Experiences with a novel method to estimate the state of charge of a lead-acid battery

This paper introduces a method, which joins a classical approach (ampere-hour balance) with a neural network (Topology-Preserving Map), to estimate the state-of-charge (SOC) of a flooded lead-acid battery. Because a normal user is not able to determine numerous parameters of a battery model, this approach uses very general data sheet information about the battery behaviour. Nevertheless, during normal system operation the new algorithm adapts to the real battery behaviour. In this method the discharge behaviour of the unstratified battery is stored in the neural network. By comparing the stored with the momentary discharge behaviour, the algorithm is able to assess the error of the SOC estimation of the ampere-hour balance and to deliver some information about the stratification of the electrolyte. The results demonstrated that this approach is able to supply the user of a battery not only with the SOC, but also with more comprehensive information about the state of the battery.