Embed Size (px)
Citation preview
Argonne National Laboratory is a U.S. Department of Energy laboratory managed by UChicago Argonne, LLC.
Blackout Analysis and Vulnerability Identification More realistic modeling of cascading blackouts: existing models have several gaps: lack of properly automated and coordinated controls, use of static system models, lack of protection system models, SPS/RAS or human intervention; thus more realistic models that address these gaps should be developed.
Interaction network and interaction model: interactions between component failures can reveal the outage propagation pattern Topology of 118-bus system Interaction network
• Propose methods to quantify effectively the interactions between component failures and obtain the interaction network capturing the propagation pattern; • Identify key links and components that play critical role in outage propagation; • Propose an interaction model based on the propagation patterns in the interaction network to greatly improve the simulation efficiency; • Develop effective key-link based mitigation strategies to prevent outage propagation and reduce the cascading risk.
Branching process: statistically describe the outage propagation by both one-type and multi-type branching processes Branching processes Marginal distribution of line outages
• Distribution of initial outage + average propagation distribution of total outage; • Average propagation: offspring mean of one type branching process or largest eigenvalue of offspring mean matrix of multi-type branching process indicate the system’s closeness to criticality; • Multi-type branching process can be used to study interdependent outages or interdependent critical infrastructure including electric power grid
UNDERSTANDING AND MITIGATING CASCADING BLACKOUTS Junjian Qi and Shrirang Abhyankar
Center for Energy, Environment, and Economics Systems Analysis (CEEESA)
Energy Systems Division
Motivation
Key Questions/Issues • Understanding blackout
propagation patterns • Blackout risk for a given
loading, weather condition • Identification of vulnerabilities. • Mitigation measures • Blackout dynamics
simulations too slow!
Research Thrust Areas • Methods to analyze, and
mitigate cascading blackouts, complementing established ANL tools, such as EPFAST
• Real-time dynamics simulation of cascading failures
• Cascading outage preventative measures
Good News ;)
Current projects Protection and Dynamic Modeling, Simulation, and Analysis of Cascading Outages. Sponsor: Grid Modernization Lab Consortium (2016-2019) Multifaceted Mathematics for Complex Energy Systems. Sponsor: Sponsor: DOE ASCR (2012-2017) Modeling of Extreme Events. Sponsor: Grid Modernization Lab Consortium (2016-2019)
References J Qi, K. Sun, and S. Mei, “An interaction model for simulation and mitigation of cascading failures,” IEEE Trans. Power Systems, vol. 30, no. 2, pp. 804-819, Mar. 2015.
J. Qi, I. Dobson, S. Mei, “Towards Estimating the statistics of simulated cascades of outages with branching processes,” IEEE Trans. Power Systems, vol. 28, no. 3, pp. 3410-3419, Aug. 2013.
J. Qi, W. Ju, and K. Sun, “Estimating the propagation of Interdependent cascading outages with multi-type branching processes,” arXiv:1405.6431v2, 2014.
S. Abhyankar, V. Rao, and M. Anitescu, “Dynamic Security Constrained Optimal Power Flow using Finite Difference Sensitivities,” Proceedings of IEEE PES General Meeting, 2014.
S. Abhyankar, and A. Flueck, “Real-time power system dynamics simulation using a parallel Block-Jacobi preconditioned Newton-GMRES scheme,” Proceedings of the 2nd International Workshop on High Performance Computing, Networking, and Analytics for the Power Grid, July 2013.
Accelerating dynamics simulations and preventive control
2003 Northeast Blackout: 55 million people affected
2011 Southwest Blackout: 7 million people affected
Blackout events on the rise!!
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
1 2 4 8 16 24
Execu1
on 1me (sec)
# Cores
Faster-than-real-time
Slower-than-real-time
5. sec dynamics simulation of a very large power grid
Faster-than-real-time dynamics simulation
Preven&ve Control
Blackout Simula&ons
Preventa&ve(Control(of(Blackouts(
Generator(3(would(trip(
Without(incorpora&ng(dynamic(scenarios(
Modified(dispatch(by(incorpora&ng(((both(scenarios(
On-going work and future steps
I Optimization with multiple dynamics scenarios (faults atdi↵erent locations).
