Soft Computing Detection Method for Remaining Capacity of Lead-Acid Battery

This paper proposes a new soft computing method based on simplified BP neural network. For the traditional fully connected BP neural network, the optimal structure 8-5-5-1 is obtained through the optimization of the hidden layer, and the output weight and the threshold are obtained. Since the input and output structure of the obtained lead-acid battery is a “black box”, a calculation method of the contribution rate of the weight and the threshold is proposed, and the “black box” is transparent according to the method. By comparing the accuracy of BP neural network before and after simplification, the accuracy has no effect, but the model is simplified, and it is easier to observe the connection strength within the model.