引入层间耦合关系表达的双层结构预测控制系统性能评估与深度诊断

There are flourishing applications of double-layered MPC in advanced control of industrial processes. The performance of double-layered MPC system would deteriorate for some reasons such as the plant ageing and load variation, which causes the loss of manufacturers. It is valuable to study the performance assessment and diagnosis methods of double-layered MPC systems for their performance maintenance. Based on the coupling mechanism between the steady-state optimization layer and the dynamic optimization layer, data-driven optimal performance criterion design method of both steady and dynamic state, and further the root diagnosis methods of performance deterioration are to be studied. The construction of steady and dynamic state optimal criterion is to be explored. The restrictive relation between the two layers is introduced by supplementing constraint condition into each other's optimization problem. The iterative solving algorithm by gradually relaxing the constraint condition is proposed to search for the feasible region of both layers. The diagnosis of mismatched model under the addition of steady state layer is to be explored. The model performance assessment algorithm and mismatched model channel location algorithm on the basis of multivariate fusion and statistical analysis technology are proposed. The diagnosis of improper controller parameters is to be explored. The diagnosis method of improper parameters based on the definition and analysis of clausal performance indices is proposed. The industrial software is to be developed and applied in the typical air separation process. The results of the project are valuable theoretically and practically. It would push the intelligent maintenance of industrial model predictive control system to some extent.

双层结构预测控制技术在工业先进控制系统中广泛应用，但设备老化、负荷变动等因素会造成系统性能衰减，使企业经济利益受损，对其进行性能评估与诊断对实现系统性能维护具有重要意义。项目基于双层结构预测控制系统稳态优化层和动态优化层的层间耦合机制，采用数据驱动方法开展稳态和动态综合最优性能评价基准设计及性能衰减根源诊断理论研究。探讨稳态和动态综合最优评估基准的构建问题，采用互补约束条件的方法引入层间耦合关系，通过约束放松反复迭代方法优化求解；研讨加入稳态优化层情况下的模型失配诊断问题，研究基于多变量融合和统计分析的模型性能评估和失配通道定位算法；探索预测控制参数设置不当的诊断问题，定义系统分项性能指标并探讨基于分项指标综合的参数设置不当诊断方法；开发双层结构预测控制系统评估与诊断工业应用软件，在工业空气分离过程形成应用案例。研究成果具有理论意义和应用价值，对工业预测控制系统智能化维护有一定推动作用。

双层结构预测控制技术在工业先进控制系统中广泛应用，但设备老化、负荷变动等因素会造成系统性能衰减，使企业经济利益受损，对其进行性能评估与诊断对实现系统性能维护具有重要意义。项目基于双层结构预测控制系统稳态优化层和动态优化层的层间耦合机制，采用数据驱动方法开展稳态和动态综合最优性能评价基准设计及性能衰减根源诊断理论研究。针对多变量预测控制系统模型失配的精确定位问题，研究了过程失配子模型深度诊断与模型修正方法、基于输入与干扰相关性分析的交叉定位方法两种方法，推进了模型性能评估理论的应用进程。针对控制器参数调节不当引起的控制性能下降问题，提出基于扰动模型的马尔可夫参数序列设计控制器参数评估指标的方法，可为控制系统维护提供控制器参数是否需要调整的指导性结论。针对双层结构 MPC 系统的性能评估问题，提出了一种基于稳态优化层和动态优化层的层间耦合机制的评估方法，在动态优化问题和稳态优化问题中分别引入耦合因子，得到了最优控制性能基准。针对所研究的各种方法开发了应用软件，应用于SHELL精馏过程、WOOD-BERRY精馏过程、TE化工过程等典型化工过程预测控制系统性能评估与性能提升，特别是实际工业空气分离过程预测控制系统性能评估，取得了良好应用效果。研究成果具有理论意义和应用价值，对工业预测控制系统智能化维护有一定推动作用。

内容获取失败，请点击重试