.9384 2.1174 0.6015 8.3826
137 0.7014 1.2966 1.2474 0.5628 0.4893 0.9675 2.1351 0.7029 6.6006
138 0.5271 1.4583 1.2426 0.411 0.4104 0.9531 2.3214 0.6969 4.632
139 0.9141 1.4433 1.2429 0.4056 0.402 0.9612 2.3772 0.6897 3.555
140 0.9993 1.3737 0.7863 0.4104 0.3993 1.0878 2.4102 0.693 4.065
141 0.9642 2.5545 0.6045 0.4251 0.6246 2.8812 2.3736 0.6939 4.1061
142 0.519 4.2294 0.6015 0.4758 3.9168 2.3157 2.4603 0.6906 1.1355
143 0.5367 1.2936 0.5994 0.6663 0.6717 1.242 2.331 0.6912 0.7275
144 0.5124 1.2612 0.5727 0.6408 0.2382 0.93 2.3877 0.6885 0.7101
231
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
145 0.87 1.3356 0.4368 0.6252 0.24 0.7872 2.3724 0.684 2.9571
146 1.1376 1.2603 0.4377 0.537 0.3054 2.2368 2.4237 0.6894 1.6086
147 1.1877 1.2966 0.5631 0.4143 0.3405 2.6745 2.3499 0.7122 0.7038
148 0.7317 1.488 1.0899 0.4209 0.3234 1.1136 2.3565 0.7314 0.7251
149 0.6999 4.4262 1.0797 0.4806 0.4182 0.9162 2.1936 0.8175 1.3755
150 0.7026 3.0378 1.0842 1.1349 0.417 1.1166 2.2377 0.9636 4.0527
151 1.086 1.4664 3.1851 2.0247 0.4632 1.3923 2.166 0.9078 0.8955
152 1.1496 1.4985 4.7025 0.9312 3.9321 3.1146 2.1735 0.8604 0.6945
153 0.9888 1.5279 3.1272 0.8811 1.3239 1.8219 2.0985 0.9129 0.6981
154 0.534 1.5255 0.6069 0.8895 0.4017 1.0551 3.1893 1.1001 1.5534
155 0.5424 1.548 0.6066 0.7488 0.399 1.2675 2.0568 1.002 3.4338
156 0.54 3.2076 0.6054 0.8676 0.3471 1.9626 2.0577 1.0764 0.7401
232
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
157 0.9309 4.5771 0.6048 1.9008 0.3708 3.6393 2.0541 1.281 0.9072
158 1.011 1.2933 0.6042 4.9983 0.2406 1.7961 2.3964 1.3095 0.7557
159 0.9975 1.398 0.6915 1.1181 0.2382 1.0839 2.163 1.2837 2.7717
160 0.5748 1.4421 1.2297 1.3224 2.4663 1.3863 2.1693 1.119 2.5725
161 0.6327 1.4463 1.2237 1.2828 2.3871 2.8689 2.1789 1.0968 0.816
162 0.7086 1.2732 1.0767 5.7111 0.3714 3.0885 2.4858 1.1823 0.8403
163 1.0131 4.7721 1.0809 6.0036 0.2334 1.2366 2.3109 1.3449 0.8055
164 1.1808 2.6751 0.69 5.5335 0.4215 1.0854 2.2371 1.152 4.2678
165 1.1706 1.4415 0.4437 5.6226 0.4152 1.5537 2.4861 0.9552 1.4847
166 0.7728 1.3521 0.4404 5.6421 0.4239 3.1707 2.2113 0.954 0.8085
167 0.6939 1.2606 1.563 5.7183 0.5592 2.0025 2.1531 0.9537 0.8046
168 0.5274 1.263 4.8009 5.7261 3.8418 1.0545 2.1435 0.9537 2.3706
233
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
169 0.9639 4.0032 4.8258 5.9844 1.5588 1.0287 2.1444 1.0833 2.8176
170 1.3356 3.2181 2.4579 6.0123 0.5382 3.0105 2.1486 1.4571 0.6336
171 1.2579 1.3098 0.6123 6.1011 0.5946 3.3657 2.9169 1.3746 0.5331
172 0.9087 1.5114 1.2516 6.0444 0.6285 1.587 2.2428 1.3725 1.9818
173 1.2822 1.5306 1.2498 5.4381 0.492 1.5693 2.1198 1.4274 3.039
174 4.083 1.5045 1.2489 5.5563 2.7549 2.4807 2.1162 1.4718 0.5616
175 1.638 4.8597 1.2381 6.4404 3.5475 3.4425 2.1339 1.2198 0.5433
176 1.6632 3.1767 0.8736 6.6582 0.3489 2.0601 2.0892 1.0323 0.5244
177 2.013 1.2519 0.5931 5.9028 0.378 0.9657 2.0655 1.134 4.0062
178 1.5687 1.2585 0.5856 5.8113 0.2409 0.963 2.1519 1.3737 0.978
179 1.3137 1.2759 0.4488 5.5149 0.3597 2.319 1.5393 1.4289 0.5082
180 1.2792 2.4585 3.3084 4.785 2.1381 3.2202 1.3197 1.368 0.5793
234
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
181 3.6339 4.842 4.8012 1.092 3.4263 1.3017 1.3467 0.858 1.0188
182 3.3555 1.2864 4.2555 3.87 0.6144 1.0116 1.3368 1.059 3.6231
183 1.5615 1.3029 0.6 1.7721 0.4101 1.2351 1.44 1.194 0.4845
184 1.0938 1.4346 0.7971 0.7692 0.4674 3.0828 1.7157 1.2315 0.333
185 1.0788 1.4427 1.0725 0.7542 0.3978 2.4789 1.7622 1.2219 0.5382
186 1.0905 3.6822 1.0752 0.7569 2.9334 0.858 1.6896 1.4271 0.5262
187 2.4387 3.4938 1.0758 0.858 2.6469 0.8625 1.3326 3.5574 1.6044
188 3.5106 1.2648 0.9174 0.8391 0.2331 1.3749 1.1073 1.2567 3.8019
189 1.2234 1.2501 0.6021 0.5883 0.219 2.9448 1.1115 1.2312 0.7395
190 0.7164 1.2507 0.6018 0.5769 0.2193 1.6425 1.302 1.2258 0.5106
191 0.7767 1.2522 0.6006 0.3636 0.2268 1.029 1.2918 1.1142 0.5109
192 0.93 1.8195 0.8811 0.3657 0.3735 1.0464 1.278 1.0845 0.5124
235
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
193 1.1721 4.7145 4.5321 0.3684 1.6761 2.3004 1.2747 0.9114 1.6134
194 2.9748 1.9857 4.6938 0.3786 3.1845 3.1995 1.2726 0.861 3.6855
195 2.3493 2.4642 2.6496 0.3858 0.4845 1.2957 1.3197 0.8607 0.69
196 0.9336 2.3235 0.4758 0.5877 0.4218 1.0953 1.2693 0.8613 0.5241
197 0.9993 4.0677 1.0575 0.6321 0.5829 1.5174 1.2528 0.864 0.4362
198 1.2501 7.407 1.0782 0.6165 0.4743 3.5622 1.2561 0.867 0.3564
199 1.8879 7.7847 1.0719 0.5967 0.5832 2.6325 1.2591 0.8601 1.9287
200 3.2346 6.7647 0.7728 4.4472 3.4695 1.1301 1.413 0.8043 3.2655
201 1.3437 6.8133 0.4371 0.9855 2.9547 1.3068 1.4109 0.8286 0.4161
202 0.8736 9.192 0.4338 0.3876 0.6231 2.646 1.4046 0.8622 0.3642
203 0.8661 9.9759 0.4362 0.3831 0.4155 3.7584 1.4073 0.7899 0.3603
204 1.1673 6.8481 0.4377 0.3882 0.3822 1.6923 1.4088 0.8892 0.4833
236
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
205 3.8631 7.0134 0.5877 1.7718 0.3861 1.4571 1.3887 1.4643 0.5721
206 3.8295 7.0494 3.9423 5.4318 2.1867 1.3569 1.4253 1.7859 4.1652
207 2.07 10.4799 5.1501 1.8759 3.7929 3.2784 4.1613 1.737 1.7382
208 1.5276 9.1482 4.1619 0.6135 0.447 2.6184 5.1321 1.1004 0.5391
209 1.5267 6.6885 0.6978 0.5301 0.3213 1.2615 1.3674 0.9678 0.5427
210 2.3097 6.621 1.026 0.3777 0.4224 1.1637 1.365 1.0806 0.5301
211 4.3344 7.7394 1.2342 1.6962 0.6399 2.478 1.3593 1.1175 0.5247
212 2.5683 9.8982 0.9603 5.1861 4.1763 3.4452 1.3545 1.0452 1.2366
213 1.8738 6.9924 0.504 1.3248 3.4572 1.5033 1.4061 0.9711 4.1454
214 1.5855 5.8716 0.4404 0.444 0.5028 1.2432 1.4685 0.9633 1.0131
215 2.2623 4.3647 0.4386 0.6288 0.4941 1.8285 1.4595 0.9618 0.5106
216 4.4469 4.6851 0.4389 1.2912 0.6306 3.4086 1.6401 0.9678 0.3375
237
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
217 2.4012 4.3467 4.6608 5.4369 0.5157 2.7036 7.2276 0.9804 0.3321
218 1.5978 1.3389 5.0694 3.0738 3.8346 1.2534 7.2732 0.999 1.0548
219 1.9269 1.2651 3.7653 0.3831 2.8251 1.2492 7.287 1.1259 4.0038
220 3.8649 1.2687 0.4791 0.3822 0.3882 2.6682 7.3071 1.0278 0.5673
221 3.7569 4.4037 0.9096 4.0062 0.465 3.3837 7.2396 0.9027 0.4551
222 1.4103 3.7968 1.1853 4.722 0.3468 1.5762 7.1019 0.8643 0.5385
223 1.3893 1.2552 1.2492 0.5562 1.9179 1.2555 7.0791 0.8622 1.0134
224 3.0483 1.2564 1.005 0.612 4.1982 2.1423 7.1445 0.8604 4.1598
225 4.0554 1.5768 0.6045 5.1597 1.185 3.3519 7.0512 0.8592 0.7449
226 1.3866 5.1906 0.6084 5.4573 0.3585 2.8125 7.0935 0.8616 0.5301
227 1.1034 3.2472 2.9892 5.4714 0.3657 1.1319 7.131 0.8631 0.4461
228 2.5602 1.401 5.2563 3.6 0.4557 1.062 7.0809 0.861 3.2142
238
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
229 4.2066 1.4325 5.3004 0.3765 3.8937 2.6403 7.0677 0.9504 1.7802
