Wide Area Monitoring of Interconnected Power Systems

Contents
About the Author
Preface
1 Wide-area monitoring and analysis systems
1.1 Introduction
1.2 Wide-area monitoring systems: a conceptual overview
1.2.1 WAMS architectures
1.2.2 Measurement and sensing technologies
1.2.3 Design of WAMS architecture
1.3 Distributed and semi-distributed monitoring frameworks
1.4 Data collection and management
1.5 Synchrophasor networks
1.5.1 Network graphs
1.5.2 Spectral graph analysis
1.5.3 Connected dominant sets
1.5.4 Geometric approaches for signal processing
1.6 Challenges of future intelligent monitoring and analysis systems
References
2 Wide-area monitoring system architectures
2.1 Introduction
2.2 WAMS architectures
2.2.1 Centralized WAMS architectures
2.2.2 Hierarchical WAMS architectures
2.2.3 Hybrid WAMS architectures
2.2.4 Distributed WAMS architectures
2.3 Issues in data fusion
2.3.1 Data
2.3.2 Intelligent synchrophasor data fusion
2.3.3 Power system data fusion strategies
2.3.4 General framework for data assimilation
2.3.5 Fusion or integration of multivariate PMU data
2.4 Relationship between multiblock and single-block models
References
3 Spatio-temporal modeling of power system dynamic processes
3.1 Introduction
3.2 Visualization of large space-time measurement data
3.3 Spatio-temporal modeling of multivariate processes
3.3.1 POD analysis
3.3.2 Singular value decomposition-based POD
3.3.3 Departure from the mean value
3.4 Spatio-temporal interpolation methods
3.4.1 Background
3.4.2 Similarity measures
3.4.3 Spatial structures
3.4.4 Derivation of weights
3.4.5 Practical issues
3.5 Dimensionality reduction
3.5.1 Projection-based reduced-order modeling
3.5.2 Proximity (similarity) measures
3.5.3 Nonlinear spectral dimensionality reduction
3.6 Motivational example
3.6.1 Small-signal response
3.6.2 Large system response
3.6.3 Statistical analysis
3.7 Sensor placement
3.7.1 Problem formulation
3.7.2 Constrained sensor placement
3.7.3 Other approximations
3.7.4 Compressed sensing
3.7.5 Sensor placement for signal (state) or process reconstruction
References
4 Advanced data processing and feature extraction
4.1 Introduction
4.2 Power oscillation monitoring
4.3 Time–frequency representations
4.3.1 Hilbert–Huang analysis
4.3.2 Wavelet analysis
4.4 Multivariate multiscale analysis
4.4.1 Multisignal Prony analysis
4.4.2 Koopman analysis and its variants
4.4.3 Dynamic mode decomposition
4.4.4 Multichannel Prony analysis
4.4.5 Deep autoencoders
4.5 Response under ambient stimulus
4.5.1 Formulation of the model
4.5.2 Modal response
4.5.3 Ensemble system response
4.6 Application to measured data
4.6.1 HHT analysis
4.6.2 Wavelet analysis
References
5 Multisensor multitemporal data fusion
5.1 Introduction
5.2 Data fusion principles
5.3 Data preprocessing and transformation
5.3.1 Bandpass filtering and denoising
5.3.2 Local-level fusion
5.4 Feature extraction and feature selection
5.4.1 Feature extraction
5.4.2 Data compression
5.4.3 Individual scales
5.4.4 Filtering and multiscale monitoring
5.5 Multisensor fusion methodologies for system monitoring
5.5.1 Single-scale analysis
5.5.2 Nonlinear PCA using autoassociative neural networks
5.5.3 Multiblock POD (PCA) analysis
5.5.4 Nonlinear PCA
5.5.5 Blind source separation
5.6 Other approaches to multisensor data fusion
5.7 Illustration
5.7.1 Multisensor (multiview) data fusion
References
6 Monitoring the status of the system
6.1 Introduction
6.2 Power system health monitoring
6.3 Disturbance and anomaly detection
6.4 Modal-based health monitoring methods
6.4.1 Filtering and data conditioning
6.4.2 Entropy and energy
6.4.3 Entropy-based detection of system changes
6.5 Wide-area inter-area oscillation monitoring
6.5.1 Case A
6.5.2 Case B
6.6 High-dimensional pattern recognition-based monitoring
6.6.1 Sparse diffusion implementation
6.6.2 Data clustering
6.6.3 Numerical example
6.6.4 Hybrid schemes
6.7 Voltage and reactive power monitoring
6.7.1 Measured data
6.7.2 Statistical approach to voltage monitoring
6.7.3 Complex POD/PCA analysis
References
7 Wide-area monitoring of power systems with high penetration of distributed generation
7.1 Introduction
7.2 Wide-area monitoring of power systems integrated with geographically distributed generation sources
7.2.1 Critical issues in system monitoring
