中文摘要：

针对模型预测控制系统中经常发生的模型失配问题,提出一种模型预测控制性能评估与模型失配诊断的系统框架,为工业系统的后期维护提供帮助。对系统的整体性能进行监控,根据已有的历史性能基准的预测控制性能评估方法,提出一种用户自定义选取历史基准数据的方法,弥补其需要专家知识的缺陷;在检测到系统整体性能下降后,采用数据驱动的思想,利用一种模型失配指标算法对可能引起系统性能变差的模型失配这一因素进行分析,包括过程模型失配和干扰模型失配,诊断模型的失配情况,完成对系统恶化源的初步定位。用Wood-Berry模型对该算法进行仿真验证,结果验证了其有效性。

英文摘要：

In view of the model mismatch problem in the model predictive control system,a system framework of model predictive control performance evaluation and model mismatch diagnosis was proposed,which was helpful for the later maintenance of industrial system.The overall performance of the system was monitored,based on the existing MPC performance evaluation method called historical benchmark,a customized historical data selection method was proposed,which selected historical reference data to make up for needing expert defects.The data driven theory was used to analyze the factors that might cause the system performance variation,including process model mismatch and disturbance model mismatch,a model mismatch index was proposed,which was used to locate the source of system deterioration.Wood-Berry model was used to simulate and verify the method.The results show that the method is effective.