230 2.337 4.5942 5.2506 1.71 3.0015 3.2631 7.2666 0.9093 0.3297
231 1.7556 4.5048 3.2877 5.2335 0.4491 1.9221 7.4094 0.8844 0.3285
232 3.1059 1.62 0.4665 3.9237 0.7908 1.1046 7.4073 0.8736 0.3969
233 4.4445 1.5996 1.0293 0.4935 0.9681 1.4814 4.0248 0.8676 4.1526
234 2.8629 3.1905 1.1061 0.6582 4.3647 3.303 1.5768 0.8637 1.2531
235 1.3353 5.3325 1.1049 5.4435 3.2826 3.1089 6.8184 0.9105 0.5262
236 2.6274 2.3673 0.8934 4.6515 0.8067 1.1286 6.7686 0.8949 0.5145
237 4.2717 1.5741 0.4656 0.6 0.8991 0.9405 6.744 0.8622 3.1227
238 2.7636 1.4013 0.465 0.3753 0.8892 2.6646 6.7461 0.8628 2.0898
239 1.3008 4.6086 4.3902 1.0521 4.1652 3.1518 6.7212 0.8616 0.3255
240 2.4912 4.1469 5.2272 5.1909 2.8497 1.3101 6.7122 0.8562 0.3171
239
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
241 4.1712 2.0277 5.2602 3.2349 0.2814 0.9336 6.7179 0.9849 0.552
242 3.0675 7.0497 1.0071 0.3792 0.3798 2.1738 6.6984 1.2132 3.9396
243 1.2636 8.8785 0.6261 0.528 1.1124 3.2325 6.5376 1.3302 1.164
244 1.2801 10.6665 1.134 5.1048 4.1241 1.8789 6.5508 1.3068 0.57
245 4.0698 8.2626 1.2591 4.3761 1.1001 1.0377 6.4947 1.3407 0.5682
246 4.1025 7.2087 1.257 0.6156 0.5172 2.1375 6.4719 1.323 0.5556
247 2.097 6.8613 1.2531 0.4251 0.4686 3.0885 6.4572 1.3605 4.0884
248 1.2657 11.0898 1.7283 0.3642 1.9236 1.8696 6.4515 1.3326 1.9809
249 1.7295 10.0887 5.001 3.2736 4.1115 0.8817 6.4728 1.3347 0.4047
250 3.9489 7.488 5.0559 3.84 0.9045 1.8549 6.5058 1.3794 0.3483
251 4.1538 6.921 4.2492 0.3756 0.5013 3.0594 7.0995 1.0506 0.3477
252 1.5405 9.423 0.6096 0.4326 0.3852 2.0883 7.4601 0.7794 2.4405
240
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
253 1.3797 10.6275 0.4665 0.6033 2.6322 0.978 7.4115 0.6882 3.3477
254 2.7384 7.1226 0.4662 1.89 3.5787 1.797 7.3095 0.6651 0.5892
255 4.254 6.972 3.4956 5.0583 0.6156 3.1497 7.2804 1.3935 0.6549
256 3.0147 8.7159 4.9155 0.7854 0.4515 2.1114 7.3977 4.6917 0.6549
257 1.5555 10.6524 4.944 0.4626 0.6003 0.9087 8.0589 3.5001 1.5909
258 1.5258 8.2068 0.912 0.5895 4.0602 1.8378 8.0649 0.702 4.1526
259 4.0158 6.996 0.9522 0.6162 3.0951 3.0237 8.0526 0.8991 1.1307
260 3.6819 6.8391 1.1277 3.4479 0.9378 1.8366 8.0295 4.4121 0.4755
261 1.3425 8.3133 0.6351 3.774 0.2694 0.8469 8.025 4.4091 0.4734
262 2.2584 10.3533 0.6342 0.8706 0.9504 1.8585 8.1336 4.4565 0.6861
263 3.9915 7.4064 0.6306 0.9753 4.0884 3.1104 8.1036 4.4874 3.9003
264 3.9408 6.8688 0.6207 0.9303 1.6851 1.9251 8.0451 4.6926 1.6278
241
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
265 3.909 7.4127 2.7222 0.8559 0.4308 0.9642 8.031 4.0344 0.6645
266 3.5457 6.9759 4.7871 2.352 1.2903 1.7754 8.6001 1.4121 0.6843
267 3.6036 8.226 3.6981 5.5575 1.9353 3.0711 8.5752 4.5258 0.6831
268 3.4956 4.9947 0.5406 1.0302 5.0202 1.9413 3.7266 5.1954 0.6768
269 3.2697 2.9076 0.9231 0.7215 2.4735 0.8373 3.186 4.1226 0.6714
270 3.2637 2.5812 1.1148 0.7287 1.3392 0.8367 3.7278 1.3848 0.6591
271 3.2574 1.5924 1.116 0.9432 1.3341 2.9295 4.1499 3.0222 0.4743
272 3.2415 4.7997 1.0527 4.9368 2.1573 2.3568 4.5714 4.914 0.4743
273 3.2388 3.6519 1.1259 5.5125 4.9758 0.9615 3.384 4.3332 0.4755
274 3.1158 4.8402 1.3584 5.4387 2.2248 0.972 3.6003 1.3131 0.4755
275 3.0597 4.7937 1.4235 5.3355 1.8291 1.6371 3.6525 2.2674 0.4761
276 3.0549 4.5777 3.7035 2.0769 4.8612 3.1026 3.6513 5.2518 1.1781
242
Table A.3 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
277 3.0579 3.2628 5.7054 1.4079 5.0004 1.7793 4.4451 4.8828 1.1775
278 3.0528 3.273 5.8104 4.9845 4.9953 0.8838 8.7486 1.3782 0.4779
279 3.0633 3.2205 3.4698 3.3642 2.2026 0.8475 8.463 1.2585 0.7155
280 3.066 3.9522 1.3302 0.606 1.1433 2.3772 7.9248 3.8475 0.6678
281 3.0585 1.1175 1.9641 0.5979 2.6886 2.5161 8.2182 4.9434 0.6261
282 3.1719 1.2042 2.052 3.57 4.7712 0.849 8.0235 2.1 0.6216
283 3.3126 1.2975 1.98 3.7083 4.2885 0.9006 8.0196 0.9156 0.6183
284 3.2604 1.3011 1.8201 0.9129 1.2555 0.9825 8.0145 3.6537 0.6162
285 3.2499 1.2984 1.3086 4.9116 2.0994 3.0744 7.9719 4.4127 0.612
286 3.2409 1.2969 2.3355 2.3574 5.4948 2.3628 8.0244 1.2405 0.6039
287 3.2328 1.1715 5.3244 5.0889 4.6413 0.978 8.0808 0.7809 0.6018
288 3.1677 1.1061 5.4663 2.1294 1.7856 0.9537 8.0337 3.3765 0.5997
243
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
244
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
245
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
246
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
247
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
248
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
249
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
250
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
251
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
252
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
253
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
254
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
255
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
256
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
257
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
258
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
259
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
260
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
261
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
262
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
263
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
264
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
265
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
266
Table A.3 Continued
Bus 29 30 31 32 33 34
Time (5-min)
267
Table A.4 34-node System 5-min Resolution Renewable Profile (kW)
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
1 0.007 0 0 0 0 0 0 0 0
2 0.007 0.0022 0 0 0 0 0 0 0
3 0.007 0.0002 0 0 0 0 0 0 0
4 0.0016 0 0 0 0 0 0 0 0
5 0 0 0 0 0 0 0 0 0
6 0 0.0028 0 0 0 0 0 0 0
7 0 0.003 0 0 0 0 0 0 0
8 0 0 0 0 0 0 0 0 0
9 0 0 0 0 0 0 0 0 0
10 0 0.0018 0 0 0 0 0 0 0
11 0 0.0036 0 0 0 0 0 0 0
12 0 0 0 0 0 0 0 0 0
268
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
13 0 0 0 0 0 0 0 0 0
14 0 0 0 0 0 0 0 0 0
15 0 0 0 0 0 0 0 0 0
16 0 0.003 0 0 0 0 0 0 0
17 0 0.0008 0 0 0 0 0 0 0
18 0 0 0 0 0 0 0 0 0
19 0 0 0 0 0 0 0 0 0
20 0 0 0 0 0 0 0 0 0
21 0 0.0006 0 0 0 0 0 0 0
22 0 0.003 0 0 0 0 0.0046 0 0
23 0 0 0 0 0 0 0.004 0 0
24 0 0 0 0 0 0 0 0 0
269
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
25 0 0 0 0 0 0 0 0 0
26 0 0 0 0 0 0 0 0 0
27 0 0 0 0 0 0 0 0 0
28 0 0.0018 0 0 0 0 0 0 0
29 0 0.002 0 0 0 0 0 0 0
30 0 0 0 0 0 0 0 0 0
31 0 0 0 0 0 0 0 0 0
32 0 0 0 0 0 0 0 0 0
33 0 0 0 0 0 0 0.0042 0 0
34 0 0 0 0 0 0 0.0048 0 0
