Data Science for Wind Energy PDF by Yu Ding

7:19 AM
Data Science for Wind Energy
By Yu Ding
Data Science for Wind Energy

Contents
Foreword xv
Preface xvii
Acknowledgments xxi
Chapter 1 _ Introduction 1
1.1 WIND ENERGY BACKGROUND 2
1.2 ORGANIZATION OF THIS BOOK 6
1.2.1 Who Should Use This Book 8
1.2.2 Note for Instructors 9
1.2.3 Datasets Used in the Book 9

Part I Wind Field Analysis
Chapter 2 _ A Single Time Series Model 17
2.1 TIME SCALE IN SHORTTERM
FORECASTING 18
2.2 SIMPLE FORECASTING MODELS 19
2.2.1 Forecasting Based on Persistence Model 19
2.2.2 Weibull Distribution 19
2.2.3 Estimation of Parameters in Weibull Distribution 20
2.2.4 Goodness of Fit 21
2.2.5 Forecasting Based on Weibull Distribution 23
2.3 DATA TRANSFORMATION AND STANDARDIZATION 24
2.4 AUTOREGRESSIVE MOVING AVERAGE MODELS 27
2.4.1 Parameter Estimation 28
2.4.2 Decide Model Order 29
2.4.3 Model Diagnostics 31
2.4.4 Forecasting Based on ARMA Model 34
2.5 OTHER METHODS 38
2.5.1 Kalman Filter 38
2.5.2 Support Vector Machine 40
2.5.3 Arti_cial Neural Network 45
2.6 PERFORMANCE METRICS 48
2.7 COMPARING WIND FORECASTING METHODS 50

Chapter 3 _ Spatiotemporal
Models 57
3.1 COVARIANCE FUNCTIONS AND KRIGING 57
3.1.1 Properties of Covariance Functions 58
3.1.2 Power Exponential Covariance Function 59
3.1.3 Kriging 60
3.2 SPATIOTEMPORAL
AUTOREGRESSIVE MODELS 65
3.2.1 Gaussian Spatio-temporal Autoregressive Model 65
3.2.2 Informative Neighborhood 68
3.2.3 Forecasting and Comparison 69
3.3 SPATIOTEMPORAL
ASYMMETRY AND SEPARABILITY 73
3.3.1 De_nition and Quanti_cation 73
3.3.2 Asymmetry of Local Wind Field 74
3.3.3 Asymmetry Quanti_cation 76
3.3.4 Asymmetry and Wake E_ect 78
3.4 ASYMMETRIC SPATIOTEMPORAL
MODELS 79
3.4.1 Asymmetric Non-separable Spatio-temporal Model 79
3.4.2 Separable Spatio-temporal Models 81
3.4.3 Forecasting Using Spatio-temporal Model 81
3.4.4 Hybrid of Asymmetric Model and SVM 83
3.5 CASE STUDY 83
Chapter 4 _ Regimeswitching
Methods for Forecasting 93
4.1 REGIMESWITCHING
AUTOREGRESSIVE MODEL 93
4.1.1 Physically Motivated Regime De_nition 94
4.1.2 Data-driven Regime Determination 96
4.1.3 Smooth Transition between Regimes 97
4.1.4 Markov Switching between Regimes 98
4.2 REGIMESWITCHING
SPACETIME
MODEL 99
4.3 CALIBRATION IN REGIMESWITCHING
METHOD 104
4.3.1 Observed Regime Changes 105
4.3.2 Unobserved Regime Changes 106
4.3.3 Framework of Calibrated Regime-switching 107
4.3.4 Implementation Procedure 111
4.4 CASE STUDY 113
4.4.1 Modeling Choices and Practical Considerations 113
4.4.2 Forecasting Results 115
Part II Wind Turbine Performance Analysis
Chapter 5 _ Power Curve Modeling and Analysis 125
5.1 IEC BINNING: SINGLEDIMENSIONAL
POWER CURVE 126
5.2 KERNELBASED
MULTIDIMENSIONAL
POWER CURVE 127
5.2.1 Need for Nonparametric Modeling Approach 128
5.2.2 Kernel Regression and Kernel Density Estimation 131
5.2.3 Additive Multiplicative Kernel Model 134
5.2.4 Bandwidth Selection 136
5.3 OTHER DATA SCIENCE METHODS 137
5.3.1 k-Nearest Neighborhood Regression 138
5.3.2 Tree-based Regression 139
5.3.3 Spline-based Regression 143
5.4 CASE STUDY 145
5.4.1 Model Parameter Estimation 145
5.4.2 Important Environmental Factors A_ecting Power
Output 147
5.4.3 Estimation Accuracy of Di_erent Models 150
Chapter 6 _ Production Efficiency Analysis and Power Curve 159
6.1 THREE EFFICIENCY METRICS 159
6.1.1 Availability 160
6.1.2 Power Generation Ratio 160
6.1.3 Power Coe_cient 161
6.2 COMPARISON OF EFFICIENCY METRICS 162
6.2.1 Distributions 164
6.2.2 Pairwise Di_erences 166
6.2.3 Correlations and Linear Relationships 168
6.2.4 Overall Insight 170
6.3 A SHAPECONSTRAINED
POWER CURVE MODEL 171
6.3.1 Background of Production Economics 172
6.3.2 Average Performance Curve 174
6.3.3 Production Frontier Function and E_ciency
Metric 177
6.4 CASE STUDY 179
Chapter 7 _ Quantification of Turbine Upgrade 187
7.1 PASSIVE DEVICE INSTALLATION UPGRADE 187
7.2 COVARIATE MATCHINGBASED
APPROACH 189
7.2.1 Hierarchical Subgrouping 189
7.2.2 One-to-One Matching 192
7.2.3 Diagnostics 193
7.2.4 Paired t-tests and Upgrade Quanti_cation 194
7.2.5 Sensitivity Analysis 196
7.3 POWER CURVEBASED
APPROACH 197
7.3.1 The Kernel Plus Method 198
7.3.2 Kernel Plus Quanti_cation Procedure 201
7.3.3 Upgrade Detection 202
7.3.4 Upgrade Quanti_cation 204
7.4 AN ACADEMIAINDUSTRY
CASE STUDY 204
7.4.1 The Power-vs-Power Method 206
7.4.2 Joint Case Study 208
7.4.3 Discussion 211
7.5 COMPLEXITIES IN UPGRADE QUANTIFICATION 213
Chapter 8 _ Wake Effect Analysis 219
8.1 CHARACTERISTICS OF WAKE EFFECT 219
8.2 JENSEN’S MODEL 220
8.3 A DATA BINNING APPROACH 222
8.4 SPLINEBASED
SINGLEWAKE
MODEL 223
8.4.1 Baseline Power Production Model 224
8.4.2 Power Di_erence Model for Two Turbines 224
8.4.3 Spline Model with Non-negativity Constraint 226
8.5 GAUSSIAN MARKOV RANDOM FIELD MODEL 230
8.6 CASE STUDY 232
8.6.1 Performance Comparison of Wake Models 232
8.6.2 Analysis of Turbine Wake E_ect 235
Part III Wind Turbine Reliability Management
Chapter 9 _ Overview of Wind Turbine Maintenance Optimization
247
9.1 COSTEFFECTIVE
MAINTENANCE 248
9.2 UNIQUE CHALLENGES IN TURBINE MAINTENANCE 249
9.3 COMMON PRACTICES 251
9.3.1 Failure Statistics-Based Approaches 251
9.3.2 Physical Load-Based Reliability Analysis 252
9.3.3 Condition-Based Monitoring or Maintenance 252
9.4 DYNAMIC TURBINE MAINTENANCE OPTIMIZATION 252
9.4.1 Partially Observable Markov Decision Process 254
9.4.2 Maintenance Optimization Solutions 256
9.4.3 Integration of Optimization and Simulation 260
9.5 DISCUSSION 263
Chapter 10 _ Extreme Load Analysis 267
10.1 FORMULATION FOR EXTREME LOAD ANALYSIS 267
10.2 GENERALIZED EXTREME VALUE DISTRIBUTIONS 270
10.3 BINNING METHOD FOR NONSTATIONARY GEV DISTRIBUTION
272
10.4 BAYESIAN SPLINEBASED
GEV MODEL 277
10.4.1 Conditional Load Model 277
10.4.2 Posterior Distribution of Parameters 280
10.4.3 Wind Characteristics Model 282
10.4.4 Posterior Predictive Distribution 284
10.5 ALGORITHMS USED IN BAYESIAN INFERENCE 285
10.6 CASE STUDY 285
10.6.1 Selection of Wind Speed Model 285
10.6.2 Pointwise Credible Intervals 285
10.6.3 Binning versus Spline Methods 289
10.6.4 Estimation of Extreme Load 293
10.6.5 Simulation of Extreme Load 294

