Zhao / Wang / Sheng Data-Driven Prediction for Industrial Processes and Their Applications

1. Auflage 2018
ISBN: 978-3-319-94051-9
Verlag: Springer International Publishing
Format: PDF
Kopierschutz: 1 - PDF Watermark

E-Book, Englisch, 453 Seiten

Reihe: Information Fusion and Data Science

This book presents modeling methods and algorithms for data-driven prediction and forecasting of practical industrial process by employing machine learning and statistics methodologies. Related case studies, especially on energy systems in the steel industry are also addressed and analyzed. The case studies in this volume are entirely rooted in both classical data-driven prediction problems and industrial practice requirements. Detailed figures and tables demonstrate the effectiveness and generalization of the methods addressed, and the classifications of the addressed prediction problems come from practical industrial demands, rather than from academic categories. As such, readers will learn the corresponding approaches for resolving their industrial technical problems. Although the contents of this book and its case studies come from the steel industry, these techniques can be also used for other process industries. This book appeals to students, researchers, and professionals within the machine learning and data analysis and mining communities.

Jun Zhao is currently a Professor with the School of Control Science and Engineering, Dalian University of Technology, China. Chunyang Sheng is currently a lecturer with the School of Electrical Engineering and Automation, Shandong University of Science and Technology, China.  Wei Wang is currently a Professor with the School of Control Science and Engineering, Dalian University of Technology, China.

Autoren/Hrsg.