35 0 0 0 0 0 0 0 0 0
36 0 0 0 0 0 0 0 0 0
270
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
37 0 0.0006 0 0 0 0 0 0 0
38 0 0.0026 0 0 0 0 0 0 0
39 0 0 0 0 0 0 0 0 0
40 0 0 0 0 0 0 0 0 0
41 0 0 0 0 0 0 0 0 0
42 0 0 0 0 0 0 0 0 0
43 0 0 0 0 0 0 0 0 0
44 0 0 0 0 0 0 0.0048 0 0.0002
45 0 0 0 0 0 0 0 0 0
46 0 0 0 0 0 0 0 0 0.0002
47 0 0 0 0 0 0 0 0 0
48 0 0 0 0 0 0 0 0 0
271
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
49 0 0 0 0 0 0 0 0 0
50 0 0 0 0 0 0 0 0 0
51 0 0 0 0 0 0 0 0 0
52 0 0.0028 0 0 0 0 0 0 0
53 0 0.0016 0 0 0 0 0.0002 0 0
54 0 0 0 0 0 0 0.0038 0 0
55 0 0 0 0 0 0 0 0 0
56 0 0 0 0 0 0 0 0 0
57 0 0 0 0 0 0 0 0 0
58 0 0.003 0 0 0 0 0 0 0
59 0 0.0006 0 0 0 0 0 0 0
60 0 0 0 0 0 0 0 0 0
272
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
61 0 0 0 0 0 0 0 0 0
62 0 0 0 0 0 0 0 0 0
63 0 0 0 0 0 0 0 0 0
64 0 0.0006 0 0 0 0 0 0 0
65 0 0.002 0 0 0 0 0 0 0
66 0 0 0 0 0 0 0 0 0
67 0 0 0 0 0 0 0 0 0
68 0 0 0 0 0 0 0 0 0
69 0 0 0 0 0 0 0 0 0
70 0 0 0 0 0 0 0 0 0
71 0 0 0 0 0 0 0 0 0
72 0 0 0 0 0 0 0 0 0
273
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
73 0 0 0 0 0 0 0 0 0
74 0 0 0 0 0 0 0 0 0
75 0 0 0 0 0 0 0 0 0
76 0 0 0 0 0 0 0 0 0
77 0 0 0 0 0 0 0 0 0
78 0 0.0018 0 0 0 0 0 0 0
79 0 0 0 0 0 0 0 0 0
80 0 0 0 0 0 0 0 0 0
81 0 0 0 0 0 0 0 0 0
82 0 0.0006 0 0 0 0 0 0 0
83 0 0.0012 0 0 0 0 0 0 0
84 0 0 0 0 0 0 0 0 0
274
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
85 0 0 0 0 0 0 0 0 0
86 0 0 0 0 0 0 0 0 0
87 0 0 0 0 0 0 0 0 0
88 0 0.0018 0 0 0 0 0 0 0
89 0 0.0002 0 0 0 0 0 0 0
90 0 0 0 0 0 0 0 0 0
91 0 0 0 0 0 0 0.0072 0 0.0036
92 0 0 0 0 0 0 0.0072 0 0.0148
275
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
104 0.0298 0.0444 0.0704 0.0856 0.001 0.0788 0.167 0.0276 0.1592
106 0.0642 0.0648 0.116 0.14 0.0074 0.138 0.2642 0.0458 0.2518
107 0.1016 0.1088 0.1726 0.2206 0.0908 0.2106 0.2406 0.1642 0.2892
108 0.1132 0.1106 0.2046 0.2338 0.118 0.2428 0.1738 0.2214 0.2122
276
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
109 0.1208 0.1314 0.221 0.2714 0.13 0.27 0.2194 0.261 0.1436
110 0.063 0.0962 0.148 0.1878 0.0596 0.1816 0.259 0.1906 0.3418
111 0.0554 0.083 0.121 0.159 0.0482 0.14 0.2032 0.1648 0.1974
112 0.1508 0.1656 0.2468 0.3216 0.1438 0.3162 0.3164 0.316 0.2976
113 0.0766 0.0904 0.147 0.1784 0.0662 0.1716 0.3744 0.1822 0.2868
114 0.1456 0.1702 0.2482 0.321 0.1514 0.304 0.468 0.3264 0.4506
115 0.1352 0.1438 0.2334 0.2834 0.1458 0.2886 0.3954 0.2776 0.4234
116 0.2296 0.249 0.365 0.483 0.2546 0.4496 0.729 0.4774 0.6332
117 0.1956 0.185 0.3158 0.3848 0.2168 0.3976 0.7322 0.3622 0.6888
118 0.3742 0.3712 0.5532 0.7372 0.4428 0.7048 0.5416 0.7056 0.6914
119 0.3648 0.318 0.5568 0.6804 0.4636 0.7074 0.6186 0.6194 0.5784
120 0.35 0.3302 0.514 0.678 0.3912 0.6814 0.442 0.634 0.5774
277
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
121 0.3176 0.3124 0.4534 0.6398 0.3538 0.5964 0.3534 0.6008 0.396
122 0.2646 0.2624 0.4066 0.529 0.2866 0.5258 0.2748 0.503 0.2094
123 0.1778 0.1732 0.328 0.349 0.2316 0.3794 0.5236 0.3332 0.392
124 0.0958 0.1112 0.1754 0.21 0.095 0.2022 0.994 0.2148 0.9146
125 0.221 0.2144 0.3552 0.4338 0.2732 0.4278 0.765 0.411 0.8766
126 0.5494 0.4868 0.7828 1.0402 0.6734 1.0206 0.5142 0.9444 0.5092
127 0.4214 0.3746 0.6728 0.7762 0.5314 0.8454 0.7174 0.7136 0.6748
128 0.2552 0.2058 0.4144 0.4606 0.3042 0.5066 0.961 0.4074 0.9272
129 0.3668 0.4002 0.5404 0.7374 0.4114 0.713 1.0964 0.734 1.258
130 0.5322 0.5184 0.7446 1.043 0.6108 0.9984 0.3624 0.9856 0.5112
131 0.6908 0.5872 1.0108 1.2752 0.8882 1.3144 0.9442 1.1374 0.5636
132 0.2178 0.2568 0.3432 0.4572 0.214 0.4468 1.3128 0.457 1.3018
278
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
133 0.3944 0.3424 0.54 0.739 0.5192 0.7162 1.498 0.6552 1.4672
134 0.8072 0.6596 1.0232 1.5212 1.0128 1.4702 1.1052 1.2866 1.377
135 0.8742 0.7966 1.1258 1.6712 0.998 1.589 1.8212 1.5018 1.446
136 0.7192 0.6854 1.044 1.4028 0.8084 1.3876 1.0322 1.2864 1.6458
137 0.8782 0.711 1.2392 1.5624 1.2516 1.6264 1.0098 1.3688 0.93
138 0.8588 0.6836 1.1412 1.5214 1.0372 1.6274 1.405 1.3432 1.198
139 0.519 0.4418 0.726 0.9384 0.63 1.0102 1.8532 0.8452 1.8584
140 0.7784 0.5534 0.9718 1.3864 1.1152 1.3894 1.3168 1.102 1.352
141 1.2522 0.8314 1.5142 1.9752 1.6216 2.1618 2.2364 1.6204 2.7942
142 0.7548 0.5386 0.9502 1.2798 0.9892 1.34 0.9018 1.039 1.5276
143 1.9566 1.649 1.934 3.934 1.985 3.0828 0.6304 3.2984 0.7196
144 0.6518 0.6956 0.977 1.3138 0.694 1.3102 0.9978 1.256 1.0552
279
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
145 0.3388 0.4188 0.5282 0.7142 0.2878 0.7008 2.0756 0.7222 1.686
146 0.5628 0.6382 0.8226 1.1352 0.595 1.1134 2.0006 1.1368 2.7012
147 1.0914 0.9298 1.3512 2.0096 1.358 1.9332 0.9516 1.773 1.3606
148 2.05 1.7582 1.8626 4.0534 1.9448 3.3424 1.9388 3.3998 1.8194
149 0.6518 0.681 0.849 1.2916 0.603 1.2592 0.847 1.2414 1.1592
150 1.2132 0.8242 1.257 2.282 1.239 2.1672 0.7806 1.7072 0.6744
151 0.5062 0.5196 0.7712 0.976 0.5896 1.0176 1.4676 0.9146 1.0136
152 0.3778 0.3412 0.5336 0.6836 0.4708 0.7248 1.3518 0.6108 1.1742
153 0.6252 0.58 0.8622 1.1868 0.774 1.1518 3.4676 1.0572 2.7036
154 0.6408 0.5538 1.0784 1.2008 0.9848 1.261 3.355 1.026 3.0642
155 1.9792 1.6688 2.3968 3.6498 2.5604 3.488 3.8724 3.123 4.564
156 1.7882 1.5036 2.3782 3.1736 2.3368 3.3122 1.854 2.8786 2.9338
280
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
157 2.6322 2.1138 3.3502 4.7164 3.438 4.9104 1.7642 3.9966 1.294
158 1.7112 1.5588 2.1196 3.398 1.9806 2.9832 1.5926 2.9924 1.768
159 0.9128 0.7504 1.158 1.6142 1.1974 1.6056 0.861 1.4266 1.15
160 0.926 0.7514 1.3924 1.6392 1.37 1.733 0.8562 1.425 0.8566
161 0.6216 0.4806 0.8264 1.058 0.7494 1.1702 1.0234 0.941 0.9416
162 0.5002 0.4018 0.6464 0.9152 0.6268 0.9154 1.181 0.7816 1.109
163 0.5688 0.4934 0.7522 1.06 0.6956 1.06 1.2118 0.9292 1.2164
164 0.6314 0.5882 0.8996 1.1872 0.774 1.2092 1.236 1.09 1.187
165 0.6732 0.6288 0.969 1.253 0.812 1.284 1.6202 1.1644 1.3442
166 0.6526 0.608 0.9582 1.2232 0.795 1.2584 1.245 1.1428 1.577
167 0.7708 0.7224 1.115 1.452 0.9288 1.491 1.9794 1.349 1.1138
168 0.8096 0.8172 1.1496 1.5434 0.8862 1.6128 2.4282 1.4668 2.6064
281
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
169 0.6892 0.6972 1.027 1.3302 0.8426 1.311 2.0016 1.2934 1.907
170 1.4882 1.199 2.2246 2.6094 2.0584 2.7736 3.4082 2.2928 3.0278
171 1.0658 0.929 1.5154 1.9194 1.3348 2.0054 4.2296 1.71 4.2568
172 2.0416 1.7892 2.7192 3.72 2.53 3.5726 4.6178 3.3454 5.0648
173 2.547 2.5716 3.2366 5.0794 3.1438 4.379 1.4808 4.5732 2.6356
174 2.9668 2.6692 3.7024 5.5828 3.738 5.344 0.655 4.8222 0.8676
175 1.3392 1.3588 1.7852 2.709 1.6162 2.5314 0.9144 2.5144 0.8698
