List of Figures xvii -- List of Tables xxix -- About the Authors xxxiii -- Preface xxxv -- Abbreviations xxxvii -- Symbol Meaning xliii -- 1 Introduction 1 -- 1.1 Municipal Solid Waste Incineration (MSWI) Process and Optimal Control 1 -- 1.1.1 Description of MSWI Process 1 -- 1.1.2 Control Mode in Developed and Developing Countries 5 -- 1.1.3 Difficulties in the Implementation of Optimal Control and Application 9 -- 1.1.3.1 Description of Optimal Control for Complex Industrial Process 9 -- 1.1.3.2 Requirements of the MSWI Process in Academic Research and Industrial Applications 13 -- 1.1.3.3 Difficulties of AI Algorithm Research and Validation for MSWI Processes 14 -- 1.1.4 Development of Optimal Control Research Based on Artificial Intelligence (AI) 16 -- 1.2 AI-Based Modeling and Monitoring 17 -- 1.2.1 Numerical Simulation Modeling 17 -- 1.2.1.1 Brief Description of Numerical Simulation for MSWI Processes 17 -- 1.2.1.2 Based on Commercial Software 19 -- 1.2.1.3 Based on Self-developed Software 22 -- 1.2.1.4 Difficulties of Numerical Simulation Modeling 22 -- 1.2.1.5 Digital Twin (DT) Model Construction for MSWI 24 -- 1.2.2 Combustion Process Modeling 26 -- 1.2.2.1 Key Controlled Variables (CVs) Modeling 26 -- 1.2.2.2 Auxiliary Variables (AVs) Modeling 27 -- 1.2.3 Operational Indicators Modeling 28 -- 1.2.3.1 Environmental Indicators (EIs) Modeling 28 -- 1.2.3.2 Product Indicators (PIs) Modeling 29 -- 1.2.3.3 Economic Indicators Modeling 30 -- 1.2.4 Flame Status Monitoring 30 -- 1.2.5 Operational Abnormal Monitoring 31 -- 1.3 Control and Optimization Based on AI and DT 32 -- 1.3.1 Control in On-site 32 -- 1.3.1.1 Research of Automatic Combustion Control (ACC) System 32 -- 1.3.1.2 Research of Non-ACC System 33 -- 1.3.2 Control in Off-site 33 -- 1.3.2.1 Single Input Single Output (SISO) Control 33 -- 1.3.2.2 Multiple Input Multiple Output (MIMO) Control 34 -- 1.3.3 Optimization of Pollution Emission 35 -- 1.4 Hardware-in-Loop DT for MSWI Processes 36 -- 1.4.1 Brief Description of Simulation Platform for Industrial Process 36 -- 1.4.2 Simulation Platform in Terms of Real/Virtual Perspective 37 -- 1.4.2.1 “Real–Real” Simulation Platform 37 -- 1.4.2.2 “Real–Virtual” Simulation Platform 38 -- 1.4.2.3 “Virtual–Real” Simulation Platform 39 -- 1.4.2.4 “Virtual–Virtual” Simulation Platform 40 -- 1.4.3 Difficulties of Simulation Platform for MSWI Process 41 -- 1.5 Book’s Structure 42 -- Part I 42 -- Part II 45 -- Part III 47 -- References 48 -- Part I Modeling and Monitoring Based on AI 67 -- 2 Numerical Simulation and Modeling Analysis on Whole Industrial Process by Coupling Multiple Software 69 -- 2.1 Simulated Plant and Simulation Modeling 69 -- 2.1.1 Simulated MSWI Plant 69 -- 2.1.1.1 Process Flow Description of the Simulated MSWI Plant 70 -- 2.1.1.2 Slag, Fly Ash, and Leachate Treatment of the Simulated MSWI Plant 71 -- 2.1.2 Simulation and Modeling Requirements 72 -- 2.1.3 MSWI Process Description for Numerical Simulation 74 -- 2.1.3.1 Mechanism Oriented