Zhao, Jun

Wang, Wei

Sheng, Chunyang

Inhaltsverzeichnis

1;Preface;6
2;Audience and Goal of This Book;8
3;Acknowledgements;9
4;Contents;10
5;Chapter 1: Introduction;16
5.1;1.1 Why Prediction Is Required for Industrial Process;17
5.2;1.2 Category of Data-Based Industrial Process Prediction;18
5.2.1;1.2.1 Data Feature-Based Prediction;18
5.2.2;1.2.2 Time Scale-Based Prediction;18
5.2.3;1.2.3 Prediction Reliability-Based Prediction;19
5.3;1.3 Commonly Used Techniques for Industrial Prediction;20
5.3.1;1.3.1 Time Series Prediction Methods;20
5.3.2;1.3.2 Factor-Based Prediction Methods;21
5.3.3;1.3.3 Methods for PIs Construction;22
5.3.4;1.3.4 Long-Term Prediction Intervals Methods;23
5.4;1.4 Summary;24
5.5;References;25
6;Chapter 2: Data Preprocessing Techniques;27
6.1;2.1 Introduction;27
6.2;2.2 Anomaly Data Detection;29
6.2.1;2.2.1 K-Nearest-Neighbor;29
6.2.2;2.2.2 Fuzzy C Means;30
6.2.3;2.2.3 Adaptive Fuzzy C Means;32
6.2.4;2.2.4 Trend Anomaly Detection Based on AFCM and DTW;33
6.2.5;2.2.5 Deviants Detection Based on KNN-AFCM;36
6.2.6;2.2.6 Case Study;40
6.3;2.3 Data Imputation;43
6.3.1;2.3.1 Data-Missing Mechanism;43
6.3.2;2.3.2 Regression Filling Method;44
6.3.3;2.3.3 Expectation Maximum;44
6.3.4;2.3.4 Varied Window Similarity Measure;45
6.3.5;2.3.5 Segmented Shape-Representation Based Method;46
6.3.5.1;Key-Sliding-Window for Sequence Segmentation;46
6.3.5.2;Representation of Sequence Segmentation;48
6.3.5.3;Procedure of Data Imputation Based on Segmented Shape-Representation;50
6.3.6;2.3.6 Non-equal-Length Granules Correlation;51
6.3.6.1;Calculation for NGCC;53
6.3.6.2;NGCC-Based Correlation Analysis;55
6.3.6.3;Correlation-Based Data Imputation;55
6.3.7;2.3.7 Case Study;56
6.4;2.4 Data De-noising Techniques;60
6.4.1;2.4.1 Empirical Mode Decomposition;60
6.4.2;2.4.2 Case Study;61
6.5;2.5 Discussion;64
6.6;References;65
7;Chapter 3: Industrial Time Series Prediction;67
7.1;3.1 Introduction;67
7.2;3.2 Phase Space Reconstruction;69
7.2.1;3.2.1 Determination of Embedding Dimensionality;69
7.2.1.1;False Nearest-Neighbor Method (FNN);70
7.2.1.2;Cao Method;70
7.2.2;3.2.2 Determination of Delay Time;71
7.2.2.1;Autocorrelation Function Method;71
7.2.2.2;Mutual Information Method;72
7.2.3;3.2.3 Simultaneous Determination of Embedding Dimensionality and Delay Time;72
7.3;3.3 Linear Models for Regression;74
7.3.1;3.3.1 Basic Linear Regression;74
7.3.2;3.3.2 Probabilistic Linear Regression;76
7.4;3.4 Gaussian Process-Based Prediction;78
7.4.1;3.4.1 Kernel-Based Regression;78
7.4.2;3.4.2 Gaussian Process for Prediction;80
7.4.3;3.4.3 Gaussian Process-Based ESN;81
7.4.4;3.4.4 Case Study;85
7.5;3.5 Artificial Neural Networks-Based Prediction;89
7.5.1;3.5.1 RNNs for Regression;89
7.5.2;3.5.2 ESN for Regression;93
7.5.3;3.5.3 SVD-Based ESN for Industrial Prediction;95
7.5.4;3.5.4 ESNs with Leaky Integrator Neurons;96
7.5.5;3.5.5 Dual Estimation-Based ESN;98
7.5.6;3.5.6 Case Study;101
7.5.6.1;Extended Kalman-Filter-Based Elman Network;102
7.5.6.2;SVD-Based ESN for Industrial Prediction;104
7.5.6.3;ESN with Leaky Integrator Neurons;106
7.5.6.4;Dual Estimation-Based ESN;110
7.6;3.6 Support Vector Machine-Based Prediction;113
7.6.1;3.6.1 Basic Concept of SVM;113
7.6.2;3.6.2 SVMs for Regression;115
7.6.3;3.6.3 Least Square Support Vector Machine;118
7.6.4;3.6.4 Sample Selection-Based Reduced SVM;119
7.6.5;3.6.5 Bayesian Treatment for LSSVM Regression;124
7.6.5.1;Probabilistic Interpretation of LSSVM Regressor (Level 1): Predictive Mean and Error Bars;124
7.6.5.1.1;Calculation of Maximum Posterior;124
7.6.5.1.2;Moderated Output of LSSVM Regressor;126
7.6.5.2;Inference of Hyper-Parameters (Level 2);127
7.6.5.3;Inference of Kernel Parameters and Model Comparison;128
7.6.6;3.6.6 Case Study;128
7.7;3.7 Discussion;132
7.8;References;132
8;Chapter 4: Factor-Based Industrial Process Prediction;134
8.1;4.1 Introduction;134
8.2;4.2 Methods of Determining Factors;135
8.2.1;4.2.1 Gray Correlation;136
8.2.1.1;Case Study;137
8.2.2;4.2.2 Convolution-Based Methods;139
8.2.2.1;Case Study;140
8.2.3;4.2.3 Bayesian Technique of Automatic Relevance;141
8.3;4.3 Factor-Based Single-Output Model;144
8.3.1;4.3.1 Neural Networks-Based Model;144
8.3.2;4.3.2 T-S Fuzzy Model-Based Prediction;147