176 0.401 0.3998 0.5906 0.776 0.4468 0.8076 0.7346 0.7324 0.8468
177 0.4994 0.5012 0.7302 0.9744 0.6218 0.9398 0.5096 0.9202 0.5532
178 0.4418 0.3946 0.6754 0.814 0.5676 0.8508 0.7378 0.7376 0.6714
179 0.2784 0.2582 0.4232 0.521 0.3316 0.5396 0.5536 0.4676 0.649
180 0.36 0.3542 0.58 0.7006 0.4484 0.7182 0.2268 0.6638 0.3698
282
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
181 0.3012 0.292 0.5036 0.5806 0.3746 0.6078 0.2456 0.5606 0.2178
182 0.1564 0.151 0.2522 0.3014 0.1744 0.3128 0.2284 0.28 0.2538
183 0.1082 0.1166 0.1686 0.2204 0.105 0.2118 0.2088 0.2066 0.2136
184 0.1016 0.1254 0.1914 0.2184 0.0976 0.217 0.2118 0.2234 0.2012
185 0.0826 0.097 0.1628 0.177 0.0658 0.179 0.2188 0.1874 0.2084
186 0.08 0.0938 0.1576 0.1732 0.0608 0.1742 0.2314 0.1828 0.2244
187 0.0876 0.107 0.1646 0.1864 0.0698 0.1872 0.2518 0.1922 0.2476
188 0.097 0.1144 0.1768 0.206 0.0822 0.2086 0.3614 0.2064 0.3076
189 0.113 0.1254 0.1936 0.2394 0.1022 0.2406 0.4554 0.23 0.4024
190 0.1568 0.1644 0.2518 0.3202 0.1576 0.3222 0.5904 0.3002 0.5086
191 0.211 0.2184 0.3308 0.4178 0.2272 0.4268 0.8046 0.3856 0.694
192 0.2894 0.2812 0.422 0.559 0.324 0.5674 1.048 0.4966 0.932
283
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
193 0.4172 0.3814 0.5792 0.7904 0.4784 0.7914 1.3384 0.6846 1.1844
194 0.5592 0.5046 0.7726 1.055 0.661 1.0564 2.6812 0.9156 1.6166
195 0.6976 0.6402 0.987 1.3086 0.8512 1.3304 5.3054 1.1654 4.54
196 1.0136 1.0218 1.4868 1.9546 1.2122 1.9476 3.8724 1.8762 4.8516
197 2.6128 2.6546 3.7088 4.938 3.2698 5.1066 1.6004 4.7718 1.9564
198 2.3082 2.1704 3.824 4.089 3.1876 4.717 1.6158 4.0524 1.5874
199 0.9158 0.9252 1.5848 1.7006 1.1668 1.8212 3.8486 1.7162 2.754
200 0.8292 0.8624 1.4044 1.553 1.0242 1.6388 4.6896 1.5724 4.7162
201 1.69 1.9234 2.342 3.316 1.8208 3.23 3.0524 3.5212 3.9214
202 2.4198 2.2982 3.8912 4.3416 3.1036 4.9236 3.2022 4.2848 2.577
203 1.8334 1.9832 3.2476 3.3514 2.4482 4.009 4.2668 3.5484 3.9944
204 1.3498 1.606 2.472 2.7362 1.799 2.6374 2.7064 2.939 3.1446
284
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
205 2.037 1.6872 3.8894 3.2072 2.6892 4.185 2.0912 3.1186 2.3788
206 1.7248 1.396 2.8956 2.7938 2.21 3.5046 3.1938 2.509 3.701
207 1.314 1.5682 2.1538 2.647 1.4582 2.6532 2.829 2.8008 3.0018
208 1.8586 2.0874 2.806 3.4978 1.966 3.6448 1.7476 3.8108 2.1368
209 1.3618 1.4296 2.7082 2.4004 1.7616 2.9814 1.4966 2.5536 1.5358
210 1.03 1.0736 1.695 1.936 1.1634 2.0924 2.8142 1.8832 2.143
211 0.8512 0.986 1.3306 1.7432 0.9326 1.7054 3.2536 1.7276 2.4694
212 1.025 1.0776 2.2344 1.77 1.3558 2.0532 2.1436 1.8546 2.0336
213 1.2152 1.2932 2.7862 2.2972 1.7054 2.5982 2.3458 2.43 2.766
214 0.9362 1.0062 2.0256 1.674 1.2168 2.1778 2.7316 1.7174 3.1212
215 1.3052 1.5748 2.6662 2.6114 1.5932 2.8098 2.5294 2.9668 1.9298
216 1.4416 1.817 2.4528 2.8214 1.4032 2.9624 3.3382 3.2374 3.3506
285
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
217 0.8188 0.9982 1.9184 1.4878 1.0458 1.927 3.498 1.6826 3.151
218 1.4012 1.5818 3.3642 2.4336 1.848 3.2396 2.7356 2.7244 3.4654
219 1.352 1.708 3.2038 2.5356 1.7948 3.0982 1.6232 2.9148 2.333
220 1.5348 1.8444 3.2892 2.8462 1.9186 3.2842 1.9172 3.1432 1.8286
221 0.9622 1.3194 1.8672 1.8656 1.0592 2.1824 2.817 2.1434 2.2954
222 0.7906 1.0974 1.7324 1.5874 0.8902 1.8264 1.2046 1.822 2.1722
223 0.9472 1.2282 2.7808 1.662 1.329 2.332 2.4176 1.9256 1.2708
224 0.8534 1.2988 1.582 1.8058 0.7962 1.8672 3.136 2.0646 3.0574
225 0.6162 0.6652 1.739 1.0238 0.87 1.47 2.742 1.0442 2.846
226 1.1676 1.4406 3.1854 2.0024 1.4302 2.868 2.4292 2.274 2.4674
227 1.006 1.4238 2.9592 1.7792 1.08 2.5454 2.4342 2.1234 2.4182
228 0.8736 1.3028 2.6182 1.536 0.9144 2.2538 2.4644 1.9054 2.323
286
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
229 0.776 1.189 2.649 1.323 0.785 2.0994 2.3428 1.7322 2.2654
230 0.7574 1.1968 2.6072 1.271 0.7818 2.0566 2.0734 1.752 2.095
231 0.6844 1.1078 2.557 1.1344 0.7614 1.9868 2.0848 1.6406 2.0144
232 0.5928 1.0942 2.356 1.0132 0.5592 1.7912 1.74 1.617 1.9172
233 0.569 1.0748 2.2928 1.0034 0.4736 1.734 1.1272 1.6014 1.4694
234 0.5382 0.9916 2.0872 0.9796 0.4842 1.6248 0.7926 1.4828 0.9924
235 0.4664 0.8036 1.3802 0.9288 0.4668 1.2778 1.7144 1.2514 1.4068
236 0.4038 0.6518 0.7828 0.8314 0.4232 0.9382 1.7304 1.0098 1.7052
237 0.4542 0.8588 1.663 0.9218 0.4988 1.3316 1.1512 1.3456 1.4354
238 0.4266 0.9094 1.976 0.8536 0.4596 1.4556 0.8132 1.1 0.7868
239 0.3904 0.6818 1.379 0.7628 0.4288 1.1888 0.8702 0.7778 0.7796
240 0.2872 0.365 0.7952 0.561 0.3332 0.7234 0.9582 0.55 0.9448
287
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
241 0.277 0.393 0.9532 0.5392 0.3162 0.7286 0.8196 0.5298 0.7208
242 0.2372 0.4846 1.0588 0.481 0.2338 0.779 1.124 0.5282 0.8122
243 0.1872 0.3326 0.8472 0.363 0.188 0.6144 1.1078 0.4048 0.9674
244 0.2794 0.4588 1.0312 0.5378 0.2972 0.8032 0.9612 0.5784 0.813
245 0.264 0.5316 1.2156 0.5166 0.295 0.8748 0.8268 0.576 0.6406
246 0.195 0.4742 1.0912 0.4026 0.201 0.742 0.5324 0.4622 0.4928
247 0.1908 0.3896 0.8646 0.378 0.1882 0.643 0.3434 0.4192 0.4008
248 0.1742 0.2606 0.534 0.3468 0.1704 0.479 0.2006 0.3708 0.239
249 0.1432 0.1996 0.3848 0.2926 0.136 0.3572 0.1432 0.3108 0.1538
250 0.0756 0.1162 0.1934 0.166 0.0454 0.1938 0.096 0.1898 0.1026
251 0.045 0.075 0.117 0.1106 0.0058 0.1228 0.0598 0.1266 0.063
288
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
255 0 0.002 0 0 0 0 0 0 0
256 0 0 0 0 0 0 0 0 0
257 0 0 0 0 0 0 0 0 0
258 0 0 0 0 0 0 0 0 0
259 0 0 0 0 0 0 0 0 0
260 0 0 0 0 0 0 0.0068 0 0
261 0 0 0 0 0 0 0.009 0 0
262 0 0 0 0 0 0 0.001 0 0
263 0 0.0018 0 0 0 0 0 0 0
289
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
266 0 0 0 0 0 0 0 0 0
267 0 0.0012 0 0 0 0 0 0 0
271 0 0 0 0 0 0 0 0 0
272 0 0.0058 0 0 0 0 0 0 0
275 0 0 0 0 0 0 0.0006 0 0
276 0 0.005 0 0 0 0 0 0 0
290
Table A.4 Continued
Bus 2 3 4 5 6 7 8 9 10
Time (5-min)
283 0.0076 0 0 0 0 0 0 0 0
284 0.0062 0 0 0 0 0 0 0 0
287 0.0056 0 0 0 0 0 0 0 0
288 0.005 0 0 0 0 0 0 0 0
291
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
1 0 0 0 0 0 0 0 0 0
2 0 0 0 0.0004 0 0 0 0 0
3 0 0 0 0.0004 0 0 0 0 0
4 0 0 0 0.0006 0 0 0 0 0
5 0 0 0 0.001 0 0 0 0 0
6 0 0 0 0.0002 0 0 0 0 0
7 0 0 0 0 0 0 0 0 0
8 0 0 0 0 0 0 0 0 0
9 0 0 0 0 0 0 0 0 0
10 0 0 0 0 0 0 0 0 0
11 0 0 0 0 0 0 0 0 0
12 0 0 0 0.0004 0 0 0 0 0
292
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
13 0 0 0 0.001 0 0 0 0 0
14 0 0 0 0 0 0 0 0 0
15 0 0 0 0 0 0 0 0 0
16 0 0 0 0 0 0 0 0 0
17 0 0 0 0 0 0 0 0 0
18 0 0 0 0 0 0 0 0 0
19 0 0 0 0 0 0 0 0 0
20 0 0 0 0 0 0 0 0 0
21 0 0 0 0.0004 0 0 0 0 0
22 0 0 0 0.0008 0 0 0 0 0.0004
23 0 0 0 0 0 0 0 0 0
