基于改进PSO算法的过热汽温神经网络预测控制

Neural network predictive control for superheated steam temperature based on modified particle swarm optimization

将改进粒子群优化算法(MPSO)融合到神经网络预测控制中,提出了基于MPSO-RBF混合优化策略的模型预测器,以及基于MPSO算法的非线性优化控制器.针对过热汽温的控制。构造了基于神经网络预测控制的串级控制系统,并就该系统在实现时所涉及到的预测模型、滚动优化算法、反馈校正、仿真参数设置问题等进行了分析,给出了MPSO算法的粒子编码、操作设计和混合优化算法步骤.对某超临界600 MW直流锅炉高温过热器的过热汽温控制,进行了仿真试验,结果表明该方法具有良好的性能指标和应用前景.

Combining modified particle swarm optimization （MPSO） with neural network predictive control （NNPC）, we propose a model-prediction controller, based-on modified particle swarm optimization （MPSO） and radial basis function （RBF） hybrid optimization strategy （MPSO-RBF）, and a nonlinear optimization controller, based-on MPSO. For the super- heated steam temperature control, we construct a cascade control system based on the neural network predictive control, and analyze all related problems, including the predictive model, the rolling optimizing algorithm, the feedback adjusting and the simulation-parameter setting. We also present the particle encoded format of MPSO, operating design method, and steps in hybrid optimization algorithm. Simulation experiments of the superheated steam temperature control were done in a super-critical-600 MW direct-current boiler, demonstrating the validity, the superior performance and the application prospects.