非侵入式負荷識別技術作為智能電網(wǎng)需求側(cè)能源管理的基礎，在優(yōu)化電網(wǎng)供求關系、促進節(jié)能減排等方面具有廣闊的應用前景。本書全面介紹了非侵入式負荷識別的相關的基本理論、關鍵技術和應用實例。全書分為10章，第1章介紹非侵入式負荷識別的技術背景、相關定義和涉及的關鍵基礎問題；接著分4大部分展開，第一篇介紹非侵入式負荷分解的前處理過程，包括第2章介紹的負荷時序狀態(tài)變動檢測和第3章介紹的用電負荷差異化特征提��；第二篇介紹非侵入式負荷統(tǒng)計識別方法，重點討論基于模板匹配的負荷識別模型和基于穩(wěn)態(tài)電流分解的負荷識別模型；第三篇介紹用電負荷智能識別方法，包括基于機器學習的負荷識別模型，基于隱含馬爾可夫的負荷識別模型和基于深度學習的負荷識別模型；第四篇介紹非侵入式負荷識別在智能用電負荷預測方法中的應用，包括用電負荷時序確定性預測和用電負荷時序區(qū)間預測。各部分內(nèi)容都附有實例分析，幫助讀者深入理解相關內(nèi)容、激發(fā)創(chuàng)新靈感。
更多科學出版社服務，請掃碼獲取。
Contents
1 Introduction 1
1.1 Overview of the Non-intrusive Load Monitoring 1
1.1.1 The Non-intrusive Load Monitoring 1
1.1.2 Overview of Recent Research in Non-intrusive Load Monitoring 3
1.2 Fundamental Key Problems of Non-intrusive Load Monitoring 8
1.2.1 Event Detection in Non-intrusive Load Monitoring 8
1.2.2 Feature Extraction in Non-intrusive Load Monitoring 10
1.2.3 Load Identification in Non-intrusive Load Monitoring 13
1.2.4 Energy Forecasting in Smart Buildings 15
1.3 Scope of This Book 17
References 19
2 Detection of Transient Events in Time Series 23
2.1 Introduction 23
2.2 Cumulative Sum Based Transient Event Detection Algorithm 24
2.2.1 Mathematical Description of Change Point Detection 24
2.2.2 Parametric CUSUM Algorithm 24
2.2.3 Non-parametric CUSUM Algorithm 25
2.2.4 Sliding Windows Based on Two-Sided CUSUM Algorithm 26
2.2.5 Original Dataset 26
2.2.6 Evaluation Criteria and Results Analysis 29
2.3 Generalized Likelihood Ratio 36
2.3.1 The Theoretical Basis of GLR 36
2.3.2 Comparison of Event Detection Results 37
2.4 Sequential Probability Ratio Test 39
2.4.1 The Theoretical Basis of SPRT 39
2.4.2 Comparison of Event Detection Results 40
2.5 Experiment Analysis 42
2.5.1 The Results of Three Kinds of Algorithms. 42
2.5.2 Conclusion 42
References 43
3 Appliance Signature Extraction 45
3.1 Introduction 45
3.1.1 Background 45
3.1.2 Feature Evaluation Indices 46
3.1.3 Classification Evaluation Indices 48
3.1.4 Data Selection 49
3.2 Features Based on Conventional Physical Definition 50
3.2.1 The Theoretical Basis of Physical Definition Features 50
3.2.2 Feature Extraction 52
3.2.3 Feature Evaluation 53
3.2.4 Classification Results 54
3.3 Features Based on Time-Frequency Analysis 55
3.3.1 The Theoretical Basis of Harmonic Features 55
3.3.2 Feature Extraction 56
3.3.3 Feature Evaluation 58
3.3.4 Classification Results 59
3.4 Features Based on VI Image 62
3.4.1 The Theoretical Basis of VI Image Features 62
3.4.2 Feature Extraction 65
3.4.3 Feature Evaluation 68
3.4.4 Classification Results 70
3.5 Features Based on Adaptive Methods 73
3.5.1 The Theoretical Basis of Adaptive Features 73
3.5.2 Feature Extraction 74
3.5.3 Classification Results 74
3.6 Experimental Analysis 76
3.6.1 Comparative Analysis of Classification Performance 76
3.6.2 Conclusion 76
References 77
4 Appliance Identification Based on Template Matching 79
4.1 Introduction 79
4.1.1 Background 79
4.1.2 Data Preprocessing of the PLAID Dataset 80
4.2 Appliance Identification Based on Decision Tree 82
4.2.1 The Theoretical Basis of Decision Tree 82
4.2.2 Steps of Modeling 83
4.2.3 Classification Results 84
4.3 Appliance Identification Based on KNN Algorithm 85
4.3.1 The Theoretical Basis of KNN 85
4.3.2 Steps of Modeling 86
4.3.3 Classification Results 87
4.4 Appliance Identification Based on DTW Algorithm 88
4.4.1 The Theoretical Basis of DTW 88
4.4.2 Steps of Modeling 90
4.4.3 Classification Results 91
4.5 Experiment Analysis 91
4.5.1 Model Framework 91
4.5.2 Comparative Analysis of Classification Performance 92
4.5.3 Conclusion 100
References 102
5 Steady-State Current Decomposition Based Appliance Identification 105
5.1 Introduction 105
5.2 Classical Steady-State Current Decomposition Models 107
5.2.1 Model Framework 107
5.2.2 Classical Features of Steady-State Decomposition and the Feature Extraction Method 108
5.2.3 Classical Methods in Steady-State Current Decomposition 115
5.2.4 Performance of the Various Features and Models 116
5.3 Current Decomposition Models Based on Harmonic Phasor 120
5.3.1 Model Framework 120
5.3.2 Novel Features of Steady-State Current Decomposition 121
5.3.3 Multi-objective Optimization Methods in Steady-State Current Decomposition 124
5.3.4 Performance of the Novel Features and Multi-objective Optimization Models 126
5.4 Current Decomposition Models Based on Non-negative Matrix Factor 129
5.4.1 Model Framework 129
5.4.2 Reconstruction of the Data 129
