基于模型预测控制及智能寻优的水泥粉磨优化控制

本文提出一种基于运行状态软测量和成本软约束的多变量模型预测控制(MPC)。MPC控制与传统的专家经验控制和模糊控制相比,通过模型对系统工艺参数的预测,不断地学习更新模型,更符合水泥粉磨大时延、多工况的特性。应用中通过对水泥粉磨装置的阶跃响应实验,建立多变量预测控制模型,解决水泥粉磨系统生产过程的不确定性。在此基础上,通过增量学习和机器学习找到最优运行参数,使水泥粉磨的MPC控制一直保持在最优工况。

Optimization control of cement grinding based on model predictive control and intelligent optimization

This paper proposes a multivariate model predictive control (MPC) based on operational state soft measurement and cost soft constraints. Compared with traditional expert experience control and fuzzy control, MPC control could continuously learns and updates the model through the prediction of system process parameters, which is more in line with the characteristics of cement grinding large delay and multiple working conditions. In the application, the multi-variable predictive control model is established through the step response experiment of the cement grinding device to solve the uncertainty of the cement grinding system production process. On this basis, the optimal operating parameters are found through incremental learning and machine learning, so that the MPC control of cement grinding is always maintained at the optimal working condition.