State Of Health Prediction Of Lithium-Ion Batteries Based On Attention Mechanism

Lithium-ion battery technology is becoming the most suitable device for storing electrical energy in mobile devices and electric vehicles due to its inherent and attractive characteristics:light weight,high energy density,small size,low memory effect,long life and low pollution.However,with the rapid growth in the number of lithium-ion battery users,the demands on the safety performance of lithium-ion batteries are increasing.In particular,sudden failures of industrial and light mechanical equipment due to battery failure can cause significant economic losses in industrial production.Therefore,State of Health(SOH)prediction of lithium-ion batteries has become one of the key research areas in the field of new energy in recent years.This paper analyses the capacity degradation data of three lithium-ion batteries of the same model from the NASA PCo E public dataset,and extracts the external parameters related to voltage and temperature from the charging and discharging process of lithium-ion batteries as HI to characterize the degradation process of lithium-ion batteries.The correlation analysis was used to investigate the correlation between HI and battery capacity.A hybrid model based on Convolutional Neural Network(CNN),Bidirectional Long Short-Term Memory(BiLSTM)and Attention Mechanism(AM)is developed to achieve accurate prediction of SOH of lithium-ion batteries.By analyzing the charging and discharging process of lithium-ion batteries,this paper first extracts the indirect HI,which is highly correlated with the capacity,as the input to the CNN,and uses the convolution and pooling operations of the CNN layers to extract the features of the time series data of lithium-ion batteries.On this basis,a BiLSTM depth model is built in this paper to capture the forward and reverse dependencies of the data coming from the CNN and to further emphasize the correlation between the serial data by AM,resulting in accurate SOH prediction.The experimental results based on NASA PCo E lithium-ion battery data show that the proposed hybrid model outperforms other single models,and the root mean square error(RMSE)of SOH prediction results are less than 0.01,which can accurately predict the SOH of lithium-ion batteries.Based on the long time series prediction problem,this paper proposes a hybrid model of Bidirectional Gated Recurrent Unit(BiGRU)and Transformer with multi-headed attention mechanism for SOH prediction of lithium-ion batteries.In the prediction model,the extracted indirect HI capable of characterizing battery degradation is fed into BiGRU for learning the hidden states of the input features to further extract time series features.Based on this,multiheaded attention is given to the Transformer encoder layer and the input feature vector to enable better performance in terms of long-term dependence of the time series.BiGRU-Transformer combines the structure and advantages of BiGRU and Transformer to predict SOH of lithiumion batteries with high accuracy.The study based on NASA PCo E lithium-ion battery data shows that the BiGRU-Transformer model proposed in this paper has higher accuracy,better robustness and generalization capability.