Design and Analysis of Wide-Area Resilient Control Algorithms for Large-Scale Power Systems: Theoretical and Experimental Methods

Team Members: Prateek Jaipuria, Vineet Ghatge, Terry Benzel (USC/ISI), Jianhua Zhang (NC State), Yufeng Xin (UNC Chapel Hill)

Our team is advancing the science and technology for cyber-physical power systems by developing transformative algorithms for wide area monitoring and control of next- generation smart grid networks.

Goals

We are performing a set of interconnected activities in the following areas: •  Enabling facility integration by leveraging existing testbed infrastructure across different

institutions.

•  Advancing the underlying experimentation technologies to increase the scale and the scope of the evaluation of wide area CPS monitoring, control, and security algorithms.

•  Providing technologies to extract and maintain the theoretical and empirical understanding of stability and control algorithms in wide area CPS scenarios.

•  Developing tools and methods that can be used for experimentation by experts and non experts in addition to applications-specific and domain-specific support tools.

Key Elements

Centralized RLS Distributed Prony Experimental Approach DETERLab and RTDS-WAMS tools to create several what-if scenarios for evaluation. 1. Emulate real-time dynamic models of power system coupled with representative wide area topology. 2. Orchestrate traffic and faults that impact monitoring and estimation. 3. Compare impact of communication bottleneck on the convergence of the estimation. 4. Collect data and visualize results.

Impact on CPS Community

Theoretical Approach

Alefiya Hussain (USC/ISI) and Aranya Chakrabortty (NC State Univ)

Problem Statement

Control Actions

Internet

Real-­‐timeDispatch

Topology  Variation

Algorithms&  Decisions

Unpredicted  Disturbance  Events

Highly  Stochastic  Load  Variations

EventLevel

PDC  #  1 PDC  #  2 PDC  #  3 PDC  #  n

Power  System

SubstationLevel

PDC

PDCPDCControl  

Center  LevelPhasor  StateEstimator

Supervisory  Controller

Adaptive  AGC

ApplicationLevelModal  

Analysis  &  FFT

Event  Detection  &  Analysis

Voltage  Stability  &  Load  Modeling

PMU

PMU PMU

Region 1 Region 2 Region 3 Region n

PMU

PMUPMU

PMU PMU

PMU

PMU Clusters

Transient  Stability

Assessment

Frequency  Control

Dynamic  State  Estimation

SituationalAwareness

Wide-­‐areaDamping  Control  

&  Optimization

To translate current state-of-art centralized processing algorithms for wide-area monitoring and control of large power grids using large volumes of Synchrophasor data to a completely distributed attack-resilient cyber-physical architecture.

1.  Centralized versus distributed power flow oscillation monitoring following critical disturbances

2.  Applying distributed algorithms for ensuring resiliency against cyber and physical attacks

3.  Experimental verification using federation of DETERLab and RTDS-WAMS testbed

PDC

PMU

PDC

PMU

PDC

PMU

Super  PDC

Centralized  Data  

Processing

UnidirectionalCommunication

PDC PDC PDC PDC

PDC PDC

PMU PMU

PMU

PDC PDC PDC PDC

PDC PDC

PMU PMU

PMU

PDC PDC PDC PDC

PDC PDC

PMU

Local PDC-PDC Communication

Interarea PDC-PDC Communication

PMU PMU

State-of-art Centralized Processing Architecture:

1.  Regional PMU data sent to regional PDC

2.  Independent local analysis and storage

3.  Regional PDCs send data to super-PDC for archival

Proposed Distributed Cyber-Physical Architecture for PMU-PDC Communication:

!! "!! #$1 #

112

11

ModeInterarea

aa

aa

sss

σγβα++

+

!! "!! #$2 #

222

22

ModeInterarea

aa

aa

sss

σγβα++

+

Cluster 1

Cluster 2Cluster 3

!! "!! #$3 #

332

33

ModeInterarea

aa

aa

sss

σγβα++

+

Zero/First Order Hold !!! "!!! #$

2 #

222

22

ModeInterarea

i iaia

iaia

sss

∑ ++

+

σγ

βαIntra-cluster

Virtualization

•  Centralized Recursive Least Squares (RLS)

•  Distributed Subgradient Method

•  ADMM and other convex algorithms

))()()((min 000

1

0

2 θθθθθφθ

−−+−∑−

=K

TK

K

kK

Tkk Ry

K

Compute power flow oscillation frequencies (eigenvalues), mode shapes (eigenvectors), damping, residue, participation factors, and mode energy of electro-mechanical swing dynamics from PMU measurements. Pose a least-squares estimation for transfer function parameters:

1: Update both primal and dual estimation variables at every local control center:

2: Gather the values of at the central ISO 3: Take the average of at the central ISO 4: Broadcast the average value to local control

centers and iterate to Step 1

))(())(()()()()(1)()()1( kk

iki

Tki

ki

Tki

ki wcHIHH βρρβ +−+= −+

)()1()1()()1( +++ −+=

kki

ki

ki ww ββρ

)1( +kiβ)1( +k

iβ )1( +kiβ

Facility Integration

1.  High-fidelity dynamic models of IEEE 39-bus New England instantiated across three facilities •  PMU–based hardware-in-loop

simulation testbed at NCSU •  Federated with DETERLab

and ExoGENI 2.  Create at-scale scenarios to

explore resiliency of distributed algorithms

3.  Test fault tolerance and accuracy in the presence of communication

failures and cyber-security incidents.

Verification and validation of distributed algorithms for wide-area stability monitoring and control using real world communication protocols implemented over a nation wide network.

•  Increases the technology readiness through rigorous evaluation and validation

•  Facility integration provides an inter-disciplinary training and educational platform for power engineering and computer science

•  Tight collaboration between team members fosters an integrated research approach