Professional Documents
Culture Documents
(DYMONDS) for Sustainable Energy Services
Marija
Ilic,
Professor
and
SRC
Smart
Grid
Research
Center
Director,
milic@ece.cmu.edu
Acknowledgment
Electric Energy Systems Group (EESG)
http://www.eesg.ece.cmu.edu
ERI
4
Outline
Electronic
technology
has
fundamentally
altered
the
way
we
live;
Communica7ons
Commerce
Entertainment
…
What
has
been
the
key
to
these
changes?
InformaGon
Technology
Integrated
circuits
have
developed
in
a
manner
that
provides
ever
growing
informaGon
handling
power
at
ever
decreasing
cost
Key
to
the
development
of
this
technology
and
its
wide
adopGon:
the
ability
to
model
and
design
these
systems
the
associated
development
of
so4ware
systems
+
ApplicaGon
to
Power
Systems
The
creaGon
of
“smart
grids”
is
the
applicaGon
of
informaGon
technology
to
the
power
system
while
coupling
this
with
an
understanding
of
the
business
and
regulatory
environment
CriGcal
to
the
creaGon
of
“smart
grids”
is;
development
of
models
of
the
power
system
development
of
command
and
control
soSware
incorpora7on
of
security,
communica7ons,
and
safety
systems
BEFORE
hardware
is
deployed!
Our
Main
Approach‐‐Dynamic
Monitoring
and
Decision
Systems
(DYMONDS)
DYMONDS‐Enabled
Physical
Grid
Requires:
•
SoSware
models
•
Control
•
Security
•
Sensors
•
Communica7ons
• …
CreaGon
of
“Smart
Grids”
Clear definiGon of what “Smart Grids” means
Deep understanding of the complexity of the power system
Ability
to
not
simply
introduce/develop
technology
but
to
understand
the
effects
of
changes
Ensure,
up
front,
security,
efficiency,
reliability,
and
integraGon
with
business/regulatory
environment
CreaGng
flexibility
and
empowering
all
levels
from
producers
to
consumers
New
DYMONDS
FuncGonaliGes
Just‐in‐Time
(JIT)
‐‐predicGons;
dynamic
look‐ahead
decision
making
Just‐in‐Place (JIP) ‐‐distributed, interacGve, mulG‐layered
Just‐in‐Context
(JIC)
‐‐‐‐
performance
objecGves
funcGon
of
organizaGonal
rules,
rights,
and
responsibiliGes
(3Rs)
and
system
condiGons.
Sample
examples
of
improved
performance—on‐going
work
in
EESG
(hap://www.eesg.ece.cmu.edu)
Key technical challenges
Establish new modeling paradigms---models driven by
sensing, communications and decision making/automation
Using these new models introduce next generation
dispatch/unit commitment methods and algorithms better
suited to manage intermittency (demand active decision
maker; topology switching for efficiency)
Ensure short-term stabilization using on-line sensing and
adaptation (for the first time PMUs being deployed in large
amounts); renewal of high gain control using power
electronics switching; new models;
Revisit Automatic Generation Control, Automatic Voltage
and Flow Control to include the potential of PMU
measurements and WAMS-based regulation; demand
included as decision variable
Transformational change in objectives of future energy
systems
Deliver supply to meet given demand Deliver power to support supply and
demand schedules in which both supply
and demand have costs assigned
$
J.Y.
Joo
kWh
Elastic
demand
that responds
to time-varying
prices
8
Example
1:
AdapGve
Load
Management
‐‐scheduling
Ter5ary
level
Bid
func<on
y(λ)
Market
price
λ
…
Secondary
level
Load
aggregator
I
Load
aggregator
II
Load
aggregator
III
Demand
func<on
End‐user
rate
x(λI)
λI
Primary
level
…
End‐user
DYMONDS Simulator
Scenario 1: + Wind generation [3,4]
Le Xie
20% / 50%
penetration to
the system
6
Example
2:
Wind
predicGon,
look‐ahead
management
using
storage
Compare
the
outcome
of
ED
from
both
the
centralized
and
distributed
MPC
approaches.
IntegraGng
>50%
Wind
DYMONDS Simulator
Scenario 3: + Electric vehicles [6]
NiklasRotering
Interchange
supply /
demand mode
by time-varying
prices
10
Example
3:
OpGmal
Control
of
Plug‐in‐Electric
Vehicles:
Fast
vs.
