论文部分内容阅读
运用基于数据协方差基准的控制性能评估方法,并结合协方差矩阵的递推更新,本文提出了一种控制器性能的在线监控策略。首先给出一种协方差矩阵的递推更新算法,通过实时输出监控阶段数据相对于基准阶段数据的广义特征值及其构成的系统总体性能指标,在线监控系统性能的变化趋势。进一步使用统计推断方法得出每个时刻的特征值所对应的置信区间,得到优/劣子空间的性能指标实时变化趋势,从而判定控制器的性能变化。最后,将本方法应用在重油分馏塔预测控制的实时性能监控中,验证了本方法的有效性。
By using the control performance evaluation method based on data covariance benchmark and recursive update of covariance matrix, an online monitoring strategy for controller performance is proposed in this paper. Firstly, a recursive update algorithm of covariance matrix is presented. By monitoring the generalized eigenvalue of the data in the monitoring stage and the data of the reference stage in real time, the trend of performance of the system is monitored online. The statistical confidence method is further used to derive the confidence interval corresponding to the eigenvalue of each moment to get the real-time trend of the performance index of the superior / inferior subspace, so as to determine the performance change of the controller. Finally, the method is applied to real-time performance monitoring of heavy oil fractionation tower predictive control, which verifies the effectiveness of the method.