基于PSO-CNN的储能锂离子电池组荷电状态评估  

State of charge evaluation of energy storage lithium ion battery group based on PSO-CNN

作　　者：李肖辉 肖亚哲 田志国[1] 王京 李晶晶 LI Xiaohui;XIAO Yazhe;TIAN Zhiguo;WANG Jing;LI Jingjing(XU JI CEPRI Energy Storage Technology Co.,Ltd.,XJ Group Corporation,Xuchang Henan 461000,China)

机构地区：[1]许继集团有限公司许昌许继电科储能技术有限公司,河南许昌461000

出　　处：《电源技术》2024年第4期685-692,共8页Chinese Journal of Power Sources

基　　金：国家电网有限公司科技项目(4000-202237504A-3-0-SF)。

摘　　要：将卷积神经网络(convolutional neural network,CNN)的权值和阈值作为粒子群算法(particle swarm optimization,PSO)的粒子,将CNN的损失函数作为PSO的适应度函数,从而构建PSO-CNN算法对储能锂离子电池组的荷电状态(state of charge,SOC)进行预测。以储能系统现场采集的充放电数据为样本,分别采用本文算法、基于PSO优化的支持向量机(support vector machine,SVM)、CNN进行训练,并在完整充放电数据集上对比3种算法的预测效果。结果表明本文算法收敛性好、预测精度高。最后采用另一储能现场的数据验证本文算法具有良好的鲁棒性,可以广泛适用于储能系统锂离子电池组SOC的在线预测。With the weights and thresholds of the convolutional neural network(CNN)as the particles of the particle swarm optimization(PSO)algorithm and the loss function of CNN as the fitness func-tion of the PSO,the PSO-CNN algorithm was built to predict the state of charge(SOC)of the energy storage lithium ion battery group.The on-site charge-discharge data collected by the energy storage system were taken as samples,and the proposed algorithm,support vector machine(SVM)optimized based on PSO,and CNN were respectively used for training,and the prediction effects of the three al-gorithms were compared on the complete charge-discharge data set.The results show that the pro-posed algorithm has good convergence and high prediction accuracy.Finally,the data from another energy storage site were used to verify the robustness of the proposed algorithm.The results show that the proposed algorithm has good robustness and can be widely applied to the online prediction of lithium ionbatterygroup in energy storage system.

关 键 词：荷电状态 锂离子电池组 粒子群算法 卷积神经网络 

分 类 号：TM912.9[电气工程—电力电子与电力传动]