5.4.3 Non-negative Matrix Factorization Method of the Current Decomposition 131
5.4.4 Evaluation of the NMF Method in Current Decomposition 133
5.5 Experiment Analysis 135
5.5.1 Data Generation 135
5.5.2 Comparison Analysis of the Features Used in the Steady-State Decomposition 136
5.5.3 Comparison Analysis of the Models Used in the Steady-State Decomposition 137
5.5.4 Conclusion 138
References 141
6 Machine Learning Based Appliance Identification 141
6.1 Introduction 142
6.2 Appliance Identification Based on Extreme Learning Machine 142
6.2.1 The Theoretical Basis of ELM 143
6.2.2 Steps of Modeling 143
6.2.3 Classification Results 145
6.3 Appliance Identification Based on Support Vector Machine 145
6.3.1 The Theoretical Basis of SVM 146
6.3.2 Steps of Modeling 146
6.3.3 Classification Results 148
6.4 Appliance Identification Based on Random Forest 148
6.4.1 The Theoretical Basis of Random Forest 149
6.4.2 Steps of Modeling 149
6.4.3 Classification Results 150
6.5 Experiment Analysis 150
6.5.1 Model Framework 150
6.5.2 Feature Preprocessing for Non-intrusive Load Monitoring 150
6.5.3 Classifier Model Optimization Algorithm for Non-intrusive Load Monitoring 155
6.6 Conclusion 158
References 161
7 Hidden Markov Models Based Appliance 163
7.1 Introduction 163
7.2 Appliance Identification Based on Hidden Markov Models 164
7.2.1 Basic Problems Solved by HMM 164
7.2.2 Data Preprocessing 165
7.2.3 Determination of Load Status Information 167
7.3 Appliance Identification Based on Factorial Hidden Markov Models 170
7.3.1 The Theoretical Basis of the FHMM 170
7.3.2 Load Decomposition Steps Based on FHMM 171
7.3.3 Load Power Estimation 172
7.3.4 Decomposition Experiment Based on FHMM 173
7.3.5 Evaluation Criteria and Result Analysis 177
7.4 Appliance Identification Based on Hidden Semi-Markov Models 183
7.4.1 Hidden Semi-Markov Model 183
7.4.2 Improved Viterbi Algorithm 184
7.4.3 Evaluation Criteria and Result Analysis 184
7.5 Experiment Analysis 184
References 189
8 Deep Learning Based Appliance Identification 191
8.1 Introduction 191
8.1.1 Deep Learning 191
8.1.2 NILM Based on Deep Learning 191
8.2 Appliance Identification Based on End-to-End Decomposition 192
8.2.1 Single Feature Based LSTM Network Load Decomposition 193
8.2.2 Multiple Features Based LSTM Network Load Decomposition 196
8.3 Appliance Identification Based on Appliance Classification 199
8.3.1 Appliance Identification Based on CNN 199
8.3.2 Appliance Identification Based on AlexNet 203
8.3.3 Appliance Identification Based on LeNet-SVM Model 207
8.4 Experiment Analysis 211
8.4.1 Experimental Analysis of End-to-End Decomposition 211
8.4.2 Experimental Analysis of Appliance Classification 212
References 214
9 Deterministic Prediction of Electric Load Time Series 215
9.1 Introduction 215
9.1.1 Background 215
9.1.2 Advance Prediction Strategies 216
9.1.3 Original Electric Load Time Series 217
9.2 Load Forecasting Based on ARIMA Model 218
9.2.1 Model Framework 218
9.2.2 Theoretical Basis of ARIMA 218
9.2.3 Modeling Steps of ARIMA Predictive Model 221
9.2.4 Predictive Results 224
9.2.5 The Theoretical Basis of EMD 227
9.2.6 Optimization of EMD Decomposition Layers 227
9.2.7 Predictive Results 231
9.3 Load Forecasting Based on Elman Neural Network 234
9.3.1 Model Framework 234
9.3.2 Steps of Modeling 234
9.3.3 Predictive Results 235
9.3.4 Optimization of EMD Decomposition Layers 238
9.3.5 Predictive Results 239
9.4 Experiment Analysis 239
9.4.1 Comparative Analysis of Predictive Performance 239
9.5 Conclusion 243
References 244
10 Interval Prediction of Electric Load Time Series 247
10.1 Introduction 247
10.2 Interval Prediction Based on Quantile Regression 248
10.2.1 The Performance Evaluation Metrics 248
10.2.2 Original Sequence for Modeling 251
10.2.3 The Theoretical Basis of Quantile Regression 252
10.2.4 Quantile Regression Based on the Total Electric Load Time Series 253
10.2.5 Quantile Regression Based on Additional Time and Date Information 256
10.2.6 Quantile Regression Based on Information from Sub-meters 260
10.3 Interval Prediction Based on Gaussian Process Filtering 264
10.3.1 The Theoretical Basis of Gaussian Process Regression 264
10.3.2 Gaussian Process Regression Based on the Total Electric Load Time Series 265
10.3.3 Gaussian Process Regression Based on Different Input Features 270
10.3.4 Gaussian Process Regression Based on Feature Selection 273
10.4 Experiment Analysis 274
References 276
書單推薦
