蚁群神经网络算法在电动车用直流电机起动过程中的应用  

Application of Ant Colony Algorithm Neural Network in the Starting Process of DC Motor Used in Electric Vehicle

作　　者：王旭东[1] 刘金凤[1] 张雷[1] 

机构地区：[1]哈尔滨理工大学电气与电子工程学院,黑龙江省哈尔滨市150080

出　　处：《中国电机工程学报》2010年第24期95-100,共6页Proceedings of the CSEE

基　　金：黑龙江省科技攻关重点项目(GB08A306)~~

摘　　要：电动车用直流电机控制器系统在电动车起动过程中具有非线性、快时变的特点,常规的比例-积分-微分(proportion-integration-differentiation,PID)控制方法很难满足系统非线性、参数摄动的要求,即使采用了前向神经网络算法进行PID整定,也由于结构复杂,训练速度慢等原因而很难满足实时控制的要求。于是提出了采用蚁群神经网络(ant colony algorithm neural network,ACANN)整定PID控制策略,用蚁群算法学习多层前馈(back propagation,BP)神经网络的权系,建立了基于该算法的BP神经网络训练模型,因而兼有了神经网络的广泛映射能力和蚁群算法的快速全局收敛以及启发式学习等特点,该控制策略可以补偿系统参数摄动、非线性和外界扰动对系统控制性能的影响,达到电动车平稳快速起动的目的。仿真和实验结果证明,该控制策略对电动车起动过程中电机起动电流的控制具有快速性、稳定性和鲁棒性。Since the start process for control system of electric vehicle has the characters of nonlinearity and fast time-variety,and routine PID method is difficult to satisfy the nonlinear and variable request.Even Back Propagation,BP,neural network was applied in system,it＇s still difficult to meet the requirement of real time because its structure was complex and train speed was slow.So it applied a control strategy based on ant colony algorithm,ACA PID,to control the motor through closed-loop control.Ant colony algorithm was tried to be introduced into the BP neural network optimization training by using it to learn the weights of BP neural network and established training model based on it,which combined the extensive mapping ability of neural network with the rapid global convergence and heuristic learning characteristics of ant colony algorithm.This strategy can compensate the perturbation,nonlinearity and outside disturbance of system parameter,and achieve the purpose of a smooth start-up of electric vehicle.It is proved through the simulation and experiment that this control strategy of starting process will generate the starting current which is rapid,stable and robust.

关 键 词：电动车 直流电机 起动过程 鲁棒稳定性 蚁群神经网络 

分 类 号：TM306[电气工程—电机]