基于PSO-SA的二级倒立摆前馈补偿模糊神经网络控制  

Fuzzy Neural Network Control Combined with Feed-forward Compensation for a Double Invert Pendulum Based on PSO-SA

作　　者：张秀玲[1,2] 田力勇[1,2] 李晓辉[1,2] 

机构地区：[1]燕山大学河北省工业计算机控制工程重点实验室,河北秦皇岛066004 [2]燕山大学国家冷轧板带装备及工艺工程技术研究中心,河北秦皇岛066004

出　　处：《山东科技大学学报（自然科学版）》2012年第2期81-85,92,共6页Journal of Shandong University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(50675186)

摘　　要：以模糊神经网络为基础,结合误差前馈补偿完成了二级倒立摆系统的稳定控制,并采用模拟退火粒子群算法对控制参数进行全局寻优。与基于状态变量合成的模糊神经网络控制器相比,该控制方法不仅解决了多变量系统模糊控制器的"规则爆炸"问题,并且,由于所有状态变量直接参与控制输出,控制精度亦有所提高。仿真结果表明,该控制方案所需规则数目少,响应速度快,有良好的鲁棒性和非线性适应能力。The stability control of the double invert pendulum system was completed based on a fuzzy neural network and combined with the error feed-forward compensation in this paper and,an algorithm of particle swarm optimiza- tion combined with simulated annealing （PSO-SA） was used in global optimizing of control parameters. Compared with the fuzzy controller based on the synthesis of state variables this control scheme has not only solved the rule number explosion problem of multi-variable system fuzzy controller, but also has improved the control accuracy be- cause of the direct participation of control output by all state variables. Simulation results show that this control scheme has the advantages of less rules,fast response speed,good robustness and strong nonlinear adaptive ability.

关 键 词：二级倒立摆 模糊神经网络 前馈补偿 模拟退火粒子群算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]