Chapter 11 _ Computer SimulatorBased
Load Analysis 301
11.1 TURBINE LOAD COMPUTER SIMULATION 302
11.1.1 NREL Simulators 302
11.1.2 Deterministic and Stochastic Simulators 302
11.1.3 Simulator versus Emulator 304
11.2 IMPORTANCE SAMPLING 306
11.2.1 Random Sampling for Reliability Analysis 306
11.2.2 Importance Sampling Using Deterministic Simulator 307
11.3 IMPORTANCE SAMPLING USING STOCHASTIC SIMULATORS 311
11.3.1 Stochastic Importance Sampling Method 1 312
11.3.2 Stochastic Importance Sampling Method 2 314
11.3.3 Benchmark Importance Sampling Method 314
11.4 IMPLEMENTING STOCHASTIC IMPORTANCE SAMPLING 315
11.4.1 Modeling the Conditional POE 315
11.4.2 Sampling from Importance Sampling Densities 316
11.4.3 The Algorithm 317
11.5 CASE STUDY 319
11.5.1 Numerical Analysis 319
11.5.2 NREL Simulator Analysis 323

Chapter 12 _ Anomaly Detection and Fault Diagnosis 331
12.1 BASICS OF ANOMALY DETECTION 331
12.1.1 Types of Anomalies 331
12.1.2 Categories of Anomaly Detection Approaches 333
12.1.3 Performance Metrics and Decision Process 335
12.2 BASICS OF FAULT DIAGNOSIS 336
12.2.1 Tree-Based Diagnosis 337
12.2.2 Signature-Based Diagnosis 338
12.3 SIMILARITY METRICS 340
12.3.1 Norm and Distance Metrics 341
12.3.2 Inner Product and Angle-Based Metrics 342
12.3.3 Statistical Distance 344
12.3.4 Geodesic Distance 345
12.4 DISTANCEBASED
METHODS 347
12.4.1 Nearest Neighborhood-based Method 347
12.4.2 Local Outlier Factor 347
12.4.3 Connectivity-based Outlier Factor 348
12.4.4 Subspace Outlying Degree 349
12.5 GEODESIC DISTANCEBASED
METHOD 351
12.5.1 Graph Model of Data 351
12.5.2 MST Score 352
12.5.3 Determine Neighborhood Size 355
12.6 CASE STUDY 355
12.6.1 Benchmark Cases 355
12.6.2 Hydropower Plant Case 360
Bibliography 367
Index 387

It is US$10. To get this book send email: textileebooks@gmail.com

Share This

Related Posts

Previous
Next Post »