Power System Resilience in the Pacific Northwest Eduardo - PowerPoint PPT Presentation

Power System Resilience in the Pacific Northwest Eduardo Cotilla-Sanchez, Ph.D., Assistant Professor School of Electrical Engineering & Computer Science College of Engineering, Oregon State University Corvallis, Oregon, USA Nov 28, 2017 Funded by the Oregon Talent Council & Department of Energy Award Number DE-OE0000780 & Electric Power Research Institute

Power System Protection Models Cascading Failures Clustering and Islanding as Resiliency Strategies 1 Power System Protection Models Understanding Critical Assets Blackout Risk Models 2 Cascading Failures Building a Dynamic Cascading Failure Model Example Simulation: Poland 3 Clustering and Islanding as Resiliency Strategies Remedial Action Schemes Automated Planning and Policy Switching Clustering E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 2/27

Power System Protection Models Understanding Critical Assets Cascading Failures Blackout Risk Models Clustering and Islanding as Resiliency Strategies Western Interconnection E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 3/27

Power System Protection Models Understanding Critical Assets Cascading Failures Blackout Risk Models Clustering and Islanding as Resiliency Strategies Power System Models E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 4/27

Power System Protection Models Understanding Critical Assets Cascading Failures Blackout Risk Models Clustering and Islanding as Resiliency Strategies Topological Metrics Average path length ( L ) 1 2 Number of bus failures à E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 5/27

Power System Protection Models Understanding Critical Assets Cascading Failures Blackout Risk Models Clustering and Islanding as Resiliency Strategies Connectivity Loss Connectivity loss ( % ) 1 2 Number of bus failures à E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 6/27

Power System Protection Models Understanding Critical Assets Cascading Failures Blackout Risk Models Clustering and Islanding as Resiliency Strategies Simple Cascading Failure Model (dc power flow) Blackout size (%) 1 2 Number of bus failures à E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 7/27

Power System Protection Models Understanding Critical Assets Cascading Failures Blackout Risk Models Clustering and Islanding as Resiliency Strategies Western Electricity Coordinating Council 1996 Blackout E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 8/27

Power System Protection Models Building a Dynamic Cascading Failure Model Cascading Failures Example Simulation: Poland Clustering and Islanding as Resiliency Strategies Load Models |V|: constant Z load |V|: constant I load 1.2 1.2 1.1 1.1 |V| |V| 1 1 0.9 0.9 0.8 0.8 0 5 10 15 20 0 5 10 15 20 time (sec.) time (sec.) |V|: constant P load |V|: constant E load 1.2 1.2 1.1 1.1 |V| |V| 1 1 0.9 0.9 0.8 0.8 0 5 10 15 20 0 5 10 15 20 time (sec.) time (sec.) E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 9/27

Power System Protection Models Building a Dynamic Cascading Failure Model Cascading Failures Example Simulation: Poland Clustering and Islanding as Resiliency Strategies Relay Models https://github.com/ecotillasanchez/cosmic.git E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 10/27

Power System Protection Models Building a Dynamic Cascading Failure Model Cascading Failures Example Simulation: Poland Clustering and Islanding as Resiliency Strategies Cascading Paths Number of Branch Outages 1 10 0 10 150 200 250 300 350 400 450 Time (sec.) E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 11/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Load Shedding (LS) CHARACTERISTICS LS BENEFITS Naive and Homogenous: Affects all loads in the Direct manipulation of network. generation/load mismatch. Usually triggered by under-voltage conditions. Can be done in real-time. Adaptive and Dynamic: Ability to select loads LS CHALLENGES based on a rule. (priority, severity, Direct customer impact. electrical distance, etc.) Usually triggered by Must be done multiple conditions (UV, UF, frequency, dV dt , dF incrementally. dt ). E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 12/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Islanding CHARACTERISTICS BENEFITS Can isolate operational Intentional creation of sections from affected microgrids. sections. Typically pre-determined Can fully utilize DER capability to decrease based on common contingencies. impact on customers. CHALLENGES Usually done based on Slow calculation – usually slow-coherency, electrical done offline. distance, active/reactive Can cause portions of the power balance. grid to fully collapse. E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 13/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Markov Decision Processes (MDP) MDPs provide well-developed theory for computational solutions to controllable and observable systems with stochastic dynamics. MDP COMPONENTS System States (S): A ij , P ij ,R 2 Time Independent S2 Control Actions (A): S1 A ij , P ij ,R 1 Stationary or Non-Stationary A A ij , P ij ,R 1 i j , P Probabilistic Transition A ij , P ij ,R 3 i j , R 2 Distributions (P): Stochastic Movement A ij , P ij ,R 3 Between States S3 Rewards (R): Function of Current State E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 14/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Solutions to MDPs What is the solution to an MDP? Policy ( π ) = A mapping of States ( s ) to Actions ( a ). Optimal Policy ( π ∗ ) = Policy with max Value (V ) in any given state ( s ). Value: � ∞ � � β t R ( s t ) V π ( s ) = E t =0 E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 15/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Policy Switching Simulation Time Horizon ... π 1 π 2 π 1 π 3 π 2 π ps Formally: Policy Switching: π ps ( s ) = π i ∗ ( s ) Basic guarantee that i ∗ = arg max V ( π ps ) ≥ max i V ( π i ) V π i ( s ) i E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 16/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Implementation 1 Python/Siemens PSSe 2 Time-Domain Simulation 3 3 Different Timescales 1 120 s – PSSe step 1 1 10 s – Grid check 2 5 s – Emgcy. dispatch 3 4 Random N-2 Contingencies 5 Parallelization E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 17/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Case Study: IEEE 39-Bus Case Topology 39 Buses 10 Generators 19 Loads 46 Branches N-1 Secure Flow Limits Governor Dynamics Exciter Dynamics E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 18/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Example Simulation: N − 2 (Lines 19-20 and 2-25) E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 19/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering CREDC Activity: Towards Attack Resilient Data Analytics for Power Grid Operations E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 20/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Active Power Sensitivity Based Electrical Distance E p � δ P � δ P � � ∆ P = ∆ θ + ∆ | V | (1) δθ δ | V | E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 21/27

Power System Protection Models Remedial Action Schemes Cascading Failures Automated Planning and Policy Switching Clustering and Islanding as Resiliency Strategies Clustering Active Power Sensitivity Based Electrical Distance E p E. Cotilla-Sanchez Power System Resilience in the Pacific Northwest 22/27