24 0 0 0 0 0 0 0 0 0
293
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
25 0 0 0 0 0 0 0 0 0
26 0 0 0 0 0 0 0 0 0
27 0 0 0 0 0 0 0 0 0
28 0 0 0 0 0 0 0 0 0
29 0 0 0 0 0 0 0 0 0
30 0 0 0 0 0 0 0 0 0
31 0 0 0 0 0 0 0 0 0
32 0 0 0 0 0 0 0 0 0
33 0 0 0 0 0 0 0 0 0
34 0 0 0 0 0 0 0 0 0
35 0 0 0 0 0 0 0 0 0
36 0 0 0 0 0 0 0 0 0.0002
294
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
37 0 0 0 0 0 0 0 0 0
38 0 0 0 0 0 0 0 0 0
39 0 0 0 0 0 0 0 0 0
40 0 0 0 0 0 0 0 0 0
41 0 0 0 0.0002 0 0 0 0 0
42 0 0 0 0 0 0 0 0 0
43 0 0 0 0 0 0 0 0 0
44 0 0 0 0 0 0 0 0 0
45 0 0 0 0 0 0 0 0 0
46 0 0 0 0 0 0 0 0 0
47 0 0 0 0 0 0 0 0 0
48 0 0 0 0 0 0 0 0 0
295
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
49 0 0 0 0 0 0 0 0 0
50 0 0 0 0 0 0 0 0 0
51 0 0 0 0 0 0 0 0 0
52 0 0 0 0.0002 0 0 0 0 0
53 0 0 0 0.0002 0 0 0 0 0
54 0 0 0 0 0 0 0 0 0
55 0 0 0 0 0 0 0 0 0
56 0 0 0 0 0 0 0 0 0
57 0 0 0 0 0 0 0 0 0
58 0 0 0 0 0 0 0 0 0
59 0 0 0 0 0 0 0 0 0
60 0 0 0 0 0 0 0 0 0.0002
296
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
61 0 0 0 0 0 0 0 0 0
62 0 0 0 0 0 0 0 0 0
63 0 0 0 0 0 0 0 0 0
64 0 0 0 0 0 0 0 0 0
65 0 0 0 0.0006 0 0 0 0 0
66 0 0 0 0.0008 0 0 0 0 0
67 0 0 0 0 0 0 0 0 0
68 0 0 0 0 0 0 0 0 0
69 0 0 0 0 0 0 0 0 0
70 0 0 0 0 0 0 0 0 0
71 0 0 0 0 0 0 0 0 0
72 0 0 0 0 0 0 0 0 0
297
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
73 0 0 0 0 0 0 0 0 0
74 0 0 0 0 0 0 0 0 0
75 0 0 0 0 0 0 0 0 0
76 0 0 0 0 0 0 0 0 0
77 0 0 0 0 0 0 0 0 0
78 0 0 0 0 0 0 0 0 0
79 0 0 0 0 0 0 0 0 0
80 0 0 0 0 0 0 0 0 0
81 0.0026 0 0 0 0 0 0 0 0
82 0.0104 0 0 0 0 0 0 0 0
83 0.0114 0 0 0 0 0 0 0 0
84 0.0306 0 0 0 0 0 0 0 0.0002
298
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
85 0.0374 0 0 0 0 0 0 0 0.001
86 0.0354 0 0 0 0 0 0 0 0.0004
87 0.063 0 0 0 0 0 0 0 0
88 0.0528 0 0 0 0 0 0 0 0
89 0.057 0 0 0 0 0 0 0 0
90 0.0594 0 0 0 0 0 0 0 0
91 0.1 0 0 0 0 0 0 0 0
92 0.103 0 0 0 0 0 0 0 0.0002
299
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
106 0.7038 0.0336 0.207 0.1764 0.136 0.1656 0.0492 0.117 0.215
107 0.7434 0.1852 0.317 0.2628 0.2062 0.2322 0.1574 0.1698 0.307
108 0.7178 0.2226 0.3612 0.2958 0.2212 0.2696 0.1896 0.185 0.3452
300
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
109 0.6028 0.2786 0.402 0.3302 0.2508 0.2918 0.2186 0.2 0.3784
110 0.585 0.1544 0.2718 0.2318 0.1768 0.205 0.1318 0.1312 0.2632
111 0.4384 0.1056 0.2188 0.1882 0.1526 0.149 0.0978 0.1288 0.2256
112 0.2342 0.361 0.4668 0.3822 0.2952 0.3354 0.263 0.2194 0.4298
113 0.46 0.139 0.259 0.2174 0.169 0.1926 0.1258 0.1376 0.257
114 1.0166 0.3466 0.4534 0.3706 0.2952 0.3156 0.253 0.2186 0.4204
115 0.9184 0.3028 0.4218 0.3448 0.263 0.3144 0.2324 0.2116 0.3898
116 0.5584 0.5748 0.6748 0.5438 0.437 0.4534 0.3964 0.3542 0.6186
117 0.6996 0.4582 0.5794 0.4684 0.3502 0.417 0.3346 0.2828 0.5246
118 0.984 0.925 1.0482 0.8272 0.651 0.693 0.6336 0.5464 0.931
119 1.3654 0.8774 1.0318 0.8062 0.6062 0.7276 0.6244 0.5304 0.9086
120 0.5082 0.8698 0.9902 0.7936 0.606 0.6962 0.5986 0.4848 0.87
301
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
121 0.6962 0.788 0.8868 0.7186 0.5678 0.6048 0.5382 0.4704 0.7936
122 1.4166 0.6596 0.7684 0.6278 0.473 0.5386 0.4562 0.362 0.6828
123 1.5422 0.3942 0.5496 0.4318 0.3206 0.4146 0.308 0.2718 0.5016
124 1.4044 0.1744 0.305 0.2496 0.1996 0.2246 0.1496 0.1654 0.2954
125 1.702 0.5068 0.6382 0.5026 0.3936 0.4166 0.366 0.3642 0.5866
126 1.6658 1.3386 1.5016 1.1836 0.9072 0.9912 0.9186 0.8108 1.3308
127 0.9884 1.032 1.1992 0.9468 0.6884 0.88 0.7246 0.5756 1.0528
128 1.4158 0.5704 0.7362 0.5802 0.4252 0.5276 0.4278 0.3842 0.6656
129 2.1984 0.9568 1.0372 0.8432 0.6544 0.6738 0.6234 0.5082 0.8976
130 1.4502 1.3478 1.4534 1.172 0.91 0.9572 0.9012 0.7504 1.2688
131 2.7842 1.6806 1.8846 1.4892 1.1054 1.3212 1.1832 0.987 1.662
132 1.4572 0.525 0.6428 0.5394 0.4126 0.446 0.3534 0.3092 0.5682
302
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
133 0.6816 0.875 1.0444 0.8172 0.6396 0.6828 0.6352 0.658 0.93
134 1.062 1.8286 1.9994 1.6566 1.2244 1.5406 1.3068 1.1258 1.7628
135 1.8452 2.1158 2.237 1.8436 1.3926 1.6276 1.4496 1.1502 1.9496
136 2.6432 1.8252 1.9564 1.5892 1.188 1.3878 1.233 0.9298 1.7042
137 1.1646 2.0676 2.4312 1.8174 1.3896 1.5234 1.503 1.4774 2.2202
138 1.6702 1.9902 2.1932 1.8314 1.3164 1.7492 1.501 1.1102 2.0112
139 1.0666 1.2256 1.4088 1.112 0.8242 0.9906 0.878 0.751 1.2362
140 0.739 1.6378 1.9264 1.5118 1.1184 1.5098 1.268 1.2138 1.7128
141 1.1562 2.7764 2.9964 2.4976 1.8064 2.3486 2.059 1.9016 2.7648
142 1.312 1.6562 1.9048 1.5426 1.0932 1.4684 1.2462 1.17 1.6888
143 3.137 4.3316 4.1576 3.8634 2.9452 3.2188 3.2064 2.6034 3.9172
144 3.077 1.7328 1.838 1.5344 1.0932 1.3124 1.1372 0.7734 1.6026
303
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
145 4.8022 0.9686 1.0128 0.8386 0.6506 0.5892 0.5784 0.4514 0.8508
146 3.0246 1.5636 1.6086 1.3054 1.0166 0.9316 0.9696 0.7824 1.3634
147 1.5682 2.541 2.682 2.2462 1.6608 1.9294 1.7886 1.6032 2.412
148 1.9496 4.4994 4.1544 3.9442 3.0378 4.0064 3.151 2.1198 3.6034
149 1.2174 1.6888 1.6484 1.4644 1.0496 1.3788 1.0282 0.7104 1.381
150 0.9436 2.692 2.8838 2.4448 1.8334 2.2434 1.9948 1.5646 2.6196
151 1.0358 1.2916 1.4022 1.162 0.8424 1.0962 0.8638 0.6658 1.2142
152 1.191 0.871 1.0088 0.8192 0.6002 0.767 0.6156 0.5558 0.8686
153 1.274 1.5246 1.6754 1.3206 1.0348 1.079 1.0272 0.9476 1.4484
154 1.2432 1.5178 1.7266 1.3666 0.9874 1.3328 1.0842 0.939 1.4988
155 1.403 4.5078 4.8056 3.814 2.8784 3.5334 3.1704 2.8656 4.3256
156 1.5388 4.4166 4.765 3.7756 2.8138 3.1842 3.049 2.7274 4.2334
304
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
157 1.2022 6.3206 6.9438 5.4276 3.88 4.803 4.2916 3.6386 6.0496
158 2.9498 4.1436 3.9504 3.4826 2.61 3.2198 2.6962 1.9578 3.401
159 1.9756 2.0454 2.3198 1.8162 1.381 1.4822 1.4784 1.4598 2.083
160 3.3534 2.1022 2.4502 1.8994 1.3694 1.8486 1.5644 1.3722 2.1834
161 4.7768 1.4454 1.6966 1.306 0.9612 1.0812 1.0602 0.8388 1.4942
162 5.3012 1.0978 1.2676 1.015 0.7568 0.976 0.819 0.7242 1.1286
163 2.6286 1.344 1.5064 1.1778 0.9078 1.053 0.9404 0.82 1.2942
164 0.8678 1.586 1.7378 1.3674 1.0448 1.178 1.0808 0.8856 1.4982
165 0.9692 1.678 1.8334 1.463 1.095 1.2976 1.159 0.8992 1.6022
166 0.9182 1.6444 1.7946 1.4318 1.0714 1.268 1.1318 0.8836 1.578
167 0.602 1.9644 2.1148 1.6834 1.2744 1.4778 1.3252 1.0356 1.8468
168 0.7258 2.1252 2.2142 1.8196 1.3364 1.624 1.3888 0.962 1.8912
305
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
169 0.6728 1.7798 1.9084 1.5064 1.19 1.2042 1.1724 1.0532 1.6756
170 0.3882 3.6732 4.1798 3.1862 2.3458 2.6134 2.5848 2.1864 3.6946
171 0.247 2.605 2.888 2.239 1.6888 1.9816 1.7966 1.4508 2.4566