Smart
24
InformaGon
flow
for
integraGng
PHEVs
DYMONDS Simulator
Scenario 5: + PMU-Based Robust Control [7]
P
Zhijian
Liu
P Automated Voltage
Control (AVC) and
Automated Flow
P Control (AFC)
Design Best
Locations of PMUs
Design Feedback
Control Gains
P
26
Two
ConvenGonal
Generators
Hydroelectric
Generator
Dynamic Equations: Dynamic states for Hydro
Generator
x1 = [ωH , δHline:
Transmission , P m,H , G]
Constant
Reactance
Load: Constant power
Transmission
Line
Transmission Line
Equations
Assumptions:
Wind Generator
Dynamic
Equations: Hydroelectric
Dynamic Generator
states for Dynamic Equations:
Hydro
Generator
x1 = [ωw ,
δw, λds, λqs,
λdr
, λqr]
Transmission
Line
Transmission Line
Equations
Assumptions:
Simultaneous Equations:
2x Hydroelectric Generator Dynamic Equations:
Hydro 1 Hydro 2
Transmission Line
Equations
One
Hydro,
One
Wind
Simultaneous Equations:
Solver: Newton Raphson Method
Wind Generator
Hydroelectric
Dynamic
Generator
Equations:
Dynamic Equations:
Transmission Line
Equations
Equilibrium
Results
One Hydro, One Wind Results
Two Hydro-generator Results
Feasible Solution*
One Solution 2bus hydro-wind
2bus hydro
Wref = 1 Pmh = .8
Wsys = 1 Wref1 = 1.43 Wref2 = 1.43 PL = 1 Pmw = .2
PL = 1 Pm1 = 0.8 Pm2 = 0.2
wH 1.0404
Pe1 0.8000 wR 1.1297
Pe2 0.2000 PeH 0.9872
w1 1.0019 Pew 0.0128
w2 1.0019 flux_ds 0.0000
Steps 2.0000 flux_qs -0.9600
flux_dr 0.1404
flux_qr 0.0002
*multiple equilibria (3), one has high freq (~3 p.u.), one has negative frequency
Distributed
Equilibrium
Solver
Equality Constrained Newton Method
Objective function:
f(x) = Σkn Φk (xk) where x = [P1, … Pn] represents the power flows
for each of the n buses in the system
Hk = = Hessian Matrix
= Jacobian Matrix
Ax =b : network constraint
hk = Axk - b
Source: Jadbabaie, A. , et. Al. “A Distributed Newton Method for Network Optimization
AutomaGc
GeneraGon
Control
:
Revisited
Need Systematic Approach that accounts for Network Constraints
Inter-Area Oscillations
Insufficient Regulation Capacity
Wind Farms far away from the Load Centre
Pg (p.u.)
1.5
0.5
Hard-to-Predict Imbalances
Load Fluctuations: White Noise
Wind: Case of Non-Zero Mean
Deviations
v/s
An LQG Problem
0
-0.1
-0.2
0 100 200 300 400 500 600 700 800 900 1000
0.2
Frequency Sensitivity L4=0.25, L5=0.25 (rad/sec)/p.u.
0.1
-0.1
-0.2
0 100 200 300 400 500 600 700 800 900 1000
Network Constraints
0.2
Frequency Sensitivity L4=0.40, L5=0.40 (rad/sec)/p.u.
0.1
-0.2
0 100 200 300 400 500 600 700 800 900 1000
Time(s)
RegulaGon
Reserve
Planning:
Frequency
Bias
AutomaGc
GeneraGon
&
Demand
Control
(AGDC)
Active Demand Response Power based Control Model
Sensors embedded in Smart Appliances
Potential Source of Regulation
Flexibility in Electric Grid
Load Characterization
Trade-off between the need for very accurate sensor (e.g. to sense
frequency) and the need for communication (e.g. power based control
on load side)
We
are:
Moving
DYMONDS
concepts
forward
PresenGng
and
becoming
familiar
to
DoE,
NIST,
FERC,
EPRI,
NERC
Developing
modeling,
simulaGons,
tesGng
using
real‐system
data
Assessing
potenGal
benefits
from
implemenGng
JIT,
JIP
and
JIC
operaGng
and
planning
paradigm
Making
Smart
Grid
A
Reality!
A
huge
job
which
only
can
be
done
by
drawing
on
previous
R&D,
aoer
re‐posing
the
problems.
For
the
first
Gme
feasible!