63
Modern Power System Dynamic Performance Improvement through Big Data Analysis Yajun Wang Department of Electrical Engineering and Computer Science Committee Members: Dr. HéctorA. Pulgar, Dr. Yilu Liu, Dr. Hairong Qi, Dr. Russell Zaretzki Ph.D. Dissertation Defense April 4, 2019

Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

  • Upload
    others

  • View
    10

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Modern Power System Dynamic Performance Improvement through Big Data Analysis

Yajun WangDepartment of Electrical Engineering and Computer Science

Committee Members:Dr. Héctor A. Pulgar, Dr. Yilu Liu, Dr. Hairong Qi, Dr. Russell Zaretzki

Ph.D. Dissertation DefenseApril 4, 2019

Page 2: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Outline

• Introduction

• Inertia Distribution Estimation in Power Systems

• Actuator Placement for Enhanced System Dynamics

• Real-time Security Assessment and Cascading Failure Analysis

• Conclusion and Future Work

2

Page 3: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Outline

• Introduction

• Inertia Distribution Estimation in Power Systems

• Actuator Placement for Enhanced System Dynamics

• Real-time Security Assessment and Cascading Failure Analysis

• Conclusion and Future Work

3

Page 4: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Background and Motivation

Modern power system transformation

Voltage StabilityTransient Stability

4 8 12 16 200

0.6

0.7

0.8

0.9

1

1.1

1.2

Sec

Bus Voltage (pu)

4 8 12 16 200

0

200

400

600

800

1000

Sec

Generator Rotor Angle (Degrees)

Frequency concerns

4

New challenges

Traditional model-based methods are finding difficulties to adapt to the changes

Page 5: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Background and Motivation

Massive PMU and FDR deployed in the system

5

Huge development in AI techniques

Big data analysis extract valuable information from PMU data to enhance system dynamics

Page 6: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Main Contributions

6

• A pioneer to find the physical location of COI and create accurateinertia distribution map

Power system

data

Inertia distribution estimation

Power system

dynamics

Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Estimating inertia distribution to enhance powersystem dynamics, North American Power Symposium, Morgantown, WV, USA, Sep 2017H. Pulgar-Painemal, Y. Wang, H. Silva-Saravia, On inertia distribution, inter-area oscillations andlocation of electronically-interfaced resources, IEEE Transactions on Power Systems, Vol. 33, No.1, 2018, pp. 995-1003Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Equivalent inertia constant expression for powersystems, IEEE Transactions on Power Systems, 2019 (In preparation)

Page 7: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Main Contributions

7

• A pioneer to find the physical location of COI and create accurateinertia distribution map

Power system

data

Inertia distribution estimation

Actuator placement

Power system

dynamics

• Placed the actuators at the bus that are far away from the COI bus,increasing damping ratio to 14% and reducing the computationalcomplexity

• Created general and effective guidance for planners, considering moresystem features and more dynamic problems

H. Pulgar-Painemal, Y. Wang, H. Silva-Saravia, On inertia distribution, inter-area oscillations andlocation of electronically-interfaced resources, IEEE Transactions on Power Systems, Vol. 33, No.1, 2018, pp. 995-1003Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Actuator placement for enhanced grid dynamicperformance: A machine learning approach, IEEE Transactions on Power Systems (Early access),2019H. Silva-Saravia, Y. Wang, H. Pulgar-Painemal, Determining wide-area signals and locations ofregulating devices to damp inter-area oscillations through eigenvalue sensitivity analysis usingDIgSILENT Programming Language, Advanced Smart Grid Functionalities based on PowerFactory, Springer, 2018, pp. 153-179H. Silva-Saravia, Y. Wang, H. Pulgar-Painemal, K. Tomsovic, Oscillation energy based sensitivityanalysis and control for multi-mode oscillation systems, IEEE PES General Meeting, Portland, OR,USA, August 2018

Page 8: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Main Contributions

8

• A pioneer to find the physical location of COI and create accurateinertia distribution map

Power system

data

Inertia distribution estimation

Actuator placement

Power system

dynamics

• Placed the actuators at the bus that are far away from the COI bus,increasing damping ratio to 14% and reducing the computationalcomplexity

• Created general and effective guidance for planners, considering moresystem features and more dynamic problems

Operational studies

• Achieved the highest accuracy and lower computational time for real-time security analysis

• Identified the most critical links and generate probability tree modelfor cascading failure analysis

