论文部分内容阅读
针对实际工业过程中希望只利用过程的日常运行数据对控制系统进行监控的情况,提出了1种用历史性能基准对预测控制系统进行性能评估、监控以及分析的方法。利用运行状态良好的1段历史数据计算出历史性能基准,并将历史性能基准与实际性能的比值用于控制器性能的实时监控。根据历史性能指标的残差监控检测出性能变化时,进一步通过历史数据协方差监控及广义特征值分析,区分出性能显著变差或变好的方向和子空间。为现场工程师提供性能变化的一些原因,用于系统维护。最后通过Shell重油分馏塔仿真证明该方法可以获得很好的评估与监控结果。
Aiming at the situation that the control system needs to be monitored only by the routine operation data of the actual industrial process, a method of evaluating, monitoring and analyzing the performance of the predictive control system based on the historical performance benchmark is proposed. The historical performance benchmark is calculated using one well-run historical data and the ratio of historical performance benchmark to actual performance is used for real-time monitoring of controller performance. When detecting the performance change based on the residual performance monitoring of the historical performance indicators, the covariance monitoring and the generalized eigenvalue analysis of the historical data are further used to distinguish the directions and subspaces in which performance is significantly deteriorated or improved. Some reasons for providing performance changes to field engineers for system maintenance. Finally, the simulation of the heavy oil fractionation tower of Shell shows that the method can obtain good evaluation and monitoring results.