In order to solve the problems of difficulty in estimating the remaining service life of cells,complex testing process and high energy consumption in the cascade utilization of retired batteries.The designed basic model structure is input layer + GRU layer + full connection(full connect,FC)layer+output layer.According to the score of the health factor,the dataset for training the basic model is selected,the battery similarity level is divided,and the corresponding migration learning strategy is formulated.The experimental results show that compared with other models,the accuracy of the basic model and transfer learning model trained by using the first 40%and the first 25%of the dataset respectively increased by 42.48%and 95.28%,and the prediction stability Respectively the maximum increase of 55.38%and 93.55%.
关键词
机器学习/迁移学习/锂电池/门控循环单元神经网络/健康状态估计
Key words
machine learning/transfer learning/lithium batteries/GRU neural network/state of health estimation
维普
万方数据