Professional Documents
Culture Documents
Distributed Generation
Xian Chen, Hieu Dinh, Bing Wang
Computer Science & Engineering Department, University of Connecticut, Storrs, CT 06269
AbstractSmart grid is envisioned to incorporate local distributed power generation for better efficiency and flexibility.
Distributed generation, when not used carefully, however, may
compromise the stability of the grid. Recently, researchers have
proposed innovative architectures (e.g., microgrid, LoCal grid)
that virtualize a local generator as a constant load, source, or zero
load to the grid, thus offering great promise to connect distributed
generation into the grid without sacrificing its reliability. In fact,
intuitively, using these architectures, distributed generation may
enhance the stability of the power grid. In this paper, we develop
a simulation model to quantify how much distributed generation
can mitigate cascading failures. Applying this model to IEEE
power grid test cases, we find that local power generation, even
when only using a small number of local generators, can reduce
the likelihood of cascading failures dramatically.
I. I NTRODUCTION
The electric grid is of vital importance to our daily life
and work. With careful control and management, it has been
operating with great reliability and robustness. However, largescale cascading failures did occur, as evidenced by the most
recent in 2003 when 50 million people in the U.S. Northeast
and Southeastern Canada lost power for up to several days [1].
Maintaining power system reliability and reducing the risk of
cascading blackouts is a critical issue.
Smart grid is envisioned to improve the efficiency, reliability, and flexibility of the current grid while reducing the
rate at which additional electric utility infrastructure needs to
be built [2]. To make this vision a reality, many advanced
technologies need be incorporated into the current grid, leading
to fundamental changes in the architecture of the grid. For instance, with distributed power generation (by local generators
and renewable energy sources), the power grid is transformed
from a centralized infrastructure with one-way power flow
(from the centralized power plants to distributed loads) to a
distributed infrastructure with two-way power flows.
Distributed generation, when not used carefully, may compromise the stability of the grid [3], [4]. Recently, innovative
architectures (e.g., microgrid [5], LoCal grid [6]) have been
proposed to virtualize a local generator as a constant load,
source, or zero load to the grid, greatly simplifying its impact
on the grid, and hence offering great promise to connect
distributed generation into the grid without sacrificing the
reliability of the grid. Indeed, when using these architectures,
as loads are being served locally, less power flows inside
the grid infrastructure. Therefore, intuitively, the stability and
reliability of the grid can be enhanced. In this paper, we investigate whether this is indeed the case, and quantify how much
73
(1)
|fn |1
n
max(|f1n |, . . . , |fm
|),
m
|fin |.
i=1
(2)
(3)
TABLE I
K EY NOTATION .
Notation
n
m
Pi
Li
Qi
f
fc
fn
Definition
Number of busses
Number of transmission lines
Amount of power generated by bus i
Amount of load at bus i
Actual power injection from bus i to the grid
Amount of active power flow in the transmission
lines, f = [f1 , . . . , fm ]T
c ]T
Capacity of transmission lines, fc = [f1c , . . . , fm
n ]T
Loading of transmission lines, fn = [f1n , . . . , fm
equations provide a good approximation of their AC counterparts. For simplicity, we use DC power flow equations in this
paper.
DC power flow equations can be found in many textbooks
on power systems (e.g., [18], [19]). For completeness, we also
describe them below. Let us first define some notation:
Let Q be a vector denoting the actual power injection
from busses. That is, Q = [Q1 , . . . , Qn ]T .
Let A0 denote the edge-node incidence matrix (i.e., rows
are lines and columns are nodes), where an item aij = 1
if line i exits bus j, aij = 1 if line i enters bus j, and
aij = 0 otherwise.
Let A be a matrix reduced from A0 . It is A0 with the
column associated with the slack bus removed.
Let C denote line property matrix. It is a diagonal matrix.
More specifically, C = diag(1/xi ) where 1/xi is the line
susceptance of line i.
T
Let B denote bus susceptance matrix, B = A CA.
Using the above notation, the DC power flow equations are
f = CAB1 Q.
