基于CNN-BiGRU-MHA的锂离子电池健康状态预测OA

Accurate prediction of the state of health(SOH)for lithium-ion batteries is critical for enhancing safety and extending lifespan.However,existing methods face challenges in prediction accuracy and generalization capability under the complex degradation patterns of lithium-ion batteries.To address this,a method of lithium-ion battery SOH prediction based on convolutional neural network-bidirectional gated recurrent unit-multi-head attention(CNN-BiGRU-MHA)model is proposed.Eight features highly correlated with SOH are extracted from the NASA public battery dataset,and the CNN-BiGRU-MHA model is constructed.CNN is used to extract local features,BiGRU is used to capture bidirectional correlations between features,and MHA is introduced to focus on hidden states at critical time nodes,so as to enhance the representation capability of model to complex degradation modes such as battery capacity regeneration and nonlinear sudden drop.The experimental verification is performed on different datasets.The experimental results show that,in comaprison with the comparative models,the proposed method can realize superior performance in evaluation metrics including mean absolute error(MAE)and root mean square error(RMSE),and it possesses high prediction accuracy and strong generalization ability.