差分进化神经网络干扰观测器的设计
|
摘要
针对常规非线性干扰观测器对等效干扰估计误差大、容易受参数摄动等影响,基于差分进化(DE)算法设计了神经网络干扰观测器。首先,采用神经网络逼近未知综合干扰,通过设计状态动态方程检验干扰观测器的估计效果;其次,引入DE算法在线调节神经网络权值,使逼近误差最终一致有界;最后,在Matlab中用该方法估计一类非线性不确定性系统——倒立摆所受的外界干扰。仿真结果表明:该方法能有效地提高观测器估计干扰精度,避免了权值设置及整定的盲目性。
Aiming at the fact that the conventional nonlinear disturbance observer has large errors of equivalent disturbance estimation and is easily affected by parameter perturbation,a neural network disturbance observer is designed based on the differential evolution algorithm.First,the neural network is used to approximate the unknown integrated disturbance,and a state-dynamic equation is designed to test the effect of disturbance observer estimation.Secondly,the differential evolution algorithm is introduced to adjust the neural network weights online so that the approximation errors are ultimately uniform bounded.Finally,in Matlab,the method is used to estimate the external disturbances of a kind of nonlinear uncertain systems — inverted pendulums.The simulation results show that the proposed method can effectively improve the estimation disturbance accuracy of the observers,and avoid blindness in weight setting and tuning.
Aiming at the fact that the conventional nonlinear disturbance observer has large errors of equivalent disturbance estimation and is easily affected by parameter perturbation,a neural network disturbance observer is designed based on the differential evolution algorithm.First,the neural network is used to approximate the unknown integrated disturbance,and a state-dynamic equation is designed to test the effect of disturbance observer estimation.Secondly,the differential evolution algorithm is introduced to adjust the neural network weights online so that the approximation errors are ultimately uniform bounded.Finally,in Matlab,the method is used to estimate the external disturbances of a kind of nonlinear uncertain systems — inverted pendulums.The simulation results show that the proposed method can effectively improve the estimation disturbance accuracy of the observers,and avoid blindness in weight setting and tuning.
引文
[1]刘金琨.智能控制[M].北京:电子工业出版社,2017:134-137.
[2]黄国勇.基于神经网络干扰观测器的Terminal滑模控制[J].吉林大学学报(工学版),2011 410(06):1726-1730.
[3]刘国琴,陈谋,姜长生.基于神经网络干扰观测器的导弹飞行控制器设计[J].航空兵器,2008(03):36-41.
[4]陈谋,邹庆元,姜长生,等.基于神经网络干扰观测器的动态逆飞行控制[J].控制与决策,2008(03):283-287.
[5]胡慧,刘国荣.基于神经网络干扰观测器的一类不确定非线性MIMO系统H∞跟踪控制[J].控制与决策,2009,24(03):468-471.
[6]杨立一.基于神经网络干扰观测器的终端滑模飞行控制设计[J].机械工程与自动化,2012(04):152-154.
[7]方一鸣,李叶红,石胜利,等.液压伺服位置系统的神经网络backstepping控制[J].电机与控制学报,2014,18(06):108-115.
[8]刘金琨.滑模变结构控制MATLAB仿真[M].北京:清华大学出版社,2005:329-332.
[9]曾辰子,余旌胡,邹桢苹.基于多样变异随机搜索的差分进化算法[J].武汉大学学报(理学版),2018,64(03):211-216.
[10]王盈祥,陈民铀,程庭莉,等.基于差分进化的改进狼群算法研究[J].计算机应用研究,2018,36(08):83-91.
[2]黄国勇.基于神经网络干扰观测器的Terminal滑模控制[J].吉林大学学报(工学版),2011 410(06):1726-1730.
[3]刘国琴,陈谋,姜长生.基于神经网络干扰观测器的导弹飞行控制器设计[J].航空兵器,2008(03):36-41.
[4]陈谋,邹庆元,姜长生,等.基于神经网络干扰观测器的动态逆飞行控制[J].控制与决策,2008(03):283-287.
[5]胡慧,刘国荣.基于神经网络干扰观测器的一类不确定非线性MIMO系统H∞跟踪控制[J].控制与决策,2009,24(03):468-471.
[6]杨立一.基于神经网络干扰观测器的终端滑模飞行控制设计[J].机械工程与自动化,2012(04):152-154.
[7]方一鸣,李叶红,石胜利,等.液压伺服位置系统的神经网络backstepping控制[J].电机与控制学报,2014,18(06):108-115.
[8]刘金琨.滑模变结构控制MATLAB仿真[M].北京:清华大学出版社,2005:329-332.
[9]曾辰子,余旌胡,邹桢苹.基于多样变异随机搜索的差分进化算法[J].武汉大学学报(理学版),2018,64(03):211-216.
[10]王盈祥,陈民铀,程庭莉,等.基于差分进化的改进狼群算法研究[J].计算机应用研究,2018,36(08):83-91.