摘要
为了解决纯电动汽车电池剩余电量估算难题,采用粒子群优化神经网络方法,用于BP(BackPropagation,BP)神经网络权值和阈值优化,并把优化后的神经网络用于荷电状态(SOC)离散估算。以100AhLiFePO4电池作为实验对象采集实验数据,将温度、充放电倍率和充放电电压作为PSO-BP(ParticleSwarmOptimization,PSO)神经网络输入特征向量,将电池SOC作为输出向量进行网络学习和训练,用训练好的网络对不同充放电倍率下SOC进行离散点预测,采用插值估算实现实时预测。实验结果表明,PSO-BP算法对SOC值为20%～80%区间估算准确,能够满足电动汽车正常运行的SOC估算要求。
To solve the problem of estimating SOC（state of charge） of pure EV, a neural network method based on quantum- behaved particle swarm optimization was used to optimize BP（back propagation, BP） network values and thresholds, and then the optimized neural network was used for SOC discrete estimate. 100 Ah LiFeP04 battery was studied as the subject, taking the temperature, charge-discharge rate and charge-discharge voltage as PSO-BP （Particle swarm optimization） neural network input eigenvectors, with battery SOC as output vector to perform net learning and training. The trained net and interpolation method was used to real-time predict SOC discrete points in different charge-discharge rates. The results show that PSO-BP algorithm can accurately estimate SOC value between 20% and 80%, and meet the requirements to estimate SOC of pure EV.
出处
《电源技术》
CAS
CSCD
北大核心
2013年第5期800-803,共4页
Chinese Journal of Power Sources
基金
淮安市科技支撑计划项目(HAG09039)
关键词
电动汽车
SOC
粒子群
神经网络
electric car
SOC
, particle swarms
neural network
作者简介
王业琴（1980-）。女．黑龙江省人，讲师，博士，主要研究方向为智能控制、电动汽车、图像处理与模式识别。
