关键词: 1DCNN, LSTM, 锂电池, 多通道特征, 电池寿命

Abstract:

To improve the prediction accuracy and stability of the lithium-ion battery state of health (SOH), a battery SOH prediction method combining one-dimensional convolution (1DCNN) and long and short-term memory network (LSTM) is proposed herein to solve the problems of the complex selection of conventional features and the inability to effectively use them. First, multichannel series voltage, current, and temperature are used to construct multi-dimensional features. Second, 1DCNN is used to extract advanced data features from the sample data and input them into the LSTM to effectively utilize historical information. Finally, the SOH prediction results of the battery are output through a full connection layer. The combined algorithm is verified by the capacity attenuation data of the NASA lithium battery. The results show that compared with other prediction algorithms, the algorithm based on the 1DCNN-LSTM has a more accurate prediction result of the SOH with a mean absolute error of 0.01 and a failure point error of less than two cycles.

Key words: 1DCNN, LSTM, lithium battery, battery life, multichannel feature