7.2.2 Distributed setup
7.2.3 Data fusion
7.2.4 Sensor placement and signal (state) reconstruction
7.2.5 Spatiotemporal correlation
7.3 Frequency monitoring
7.3.1 Large-scale frequency response
7.3.2 RoCoF estimation
7.3.3 Frequency response monitoring: the COG approach
7.3.4 Distributed monitoring
7.3.5 Oscillation detection analysis
7.4 Voltage and reactive power monitoring
7.4.1 Voltage monitoring
7.4.2 Dimensionality reduction and visualization
7.4.3 Islanding and coherency
7.4.4 Statistical quality control
7.5 Data correlation
7.5.1 Data correlation and distance measures
7.5.2 Oscillation detection analysis
7.6 Tensor representations for system monitoring
7.6.1 Background
7.6.2 Tensor decompositions
7.6.3 The CP decomposition
7.6.4 The Tucker decomposition
7.6.5 Tucker reconstruction
7.7 A case study involving simultaneous analysis of multiple datasets
7.7.1 Base case and modified case description
7.7.2 Time series representation
7.7.3 Tensor representation
7.7.4 Implementation and results
References
8 Distributed wide-area oscillation monitoring
8.1 Introduction
8.2 Hierarchically distributed, wide-area oscillation monitoring
8.3 Hierarchical-distributed WAMS architectures
8.3.1 Background and motivation
8.3.2 Partitioning of data blocks
8.3.3 Partitioning of sensor blocks
8.4 Single-block methods
8.4.1 Single-block PCA (SB-PCA) models
8.5 MB-PCA methods
8.5.1 MB-PCA models
8.5.2 Numerical implementation
8.6 Partial least squares regression
8.7 MB canonical correlation analysis
8.7.1 Two-block CCA
8.7.2 Multiset canonical correlation analysis
8.8 Tensor representations
8.8.1 Tensor-based multiblock representations
8.8.2 The Tucker decomposition
8.8.3 HO-SVD analysis
8.8.4 Tensor rank
8.8.5 Dynamic system reconstruction from tensor approximations
8.8.6 Three-way clustering
8.8.7 Sequentially truncated higher-order SVD (ST-HO-SVD)
8.8.8 Numerical considerations
8.8.9 Reconstruction error
8.9 Application to simulated data
8.9.1 The 14-machine Australian test power system
8.9.2 Application to a large-scale power system
8.10 Recent work
References
9 Near real-time analysis and monitoring
9.1 Introduction
9.2 Toward near real-time monitoring of system behavior
9.3 Data processing and conditioning
9.3.1 Wavelet denoising and filtering
9.3.2 EMD-based filtering
9.4 Anomaly detection from changes in system behavior
9.4.1 Event trigger
9.4.2 Event detection based on linear filtering
9.4.3 An illustration
9.5 Time-series approaches to detection of abnormal operation
9.5.1 Near real-time implementations
9.5.2 Near real-time implementation of the Hilbert transform (HT)
9.5.3 Local mean speed
9.5.4 Online applications of wavelet transform
9.5.5 Other approaches
9.6 Pattern recognition-based disturbance detection
9.7 Sliding window-based methods
9.7.1 Local HHT analysis
9.7.2 Numerical example
9.7.3 Sliding window-based Koopman mode analysis
9.8 Recursive processing methods
9.8.1 State-space model for linear regression
9.8.2 Adaptive tracking of system oscillatory modes
9.9 Data-driven prognosis
References
10 Interpretation and visualization of wide-area phasor measurement unit measurements
10.1 Introduction
10.2 Loss of generation oscillation event
10.2.1 Operational context
10.2.2 Recorded measurements
10.3 Analysis and visualization of recorded data
10.3.1 Mode shape characterization
10.3.2 Damping estimation
10.3.3 Instantaneous parameters
10.3.4 Multitemporal, multiscale analysis of measured data
10.3.5 Performance evaluation
10.4 Pattern recognition analysis
10.4.1 Diffusion map analysis
10.4.2 Comparison with other approaches
10.5 POD/BSS analysis
10.6 Distributed monitoring
10.6.1 Single-block representation
10.6.2 Multiblock PCA of PMU data
10.6.3 C-means analysis
10.7 Validation of power system model
10.7.1 Frequency scanning analysis
10.7.2 Single-machine infinite bus scan
10.7.3 Large system performance
10.8 Evaluation of control performance
References
Appendix A Physical meaning of proper orthogonal modes
A.1 Eigenvalue-based decomposition
A.2 SVD-based POD
References
Appendix B Data for the 5-machine 10-bus test system
B.1 System data
B.2 Base case load flow condition
B.3 Modeling of wind and solar PV farms
References
Appendix C Masking techniques to improve empirical mode decomposition
C.1 Energy-based masking technique
References
Index