172 0.2638 5.061 5.379 4.1748 3.2532 3.0888 3.3214 2.97 4.7988
173 0.2206 6.6134 6.5036 5.2178 4.4134 4.1516 3.963 3.5652 5.937
174 0.2102 7.4846 7.7352 5.9134 4.805 5.2354 4.3746 3.78 6.4142
175 0.2216 3.3708 3.4104 2.8284 2.1754 2.5414 2.1558 1.589 2.8508
176 0.2396 1.0088 1.1176 0.9288 0.6746 0.841 0.6794 0.5344 0.9606
177 0.2676 1.2622 1.3958 1.1012 0.8646 0.883 0.8532 0.7606 1.2284
178 0.3386 1.064 1.2338 0.9718 0.7228 0.8706 0.758 0.6482 1.1058
179 0.439 0.6354 0.7738 0.617 0.4624 0.566 0.4604 0.4092 0.7014
180 0.5628 0.9276 1.0576 0.83 0.6348 0.714 0.645 0.5284 0.9544
306
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
181 0.7688 0.7532 0.8788 0.708 0.525 0.6644 0.5458 0.4302 0.8198
182 1.0238 0.3068 0.4462 0.3676 0.2714 0.3662 0.2568 0.236 0.4308
183 1.2944 0.185 0.3128 0.2586 0.2026 0.2498 0.168 0.1718 0.3064
184 1.739 0.2128 0.3226 0.2734 0.206 0.2664 0.181 0.1538 0.3234
185 4.8498 0.1634 0.2688 0.2282 0.1738 0.2134 0.1396 0.125 0.2698
186 5.1754 0.1576 0.2616 0.2206 0.171 0.2022 0.133 0.1222 0.2602
187 2.094 0.1744 0.2796 0.2322 0.181 0.211 0.1436 0.1334 0.2726
188 1.7238 0.2016 0.311 0.2544 0.1986 0.2238 0.1602 0.1498 0.2914
189 2.7106 0.2472 0.3542 0.2874 0.2214 0.2382 0.1858 0.1774 0.3202
190 5.0676 0.3632 0.47 0.3748 0.288 0.3042 0.2588 0.24 0.4146
191 4.0544 0.5054 0.6124 0.4874 0.3724 0.4036 0.3506 0.3168 0.5366
192 2.7844 0.695 0.8108 0.6366 0.4912 0.5262 0.472 0.4306 0.6916
307
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
193 4.3334 1.0032 1.146 0.8824 0.6896 0.7274 0.6724 0.6072 0.9592
194 3.535 1.3682 1.5318 1.1762 0.9248 0.9758 0.9114 0.8078 1.2868
195 2.6614 1.7552 1.93 1.4918 1.1618 1.2316 1.1674 0.9984 1.6386
196 3.6714 2.6868 2.8798 2.2138 1.7164 1.76 1.7334 1.3786 2.4808
197 3.1156 7.5194 7.5308 5.711 4.6394 4.2246 4.2796 3.613 6.3896
198 2.1786 6.71 7.13 5.5556 4.0704 3.9144 4.1642 3.2262 6.014
199 1.5798 2.4972 2.7182 2.155 1.5984 1.6466 1.6242 1.3348 2.3874
200 2.2666 2.2776 2.4328 1.9294 1.4484 1.4486 1.4748 1.183 2.1256
201 2.8918 4.9886 4.8658 3.9362 3.0728 2.615 2.9742 2.1876 4.0886
202 2.2718 7.279 7.5216 5.824 4.3614 3.7714 4.4054 3.3264 6.3422
203 2.8222 5.4348 5.9778 4.307 3.1632 3.0288 3.3422 2.5826 4.9924
204 3.0286 3.9976 4.1066 3.0804 2.5354 2.0486 2.3448 1.9072 3.4438
308
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
205 2.0188 6.3618 6.4198 5.0182 3.534 3.0228 3.7678 2.7834 5.3278
206 3.515 5.0876 5.3452 4.0092 2.9036 2.618 3.0944 2.3732 4.4012
207 3.3742 3.991 3.867 3.1616 2.4332 2.031 2.3034 1.6612 3.1548
208 3.6118 5.8866 5.7308 4.5886 3.616 2.596 3.4272 2.5708 4.9788
209 2.3572 4.2588 4.6616 3.3902 2.5126 1.9882 2.6064 1.9506 3.8214
210 1.9382 2.987 3.0678 2.3914 1.8632 1.6688 1.7848 1.3488 2.5198
211 2.5516 2.4396 2.5454 1.9582 1.5888 1.3228 1.4746 1.1858 2.168
212 2.0336 3.1844 3.0806 2.509 1.8824 1.585 1.8422 1.3834 2.644
213 1.555 4.032 4.3962 3.0896 2.5524 1.6098 2.3618 1.8956 3.7642
214 3.1966 3.0632 3.2646 2.4142 1.7696 1.4684 1.7774 1.2228 2.5778
215 2.9198 4.5096 4.5984 3.4424 2.7752 1.7046 2.5804 1.9068 3.9456
216 2.5632 5.0644 4.8144 3.8084 3.0382 1.7878 2.7102 1.882 4.113
309
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
217 2.4788 2.8034 3.0928 2.2234 1.6384 1.159 1.6138 1.1094 2.4912
218 2.4 5.1046 5.2966 3.9032 2.9022 1.8448 2.83 1.918 4.3792
219 2.3572 4.9226 5.0478 3.7352 2.86 1.8098 2.6928 1.8388 4.1414
220 2.1338 5.3982 5.3262 4.1314 3.1494 2.0414 2.9794 2.0726 4.5654
221 2.0538 3.2016 3.348 2.5348 1.8696 1.3856 1.8166 1.1618 2.6422
222 1.934 2.7562 2.9322 2.1388 1.7272 1.0976 1.5486 1.1098 2.431
223 1.421 3.6506 4.0024 2.7308 2.064 1.1162 1.9804 1.3822 3.293
224 1.0186 3.2186 2.9222 2.4048 1.9218 1.0502 1.625 1.0084 2.395
225 1.4762 2.1386 2.461 1.6992 1.2388 0.8514 1.246 0.9534 2.0858
226 1.7212 4.597 4.7626 3.4724 2.5758 1.4822 2.473 1.5266 3.9414
227 1.3876 4.212 4.271 3.1812 2.32 1.2072 2.1872 1.2768 3.558
228 0.815 3.7668 3.8388 2.8096 2.0648 0.9526 1.9156 1.141 3.1878
310
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
229 0.8726 3.598 3.6444 2.6596 1.9246 0.791 1.682 0.9526 2.9922
230 0.9448 3.5078 3.5702 2.6086 1.9004 0.7478 1.5552 0.937 2.9298
231 0.7316 3.2812 3.4584 2.4626 1.7742 0.757 1.4572 0.8772 2.8366
232 0.8824 3.1328 3.172 2.2948 1.624 0.5672 1.2962 0.7538 2.5742
233 1.055 3.0226 3.0808 2.2358 1.511 0.5538 1.2702 0.7312 2.4866
234 0.9374 2.701 2.8364 2.0714 1.4312 0.5854 1.1966 0.7064 2.2562
235 0.8368 2.0624 2.173 1.6408 1.2348 0.6338 1.0116 0.6876 1.8316
236 0.585 1.4336 1.4498 1.1966 0.9446 0.6608 0.7736 0.5776 1.1924
237 0.446 2.0578 2.3594 1.7172 1.3342 0.674 1.0642 0.6904 1.9176
238 0.2488 2.2046 2.659 1.8746 1.4172 0.602 1.0994 0.6428 2.0396
239 0.162 1.7134 2.0222 1.4784 1.0958 0.644 0.9068 0.5502 1.555
240 0.108 0.98 1.163 0.8644 0.6414 0.5402 0.5886 0.4238 0.9726
311
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
241 0.0662 1.041 1.2784 0.9006 0.6966 0.4964 0.599 0.412 1.0576
242 0.0306 1.1354 1.4514 1.0116 0.7958 0.405 0.5928 0.3332 1.1102
243 0.004 0.8088 1.1036 0.757 0.5686 0.3404 0.4436 0.2676 0.8264
312
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
256 0 0 0 0 0 0 0 0 0
257 0 0 0 0 0 0 0 0 0
258 0 0 0 0 0 0 0 0 0
259 0 0 0 0 0 0 0 0 0
260 0 0 0 0 0 0 0 0 0
261 0 0 0 0 0 0 0 0 0
262 0 0 0 0.0004 0 0 0 0 0
263 0 0 0 0 0 0 0 0 0
264 0 0 0 0.0004 0 0 0 0 0
313
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
265 0 0 0 0 0 0 0 0 0
266 0 0 0 0 0 0 0 0 0
267 0 0 0 0 0 0 0 0 0
268 0 0 0 0 0 0 0 0 0
269 0 0 0 0 0 0 0 0 0
270 0 0 0 0 0 0 0 0 0
271 0 0 0 0 0 0 0 0 0
272 0 0 0 0 0 0 0 0 0
273 0 0 0 0 0 0 0 0 0
274 0 0 0 0 0 0 0 0 0
275 0 0 0 0.0002 0 0 0 0 0
276 0 0 0 0.0002 0 0 0 0 0
314
Table A.4 Continued
Bus 11 12 13 14 15 16 17 18 19
Time (5-min)
277 0 0 0 0 0 0 0 0 0
278 0 0 0 0.001 0 0 0 0 0
279 0 0 0 0.001 0 0 0 0 0
280 0 0 0 0.0004 0 0 0 0 0
281 0 0 0 0.0004 0 0 0 0 0
282 0 0 0 0.001 0 0 0 0 0
283 0 0 0 0.0008 0 0 0 0 0
285 0 0 0 0.001 0 0 0 0 0
286 0 0 0 0.0006 0 0 0 0 0
287 0 0 0 0 0 0 0 0 0
288 0 0 0 0 0 0 0 0 0
315
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
1 0 0 0 0 0 0.004 0 0 0
2 0 0 0 0 0 0.004 0 0 0
3 0 0 0 0 0 0.004 0 0 0
4 0 0 0 0 0 0.004 0 0 0
5 0 0 0 0 0 0.004 0 0 0
6 0 0 0 0 0 0.0012 0 0 0
7 0 0.0006 0 0 0 0.0034 0 0 0
8 0 0.0002 0 0 0 0.004 0 0 0
9 0 0 0 0 0 0.004 0 0 0
10 0 0 0 0 0 0.004 0 0 0
11 0 0 0 0 0 0.004 0 0 0
12 0 0 0 0 0 0.004 0 0 0
316
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
13 0 0 0 0 0 0.0012 0 0 0
14 0 0.0002 0 0 0 0.0032 0 0 0
15 0 0.0002 0 0 0 0.004 0 0 0
16 0 0 0 0 0 0.004 0 0 0
17 0 0 0 0 0 0.004 0 0 0
18 0 0 0 0 0 0.004 0 0 0
19 0 0 0 0 0 0.0032 0 0 0