Y. Wang, H. Pulgar-Painemal, K. Sun, Online analysis of voltage security in a microgrid usingconvolutional neural networks, IEEE PES General Meeting, Chicago, IL, USA, July 2017Y. Wang, W. Ju, H. Pulgar-Painemal, Cascading failure key link identification and tree modelgeneration: A data driven approach, IEEE PES General Meeting, 2020 (In preparation)

Page 9: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Main Contributions

9

• A pioneer to find the physical location of COI and create accurateinertia distribution map

Power system

data

Inertia distribution estimation

Actuator placement

Power system

dynamics

• Placed the actuators at the bus that are far away from the COI bus,increasing damping ratio to 14% and reducing the computationalcomplexity

• Created general and effective guidance for planners, considering moresystem features and more dynamic problems

Operational studies

• Achieved the highest accuracy and lower computational time for real-time security analysis

• Identified the most critical links and generate probability tree modelfor cascading failure analysis

Power system dynamics is improved through big data analysis

Page 10: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Outline

• Introduction

• Inertia Distribution Estimation in Power Systems

• Actuator Placement for Enhanced System Dynamics

• Real-time Security Assessment and Cascading Failure Analysis

• Conclusion and Future Work

10

Page 11: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Research Gap and Highlighted Contribution

11

Inertia Definition• Provided by spinning mass of directly synchronized

electrical machines• Defines reaction of frequency to a sudden power

imbalance

Page 12: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Research Gap and Highlighted Contribution

12

Inertia Definition• Provided by spinning mass of directly synchronized

electrical machines• Defines reaction of frequency to a sudden power

imbalance

Low inertia

High inertia

System with higher inertia has more ability to resist change

Page 13: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Research Gap and Highlighted Contribution

13

Current Research Gap

G1G2

G3Inertia constant

Accelerating power

Machine speed

• Center of Inertia (COI) reference defines aweighted function of all machine rotorangle and speed

No physical location of COINo inertia distribution estimation

Page 14: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Research Gap and Highlighted Contribution

14

G1

Heavier inertia

Lighter inertia

Page 15: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Research Gap and Highlighted Contribution

15

G2

Heavier inertia

Lighter inertia

Page 16: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Research Gap and Highlighted Contribution

16

G3Heavier inertia

Lighter inertia

Page 17: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Research Gap and Highlighted Contribution

17

G1G2

G3Find physical location of COI

Explore inertia distribution in the system for every location

Show potential application of inertia distribution results

Main Contribution

H. Pulgar-Painemal, Y. Wang, H. Silva-Saravia,On inertia distribution, inter-area oscillationsand location of electronically-interfacedresources, IEEE Transactions on PowerSystems, Vol. 33, No. 1, 2018, pp. 995-1003Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal,Equivalent inertia constant expression for powersystems, IEEE Transactions on Power Systems,2019 (In preparation)

Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Estimating inertia distribution toenhance power system dynamics, NorthAmerican Power Symposium, Morgantown,WV, USA, Sep 2017

Page 18: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Find Physical Location of COI

18

Mathematical Proof

Equivalent swing equation

Equivalent inertia constant

SG Acceleration power

Page 19: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Find Physical Location of COI

19

Mathematical Proof

Equivalent swing equation

Equivalent inertia constant

SG Acceleration power

Page 20: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Find Physical Location of COI

20

Mathematical Proof Equivalent swing equation

Equivalent inertia constant

SG Acceleration power

COI physical location

At COI bus

Expression for COI

H. Pulgar-Painemal, Y. Wang, H. Silva-Saravia, On inertia distribution, inter-area oscillations andlocation of electronically-interfaced resources, IEEE Transactions on Power Systems, Vol. 33, No. 1,2018, pp. 995-1003

Page 21: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Find Physical Location of COI

21

Mathematical Proof Equivalent swing equation

Equivalent inertia constant

SG Acceleration power

COI physical location

At COI bus

Expression for COI

H1=5s H2=1s-25s

Page 22: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Find Physical Location of COI

22

Mathematical Proof

H1=5s H2=1s

COICOI physical location

At COI bus

Expression for COI

=0.167

Page 23: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Find Physical Location of COI

23

Mathematical Proof

H1=5s H2=2s

COICOI physical location

At COI bus

Expression for COI

=0.286

Page 24: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Find Physical Location of COI

24

Mathematical Proof

H1=5s H2=5s

COICOI physical location

At COI bus

Expression for COI

=0.5

Page 25: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Find Physical Location of COI