8.3.3;4.3.3 Multi-Kernels Least Square Support Vector Machine;148
8.3.4;4.3.4 Case Study;153
8.4;4.4 Factor-Based Multi-Output Model;156
8.4.1;4.4.1 Multi-Output Least Square Support Vector Machine;156
8.4.2;4.4.2 Case Study;159
8.5;4.5 Discussion;166
8.6;References;168
9;Chapter 5: Industrial Prediction Intervals with Data Uncertainty;171
9.1;5.1 Introduction;171
9.2;5.2 Commonly Used Techniques for Prediction Intervals;173
9.2.1;5.2.1 Delta Method;173
9.2.2;5.2.2 Mean and Variance-Based Estimation;175
9.2.3;5.2.3 Bayesian Method;177
9.2.4;5.2.4 Bootstrap Technique;179
9.3;5.3 Neural Networks-Based PIs Construction for Time Series;181
9.3.1;5.3.1 ESNs Ensemble-Based Prediction Model;181
9.3.2;5.3.2 Bayesian Estimation of the Uncertainties;182
9.3.3;5.3.3 Model Selection and Structural Optimization;185
9.3.4;5.3.4 Theoretical Analysis of the Prediction Performance;187
9.3.5;5.3.5 Case Study;188
9.3.5.1;Multiple Superimposed Oscillator;189
9.3.5.2;Application on Prediction for Generation Flow of BFG;193
9.4;5.4 Non-iterative NNs for PIs Construction;195
9.4.1;5.4.1 A Non-iterative Prediction Mode;196
9.4.2;5.4.2 Interval-Weighted ESN and Its Iterative Prediction;197
9.4.3;5.4.3 Gamma Test-Based Model Selection;199
9.4.4;5.4.4 Case Study;202
9.4.4.1;Two Benchmark Prediction Problems;202
9.4.4.2;Interval Prediction on the By-Product Gas System in Steel Industry;204
9.5;5.5 Gaussian Kernel-Based Causality PIs Construction;207
9.5.1;5.5.1 Mixed Gaussian Kernel for Regression;208
9.5.2;5.5.2 Mixed Gaussian Kernel for PIs Construction;209
9.5.3;5.5.3 Estimation of Effective Noise-Based Hyper-Parameters;211
9.5.4;5.5.4 Case Study;212
9.6;5.6 Prediction Intervals Construction with Noisy Inputs;214
9.6.1;5.6.1 Bayesian Estimation of the Output Uncertainty;215
9.6.2;5.6.2 Estimation of the External Input Uncertainties;217
9.6.3;5.6.3 Estimation of the Output Feedback Uncertainties;219
9.6.4;5.6.4 Estimation of the Total Uncertainties and PIs Construction;221
9.6.5;5.6.5 Case Study;223
9.7;5.7 Prediction Intervals with Missing Input;225
9.7.1;5.7.1 Kernel-Based DBN Prediction Model;225
9.7.2;5.7.2 Approximate Inference and PI Construction;226
9.7.3;5.7.3 Learning a Kernel-Based DBN;228
9.7.4;5.7.4 Case Study;229
9.8;5.8 Discussion;231
9.9;References;232
10;Chapter 6: Granular Computing-Based Long-Term Prediction Intervals;235
10.1;6.1 Introduction;235
10.2;6.2 Techniques of Granularity Partition;236
10.2.1;6.2.1 Partition of Equal Length;237
10.2.2;6.2.2 Partition of Unequal Length;238
10.2.2.1;Standard Granule Selection;241
10.2.2.2;Time Warping Normalization;241
10.3;6.3 Long-Term Prediction Model;243
10.3.1;6.3.1 Granular Model for Time Series Prediction;243
10.3.1.1;Case Study;245
10.3.1.1.1;BFG System Experiments;245
10.3.1.1.2;COG System Experiments;248
10.3.1.1.3;LDG System Experiments;250
10.3.2;6.3.2 Granular Model for Factor-Based Prediction;253
10.3.2.1;Case Study;259
10.3.2.1.1;Comparing with the Single-Output Model;259
10.3.2.1.2;Comparing with the Multi-output Model (Iteration Mechanism,);261
10.4;6.4 Granular-Based Prediction Intervals;263
10.4.1;6.4.1 Initial PIs Construction;263
10.4.2;6.4.2 PIs Optimization;264
10.4.3;6.4.3 Computing Procedure;265
10.4.4;6.4.4 Case Study;267
10.4.4.1;BFG Consumption Flow of Hot Blast Stove;267
10.4.4.2;BFG Generation Flows;269
10.5;6.5 Multi-dimension Granular-Based Long-Term Prediction Intervals;272
10.5.1;6.5.1 Case Study;273
10.6;6.6 Discussion;276
10.7;References;279
11;Chapter 7: Parameter Estimation and Optimization;280
11.1;7.1 Introduction;280
11.2;7.2 Gradient-Based Methods;281
11.2.1;7.2.1 Gradient Descent;282
11.2.1.1;Batch Gradient Descent;284
11.2.1.2;Stochastic Gradient Descent;284
11.2.1.3;Mini-batch Gradient Descent;285
11.2.2;7.2.2 Newton Method;285
11.2.3;7.2.3 Quasi-Newton Method;288
11.2.3.1;Broyden-Fletcher-Goldfarb-Shanno (BFGS) Algorithm;289
11.2.3.2;L-BFGS Method;291
11.2.4;7.2.4 Conjugate Gradient Method;293
11.2.4.1;Nonlinear Conjugate Gradient Method;295
11.2.5;7.2.5 Illustration: A Gradient Grid Search Algorithm;297
11.3;7.3 Intelligent Optimization Algorithms;299
11.3.1;7.3.1 Genetic Algorithm;300
11.3.1.1;Selection Operator;301
11.3.1.2;Crossover Operator;302