20 0 0 0 0 0 0.002 0 0 0
21 0 0 0 0 0 0.0034 0 0 0
22 0 0.0006 0 0 0 0.004 0 0 0
23 0 0.001 0 0 0 0.004 0 0 0
24 0 0.001 0 0 0 0.0024 0 0 0
317
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
25 0 0.0008 0 0 0 0.0018 0 0 0
26 0 0 0 0 0 0.004 0 0 0
27 0 0 0 0 0 0.004 0 0 0
28 0 0 0 0 0 0.004 0 0 0
29 0 0 0 0 0 0.004 0 0 0
30 0 0 0 0 0 0.0034 0 0 0
31 0 0 0 0 0 0.0022 0 0 0
32 0 0 0 0 0 0.004 0 0 0
33 0 0 0 0 0 0.004 0 0 0
34 0 0 0 0 0 0.004 0 0 0
35 0 0 0 0 0 0.004 0 0 0
36 0 0 0 0 0 0.0026 0 0 0
318
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
37 0 0 0 0 0 0.0018 0 0 0
38 0 0 0 0 0 0.004 0 0 0
39 0 0 0 0 0 0.004 0 0 0
40 0 0 0 0 0 0.004 0 0 0
41 0 0 0 0 0 0.004 0 0 0
42 0 0 0 0 0 0.0028 0 0 0
43 0 0 0 0 0 0.0024 0 0 0
44 0 0 0 0 0 0.004 0 0 0
45 0 0 0 0 0 0.004 0 0 0
46 0 0 0 0 0 0.004 0 0 0
47 0 0 0 0 0 0.004 0 0 0
48 0 0 0 0 0 0.0024 0 0 0
319
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
49 0 0 0 0 0 0.003 0 0 0
50 0 0 0 0 0 0.004 0 0 0
51 0 0 0 0 0 0.004 0 0 0
52 0 0 0 0 0 0.004 0 0 0
53 0 0 0 0 0 0.004 0 0 0
54 0 0 0 0 0 0.0022 0 0 0
55 0 0 0 0 0 0.0032 0 0 0
56 0 0 0 0 0 0.003 0 0 0
57 0 0 0 0 0 0.0034 0 0 0
58 0 0 0 0 0 0.004 0 0 0
59 0 0 0 0 0 0.004 0 0 0
60 0 0 0 0 0 0.0026 0 0 0
320
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
61 0 0 0 0 0 0.0014 0 0 0
62 0 0 0 0 0 0.004 0 0 0
63 0 0 0 0 0 0.004 0 0 0
64 0 0 0 0 0 0.004 0 0 0
65 0 0 0 0 0 0.004 0 0 0
66 0 0 0 0 0 0.004 0 0 0
67 0 0 0 0 0 0.0026 0 0 0
68 0 0 0 0 0 0.002 0 0 0
69 0 0 0 0 0 0.004 0 0 0
70 0 0 0 0 0 0.004 0 0 0
71 0 0 0 0 0 0.004 0 0 0
72 0 0 0 0 0 0.004 0 0 0
321
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
73 0 0 0 0 0 0.004 0 0 0
74 0 0 0 0 0 0.0024 0 0 0
75 0 0 0 0 0 0.003 0 0 0
76 0 0 0 0 0 0.004 0 0 0
77 0 0 0 0 0 0.004 0 0 0
78 0 0 0 0 0 0.004 0 0 0
79 0 0 0 0 0 0.004 0 0 0
80 0 0 0 0 0 0.0038 0 0 0
81 0 0 0 0 0 0.0014 0 0 0
82 0 0.0014 0 0 0 0.0036 0 0 0
83 0 0.0024 0 0 0 0.004 0 0 0
84 0 0 0 0 0 0.004 0 0 0
322
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
85 0 0 0 0 0 0.004 0 0 0
86 0 0 0 0 0 0.004 0 0 0
87 0 0 0 0 0 0.0026 0 0 0
88 0 0.0002 0 0 0 0.0014 0 0 0
89 0 0 0 0 0 0.004 0 0 0
90 0 0 0 0 0 0.004 0 0 0
91 0 0 0 0 0 0.004 0 0 0
92 0 0 0 0 0 0.0094 0 0 0
323
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
324
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
325
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
326
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
327
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
328
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
161 1.0788 0.7808 1.0074 0.8616 0.1046 1.0714 0.9832 1.1312 0.988
162 0.9146 0.6878 0.7648 0.713 0.6512 0.9404 0.7896 0.855 0.803
163 1.0232 0.7892 0.921 0.8582 0.7844 1.0142 0.9592 0.9598 0.9004
164 1.112 0.8942 1.1104 0.9924 0.9248 1.1314 1.1922 1.1202 1.0408
165 1.1414 0.9632 1.1842 1.0462 0.9892 1.2202 1.3222 1.2322 1.1328
166 1.118 0.9372 1.1752 1.0264 0.9726 1.1922 1.2966 1.2044 1.097
167 1.3058 1.1156 1.374 1.212 1.15 1.3612 1.5406 1.3994 1.2884
168 1.2278 1.2432 1.4182 1.2958 1.2322 1.421 1.8008 1.5408 1.4058
329
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
169 1.2316 0.9568 1.295 1.181 1.0842 1.1638 1.321 1.239 1.0926
170 2.7958 1.9074 2.6686 2.089 2.0354 3.0176 2.486 2.7052 2.8122
171 1.838 1.45 1.7536 1.596 1.51 2.059 1.9356 1.9932 1.9184
172 3.5936 2.6696 3.428 3.0982 2.922 2.8346 3.1114 3.0872 2.6568
173 4.7418 3.4764 4.5618 4.2384 4.0124 3.8176 4.0452 3.8476 3.407
174 5.0552 4.1566 4.8306 4.5178 4.3058 5.301 5.0662 5.2286 5.1776
175 2.029 2.0652 2.0612 2.2272 2.0158 2.646 2.5898 2.3888 2.513
176 0.6704 0.623 0.7264 0.6598 0.6056 0.789 0.8824 0.7904 0.7408
177 0.9352 0.703 0.9302 0.8452 0.7692 0.8834 0.8916 0.8426 0.7766
178 0.8272 0.6146 0.8098 0.6762 0.6358 0.8726 0.8162 0.8306 0.7632
179 0.5288 0.3994 0.5088 0.4374 0.405 0.5646 0.531 0.5184 0.4774
180 0.6752 0.5304 0.7086 0.6044 0.5708 0.6952 0.7434 0.6794 0.612
330
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
181 0.5588 0.4544 0.6024 0.497 0.4776 0.6296 0.6616 0.6142 0.5518
182 0.3168 0.234 0.302 0.2516 0.2344 0.357 0.3218 0.313 0.2858
183 0.2378 0.1624 0.2064 0.1904 0.1706 0.244 0.216 0.1994 0.1786
184 0.2158 0.1698 0.2276 0.1988 0.1848 0.2482 0.2712 0.2268 0.1908
185 0.1794 0.1376 0.1948 0.169 0.1546 0.1998 0.2352 0.1834 0.153
186 0.175 0.1346 0.1896 0.166 0.1528 0.1886 0.2292 0.175 0.146
187 0.1856 0.145 0.2 0.1766 0.1604 0.1986 0.2376 0.1836 0.1566
188 0.202 0.1586 0.2168 0.1926 0.1738 0.2132 0.2536 0.1976 0.1718
189 0.2354 0.1816 0.2422 0.2184 0.1964 0.2332 0.274 0.214 0.1868
190 0.3104 0.2418 0.3126 0.2846 0.2578 0.3 0.347 0.2798 0.2534
191 0.402 0.3202 0.4004 0.3654 0.3326 0.396 0.4494 0.3782 0.351
192 0.538 0.4242 0.5112 0.4754 0.4308 0.52 0.564 0.4954 0.4722
331
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
193 0.758 0.5892 0.7042 0.6604 0.6008 0.7274 0.7598 0.6924 0.6666
194 1.0062 0.7796 0.9524 0.8768 0.8038 0.9784 1.0052 0.9346 0.8988
195 1.2476 0.9708 1.234 1.0934 1.017 1.229 1.2812 1.1984 1.1344
196 1.7194 1.4428 1.8734 1.6572 1.5562 1.668 2.0306 1.7418 1.59
197 4.4876 3.4694 5.4176 4.425 4.2566 4.4052 5.3894 4.8368 4.5472
198 4.185 3.1106 5.056 3.7286 3.6328 4.6554 4.9478 4.7912 4.8614
199 1.6444 1.2964 2.0484 1.5562 1.4738 1.7984 2.0166 1.779 1.7208
200 1.4534 1.187 1.8102 1.4196 1.3522 1.4442 1.8032 1.5366 1.3292
201 2.6056 2.4238 3.3584 3.1276 2.9504 2.1872 3.6608 2.7614 2.191
202 4.223 3.2726 5.448 3.943 3.9616 4.2764 5.308 4.7102 4.5492
203 3.2574 2.5696 4.1864 3.098 2.8886 3.655 4.0944 3.7642 4.0258
204 2.4372 1.9504 3.0058 2.6744 2.433 2.1632 2.824 2.633 2.2974
332
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
205 3.5498 2.4214 4.6472 3.015 3.04 3.536 4.6006 4.086 3.8882
206 3.1324 2.2004 3.7718 2.426 2.5452 3.2238 3.5604 3.4108 3.4304
207 1.9924 2.0008 2.6922 2.5332 2.357 1.7302 3.1154 2.2694 1.866
208 2.9838 2.6048 4.4038 3.443 3.3478 2.2278 4.1344 2.8542 2.4102
209 2.4778 1.8438 3.4056 2.3116 2.2408 2.6382 3.2356 2.6388 2.9108
210 1.6632 1.5054 2.073 1.777 1.6722 1.8494 2.2268 1.9356 1.9118
211 1.3472 1.2246 1.7648 1.6224 1.536 1.4718 1.8102 1.4484 1.5176
212 1.7304 1.4484 2.2774 1.695 1.6616 2.16 2.592 2.1114 2.306
213 2.3748 1.6582 3.5158 2.3184 2.3244 2.3354 2.8474 2.6202 2.6008
214 1.6018 1.315 2.343 1.6496 1.575 1.8036 2.6788 2.0734 1.9874
215 2.2802 1.9174 3.7632 2.7756 2.6596 2.064 3.2866 2.5632 2.3178
216 2.1656 2.0782 3.9446 3.0528 2.829 2.1726 3.813 2.4754 2.4678
333
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
217 1.4436 1.1314 2.4454 1.6188 1.5172 1.7696 2.424 1.9082 1.9162