25

Mathematical Proof

H1=5s H2=25s

COI

COI physical location

At COI bus

Expression for COI

=0.833

Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Equivalent inertia constantexpression for power systems, IEEE Transactions on Power Systems, 2019 (Inpreparation)

Page 26: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Find Physical Location of COI

26

Mathematical Proof

H1=5s H2=5s

COI

COI is the bus that its changes in angle and frequency are minimal

=0.5At COI

Page 27: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Measurement-based Inertia Distribution Estimation

27

Center of Frequency Index

• Reports how far away a particular bus isto the COI bus

COI

Furthest to COI

Closest to COI

Page 28: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

H1=H3=6s, H3=H4=4.25s, length of

the lines between buses X1 and buses X3 is five times the value

in the base case

Measurement-based Inertia Distribution Estimation

28

Case 1: A meshed system

COI is near the machine with a larger inertia constant and larger impedance parameters

H1=H2=H3=H4=4.25s

Closest to COI

H3=8.5s, H1=H2=H4=4.25s

Page 29: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Measurement-based Inertia Distribution Estimation

29

Case 2: IEEE 39-bus test system

Bus 1 is closest to the COI bus, Bus 20 is the furthest point from the COI bus

Closest to COI

Furthest to COI

Page 30: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Measurement-based Inertia Distribution Estimation

30

Case 3: Real systems: Chilean power system and a real system

Closest to COI

Inertia distribution map is created for real system condition

Page 31: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Applications of Inertia Distribution

31

• Actuator placement

• Optimal placement of PMUs

• Power system dynamic performances can be improved by understanding of the system inertia distribution

PMU

PMUPMU

PMU

Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Estimatinginertia distribution to enhance power system dynamics, NorthAmerican Power Symposium, Morgantown, WV, USA, Sep 2017

Page 32: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Summary

• Obtained physical location of COI from theoretical proof and validated them indynamic simulation

• Created system inertia distribution map through a measurement based estimationapproach

• The understanding of the system inertia distribution can improve power systemdynamic performances in problems such as optimal PMU and control actuatorplacement

32

Page 33: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Outline

• Introduction

• Inertia Distribution Estimation in Power Systems

• Actuator Placement for Enhanced System Dynamics

• Real-time Security Assessment and Cascading Failure Analysis

• Conclusion and Future Work

33

Page 34: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Research Gap and Highlighted Contribution

34

Traditional ESS placement methods

Focused on static state or economic cost

Formed asoptimization

problem

• Control component to damp oscillations• Mainly for Energy storage system, can also be enabled with wind turbine or other converter/inverter based

machines

Control actuator

Fail to fully capture some important aspects of the system dynamics

Solved byheuristic, stochastic

optimization approaches

Current Research Gap

Page 35: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Research Gap and Highlighted Contribution

35

Main ContributionUse inertia distribution to allocate control actuators

Extract most important features to create guidelines to allocate control actuators System

featuresH. Pulgar-Painemal, Y. Wang, H. Silva-Saravia, On inertiadistribution, inter-area oscillations and location ofelectronically-interfaced resources, IEEE Transactions onPower Systems, Vol. 33, No. 1, 2018, pp. 995-1003

H. Pulgar-Painemal, Y. Wang, H. Silva-Saravia, On inertiadistribution, inter-area oscillations and location ofelectronically-interfaced resources, IEEE Transactions onPower Systems, Vol. 33, No. 1, 2018, pp. 995-1003H. Silva-Saravia, Y. Wang, H. Pulgar-Painemal, Determiningwide-area signals and locations of regulating devices to dampinter-area oscillations through eigenvalue sensitivity analysisusing DIgSILENT Programming Language, Advanced SmartGrid Functionalities based on Power Factory, Springer,2018, pp. 153-179

Page 36: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using Inertia Distribution Index

36

Residue when in ESS is connected

At COI bus

• The residue calculated from the transfer function for the dominant mode is quadratic to thedistance from a bus to the COI bus

• Less residue indicates less controllability and observability

Mathematical Proof

Actuator should be placed at the higher residue locations

Page 37: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using Inertia Distribution Index

37

Residue when in ESS is connected

Simulation Validation

Inertia distribution index

Actuator should be placed at the bus that is further to COI

Page 38: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using Inertia Distribution Index

38

Case 1: IEEE 39-bus test system

Location No actuator Closest to COI bus

Furthest to COI bus

Damping ratio 5% 8.04% 19.3%

Actuator should be placed at the bus that is further to the COI bus

Page 39: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using Inertia Distribution Index