11.3.1.3;Mutation Operator;302
11.3.2;7.3.2 Differential Evolution Algorithm;304
11.3.2.1;Initialization of the Population;305
11.3.2.2;Mutation with Differential Operators;305
11.3.2.3;Crossover;306
11.3.2.4;Selection;306
11.3.3;7.3.3 Particle Swarm Optimization Algorithm;308
11.3.4;7.3.4 Simulated Annealing Algorithm;311
11.4;7.4 Nonlinear Kalman-Filter Estimation;314
11.4.1;7.4.1 Extended Kalman-Filter;315
11.4.2;7.4.2 Unscented Kalman-Filter;317
11.4.3;7.4.3 Cubature Kalman-Filter;319
11.4.4;7.4.4 Nonlinear Kalman-Filters-Based Dual Estimation;321
11.4.5;7.4.5 Dual Estimation of Linear/Nonlinear Kalman-Filter;323
11.4.6;7.4.6 Case Study;325
11.5;7.5 Probabilistic Methods;328
11.5.1;7.5.1 Maximum Likelihood Method;328
11.5.1.1;MLE for Linear Regression Model;330
11.5.2;7.5.2 Bayesian Method;332
11.5.3;7.5.3 Variational Inference;334
11.5.4;7.5.4 Variational Relevance Vector Machine Based on Automatic Relevance Determination Kernel Functions;340
11.5.4.1;Preliminaries on Variational RVM (VRVM);340
11.5.4.2;Model Specification for the VRVM-ARDK;342
11.5.4.3;Variational Inference for RVM-ARDK;343
11.5.4.4;Predictive Distribution and Model Training;347
11.5.5;7.5.5 Case Study;348
11.6;7.6 Parameter Optimization for LS-SVM Based on Noise Estimation;350
11.6.1;7.6.1 Hyper-parameters Optimization Based on the CG Method;351
11.6.2;7.6.2 Case Study;353
11.6.2.1;The Sinc Function;353
11.6.2.2;Industrial Application: Prediction Modeling for Industrial Gas Flow;354
11.7;7.7 Discussion;358
11.8;References;359
12;Chapter 8: Parallel Computing Considerations;362
12.1;8.1 Introduction;362
12.2;8.2 CUDA-Based Parallel Acceleration;364
12.2.1;8.2.1 What´s CUDA?;364
12.2.2;8.2.2 Computing Architecture of CUDA;365
12.2.3;8.2.3 CUDA Libraries;366
12.3;8.3 Hadoop-Based Distributed Computation;367
12.3.1;8.3.1 What´s Hadoop?;367
12.3.2;8.3.2 Computing Architecture of Hadoop;368
12.3.3;8.3.3 MapReduce;370
12.4;8.4 GPU Acceleration for Training EKF-Based Elman Networks;372
12.4.1;8.4.1 Case Study;374
12.4.1.1;BFG System Prediction;374
12.4.1.2;COG System Prediction;376
12.5;8.5 Online Parameter Optimization-Based Prediction by GPU Acceleration;378
12.5.1;8.5.1 Initialization and Sub-swarm Separation;379
12.5.1.1;Update of Velocity and Position;380
12.5.1.2;Parameter Adaptation and Process Communication;380
12.5.2;8.5.2 Case Study;381
12.5.2.1;Parallelization for Parameter Validation;381
12.5.2.2;Parallelization for Parameter Selection;383
12.5.2.3;Online Prediction for LDG System;384
12.6;8.6 Parallelized EKF Based on MapReduce Framework;387
12.6.1;8.6.1 Case Study;391
12.7;8.7 Discussion;393
12.8;References;394
13;Chapter 9: Data-Based Prediction for Energy Scheduling of Steel Industry;395
13.1;9.1 Introduction;395
13.2;9.2 A Prediction and Adjustment Method for By-product Gas Scheduling;396
13.2.1;9.2.1 Holder Level Prediction Models and Adjustment Method;398
13.2.1.1;Holder Level Prediction Model;399
13.2.1.2;Optimization of Holder Level Prediction Model;402
13.2.1.3;Optimal Adjustment Quantity Determination;405
13.2.2;9.2.2 Case Study;406
13.3;9.3 Interval Predictive Optimization for By-product Gas System in Steel Industry;413
13.3.1;9.3.1 Mixed Gaussian Kernel-Based PIs Construction;416
13.3.1.1;Mixed Gaussian Kernel-Based Method for Regression;416
13.3.1.2;Construction of Mixed Gaussian Kernel-Based PIs;417
13.3.2;9.3.2 A Novel Predictive Optimization Method;419
13.3.2.1;Prediction for Optimized Objectives;420
13.3.2.2;Rolling Optimization for Adjusting Solution;422
13.3.3;9.3.3 Case Study;424
13.4;9.4 A Two-Stage Method for Predicting and Scheduling Energy in an Oxygen/Nitrogen System of the Steel industry;429
13.4.1;9.4.1 GrC-Based Long-Term Prediction Model;431
13.4.1.1;Data Granulation;431
13.4.1.2;Long-Term Prediction Modeling;432
13.4.2;9.4.2 MILP-Based Scheduling Model for Oxygen/Nitrogen System;434
13.4.2.1;Objective;434
13.4.2.2;Constraints;434
13.4.3;9.4.3 Case study;437
13.4.3.1;Long-Term Prediction on Oxygen/Nitrogen Requirements;437
13.4.3.2;Energy Scheduling and Optimization;439
13.4.3.2.1;Oxygen Shortage/Nitrogen Shortage;440
13.4.3.2.2;Oxygen Surplus/Nitrogen Shortage;442
13.5;9.5 Discussion;442
13.6;References;444
14;Index;447