模型预测控制约束自适应研究

Research on constraint adaptation of model predictive control

针对目前模型预测控制(Model predictive control,MPC)算法处理约束问题普遍以解决定值约束为主,面对大幅干扰及设定值改变等工况,由于定值约束的局限性,系统动态性能、鲁棒性受到影响,系统被控变量波动较大,且运算速度较低。以单层动态矩阵控制(Dynamic matrix control,DMC)算法为核心,提出一种约束自适应控制策略。简化约束在算法预测时域的计算长度,提升算法执行效率;针对控制变量约束引入松弛因子实现约束动态调整,降低动态规划数值求解时间;利用模型结构及反馈原理补偿模型误差。通过单变量系统仿真验证,结果表明:约束自适应算法可自动计算初始约束及过程约束,并对约束动态调整,相对于固定约束系统具有更好的快速性和鲁棒性。

At present,model predictive control(MPC)algorithm generally focuses on solving the constraint problem.In the face of large disturbance and set value change,due to the limitations of set value constraint,the dynamic performance and robustness of the system are affected,the controlled variables of the system fluctuate greatly and the operation speed is low.Taking the single-layer dynamic matrix control(DMC)algorithm as the core,a constrained adaptive control strategy is proposed.The calculation length of constraints in the prediction time domain of the algorithm is simplified to improve the execution efficiency of the algorithm.The relaxation factor is introduced for the control variable constraints to realize the dynamic adjustment of constraints and reduce the numerical solution time of dynamic programming.The model structure and feedback principle are used to compensate the model error.Through the simulation of multivariable systems,the results show that the constraint adaptive algorithm can automatically calculate the initial constraints and process constraints,and dynamically adjust the constraints.Compared with the fixed constraint system,it has better rapidity and robustness.