Data Mining and Machine Learning in Building Energy Analysis E-Book

Frédéric Magoules

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Sprache: Englisch

Seitenzahl: 268

Veröffentlichungsjahr: 2016

Beschreibung

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.

Leseprobe

Table of Contents

Cover

Title

Copyright

Preface

Introduction

1 Overview of Building Energy Analysis

1.1. Introduction

1.2. Physical models

1.3. Gray models

1.4. Statistical models

1.5. Artificial intelligence models

1.6. Comparison of existing models

1.7. Concluding remarks

2 Data Acquisition for Building Energy Analysis

2.1. Introduction

2.2. Surveys or questionnaires

2.3. Measurements

2.4. Simulation

2.5. Data uncertainty

2.6. Calibration

2.7. Concluding remarks

3 Artificial Intelligence Models

3.1. Introduction

3.2. Artificial neural networks

3.3. Support vector machines

3.4. Concluding remarks

4 Artificial Intelligence for Building Energy Analysis

4.1. Introduction

4.2. Support vector machines for building energy prediction

4.3. Neural networks for fault detection and diagnosis

4.4. Concluding remarks

5 Model Reduction for Support Vector Machines

5.1. Introduction

5.2. Overview of model reduction

5.3. Model reduction for energy consumption

5.4. Model reduction for single building energy

5.5. Model reduction for multiple buildings energy

5.6. Concluding remarks

6 Parallel Computing for Support Vector Machines

6.1. Introduction

6.2. Overview of parallel support vector machines

6.3. Parallel quadratic problem solver

6.4. MPI-based parallel support vector machines

6.5. MapReduce-based parallel support vector machines

6.6. MapReduce-based parallel support vector regression

6.7. Concluding remarks

Summary and Future of Building Energy Analysis

Bibliography

Index

End User License Agreement

List of Tables

1 Overview of Building Energy Analysis

Table 1.1. Brief review of commonly used methods for the prediction of building energy consumption

Table 1.2. Comparative analysis of commonly used methods for the prediction of building energy consumption

2 Data Acquisition for Building Energy Analysis

Table 2.1. Description of a single building (in metric units)

Table 2.2. Building materials used in simulation

4 Artificial Intelligence for Building Energy Analysis

Table 4.1. Sample of dataset with the description of prediction

Table 4.2. Training, testing and result

Table 4.3. Consumption period for the training and testing datasets

Table 4.4. Description of a single building (in metric units)

Table 4.5. Faults introduced to the building

Table 4.6. Number of samples in the datasets

Table 4.7. Results of recursive deterministic perceptron model in two experiments

Table 4.8. Recursive deterministic perceptron model accuracy (%) on datasets in four cases. Ten months for training and two months for testing. Each dataset contains normal consumption with faults caused by one equipment, indicated in the first column

Table 4.9. Recursive deterministic perceptron model accuracy in the diagnostic procedure with the example inputs

5 Model Reduction for Support Vector Machines

Table 5.1. Twenty-three features for the model training and testing on one building’s consumption

Table 5.2. Scores of features evaluated by RGS and CC selection methods. The stars indicate selected features in that case.

Table 5.3. Comparison of model performance on different feature sets. NF: Number of features, MSE: Mean squared error, SCC: Squared correlation coefficient

Table 5.4. Prediction results of support vector regression with two kernel methods on three data sets. BF: Before feature selection, AF: After feature selection, MSE: Mean squared error, SCC: Squared correlation coefficient

6 Parallel Computing for Support Vector Machines

Table 6.2. Characteristics of the multi-core systems

Table 6.3. Description of the five datasets and the two parameters of support vector machines on each dataset

Table 6.4. Training time and accuracy of the three systems on five datasets performed on computer I. The unit of time is second

Table 6.5. Description of the three datasets and the three parameters of support vector regression on each dataset. #tr: number of training samples, #te: number of testing samples

