Improved lead -acid battery modelling for photovoltaic application by Recurrent Neural Networks

The paper presents some multilayer recurrent neural networks (RNNs) to improve the highly non linear behaviour modelling of lead acid cell battery. Dynamic RNNs, keeping into account the non-linear dynamic behavior of both input-output variables of the battery charge-discharge processes, provide a powerful tools in the above mentioned problem, despite the higher computational burden involved respect to the feed-forward networks. Since the electric current supplied by the battery is dependent upon the user application, it can be regarded as the only effective external input of the dynamic system described by the equations, and then of the RNNs. The basic variables of the discharge process are in fact both voltage and SOC.