基于PSO-BP的电动汽车锂离子电池SOC估算  

SOC estimation of lithium battery for electric vehicle based on PSO-BP neural network

作　　者：赵钢[1] 朱芳欣 窦汝振[2] ZHAO Gang;ZHU Fang-xin;DOU Ru-zhen(Tianjin University of Technology,Tianjin Key Laboratory of Control Theory&Application in Complicated System,Tianjin 300384,China;China Automotive Technology&Research Center,Tianjin 300300,China)

机构地区：[1]天津理工大学天津市复杂控制理论与应用重点实验室,天津300384 [2]中国汽车技术研究中心,天津300300

出　　处：《电源技术》2018年第9期1318-1320,共3页Chinese Journal of Power Sources

摘　　要：针对BP神经网络算法对电动汽车锂离子电池荷电状态(SOC)估算的缺陷,提出粒子群(PSO)优化BP神经网络的方法,采用温度、电压、电流、充放电倍率作为PSO-BP神经网络的输入向量,以SOC作为输出向量,进行网络学习和训练,并不断进行神经网络权值、阈值的调整优化。在Matlab中进行仿真验证,实验结果表明BP神经网络算法和PSO-BP神经网络算法均可以使误差减小,但是使用PSO-BP神经网络算法估算SOC效果更优、误差更小、收敛性更佳,可将误差减小到4%以内。In view of the defect of BP(back propagation)neural network algorithm of electric vehicle lithium ion battery charge state SOC(state of charge)estimation,the PSO(particle swarm optimization)optimization of BP neural network method is proposed.Using temperature,voltage,current,charge and discharge rate as the input vectors of PSO-BP neural network and using SOC as the output vector to learn and train the network and to optimize and adjust the weight and the threshold of the neural network.In the Matlab simulation,the experimental results show that the errors of BP neural network algorithm as well as PSO-BP neural network algorithm both can be reduced less than 5%.Experimental results show that it's a better way to use PSO-BP neural network algorithm to estimate SOC,because the error can be reduced less than 4%.

关 键 词：PSO-BP神经网络 电动汽车 锂离子电池 荷电状态 

分 类 号：TM912[电气工程—电力电子与电力传动]