Residential micro-grid load management through artificial neural networks

This paper presents an innovative load management tool for a micro-grid composed by a photovoltaic (PV) system and an energy storage device installed at a residential user. The objective is to develop a suitable residential load management to maximize the PV plant exploitation through the storage system in order to achieve a greater energy independence of the micro-grid (MG) from the electric grid. For this purpose a MG dynamic model was developed in Matlab Simulink environment useful to analyse and optimize the MG energy performance. On the modelling results, through artificial neural networks (ANN) technique, a hierarchy load management that takes into account of the load demand, battery state of charge and weather forecast was defined. Specifically the aim of the ANN model here proposed is to predict the scheduling of programmable loads considering the weather conditions relative to the current day and the previous one, beyond that on the weather forecast for the day after. The obtained results, considering the relatively small dataset, are to be considered strongly encouraging. Greater performance is expected in the case the data set is enlarged.