Process Description 74 -- 2.1.3.2 Mechanism Model Description 77 -- 2.1.4 Numerical Simulation and Modeling Analysis Framework 91 -- 2.2 Modeling Strategy with Virtual Data-driven 92 -- 2.3 Modeling Implementation for Whole Process 94 -- 2.3.1 Multi-Software-Coupled Whole-Process Numerical Simulation Module Under Benchmark Conditions 94 -- 2.3.1.1 Solid-Phase Combustion Simulation on the Grate Based on FLIC 94 -- 2.3.1.2 Gas-Phase Combustion in the Furnace Based on Fluent 96 -- 2.3.1.3 Non-grate Solid-Phase Combustion Simulation in MSWI Process Based on Aspen Plus 97 -- 2.3.2 Simulation Mechanism Data Acquisition Module Under Multiple Operating Conditions 100 -- 2.3.3 Exhaust Emission Model Construction Module Based on Mimo-lrdt 101 -- 2.3.4 Exhaust Emissions Analysis Module Based on Single/Dual Factors 103 -- 2.4 Numerical Simulation and Modeling Results 103 -- 2.4.1 Benchmark Condition Simulation Results 103 -- 2.4.1.1 Data Description 103 -- 2.4.1.2 Results of Solid MSW Combustion on the Grate 104 -- 2.4.1.3 Results of Gas-Phase Combustion in the Furnace 106 -- 2.4.1.4 Results of Non-grate Solid-Phase Combustion in MSWI Process 107 -- 2.4.1.5 Comparison with the Actual Data 108 -- 2.4.2 Results and Analysis of Multiple Operating Conditions 109 -- 2.4.2.1 Typical Non-benchmark Condition Description 109 -- 2.4.2.2 Typical Solid-Phase MSW Combustion Results Based on FLIC 109 -- 2.4.2.3 Typical Gas-Phase Combustion Results Based on Fluent 110 -- 2.4.2.4 Typical Non-grate Solid-Phase Combustion Results Based on Aspen Plus 116 -- 2.4.2.5 Multiple Operating Conditions Results Based on Orthogonal Experimental Design 118 -- 2.4.3 Construction Results of Exhaust Emission Model Based on Mimo-lrdt 118 -- 2.4.4 Exhaust Emissions Cause and Effect Analysis Based on Single/Dual Factor 120 -- 2.4.5 Discussion 124 -- 2.5 Conclusion 124 -- References 125 -- 3 Conventional Pollutant Deep Modeling Using Virtual Data and Real Data Hybrid-Driven 129 -- 3.1 Virtual–Real Data-Driven Conventional Pollutant Modeling 129 -- 3.1.1 Motivation 129 -- 3.1.2 CO Description 130 -- 3.1.3 CO Prediction Strategy 131 -- 3.2 Real Data Hybrid-Driven Modeling Implementation 133 -- 3.2.1 Offline Training Verification Phase 133 -- 3.2.1.1 Multi-condition Virtual Mechanism Data Generation Module 133 -- 3.2.1.2 Mechanism Mapping Model Module Based on LRDT 135 -- 3.2.1.3 Real Data-Driven Model Module Based on LSTM 137 -- 3.2.1.4 Heterogeneous Ensemble Module 140 -- 3.2.2 Online Testing Verification Phase 142 -- 3.2.2.1 Mechanism Mapping Model Module Based on LRDT 142 -- 3.2.2.2 Real Data-Driven Model Module Based on LSTM 142 -- 3.2.2.3 Heterogeneous Ensemble Module 142 -- 3.3 Deep Modeling Results and Discussion 142 -- 3.3.1 Data Description 142 -- 3.3.2 Evaluation Indexes 143 -- 3.3.3 Modeling Results and Discussion 143 -- 3.3.3.1 Offline Training Verification Phase Results 143 -- 3.3.3.2 Online Testing Verification Phase Results 152 -- 3.3.4 Discussion on Model Hyperparameter 152 -- 