一种神经网络分数阶PID控制器的实现dtm∫.PDFVIP

Journal of Computer Applications ISSN 1001-9081 2014-12-15 计算机应用,2014,34(S2 :166-168,199 CODENJYIIDU http:/ / 文章编号：1001-9081（2014）S2-0166-03 一种神经网络分数阶PID控制器的实现 * 毛书军 ，盛贤君 （大连理工大学电气工程学院，辽宁大连116024） ( 通信作者电子邮箱maoshujun@ * 要：为解决分数阶PID控制器参数难于整定的问题，设计了一种基于神经网络的分数阶PID控制器。通过采 用反向传播（BP）神经网络的参数调节策略，可以实现一种五维参数自学习的PID控制器。将分数阶PID控制器数字 化，通过BP算法调节神经网络突触权值，经过调整的神经网络输出作为分数阶PID控制器的参数。经过仿真验证， 神经网络分数阶PID控制器比传统PID控制器精度提高6倍且控制更加稳定。 关键词：分数阶PID；神经网络；反向传播算法；最优解；自学习 中图分类号：TP183；TP273 文献标志码：A mplementation of an neural network fractional order P D controller * MAO Shujun , SHENG Xianjun (School of Electrical Engineering, Dalian University of Technology, DalianLiaoning 116024, hina Abstract: In order to solve the challenging problem of determing the parameters in fractional order PID controller, a fractional PID order controller based on artificial neural network was proposed. A self-learning PID controller with five- dimension parameterswas realizedbyusingparameteradjustment strategyof BackPropagation(BP neuralnetwork. Afterthe fractional order PID controller was digitized andthe synapticweightswere adjustedby using BP strategy, the adjustedoutputs of theneural networkwereusedastheparametersof thefractional orderPIDcontroller. A seriesof experimentsverifythatthe artificial neural networkfractional order PID controller can increase the accuracy by six times than the traditional PID order controller and is more stable. Key words: fra