Table 6.6. Characteristics of the experimental environment

Table 6.7. Training time and performance of the three predictors on three datasets performed on computer I. bd: building, nSVs: number of support vectors, MSE: mean squared error, SCC: squared correlation coefficient. The unit of time is second

Table 6.8. Training time of the three predictors performed on computer II. Time unit is second

List of Illustrations

1 Overview of Building Energy Analysis

Figure 1.1. Annual energy consumption in each sector of France

2 Data Acquisition for Building Energy Analysis

Figure 2.1. Overview of EnergyPlus software

Figure 2.2. Dry bulb temperature in the first 20 days of January and July. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 2.3. Relative humidity in the first 20 days of January and July. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 2.4. Hourly electricity consumptions of the single building in the first simulation month (November)

Figure 2.5. Hourly electricity consumptions of two buildings in November. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 2.6. Data uncertainty

3 Artificial Intelligence Models

Figure 3.1 Example of one neuron a) and a single-layer perceptron b)

Figure 3.2. Feed forward neural network

Figure 3.3. Simple radial basis functions neural network architecture

Figure 3.4. Simple recurrent neural network

Figure 3.5. Flowchart of the incremental recursive deterministic perceptron model training

Figure 3.6. Linearly separable classification problem

Figure 3.7. ε-tube for support vector regression

4 Artificial Intelligence for Building Energy Analysis

Figure 4.1. Flow chart of a learning process

Figure 4.2. Dataset

Figure 4.3. Measured and predicted district heating demand in heating season. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 4.4. Measured and predicted electricity consumption in randomly selected 48 hr. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 4.5. Prediction error of the model on the consumption of March, May, July and September. The model is trained on the data of January. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 4.6. Prediction error of the model on the consumption of June, August, October and December. The model is trained on the data from January to April. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 4.7. Prediction error of the model on the consumption of September, October, November and December. The model is trained on the data from January to August. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 4.8. Mean squared error of the three models on the designed testing months. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 4.9. Squared correlation coefficient of the three models on the designed testing months. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 4.10. Measured and predicted electricity consumption for a totally new building. The model is trained on 99 buildings. For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 4.11. Normal and faulty facility electric consumption in 1 year (the unit is Joules (J)). For a color version of this figure, see www.iste.co.uk/magoules/mining.zip

Figure 4.12. Flow chart of fault diagnosis. M

i

is the

i

th model,

A

i

is the prediction accuracy through the

i

th model and

E

i

stands for the

i

th equipment

5 Model Reduction for Support Vector Machines

Figure 5.1. Comparison of measured and predicted daily electricity consumption for a particular building on working days, with feature selection performed

Figure 5.2. Relative error for the prediction

Figure 5.3. Dry bulb temperature in the first 11 days of January. For a color version of the figure, www.iste.co.uk/magoules/mining.zip

Figure 5.4. Comparison of model performance from the standpoint of SCC before and after feature selection for radial basis function kernel

Figure 5.5. Comparison of training time before and after FS for RBF kernel

6 Parallel Computing for Support Vector Machines

Figure 6.1. Architecture of the parallelization in one iteration

Figure 6.2. Speedup of Pisvm and MRPsvm over Libsvm when running on computer II

Figure 6.3. Speedup of Pisvm and MRPsvm on one building’s data

Guide

Cover

Table of Contents

Begin Reading

Pages

C1

iii

iv

v

ix

x

xi

xii

xiii

xiv

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

Data Mining and Machine Learning in Building Energy Analysis

Frédéric Magoulès

Hai-Xiang Zhao

Series EditorSerge Petiton

First published 2016 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc.

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the undermentioned address:

ISTE Ltd27-37 St George’s RoadLondon SW19 4EUUK

www.iste.co.uk

John Wiley & Sons, Inc.111 River StreetHoboken, NJ 07030USA

www.wiley.com

© ISTE Ltd 2016

The rights of Frédéric Magoulès and Hai-Xiang Zhao to be identified as the authors of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act 1988.

Library of Congress Control Number: 2015958612

British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library

ISBN 978-1-84821-422-4