Estimation of the residual capacity of sealed lead-acid batteries by neural network

This paper presents a method for estimating the remaining capacity of sealed type lead-acid batteries. The approach can be divided into three parts, first a survey on battery properties over a long period of time was conducted. This data was used in the second phase to train a feedforward neural network. Finally, the third phase tested the accuracy of prediction of this network using real data. It was found that using this method, a maximum error of prediction of 10% and an average mean error of 3% could be obtained