39

Case 2: Chilean power system

Location No actuator Closest to COI bus

Furthest to COI bus

Damping ratio 0.53% 6.0% 13.1%

H. Pulgar-Painemal, Y. Wang, H. Silva-Saravia, On inertia distribution, inter-area oscillations and location ofelectronically-interfaced resources, IEEE Transactions on Power Systems, Vol. 33, No. 1, 2018, pp. 995-1003H. Silva-Saravia, Y. Wang, H. Pulgar-Painemal, Determining wide-area signals and locations of regulatingdevices to damp inter-area oscillations through eigenvalue sensitivity analysis using DIgSILENT ProgrammingLanguage, Advanced Smart Grid Functionalities based on Power Factory, Springer, 2018, pp. 153-179

Page 40: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Summary

• Verified the physical location of COI through analytical proof• Find a simple rule to place the actuators, which can effectively increase the system

dynamics and hugely reduce the calculation complexity

40

Page 41: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

41

More system features

More system dynamics

Proposed Idea

Page 42: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

42

• Extracts the significant features of a potential location/bus that interpret system dynamic behavior• Obtain general guidelines from the model for placing control actuators to improve system

dynamic performance

Flowchart

Page 43: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

43

Overall performance indexInput Output

Stage 2: Data set construction

Topological data

Physical data

Operational data

System integrated data

Data source Feature layers

Page 44: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

44

Stage 2: Data set construction (IEEE 39-bus system visualization)

Topological feature Inertia distribution feature

Voltage magnitude feature

Input: System features

Page 45: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

45

Stage 2: Data set construction (IEEE 39-bus system visualization)Output: System dynamic response

• The use of all three dynamic performanceindices is needed to asses the overallsystem performance

• Labels K ={1,2,3,4} are marked for all thebuses with k-means clustering algorithm

Voltage

Tran

sien

t

Page 46: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

46

Stage 3: Feature Extraction and Analytical Model Generation

Page 47: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

47

Stage 3: Feature Extraction and Analytical Model Generation

LASSO (Least Absolute Shrinkage and Selection Operator )

Output Input

the parameter that controls the strength of the penalty

• Use linear regression model to find the coefficients of features• When increases, coefficients of less important features will

be shrank to 0• Extract most important features by the shrinkage

Page 48: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

48

Stage 3: Feature Extraction and Analytical Model Generation

• The most influential features are inertia distribution, topologydistribution and voltage angle

• The operational features are not significant with higher RE level

Prediction model

Evaluation indices

Case 1: 39-bus system

Page 49: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

49

Stage 3: Feature Extraction and Analytical Model GenerationCase 2: 118-bus system

• The most influential features are voltage angle, faultlocation and impedance distribution

• Three dynamic indices are needed for overall systemdynamics

• Cluster number K will not affect the overall accuracy

Overall performance index

Frequency Voltage

Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Actuator placement for enhanced griddynamic performance: A machine learning approach, IEEE Transactions on PowerSystems (Early access) , 2019

Page 50: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Actuator Placement using A machine Learning Approach

50

Stage 4: Guidelines for System Planners

• System physical features play a more important role than topological and operationalfeatures

• Among all system topological and physical features, inertia distribution index, topologydistribution index and impedance distribution index are the most critical ones

• Among all operational features:1) Bus voltage angle is the most significant one2) Fault location affects the performance of the buses when RE penetration level is low.

When RE penetration level is above 20%, the impact of fault location or duration time is lessrelevant

3) The RE penetration levels and loading levels do not affect the bus performance as muchas active and reactive power output

Page 51: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Summary

• Determine the most significant features in the placement of control actuatorsconsidering dynamic performance

• Provide general criteria and guidance for system planners to place the controlresources in a given system for improving oscillation damping, voltage andtransient stability

51

Page 52: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Outline

• Introduction

• Inertia Distribution Estimation in Power Systems

• Actuator Placement for Enhanced System Dynamics

• Real-time Security Assessment and Cascading Failure Analysis

• Conclusion and Future Work

52

Page 53: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Real-time Voltage Security Assessment Voltage security assessment in a micro-grid

Flow chart of voltage security analysis

Load switching Line outage Three phase short circuit

Stable examples

Unstable examples

53

Real time data samples

Page 54: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Real-time Voltage Security Assessment Case Study in IEEE 14-bus system

