复杂工业过程的多模型辨识及控制应用研究

发布时间：2018-03-01 17:53

本文关键词： 复杂工业过程非线性多模型辨识智能体熵自适应控制广义预测控制　出处：《华北电力大学(北京)》2017年博士论文　论文类型：学位论文

【摘要】：复杂工业过程存在系统变量多、非线性作用强及环境随机干扰影响大等问题,传统的建模方法由于模型形式单一往往难以描述系统全貌,导致控制效果不理想,因此多模型辨识及相应的控制策略成为复杂工业过程控制领域的研究热点和难点。本文在现有研究基础上,研究了复杂工业过程多模型辨识及控制的有关问题,提出了智能体多模型的辨识方法和在线聚类多模型的辨识方法,为复杂工业过程多模型辨识提供了新的思路。此外,在建模的基础上研究了间接自适应模糊控制策略和广义预测控制策略,对不同辨识方法及控制策略分别给出了相应的算例及应用仿真验证。最后,利用所研究算法对某电站660MW机组旋流燃烧器进行了整体的多模型辨识及控制系统的设计。本文的主要研究工作包括:1.提出了一种智能体和T-S模糊模型相结合的智能体模糊多模型构建方法,并从理论上证明了智能体多模型系统可以任意精度逼近任意线性或非线性系统。不同于传统多模型策略,智能体多模型系统中的每个智能体是一个动态方程,可以独立表征一种工况,也可以与其他智能体协同表征新的工况,使得该方法在描述复杂系统时有相当大的灵活性。电热水器系统的辨识仿真实验被用来验证该方法的可行性和有效性。2.针对存在随机噪声干扰的多变量非线性系统,提出了一种基于最小熵聚类的多模型在线辨识算法。通过最小熵模糊减聚类算法在线确定多模型中子模型个数及其相应隶属度权值,聚类过程中同时考虑了子系统规则化程度。给出了求取子模型参数的加权最小二乘递推表达,实现了子模型参数的在线辨识。数值算例的仿真实验验证了算法的有效性。此外,本文基于在线辨识策略进一步讨论了辨识过程中模型参数初调算法,对余热利用系统中冷却水回路电动阀的模型辨识检验了算法对环境随机干扰的自适应能力。3.设计了基于李雅普诺夫函数的间接自适应模糊控制器,针对智能体与模糊逻辑相结合的多模型被控对象,给出了保证系统稳定输出的最优模糊自适应控制律。针对烟气余热利用过程进行了仿真实验,为了验证算法的性能,同传统PID控制器的控制结果进行了对比。4.基于上述的辨识方法及广义预测控制器对某电站660MW机组旋流燃烧器进行了辨识及控制实验。通过分析旋流燃烧器结构及运行机理,确定了燃烧过程的主要变量,在此基础上提出一种模型参数可随过程调节的模糊多模型系统,给出了相应的模型参数辨识方法。构造了旋流燃烧器系统的多入多出广义预测控制策略。为验证算法的有效性,应用以上算法对电站旋流燃烧器进行了仿真实验。
[Abstract]:There are many problems in complex industrial processes, such as system variables, strong nonlinear effects and random environmental disturbances. The traditional modeling method is often difficult to describe the whole situation of the system because of the single model form, which leads to the unsatisfactory control effect. Therefore, multi-model identification and corresponding control strategies have become a hot and difficult point in the field of complex industrial process control. Based on the existing research, this paper studies the problems related to multi-model identification and control of complex industrial processes. The identification method of agent multi-model and on-line clustering multi-model is proposed, which provides a new idea for multi-model identification of complex industrial processes. On the basis of modeling, indirect adaptive fuzzy control strategy and generalized predictive control strategy are studied. Corresponding examples and simulation results are given for different identification methods and control strategies. The whole multi-model identification and control system of swirl burner of 660MW unit in a power station is designed by using the algorithm studied in this paper. The main research work in this paper includes: 1. A kind of intelligence combining agent and T-S fuzzy model is proposed. Energy volume fuzzy multi-model construction method, It is proved theoretically that the agent multi-model system can approach any linear or nonlinear system with arbitrary precision. Different from the traditional multi-model strategy, each agent in the agent multi-model system is a dynamic equation. You can represent one condition independently, or you can represent new conditions in collaboration with other agents. The simulation experiment of the electric water heater system is used to verify the feasibility and effectiveness of the method. 2. For the multivariable nonlinear system with random noise disturbance, A multi-model on-line identification algorithm based on minimum entropy clustering is proposed. The number of multi-model neutron models and their corresponding membership weights are determined online by the minimum entropy fuzzy clustering algorithm. In the process of clustering, the regularization degree of subsystem is considered at the same time. The weighted least square recursive expression of the parameters of submodel is given, and the on-line identification of the parameters of submodel is realized. The simulation results of numerical examples verify the validity of the algorithm. Based on the on-line identification strategy, this paper further discusses the initial tuning algorithm of model parameters in the process of identification. The model identification of the electric valve in the cooling water loop in the waste heat utilization system tests the adaptive ability of the algorithm to the random disturbance in the environment. 3. An indirect adaptive fuzzy controller based on Lyapunov function is designed. The optimal fuzzy adaptive control law is presented for the multi-model controlled object which combines agent and fuzzy logic. The simulation experiment is carried out to verify the performance of the algorithm. Compared with the control results of the traditional PID controller. Based on the above identification method and the generalized predictive controller, the swirl burner of a 660MW unit in a power station is identified and controlled. The structure and operation mechanism of the swirl burner are analyzed. On the basis of determining the main variables of combustion process, a fuzzy multi-model system with adjustable model parameters is proposed. The corresponding model parameter identification method is given, and the multi-input and multi-output generalized predictive control strategy of swirl burner system is constructed. To verify the validity of the algorithm, the simulation experiments are carried out on the swirl burner in power station.
【学位授予单位】：华北电力大学(北京)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP273

【相似文献】