218 2.4406 1.9084 4.321 2.6288 2.6144 2.603 4.1662 3.18 2.657
219 2.3846 1.9302 4.1092 2.802 2.6718 2.5712 3.9578 3.0474 2.4784
220 2.5868 2.148 4.426 3.0412 2.9334 2.383 4.0718 3.0622 2.2936
221 1.494 1.403 2.4258 2.0366 1.8168 1.463 2.8488 1.985 1.5282
222 1.3206 1.1304 2.425 1.8134 1.5542 1.3926 2.2242 1.6814 1.4726
223 1.798 1.282 3.4794 1.9556 1.6346 1.8676 2.972 2.2472 1.7014
224 1.1956 1.3754 2.2976 2.1006 1.847 1.228 2.7452 1.6198 1.241
225 1.261 0.7814 2.1546 1.1526 1.123 1.213 1.7818 1.4204 1.2018
226 2.017 1.5632 4.0756 2.4022 2.2694 2.1928 3.9914 2.8574 2.2732
227 1.6802 1.3266 3.8268 2.3076 2.0288 1.9062 3.7704 2.5876 2.0078
228 1.4954 1.1732 3.4922 2.1022 1.8108 1.4968 3.396 2.2056 1.611
334
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
229 1.285 0.9818 3.4334 1.9466 1.6874 1.4426 3.338 2.163 1.6526
230 1.268 0.938 3.3832 1.9406 1.6878 1.324 3.2724 2.0922 1.6008
231 1.1988 0.88 3.3392 1.7942 1.6306 1.2592 3.1604 2.0732 1.585
232 1.0018 0.7546 3.1582 1.7448 1.5428 0.9404 3.0404 1.8198 1.258
233 0.9592 0.761 3.0974 1.707 1.5356 0.821 2.9612 1.7348 1.2468
234 0.9196 0.753 2.7384 1.556 1.441 0.8134 2.7522 1.6128 1.1988
235 0.8348 0.712 2.0796 1.3116 1.2282 0.7458 1.9954 1.1612 0.953
237 0.8506 0.7248 2.2756 1.3834 1.3164 0.722 2.0272 1.2212 0.966
238 0.799 0.7022 2.6154 1.4358 1.3902 0.697 2.5304 1.4428 1.1544
239 0.6968 0.6276 1.7572 1.066 1.0548 0.7058 2.0352 1.2056 1.0138
240 0.5444 0.4446 0.965 0.606 0.5878 0.5646 0.987 0.7112 0.6366
335
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
241 0.5288 0.417 1.2178 0.6716 0.6484 0.5558 1.0074 0.7168 0.6518
242 0.4254 0.3882 1.416 0.78 0.7636 0.4538 1.421 0.7692 0.6616
243 0.3558 0.2952 0.9938 0.5258 0.517 0.3972 1.1308 0.6562 0.5956
244 0.4982 0.4238 1.3048 0.7112 0.7204 0.526 1.354 0.7974 0.672
245 0.4886 0.4138 1.5018 0.68 0.7472 0.5384 1.6216 0.8864 0.7156
246 0.3804 0.3154 1.357 0.5374 0.5548 0.4038 1.4882 0.7366 0.4786
247 0.3612 0.2924 1.154 0.4774 0.4728 0.3774 1.306 0.6488 0.4322
248 0.3352 0.2736 0.7012 0.4072 0.3868 0.3476 0.8334 0.4734 0.37
249 0.2876 0.2292 0.5108 0.3274 0.306 0.2972 0.5126 0.3382 0.2884
250 0.1614 0.1352 0.257 0.1868 0.1694 0.1852 0.288 0.1868 0.1596
251 0.1152 0.087 0.1554 0.1206 0.1068 0.134 0.1812 0.1154 0.0962
252 0.078 0.0528 0.0938 0.0748 0.0634 0.0922 0.1144 0.067 0.052
336
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
253 0.0394 0.0218 0.0534 0.041 0.0332 0.0514 0.071 0.0248 0.0168
256 0 0 0 0 0 0.019 0 0 0
257 0 0 0 0 0 0.0252 0 0 0
258 0 0 0 0 0 0.0254 0 0 0
259 0 0 0 0 0 0.0242 0 0 0
260 0 0 0 0 0 0.027 0 0 0
261 0 0 0 0 0 0.0128 0 0 0
262 0 0 0 0 0 0.0026 0 0 0
263 0 0 0 0 0 0.001 0 0 0
264 0 0 0 0 0 0.0022 0 0 0
337
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
269 0 0 0 0 0 0.004 0 0 0
270 0 0 0 0 0 0.004 0 0 0
271 0 0 0 0 0 0.001 0 0 0
273 0 0 0 0 0 0.0034 0 0 0
274 0 0 0 0 0 0.004 0 0 0
275 0 0 0 0 0 0.0028 0 0 0
276 0 0 0 0 0 0.001 0 0 0
338
Table A.4 Continued
Bus 20 21 22 23 24 25 26 27 28
Time (5-min)
277 0 0 0 0 0 0.0028 0 0 0
278 0 0 0 0 0 0.004 0 0 0
279 0 0 0 0 0 0.004 0 0 0
280 0 0 0 0 0 0.0018 0 0 0
281 0 0 0 0 0 0.002 0 0 0
282 0 0 0 0 0 0.004 0 0 0
283 0 0 0 0 0 0.004 0 0 0
284 0 0 0 0 0 0.004 0 0 0
285 0 0 0 0 0 0.001 0 0 0
286 0 0 0 0 0 0.002 0 0 0
287 0 0 0 0 0 0.004 0 0 0
288 0 0 0 0 0 0.004 0 0 0
339
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
1 0 0 0 0 0 0
2 0 0 0 0 0 0
3 0 0 0 0 0 0
4 0 0 0 0 0 0
5 0 0 0 0 0 0
6 0 0 0 0 0 0
7 0 0 0 0 0 0
8 0 0 0 0 0 0
9 0 0 0 0 0 0
10 0 0 0 0 0 0
11 0 0 0 0 0 0
12 0 0 0 0 0 0
340
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
13 0 0 0 0 0 0
14 0 0 0 0 0 0
15 0 0 0 0 0 0
16 0 0 0 0 0 0
17 0 0 0 0 0 0
18 0 0 0 0 0 0
19 0 0 0 0 0 0
20 0 0 0 0 0 0
21 0 0 0 0 0 0
22 0 0 0 0 0 0
23 0 0 0 0 0 0
24 0 0 0 0 0 0
341
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
25 0 0 0 0 0 0
26 0 0 0 0 0 0
27 0 0 0 0 0 0
28 0 0 0 0 0 0
29 0 0 0 0 0 0
30 0 0 0 0 0 0
31 0 0 0 0 0 0
32 0 0 0 0 0 0
33 0 0 0 0 0 0
34 0 0 0 0 0 0
35 0 0 0 0 0 0
36 0 0 0 0 0 0
342
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
37 0 0 0 0 0 0
38 0 0 0 0 0 0
39 0 0 0 0 0 0
40 0 0 0 0 0 0
41 0 0 0 0 0 0
42 0 0 0 0 0 0
43 0 0 0 0 0 0
44 0 0 0 0 0 0
45 0 0 0 0 0 0
46 0 0 0 0 0 0
47 0 0 0 0 0 0
48 0 0 0 0 0 0
343
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
49 0 0 0 0 0 0
50 0 0 0 0 0 0
51 0 0 0 0 0 0
52 0 0 0 0 0 0
53 0 0 0 0 0 0
54 0 0 0 0 0 0
55 0 0 0 0 0 0
56 0 0 0 0 0 0
57 0 0 0 0 0 0
58 0 0 0 0 0 0
59 0 0 0 0 0 0
60 0 0 0 0 0 0
344
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
61 0 0 0 0 0 0
62 0 0 0 0 0 0
63 0 0 0 0 0 0
64 0 0 0 0 0 0
65 0 0 0 0 0 0
66 0 0 0 0 0 0
67 0 0 0 0 0 0
68 0 0 0 0 0 0
69 0 0 0 0 0 0
70 0 0 0 0 0 0
71 0 0 0 0 0 0
72 0 0 0 0 0 0
345
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
73 0 0 0 0 0 0
74 0 0 0 0 0 0
75 0 0 0 0 0 0
76 0 0 0 0 0 0
77 0 0 0 0 0 0
78 0 0 0 0 0 0
79 0 0 0 0 0 0
80 0 0 0 0 0 0
81 0 0 0 0 0 0
82 0 0 0 0 0 0
83 0 0 0 0 0 0
84 0 0 0 0 0 0
346
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
85 0 0 0 0 0 0
86 0 0 0 0 0 0
87 0 0 0 0 0 0
88 0 0 0 0 0 0
89 0 0 0 0 0 0
90 0 0 0 0 0 0
91 0 0 0 0 0 0
92 0 0 0 0.0006 0 0
347
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
348
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
349
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
350
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
351
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
352
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
353
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
354
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
355
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
356
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
357
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
358
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
359
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
360
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
256 0 0 0 0 0 0
257 0 0 0 0 0 0
258 0 0 0 0 0 0
259 0 0 0 0 0 0
260 0 0 0 0 0 0
261 0 0 0 0 0 0
262 0 0 0 0 0 0
263 0 0 0 0 0 0
264 0 0 0 0 0 0
361
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
265 0 0 0 0 0 0
266 0 0 0 0 0 0
267 0 0 0 0 0 0
268 0 0 0 0 0 0
269 0 0 0 0 0 0
270 0 0 0 0 0 0
271 0 0 0 0 0 0
272 0 0 0 0 0 0
273 0 0 0 0 0 0
274 0 0 0 0 0 0
275 0 0 0 0 0 0
276 0 0 0 0 0 0
362
Table A.4 Continued
Bus 29 30 31 32 33 34
Time (5-min)
277 0 0 0 0 0 0
278 0 0 0 0 0 0
279 0 0 0 0 0 0
280 0 0 0 0 0 0
281 0 0 0 0 0 0
282 0 0 0 0 0 0
283 0 0 0 0 0 0
284 0 0 0 0 0 0
285 0 0 0 0 0 0
286 0 0 0 0 0 0
287 0 0 0 0 0 0
288 0 0 0 0 0 0
363