基于IPSO-LSTM的新能源汽车锂电池健康状态监测

SOH monitoring of new energy vehicle lithium batteries based on IPSO-LSTM

为监测新能源汽车锂电池的健康状态(SOH),防范电池故障引发安全事故风险,提出改进粒子群算法(IPSO)和长短期记忆(LSTM)神经网络相结合的模型,监测锂电池的SOH。首先,采用Spearman相关性分析法,提取锂电池SOH监测的健康因子;其次,采用线性惯性权重和非对称学习因子改进传统粒子群算法(PSO),利用IPSO算法对LSTM模型的隐含层神经元个数、神经元失活率、批处理值进行关键参数寻优,进一步优化LSTM模型,建立IPSO-LSTM锂电池SOH监测模型;最后,以新能源汽车主流采用的18650锂电池数据集验证IPSO-LSTM模型,并对比分析BP、LSTM和PSO-LSTM这3种模型。结果表明:IPSO-LSTM模型的平均绝对误差(MAE)在0.02以内、均方根误差(RMSE)在0.03以内,监测误差在15%以内,相较于BP、LSTM、PSO-LSTM模型,IPSO-LSTM模型的误差指标值均最小,模型具有更高的精度和稳定性。

In order to monitor the SOH of lithium batteries of new energy vehicles and prevent the risk of safety accidents caused by battery failures,a model combining IPSO and LSTM was proposed.First,Spearman correlation analysis method was used to extract the SOH of lithium batteries.Secondly,the linear inertia weight and asymmetric learning factor were used to improve the traditional particle swarm optimization(PSO)algorithm,and the IPSO algorithm was used to optimize the key parameters of the hidden layer neuron number,neuron inactivation rate and batch processing value of the LSTM model,and then the LSTM model was further optimized,and the IPSO-LSTM lithium battery SOH monitoring model was established.Finally,the model was verified with the 18650 lithium battery data set which was the mainstream of new energy vehicles,and compared with BP,LSTM and PSO-LSTM models.The results show that the mean absolute error(MAE)of IPSO-LSTM model is less than 0.02,the root mean square error(RMSE)is less than 0.03,and the monitoring error is less than 15%.Compared with BP,LSTM and PSO-LSTM model,the error index value of IPSO-LSTM model is the smallest,and the model has higher accuracy and stability.