基于贝叶斯优化-卷积神经网络-双向长短期记忆神经网络的锂电池健康状态评估OA

Accurate estimation of battery state of health(SOH)is the key to the stable operation of the device.In order to solve the problems in the current SOH research,such as the difficulty to measure the volume directly and the time required to adjust the model parameters,a prediction model based on the multi-health features of Bayesian optimization(BO)optimized convolution neural network(CNN)and bi-directional long short term memory(BiLSTM)neural network is proposed.Based on NASA's publicly available lithium battery data,three health characteristics are extracted.The combination of CNN and BiLSTM improves the processing ability of time series data,and adds BO algorithm to automatically search the optimal parameter set,which avoids the combination network model falling into the local optimal and reduces the estimation time.The results show that the proposed method has the highest prediction accuracy and can aeffectively estimate the SOH of lithium batteries.The mean absolute error and root mean square error are both within 1%.