【作者】 孙乐文；

【导师】 于佐军；

【作者基本信息】 中国石油大学 ， 控制理论与控制工程， 2009， 硕士

【摘要】 针对复杂非线性系统难以建模从而难以控制以及约束预测控制的在线优化计算量大的问题,本文研究了基于T-S模糊模型的非线性系统辨识算法,并在此基础上研究非线性系统的模糊预测控制问题。对比研究了约束预测控制的在线优化几种求解算法,仿真表明本文所提算法的有效性。论文的主要研究工作内容有以下几个方面:(1)针对复杂非线性动态系统的模糊建模问题,本文采用一种基于T-S模型在线自学习的模糊辨识算法。通过模糊聚类算法对T-S模糊模型的前提部分进行辨识;对于T-S模糊模型的后件参数采用在线学习的方法进行辨识。仿真研究表明该算法较好的辨识效果。(2)有约束的预测控制滚动优化可以转化成一个标准的二次规划问题。对比研究投影最小二乘算法、有效集算法和Dantzig-Wolfe算法在预测控制在线优化中的求解效果,Shell塔仿真表明投影最小二乘算法在约束预测控制在线优化求解中有较好的效果。(3)研究了基于T-S模糊模型的DMC和GPC两种模糊预测控制算法,并且与普通的PID的控制效果进行比较,pH中和过程表明基于T-S模型模糊预测控制较普通的PID有较好的效果。将非线性系统预测控制的目标函数转化为线性二次规划问题,避免了非线性规划巨大的计算量,同时利用投影最小二乘算法进行在线求解。更多还原

【Abstract】 In this dissertation, for complex nonlinear system difficult modeling and the large online optimization calculation of constrained predictive control, the identification algorithm of nonlinear system based on T-S fuzzy model is researched. On the basis of these, the fuzzy predictive control based on T-S model is researched. Several algorithms of the online optimization of constrained predictive control are studied, the simulation results show that the new method are of effectiveness. The main contents are concluded as followings:(1) In view of modeling problems of nonlinear and dynamic system, a self-learning fuzzy identification algorithm is presented based on T-S model in this paper. The premise parameter is identified by fuzzy clustering and the consequent parameter is identified by self-learning. The simulation result shows that the algorithm has high accuracy and can be used in online modeling.(2) The online optimization of constrained model predictive control could be converted into quadratic programming. Projected least-squares algorithm, active set method and Dantzig-Wolfe algorithm are contrastively researched in the quadratic programming of model predictive control. The simulation of Shell tower predictive control shows the projected least-squares algorithm is effective to solve the online optimization problem.(3) Two nonlinear fuzzy predictive control methods, namely fuzzy DMC and GPC based on T-S fuzzy model, are contrastively researched in the paper. And comparing the performance of fuzzy predictive control with the common PID control, the simulation result on pH neutralization demonstrates the performance of fuzzy predictive control performance is superior to conventional PID controller. Objective function is converted into linear programming and thus the huge computational burden is avoided. The projected least squares algorithm is also used in the online optimization of fuzzy predictive control.更多还原