To address the insufficient accuracy and stability of multiple temperature control points in multi-objective temperature control, this study establishes a heating model based on model predictive control (MPC) and designs multiple MPC controllers. To collaboratively optimize the cost functions of multiple MPCs, a multi-objective grey wolf particle swarm optimization algorithm (MOGWO-PSO) is proposed. Simulation experiments on multi-objective temperature tracking demonstrate that the improved MOGWO-PSO algorithm demonstrates overall advantages in both control accuracy and computational efficiency when optimizing multiple MPC cost functions to track multiple temperature targets. Meanwhile, test problems have verified that MOGWO-PSO can provide high-quality and competitive Pareto solution sets. Furthermore, to address the time-varying characteristics of parameters in the MPC heating model, a backpropagation (BP) neural network is employed to fit and predict the model parameters. Compared with the MPC heating model identification method based on the least squares approach, the MPC heating model identified using a BP neural network can achieve lower tracking errors and reduce CPU computational overhead in multi-objective temperature tracking. This indicates that the MPC heating model fitted by the BP neural network can adaptively adjust to accommodate real-world conditions. In terms of engineering applications, this study provides an effective solution for complex multi-objective temperature tracking control.
Liu et al. (Fri,) studied this question.