Lithium Battery Capacity Estimation Based on Improved Convolutional Neural Network with Self-Attention Mechanism

At present, lithium batteries are very significant in the field of new energy, so the performance evaluation of lithium batteries in industry has become very important. Capacity estimation as one of the important indicators, its accuracy not only determines the service life and health of batteries, but also enables better dynamic monitoring and management of battery state, and optimization of energy use strategies. Among them, BP neural network is a classic algorithm applied in lithium battery capacity estimation systems, but this algorithm cannot effectively capture key information in input data and its learning ability is relatively weak. Therefore, an algorithm based on one-dimensional convolutional neural network and self-attention mechanism is proposed. One dimensional convolutional neural network can effectively capture the local features in the input sequence, and the self-attention mechanism can filter the information with high relevance to the target by calculating the weight of different positions. The combination of the two can improve the overall understanding ability of the model to the input data. Finally, two lithium battery data sets are used for experiments. The experimental results show that the algorithm of one-dimensional convolutional neural network combined with self-attention mechanism significantly improves the accuracy of prediction.