Data Fusion and Data Mining for Power System Monitoring

Cover
Half Title
Title Page
Copyright Page
Table of Contents
Preface
Author
List of Abbreviations
Section I: Overview of Modern Wide-Area Monitoring Systems
	1: Introduction
		1.1 Introduction to Power System Monitoring
		1.2 Wide-Area Power System Monitoring
			1.2.1 WAMS Architectures
			1.2.2 Synchronized Sensor Technology
			1.2.3 Integrated Communication Across the Grid
		1.3 Data Fusion and Data Mining for Power Health Monitoring
		1.4 Dimensionality Reduction
		1.5 Distribution System Monitoring
		1.6 Power System Data
		1.7 Sensor Placement for System Monitoring
		References
	2: Data Mining and Data Fusion Architectures
		2.1 Introduction
		2.2 Trends in Data Fusion and Data Monitoring
		2.3 Data Mining and Data Fusion for Enhanced Monitoring
			2.3.1 Data Fusion
			2.3.2 Time-Series Data Mining
		2.4 Data Fusion Architectures for Power System Health Monitoring
			2.4.1 Centralized Data Fusion Architectures
			2.4.2 Distributed Data Fusion Architectures
			2.4.3 Hierarchical Data Fusion Architectures
			2.4.4 Next-Generation Data Fusion Architectures
		2.5 Open Issues in Data Fusion
		References
Section II: Advanced Projection-Based Data Mining and Data Fusion Techniques
	3: Data Parameterization, Clustering, and Denoising
		3.1 Introduction: Background and Driving Forces
		3.2 Spatiotemporal Data Set Trajectories and Spatial Maps
			3.2.1 Dynamic Trajectories
			3.2.2 Proximity Measures
			3.2.3 Spectral Graph Theory
		3.3 Power System Data Normalization and Scaling
			3.3.1 Departure from Mean Value
			3.3.2 Standardization
		3.4 Nonlinear Dimensionality Reduction
			3.4.1 Diffusion Maps
				3.4.1.1 Numerical Issues
				3.4.1.2 Diffusion Distances
				3.4.1.3 Sparse Diffusion Maps Representation
				3.4.1.4 Time-Domain Decompositions
				3.4.1.5 Physical Interpretation of Diffusion Coordinates
			3.4.2 The Meila–Shi Algorithm
			3.4.3 Distributed Stochastic Neighbor Embedding (t-SNE) Method
			3.4.4 Multiview Diffusion Maps
		3.5 Clustering Schemes
			3.5.1 K-Means Clustering
			3.5.2 Dispersion
			3.5.3 C-Means Clustering
				3.5.3.1 Trajectory Clustering: Diffusion Maps Analysis
				3.5.3.2 C-Means Clustering
		3.6 Detrending and Denoising of Power System Oscillations
		References
	4: Spatiotemporal Data Mining
		4.1 Introduction
		4.2 Data Mining and Knowledge Discovery
		4.3 Spatiotemporal Modeling of Dynamic Processes
			4.3.1 Conceptual Framework
			4.3.2 Proper Orthogonal Decomposition (POD) Analysis
			4.3.3 Physical Interpretation of the POD Modes
			4.3.4 SVD-Based Proper Orthogonal Decomposition
			4.3.5 Principal Components Analysis
			4.3.6 Koopman-Based Approximations
				4.3.6.1 Mode Selection
				4.3.6.2 Spatiotemporal Clustering and Classification
				4.3.6.3 Application as a Predictive Tool
		4.4 Space-Time Prediction and Forecasting
			4.4.1 Interpolation and Extrapolation
			4.4.2 Compressive Sampling
			4.4.3 Spatiotemporal Clustering and Classification
		4.5 Spatiotemporal Data Mining and Pattern Evaluation
			4.5.1 Pattern Mining
			4.5.2 Spatiotemporal Autocorrelation
			4.5.3 Canonical Correlation Analysis
		References
	5: Multisensor Data Fusion
		5.1 Introduction and Motivation
		5.2 Spatiotemporal Data Fusion
		5.3 Data Fusion Principles
			5.3.1 Problem Formulation
			5.3.2 Unimodal (Single Type) Analysis Approaches
			5.3.3 Feature-Level Data Fusion
		5.4 Multisensor Data Fusion Framework
			5.4.1 Blind Source Separation-Based Data Fusion
			5.4.2 Feature-Based Fusion of Multimodal Data
		5.5 Multimodal Data Fusion Techniques
			5.5.1 Multiscale PCA/PLS Methods
			5.5.2 Multiview Diffusion Maps
			5.5.3 Partial Least Squares
			5.5.4 Other Approaches
		5.6 Case Study
			5.6.1 Partial Least Squares Correlation
			5.6.2 Multiblock PCA Analysis of Measured Data
		References
	6: Dimensionality Reduction and Feature Extraction and Classification
