基于累积平方误差–总平方波动指标的模型预测控制器性能评价及自愈

Integral squared error–total squared variation index based model predictive controller performance assessment and self-healing

中文关键词  模型预测控制器  控制器性能评价  控制器参数自愈  累积平方误差–总平方波动指标  鲁棒性

英文关键词  model predictive controller  performance assessment  self-healing  integral squared error (ISE)–total squared variation (TSV) index  robustness

中文摘要

      本文针对模型预测控制器实际投运中遇到性能下降问题, 提出了一种基于累积平方误差(ISE)–总平方波动 (TSV)指标的模型预测控制器性能评价及自愈方法. 先基于累积平方误差(ISE)和总平方波动(TSV)指标对模型预测 控制器进行实时性能评价, 再根据无限时域模型预测控制器(MPC)的逆特性, 基于ISE–TSV指标的分析, 提出了一 种MPC控制器的鲁棒自愈方法. 在二级倒立摆的模型预测控制仿真与实验结果证明了所提自愈方法的可行性及有 效性.

英文摘要

      Model predictive control (MPC) as an advanced process control method are widely applied to many industrial processes. The performance of the MPC controller may gradually decrease due to various reasons. To solve the problem that the performance of the model predictive controller decreasing in real applications, a performance assessment and selfhealing method for the model predictive controller based on the integral squared error (ISE) and total squared variation (TSV) index is proposed in this work. First, the integral squared error (ISE) and total squared variation (TSV) index are proposed to evaluate the performance of model predictive controller and then the ISE–TSV indicator is converted into a linear matrix inequality form according to the process constraints. After that, a self-healing method for MPC controller based on the time domain MPC inverse property is derived to resume the MPC controller performance and improve its on-line robustness. Considering that the controlled object contains uncertain terms and the range of model mismatch is in an interval, the parameters of the MPC controller are updated with the proposed self-healing algorithm to make the optimized controller parameters have stronger robustness. The application results with the proposed method in the linear double inverted pendulum model predictive control experiments have verified its feasibility and effectiveness.