Total cost = $6216.08Generator Bus Number MW
Gen1 1 162.71Gen2 2 103.16Gen3 3 51.53
0 0.5 1 1.5 2 2.5 358.5
59
59.5
60
60.5
61
61.5
Time (sec)
Fre
qu
en
cy
= ω
/2π
Figure: Generator frequencies forfaults at Bus 7 and Bus 9
I Low-level implementation (PETSc + IPOPT)I Mixed-BFGS approach for computing HessianI Parallelizing dynamics scenariosI Transiently unstable scenarios
Without dynamic constraints
Total cost = $5297.41
Table: Generation schedule without dynamic constraints
Generator Bus Number MW
Gen1 1 89.81Gen2 2 134.33Gen3 3 94.20
0 0.2 0.4 0.6 0.8 158.5
59
59.5
60
60.5
61
61.5
Time (sec)
Fre
qu
en
cy
= ω
/2π
Gen 1
Gen 2
Gen 3
Figure: Generator frequencies forfault at Bus 7
0 0.2 0.4 0.6 0.8 158.5
59
59.5
60
60.5
61
61.5
62
Time (sec)
Fre
qu
en
cy
= ω
/2π
Gen 1
Gen 2
Gen 3
Figure: Generator frequencies forfault at Bus 9
Generator(2(would(trip(
Preventa&ve(control(of(Power(Blackouts:(The(Gotham(Analogy(
h(
++++++++JOKER’S+NO+FLY+ZONE+
Unacceptable generator frequency deviations (would cause generator tripping)
Preventa&ve(Control(of(Blackouts(
Generator(3(would(trip(
Without(incorpora&ng(dynamic(scenarios(
Modified(dispatch(by(incorpora&ng(((both(scenarios(
On-going work and future steps
I Optimization with multiple dynamics scenarios (faults atdi↵erent locations).
Total cost = $6216.08Generator Bus Number MW
Gen1 1 162.71Gen2 2 103.16Gen3 3 51.53
0 0.5 1 1.5 2 2.5 358.5
59
59.5
60
60.5
61
61.5
Time (sec)
Fre
qu
en
cy
= ω
/2π
Figure: Generator frequencies forfaults at Bus 7 and Bus 9
I Low-level implementation (PETSc + IPOPT)I Mixed-BFGS approach for computing HessianI Parallelizing dynamics scenariosI Transiently unstable scenarios
Without dynamic constraints
Total cost = $5297.41
Table: Generation schedule without dynamic constraints
Generator Bus Number MW
Gen1 1 89.81Gen2 2 134.33Gen3 3 94.20
0 0.2 0.4 0.6 0.8 158.5
59
59.5
60
60.5
61
61.5
Time (sec)
Fre
qu
en
cy
= ω
/2π
Gen 1
Gen 2
Gen 3
Figure: Generator frequencies forfault at Bus 7
0 0.2 0.4 0.6 0.8 158.5
59
59.5
60
60.5
61
61.5
62
Time (sec)
Fre
qu
en
cy
= ω
/2π
Gen 1
Gen 2
Gen 3
Figure: Generator frequencies forfault at Bus 9
Generator(2(would(trip(
Dispatch decision
Blackout indicator
Preventa&ve(control(of(Power(Blackouts:(The(Gotham(Analogy(
h(
Thank(you(Mr.(Fox!(
Preventa&ve(Control(of(Blackouts(
Generator(3(would(trip(
Without(incorpora&ng(dynamic(scenarios(
Modified(dispatch(by(incorpora&ng(((both(scenarios(
On-going work and future steps
I Optimization with multiple dynamics scenarios (faults atdi↵erent locations).
Total cost = $6216.08Generator Bus Number MW
Gen1 1 162.71Gen2 2 103.16Gen3 3 51.53
0 0.5 1 1.5 2 2.5 358.5
59
59.5
60
60.5
61
61.5
Time (sec)
Fre
qu
ency
= ω
/2π
Figure: Generator frequencies forfaults at Bus 7 and Bus 9
I Low-level implementation (PETSc + IPOPT)I Mixed-BFGS approach for computing HessianI Parallelizing dynamics scenariosI Transiently unstable scenarios
Without dynamic constraints
Total cost = $5297.41
Table: Generation schedule without dynamic constraints
Generator Bus Number MW
Gen1 1 89.81Gen2 2 134.33Gen3 3 94.20
0 0.2 0.4 0.6 0.8 158.5
59
59.5
60
60.5
61
61.5
Time (sec)
Freq
uenc
y =
ω/2
π
Gen 1
Gen 2
Gen 3
Figure: Generator frequencies forfault at Bus 7
0 0.2 0.4 0.6 0.8 158.5
59
59.5
60
60.5
61
61.5
62
Time (sec)
Freq
uenc
y =
ω/2
π
Gen 1
Gen 2
Gen 3
Figure: Generator frequencies forfault at Bus 9
Generator(2(would(trip(
Generator frequencies with modified dispatch (initial conditions).
Without preventive control With preventive control
Fault at Bus 9 Fault at Bus 7
Dynamics simulation accelerated by HPC and advanced numerical solvers (adaptive time-stepping and Schwarz-preconditioned iterative linear solver)