		6.1 Introduction: Background and Driving Forces
		6.2 Fundamentals of Dimensionality Reduction
			6.2.1 Formal Framework
			6.2.2 Numerical Considerations
		6.3 Data-Driven Feature Extraction Procedures
			6.3.1 Feature Extraction
			6.3.2 Feature Selection
		6.4 Dimensionality Reduction Methods
			6.4.1 Projection-Based Methods
			6.4.2 Coarse Graining of Dynamic Trajectories
			6.4.3 Koopman and DMD Analysis
			6.4.4 Physical Interpretation
		6.5 Dimensionality Reduction for Classification and Cluster Validation
			6.5.1 Cluster Validation
			6.5.2 Temporal Correlations
			6.5.3 Time-Scale Separation of System Dynamics
		6.6 Markov Dynamic Spatiotemporal Models
			6.6.1 First-Order Gauss-Markov Process
			6.6.2 First-Order Markov Differential Equation
			6.6.3 Markov Clustering Algorithm
			6.6.4 Predictive Modeling
		6.7 Sensor Selection and Placement
			6.7.1 Sensor Placement
			6.7.2 Mode Reconstruction
			6.7.3 Extrapolation
			6.7.4 Data Classification
		6.8 Open Problems in Nonlinear Dimensionality Reduction
			6.8.1 Effect of Noise and Dropped Data Points
			6.8.2 Missing Data and Uncertain Value
		References
	7: Forecasting Decision Support Systems
		7.1 Introduction
		7.2 Background: Early Warning and Decision Support Systems
		7.3 Data-Driven Prognostics
		7.4 Space-Time Forecasting and Prediction
			7.4.1 Decomposition-Based Methods
			7.4.2 Exponential Smoothing Methods
			7.4.3 State-Space Models
		7.5 Kalman Filtering Approach to System Forecasting
			7.5.1 The Kalman Filter
			7.5.2 Forecasting
			7.5.3 Time-Varying Trend Estimation
		7.6 Dynamic Harmonic Regression
			7.6.1 State Space Representation
			7.6.2 The Kalman Filter and Smoothing Algorithms
		7.7 Damage Detection
		7.8 Power Systems Time Series Forecasting
		7.9 Anomaly Detection in Time Series
			7.9.1 Extraction of Dynamic Trends
			7.9.2 Second Derivative Criterion
		References
Section III: Challenges and Opportunities in the Application of Data Mining and Data Fusion Techniques
	8: Data Fusion and Data Mining Analysis and Visualization
		8.1 Introduction
		8.2 Advanced Visualization Techniques
			8.2.1 Geospatial and Spatiotemporal Maps
			8.2.2 Direct Visualization of Power System Data
			8.2.3 Cluster-Based Visualization of Multidimensional Data
			8.2.4 Moving Clusters
		8.3 Multivariable Modeling and Visualization
			8.3.1 Formal Framework
			8.3.2 Multiblock Analysis Techniques
			8.3.3 Open Issues
			8.3.4 Propagation Faults: Consensus PCA
			8.3.5 Bandwidth of the Communication Networks
		8.4 Cluster-Based Visualization of Multidimensional Data
			8.4.1 Motivation
			8.4.2 Wide-Area Voltage Visualization
			8.4.3 Data Mining-Based Visualization Methods
		8.5 Spatial and Network Displays
		References
	9: Emerging Topics in Data Mining and Data Fusion
		9.1 Introduction
		9.2 Dynamic Spatiotemporal Modeling
			9.2.1 Tensors for Data Mining and Data Fusion
			9.2.2 Tensors for Data Mining and Data Fusion
		9.3 Challenges for the Analysis of High-Dimensional Data
		9.4 Distributed Data Mining
		9.5 Dimensionality Reduction
			9.5.1 Limitations of Existing Dimensionality Reduction Methods
			9.5.2 Deep Learning Multidimensional Projections
		9.6 Bio-Inspired Data Mining and Data Fusion
		9.7 Other Emerging Issues
		9.8 Application to Power System Data
			9.8.1 Wind and Renewable Energy Generation Forecasting
			9.8.2 Application to Distribution Systems
		References
	10: Experience with the Application of Data Fusion and Data Mining for Power System Health Monitoring
		10.1 Introduction
		10.2 Background
			10.2.1 Wide-Area Measurement Scenarios
			10.2.2 Basic Modal Properties
		10.3 Sensor Placement
		10.4 Cluster-Based Visualization of Transient Performance
			10.4.1 Case 1
			10.4.2 Case 2
		10.5 Multimodal Fusion of Observational Data
			10.5.1 Multiblock PCA Analysis
			10.5.2 Multiview Diffusion Maps
		References
	Index