This book focuses on the performance optimization of fault diagnosis methods for power systems including both model-driven ones, such as the linear parameter varying algorithm, and data-driven ones, such as random matrix theory. Studies on fault diagnosis of power systems have long been the focus of electrical engineers and scientists. Pursuing a holistic approach to improve the accuracy and efficiency of existing methods, the underlying concepts toward several algorithms are introduced and then further applied in various situations for fault diagnosis of power systems in this book. The primary audience for the book would be the scholars and graduate students whose research topics including the control theory, applied mathematics, fault detection, and so on.
Author(s): Dinghui Wu, Juan Zhang, Junyan Fan, Dandan Tang
Series: Engineering Applications of Computational Methods, 9
Publisher: Springer
Year: 2022
Language: English
Pages: 133
City: Singapore
Preface
Contents
Acronyms
1 Introduction
1.1 Background and Significance
1.1.1 Background
1.1.2 Significance
1.2 Book Organization
1.3 Preliminaries
1.3.1 Gaussian Fitting
1.3.2 Random Matrix Theory
1.3.3 Federated Learning
References
2 Short-Term Wind Power Prediction Based on DP-RVESN Model
2.1 Introduction
2.2 Methodology
2.2.1 Kernel Principal Component Analysis (KPCA)
2.2.2 Network Structure
2.2.3 Robust Variational Reasoning
2.3 Proposed Hybrid Forecasting Model
2.4 Experimental Results and Analysis
2.4.1 Data Process (DP)
2.4.2 Prediction Results
2.5 Conclusion
References
3 A Fault Detection Method for Power Systems Based on Random Matrix Theory
3.1 Introduction
3.2 Fault Detection Method Based on Random Matrix Theory
3.2.1 Construction of Random Matrix and Data Processing
3.2.2 Fault Detection Method Based on the Maximum Eigenvalue Variation Rate
3.2.3 Discussions with the Other Methods
3.3 Case Studies
3.3.1 Case 1: Effectiveness Verification of the MEVR Method Under Three-Phase Short-Circuit on Bus 9
3.3.2 Case 2: Effectiveness Verification of the MEVR Method Under Load Jump of Bus 29
3.4 Conclusion
References
4 A Fault Location Method for Power Systems Based on Random Matrix Theory
4.1 Introduction
4.2 Fault Location Method Based on Random Matrix Theory
4.2.1 Fault Location Based on the Largest Eigenvector
4.2.2 Discussions with the Other Methods
4.3 Case Study
4.3.1 Case 1: Effectiveness Verification of the Proposed Method for Simple Fault
4.3.2 Case 2: Effectiveness Verification of the Proposed Method for Simultaneous Multiple Faults
4.3.3 Case 3: Effectiveness Verification of the Proposed Method for Successive Multiple Faults
4.4 Conclusion
References
5 Joint Weighted Domain Adaptation Network for Bearing Fault Diagnosis
5.1 Introduction
5.2 Preliminaries
5.2.1 Transfer Learning Problem
5.2.2 MMD
5.2.3 JWDD
5.3 JWDAN for Transfer Fault Diagnosis
5.3.1 JWDAN
5.3.2 Training JWDAN
5.4 Experiments
5.4.1 Data Description
5.4.2 Experimental Setup
5.4.3 Results
5.4.4 Hyper-Parameters Analysis
5.4.5 Ablation Study
5.5 Conclusion
References
6 Bearing Fault Diagnosis of Wind Motor Based on Shared-FedAvg with Non-IID Data
6.1 Introduction
6.2 Related Knowledge
6.2.1 Federated Learning
6.2.2 IID and Non-IID
6.2.3 Fault Diagnosis on Client Side Based on WPD-CNN
6.3 Fault Diagnosis Algorithm Based on Shared-FedAvg
6.4 Experimental Analysis
6.4.1 Experimental Data
6.4.2 Different Diagnostic Strategies
6.4.3 Experimental Analysis
6.5 Conclusion
References
7 Federated Learning for Rotating Machinery Fault Diagnosis with Knowledge Distillation
7.1 Introduction
7.2 Related Knowledge
7.2.1 Federated Learning
7.2.2 Knowledge Distillation
7.3 Proposed Method
7.3.1 Dirichlet Data Splitting Strategy
7.3.2 CNN for Fault Diagnosis
7.3.3 Preposed Efficient Fault Diagnosis Methods
7.4 Experiments
7.4.1 Setup
7.4.2 Performance Overview of FedAvg
7.4.3 Evaluate Performance of KD-FedAvg
7.4.4 Generalization Analysis
7.5 Conclusion
References