(4)
B. Cascading failures
One transmission line outage may lead to a cascading
sequence of line outages since when one line trips, its power
flow redistributes to other lines in the grid, and may cause
them to be congested and tripped. We obtain the evolution
of cascading failures as follows [9]. Suppose a line, l, that
connects bus i to j is tripped. Let f be the vector of line
flows after the outage of line l. Then
f = C AB1 Q,
(5)
74
maximize:
n
Pi
(6)
or
i=1
subject to:
1
f = CAB Q
Q=PL
L = yL
y>0
|f| fc
n
n
Pi =
Li
i=1
(7)
(8)
(9)
(10)
(11)
(12)
i=1
(13)
(14)
Fig. 1. Determine the amount of load and generation at each bus in the
original power grid (i.e., before adding local generators), where > 0, >
0, Li (0, 1], i are given beforehand.
minimize: |fn |
(15)
minimize:
|fn |1
m
subject to:
(16)
fn = f/fc
f = CAB1 Q
(17)
(18)
Q=PL
|f| fc
n
n
Pi =
Li
(19)
(20)
i=1
0 Pi
i=1
Pi0 , bus
(21)
i is a central generator (22)
(23)
We make multiple simulation runs using independently generated seeds and use the percentage of runs ending up in grid
breakup as the metric to measure the vulnerability of the grid
with and without local generators.
We next describe the two LP problems in more detail.
The LP problem in Fig. 1 determines the load and power
injection at each bus in the original grid. In this problem,
> 0, > 0, and Li for each bus i are given beforehand.
The variables to be determined are y (and hence the actual
load Li = yLi at each bus i), and the amount of power
generated by each central generator. In our simulations, we
set close to 1.25 so that the most loaded line is close to
its protection trip point since we are interested in cascading
failures. The objective function maximizes the total amount of
power that is generated by the central generators. Since power
generation and demand have to be balanced, this optimization
problem also maximizes the total amount of load that is
supported by the grid. Constraints (7) to (10) specify how to
calculate the flows inside the grid. Constraint (11) states that
the amount of flow on each line is below times its capacity.
Constraint (12) specifies the relationship between the total
amount of power generation and load; the difference between
these two quantities is compensated through the slack bus.
Last, constraints (13) and (14) state that only central generators
generate power.
75
TABLE II
IEEE POWER SYSTEM TEST CASES .
Test case
IEEE 14
IEEE 57
IEEE 118
busses
14
57
118
Lines
20
80
186
Load busses
9
50
64
Central generators
4
6
53
76
1.2
Max loading
0.6
ideal
eps=0
eps=0.1
eps=0.2
0.8
0.6
0.4
0.2
0
0.5
0.4
0.3
0.2
0.1
5 10 15 20 25 30 35 40 45 50
1.5
0.9
Percentage of grid breakup
1.15
5 10 15 20 25 30 35 40 45 50
Simulation results for minimizing the maximum line loading using the IEEE 57 bus test case.
ideal
eps=0
eps=0.1
eps=0.2
1.2
1.1
1.05
1
0.95
0.9
12
ideal
eps=0.2
0.8
10
0.7
0.6
0.5
0.4
9
8
7
6
5
4
3
0.3
0.85
ideal
eps=0.2
11
Number of tripped lines
1.25
5 10 15 20 25 30 35 40 45 50
Number of local generators
Max loading
2.5
1
0
Fig. 3.
ideal
eps=0.2
0
0
0.8
0.2
0
10
20
30
40
50
60
70
1
0
0.25
30
40
50
60
70
0.2
0.15
0.1
0.05
0
5 10 15 20 25 30 35 40 45 50
Number of local generators
10
20
30
40
50
60
70
0.6
3.5
ideal
eps=0.2
0.5
0.4
0.3
0.2
0.1
0
Simulation results for minimizing the maximum line loading using the IEEE 118 bus test case.
ideal
eps=0
eps=0.1
eps=0.2
0.3
20
0.35
10
Fig. 4.
Average loading
3.5
ideal
eps=0.2
Number of tripped lines
1.4
ideal
eps=0.2
2.5
1.5
1
0
5 10 15 20 25 30 35 40 45 50
Number of local generators
5 10 15 20 25 30 35 40 45 50
Number of local generators
Simulation results for minimizing the average line loading using the IEEE 57 bus test case.
77
0.4
0.35
0.3
0.25
0.2
0.15
14
ideal
eps=0.2
0.9
Percentage of grid breakup
0.45
Average loading
ideal
eps=0
eps=0.1
eps=0.2
0.5
0.8
0.7
0.6
0.5
0.4
0.05
0.2
0
10
20
30
40
50
60
70
8
6
2
0
10
0.3
0.1
ideal
eps=0.2
12
Number of tripped lines
0.55
10
20
30
40
50
60
70
10
20
30
40
50
60
70
Simulation results for minimizing the average line loading using the IEEE 118 bus test case.
78