Research On SOH Estimation Algorithm Of Lithium Battery Based On CNN-BiGRU Neural Network

With the rapid development of electric vehicles today,the prediction of state-of-health(SOH)of the lithium-ion battery,the core of its power system,is increasingly critical for battery management systems.When the lithium ion battery cycle reaches a certain degree of aging,there may be unpredictable risks,and the health status of the lithium ion battery can directly provide a quantitative reference for the degree of battery aging and replacement.Therefore,it is of great significance to accurately predict SOH of lithium battery for personal safety.Based on the development of big data and the maturity of neural network technology,this paper designed a composite neural network structure combining Convolutional Neural Network(CNN)and Bidirectional Gated Recurrent Unit(BiGRU).Whale Optimization Algorithm(WOA)was used to optimize the parameters,and appropriate SOH-related characterization parameters were extracted from the constant voltage discharge process of lithium batteries as input.The SOH model is built for the output,and then the SOH of lithium battery is predicted.The main contents of this paper include:(1)Introduce the relevant background and significance of SOH prediction research of lithium ion battery,and then analyze the current research status at home and abroad,expound four different SOH definition methods,and divide the mainstream SOH prediction methods of lithium battery into experimental analysis,model-based method and data-driven method.The advantages and disadvantages of the three methods are discussed with examples.(2)Make an in-depth analysis of the lithium-ion battery,analyze the changes of the charge-discharge curve of the lithium battery,extract the average current,charging capacity,duration and isocurrent drop time of four characteristic parameters for the current curve of the constant voltage discharge stage.Through the gray correlation analysis method,the four parameters have a greater correlation with SOH,which can be used as the model input to predict SOH.(3)The CNN-BiGRU neural network structure was designed,and the WOA was used to optimize the model parameters and construct the theoretical flow of prediction.In view of the large amount of data required by CNN-BiGRU model,Gaussian white noise and bias were added to the initial data to enhance it.Using python language to complete the model architecture,input data for prediction.At the same time,it is compared with the common recurrent neural network,and then the prediction before and after data enhancement is compared with four different single feature input and multi-feature input.The experimental results show that the multi-feature input WOA-CNN-BiGRU model based on CV stage extraction can effectively predict the SOH of lithium battery,and the prediction result of multi-feature input is obviously better than that of single feature input.After data enhancement,the model has higher accuracy and robustness.Figure [38] Table [7] reference [70]...