Data Mining and Machine Learning in Building Energy Analysis

The energy consumption of a building has, in recent years, become a determining factor during its design and construction. With carbon footprints being a growing issue, it is important that buildings be optimized for energy conservation and CO2 reduction. This book therefore presents AI models and optimization techniques related to this application.

The authors start with a review of recent models for the prediction of building energy consumption: engineering methods, statistical methods, artificial intelligence methods, ANNs and SVMs in particular. The book then focuses on SVMs, by first applying them to building energy consumption, then presenting the principles and various extensions, and SVR. The authors then move on to RDP, which they use to determine building energy faults through simulation experiments before presenting SVR model reduction methods and the benefits of parallel computing. The book then closes by presenting some of the current research and advancements in the field.

Autorentext

Frédéric Magoulès is Professor at the Ecole Centrale Paris in France and Honorary Professor at the University of Pècs in Hungary. His research focuses on parallel computing, numerical linear algebra and machine learning. Hai-Xiang Zhao is Senior Researcher at Amadeus in France. His research focuses on parallel computing, data mining and machine learning.

Inhalt
Preface ix

Introduction xi

Chapter 1. Overview of Building Energy Analysis 1

1.1. Introduction 1

1.2. Physical models 3

1.3. Gray models 6

1.4. Statistical models 6

1.5. Artificial intelligence models 8

1.5.1. Neural networks 8

1.5.2. Support vector machines 13

1.6. Comparison of existing models 14

1.7. Concluding remarks . 16

Chapter 2. Data Acquisition for Building Energy Analysis 17

2.1. Introduction 17

2.2. Surveys or questionnaires 18

2.3. Measurements 21

2.4. Simulation 25

2.4.1. Simulation software 26

2.4.2. Simulation process 28

2.5. Data uncertainty 34

2.6. Calibration 35

2.7. Concluding remarks 37

Chapter 3. Artificial Intelligence Models 39

3.1. Introduction 39

3.2. Artificial neural networks 40

3.2.1. Single-layer perceptron 41

3.2.2. Feed forward neural network 43

3.2.3. Radial basis functions network 44

3.2.4. Recurrent neural network 47

3.2.5. Recursive deterministic perceptron 49

3.2.6. Applications of neural networks 51

3.3. Support vector machines 53

3.3.1. Support vector classification 54

3.3.2. -support vector regression 59

3.3.3. One-class support vector machines 62

3.3.4. Multiclass support vector machines 63

3.3.5. v-support vector machines 64

3.3.6. Transductive support vector machines 65

3.3.7. Quadratic problem solvers . 67

3.3.8. Applications of support vector machines 75

3.4. Concluding remarks 76

Chapter 4. Artificial Intelligence for Building Energy Analysis 79

4.1. Introduction 79

4.2. Support vector machines for building energy prediction 80

4.2.1. Energy prediction definition 80

4.2.2. Practical issues 81

4.2.3. Support vector machines for prediction 85

4.3. Neural networks for fault detection and diagnosis 91

4.3.1. Description of faults 94

4.3.2. RDP in fault detection 95

4.3.3. RDP in fault diagnosis 100

4.4. Concluding remarks 102

Chapter 5. Model Reduction for Support Vector Machines 103

5.1. Introduction 103

5.2. Overview of model reduction 104

5.2.1. Wrapper methods 105

5.2.2. Filter methods 106

5.2.3. Embedded methods 107

5.3. Model reduction for energy consumption 108

5.3.1. Introduction 108

5.3.2. Algorithm 109

5.3.3. Feature set description 111

5.4. Model reduction for single building energy 112

5.4.1. Feature set selection 112

5.4.2. Evaluation in experiments 114

5.5. Model reduction for multiple buildings energy 116

5.6. Concluding remarks 119

Chapter 6. Parallel Computing for Support Vector Machines 121

6.1. Introduction 121

6.2. Overview of parallel support vector machines 122

6.3. Parallel quadratic problem solver 123

6.4. MPI-based parallel support vector machines 127

6.4.1. Message passing interface programming model 127

6.4.2. Pisvm 129

6.4.3. Psvm 130

6.5. MapReduce-based parallel support vector machines 130

6.5.1. MapReduce programming model 131

6.5.2. Caching technique 133

6.5.3. Sparse data representation 133

6.5.4. Comparison of MRPsvm with Pisvm 134

6.6. MapReduce-based parallel -support vector regression 138

6.6.1. Implementation aspects 138

6.6.2. Energy consumption datasets 139

6.6.3. Evaluation for building energy prediction 140

6.7. Concluding remarks 142

Summary and Future of Building Energy Analysis 145

Bibliography 149

Index 163