3.4 Conclusion 157 -- References 160 -- 4 Trace Pollutant Modeling Using the Selective Ensemble Algorithm 163 -- 4.1 Selective Ensemble Modeling Strategy 163 -- 4.1.1 Motivation 163 -- 4.1.2 DXN Generation Description of MSWI Process 164 -- 4.1.3 DXN Soft Sensing Strategy 166 -- 4.2 Trace Pollutant Modeling Implementation 168 -- 4.2.1 Ensembled Submodel Building Module 168 -- 4.2.1.1 BT Candidate Submodel Construction and Prediction Submodule 168 -- 4.2.1.2 Candidate Submodel Bayesian Information Acquisition Submodule 170 -- 4.2.1.3 Ensembled Submodel Selection Submodule 174 -- 4.2.2 Ensembled Submodel Weighted Fusion Module 175 -- 4.3 Data-Driven Ensemble Modeling Results and Discussion 176 -- 4.3.1 Data Description 176 -- 4.3.2 Evaluation Indicators 176 -- 4.3.3 Benchmark Data Verification 180 -- 4.3.3.1 Ensembled Submodel Construction Results 181 -- 4.3.3.2 Weighted Prediction Results of Ensembled Submodel 183 -- 4.3.3.3 Comparison of Experimental Results 184 -- 4.3.4 Industrial Data Validation 186 -- 4.3.4.1 Ensembled Submodel Construction Results 187 -- 4.3.4.2 Weighted Prediction Results of the Ensembled Submodel 189 -- 4.3.4.3 Comparison of Experimental Results 189 -- 4.3.5 Hyperparameter Analysis for Different Datasets 196 -- 4.4 Conclusion 201 -- References 201 -- 5 Trace Pollutant Modeling Based on Semi-supervised Random Forest Optimization 205 -- 5.1 Data-Driven Trace Pollutant Semi-supervised Random Forest Optimization Modeling Strategy 205 -- 5.1.1 Semi-supervised Random Forest Optimization Modeling 205 -- 5.1.2 Semi-supervised Regression Modeling and Optimization Research 206 -- 5.1.3 Dioxin (DXN) Semi-supervised Soft Sensing Strategy 209 -- 5.2 Data-Driven Trace Pollutant Modeling Implementation 212 -- 5.2.1 Parameter Coding Design Module for Hybrid Optimization 212 -- 5.2.2 Initialization and Decoding Module of the Hybrid Parameter 213 -- 5.2.3 Fitness Evaluation Module for Multi-objective 215 -- 5.2.3.1 Train the Random Forest (RF) Model Based on the Labeled Samples 215 -- 5.2.3.2 Get the Pseudo-labeled Samples 217 -- 5.2.3.3 Select the Pseudo-labeled Samples 217 -- 5.2.3.4 Get the Mixed Sample Set 218 -- 5.2.3.5 Train the RF Model Based on the Mixed Sample set 218 -- 5.2.3.6 Evaluate the Fitness and Optimal Archive 218 -- 5.2.4 Iterative Optimization and Optimal Solution

Acquisition Module 218 -- 5.2.4.1 Optimization Termination and Update 219 -- 5.2.4.2 Optimal Solution Acquisition Module Based on the Pareto Solution Set 220 -- 5.2.5 RF Model Construction Module Based on the Mixed Sample Set 220 -- 5.3 Experimental Verification 221 -- 5.3.1 Benchmark Dataset 221 -- 5.3.1.1 Data Description 221 -- 5.3.1.2 Experimental Results 221 -- 5.3.1.3 Comparison with Other Methods 225 -- 5.3.2 DXN Dataset 227 -- 5.3.2.1 Data Description 227 -- 5.3.2.2 Experimental Results 227 -- 5.3.2.3 Comparison with Other Methods 231 -- 5.3.3 Parameter Sensitivity Analysis 231 -- 5.4 Conclusion 238 -- References 239 -- 6 Combustion State Identificatio ...