Abstract: In the view of the problem that the current performance evaluation method of MPC control loop cannot accurately locate the source causing performance degradation. A performance diagnosis method for control loop is proposed based on weighted L₂-Hausdorff subspace distance. The method characterizes the loop performance features under different poor operating conditions by using deteriorating performance subspace, and monitors the control loop performance in real time through the model prediction residual based closed-loop potential index. When the performance degradation is detected, weighted L₂-Hausdorff subspace distance is constructed to measure the similarity between the current loop performance pattern and the known ones, and the source causing performance degradation is then positioned by distance clustering. Finally, the simulation experiment on the continuous stirred tank heater(CSTH) validates the effectiveness and reliability of the proposed method.

Key words: model predictive control, performance diagnosis, model prediction residual, covariance, weighted subspace distance

摘要： 针对当前MPC控制回路性能评价方法无法准确定位性能下降源的问题,提出一种基于加权L₂-Hausdorff子空间距离的控制回路性能诊断方法.该方法引用恶化性能子空间表征各个性能恶化工况下的回路特征,通过基于模型预测残差的闭环潜能指标实时监测控制回路性能.当发现性能下降时,构造加权L₂-Hausdorff子空间距离来度量当前回路性能模式与已知性能模式的相似度,通过距离聚类来定位回路性能下降的恶化源.最后,通过连续搅拌加热器(continuous stirred tank heater, CSTH)上的仿真实验,验证了所提方法的有效性及可靠性.

关键词: 模型预测控制, 性能诊断, 模型预测残差, 协方差, 加权子空间距离