基于性能评估的动态过程建模和控制方法研究

Aim at typical industrial processes with unstable operative conditions, and strong disturbances, the project studies methodology on dynamical modeling and control for processes with performance estimation, in which dynamical modeling, adaptive fuzzy control, and performance estimation for control system are investigated based on real-time data. The project studies modeling method combining association analysis and data-driven modeling, in order to establish a sintering bum-through point (BTP) model with subspace modeling based on principal component analysis. Such model will serve as reference model, by which a modeling reference adaptive fuzzy control method based on the differential evolution algorithm is studied in order to realize intelligent control for BTP. With real-time performance of control taken account, a performance estimation method for process model with data-driven subspace modeling and adaptive fuzzy control is studied, in which the comprehensive performance will be assessed by using expert analyzing method. Thus based on this estimation method, the project studies an on-line coordination optimization strategy for adjusting parameters of process model and controller, so that the system will be controlled stably, and good at adapting different operative conditions. This method will be testified its correctness, validity, and superiority through both stimulation platform and real sintering bum-through point process in industrial process. The study will synthesize modeling, control and performance assessment to provide a comprehensive design method for some typical dynamic industrial processes. It is of significant research and application value.

针对工况波动大、干扰因素多的典型工业过程，本项目研究以实时数据为驱动，过程动态建模、自适应模糊控制和控制性能评估三者协调运行的设计与实现，提出一种基于性能评估的动态过程建模和控制方法。以烧结终点过程控制为研究对象，研究基于关联性分析和数据驱动建模方法，建立基于主元分析的子空间数据驱动模型；以该模型为参考模型，研究基于差分进化算法的模型参考自适应模糊控制方法；根据过程的实时控制效果，设计子空间数据驱动模型和自适应模糊控制器的性能评估方法，运用专家分析法获得系统的综合性能评估结果，研究过程模型参数和控制器参数的在线协调优化策略，使系统有效适应工况的变化，实现系统的长期稳定控制。通过已有的烧结终点过程仿真实验系统和实际工业应用探讨，验证所提方法的正确性、有效性和优越性。本项目提出数据驱动的建模、控制与评估的综合化设计，为复杂动态工业过程提供一种实时有效的新方法，具有重要的研究意义和应用价值。

针对机理复杂、工况波动大且干扰因素多的钢铁烧结过程，本项目以大量的过程数据为基础，通过建立基于子空间和模型评估的动态过程模型、可变论域参数自整定模糊控制器设计、控制器性能评估，提出了一种动态协调建模、性能评估和控制器参数调整运行的控制结构，实现了复杂生产中过程模型和控制器参数实时调整。本项目采用基于主元分析的方法，分析烧结过程生产数据中关键影响因素，提出一种基于动态子空间的过程建模方法，结合网格搜索算法对算法参数进行寻优，得到了准确的生产过程模型；采用数据挖掘算法和基于冲突规则加权的方法，提取模糊规则和确定模糊论域，设计了定论域模糊控制器，基于单纯形法设计了变论域下控制器参数的自调整策略，提出了一种可变论域参数自整定模糊控制器设计方法，提高了控制器的控制精度和自适应性；根据模型和控制器的运行结果评价两者的性能状况，提出了一种基于支持向量机的性能诊断方法，对模型和控制器参数进行协调优化；建立了烧结过程的仿真系统，应用了所提出的提出的建模、控制和性能评估的算法并验证了其正确性和有效性；同时项目组将相关方法推广应用到钢铁工业煤气混合加压、加热炉等生产过程，并进行实践，实现了煤气混合加压过程准确建模和加热炉温度的精确控制，表明本项目所提出的复杂工业过程建模、控制和性能评估方法具有良好的应用价值。