Data-Driven Dynamic Multiobjective Optimization With Response to Stochastic Changes for Municipal Solid Waste Incineration Process

Simultaneous optimization of both pollutant emissions and power generation holds significant importance for the municipal solid waste incineration (MSWI) process. Nevertheless, the inherent uncertainty and time-varying dynamics of the MSWI process induce the optimization problem into a complicated dynamic multiobjective optimization problem (DMOP) with stochastic changes, challenging the acquisition of optimal set-points. To address this problem, a novel data-driven dynamic multiobjective optimization scheme is proposed with response to stochastic changes. First, surrogate models for performance indices are established using fuzzy neural networks, assisted with an online learning method developed by an adaptive Levenberg-Marquardt algorithm to ensure model applicability. Second, an adaptive multiobjective triple competitive swarm optimization algorithm is proposed specifically designed for the DMOP of the MSWI process, incorporating comprehensive triple competition and a multi-mode learning strategy to effectively balance global exploration and local exploitation. Furthermore, in response to the stochastic changes of optimization environment, a stochastic dynamic response strategy is designed to establish dynamic mapping relationships for Pareto optimal solutions, thereby enhancing optimization efficiency in evolving environments. Finally, experimental validation using operational data from a real-world MSWI plant demonstrates the effectiveness of the proposed method, with comparative studies confirming its superior performance.