Structure of proposed CNN

Performance of Back-Propagation Neural Networks, Decision Tree, Support Vector

Machine, and Convolutional Neural Networks

• CNN has the best performance due the 2-D data transformationand hidden layer construction

• Deep learning algorithm has great potential in the modern powersystem studies

54

Y. Wang, H. Pulgar-Painemal, K. Sun, Online analysis of voltage security in a microgrid usingconvolutional neural networks, IEEE PES General Meeting, Chicago, IL, USA, July 2017

Page 55: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Critical Link Identification in Cascading Failure Analysis

55

Key Link Identification in WECC system

Distance matrix

Page 56: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Critical Link Identification in Cascading Failure Analysis

56

Cascading Failure Tree Model

• Key links are obtained using 400,000 data samples

• Cascading failure tree model based on probability calculation has been generated

Y. Wang, W. Ju, H. Pulgar-Painemal, Cascading failure key link identification and tree modelgeneration: A data driven approach, IEEE PES General Meeting, 2020 (In preparation)

Page 57: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Summary

• Achieve higher voltage security assessment accuracy using a CNN-based method

• Understand the cascading procedure better with statistic based model

57

Page 58: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Outline

• Introduction

• Inertia Distribution Estimation in Power Systems

• Actuator Placement for Enhanced System Dynamics

• Real-time Security Assessment and Cascading Failure Analysis

• Conclusion and Future Work

58

Page 59: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Conclusion

• The understanding of the system inertia distribution can improve power system dynamic performance

• Allocation of actuators can enhance system dynamic response. Furthermore, For system with single-dominant oscillation mode:

1) Inertia distribution index can help to allocate control actuators in the; the actuators shouldbe placed at the bus further to the COI bus

For system with multi oscillation mode and multi stability problems:1) System physical features play a more important role than topological and operational

features2) Among all system topological and physical features, inertia distribution index, topology

distribution index and impedance distribution index are the most critical ones

• Higher security assessment accuracy can be achieved using a deep learning method and critical linkscan be identified using a tree model

59

Page 60: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Future Work

60

• Study active power control enabled with inverter-based machine and evaluate the emulated inertiaresponse

• Design model-free and data-driven damping controller

Page 61: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

PublicationY. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Actuator placement for enhanced grid dynamic performance: Amachine learning approach, IEEE Transactions on Power Systems, (Early access), 2019H. Pulgar-Painemal, Y. Wang, H. Silva-Saravia, On inertia distribution, inter-area oscillations and location ofelectronically-interfaced resources, IEEE Transactions on Power Systems, Vol. 33, No. 1, 2018, pp. 995-1003Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Equivalent inertia constant expression for power systems, IEEETransactions on Power Systems, 2019 (In preparation)Y. Wang, H. Silva-Saravia, H. Pulgar-Painemal, Estimating inertia distribution to enhance power system dynamics,North American Power Symposium, Morgantown, WV, USA, Sep 2017Y. Wang, H. Pulgar-Painemal, K. Sun, Online analysis of voltage security in a microgrid using convolutional neuralnetworks, IEEE PES General Meeting, Chicago, IL, USA, July 2017H. Silva-Saravia, Y. Wang, H. Pulgar-Painemal, K. Tomsovic, Oscillation energy based sensitivity analysis andcontrol for multi-mode oscillation systems, IEEE PES General Meeting, Portland, OR, USA, August 2018Y. Wang, W. Ju, H. Pulgar-Painemal, Cascading failure key link identification and tree model generation: A datadriven approach, IEEE PES General Meeting, 2020 (In preparation)H. Silva-Saravia, Y. Wang, H. Pulgar-Painemal, Determining wide-area signals and locations of regulating devicesto damp inter-area oscillations through eigenvalue sensitivity analysis using DIgSILENT Programming Language,Advanced Smart Grid Functionalities based on Power Factory, Springer, 2018, pp. 153-179

61

Page 62: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data

Acknowledgements

This work was supported primarily by the National Science Foundation under Grant No. 1509114, the Engineering Research

Center Program of the National Science Foundation and the Department of Energy under NSF Award No. EEC-1041877 and the

CURENT Industry Partnership Program.

Other US government and industrial sponsors of CURENT research are also gratefully acknowledged.

62

Page 63: Modern Power System Dynamic Performance Improvement ...web.eecs.utk.edu/~hpulgar/Material/Thesis_Slides_YWang.pdfModern Power System Dynamic Performance Improvement through Big Data