Study of forecasting renewable energies in smart grids using linear predictive filters and neural networks

Accurate forecasting of renewable energies such as wind and solar has become one of the most important issues in developing smart grids. Therefore introducing suitable means of weather forecasting with acceptable precision becomes a necessary task in today's changing power world. In this work, an intelligent way for hourly estimation of both wind speed and solar radiation in a typical smart grid has been proposed and its superior performance is compared to those of conventional methods and neural networks (NNs). The methodology is based on linear predictive coding and digital image processing principles using two dimensional (2-D) finite impulse response filters. Meteorological data have been collected during the period 1 January 2009 to 31 December 2009 from Casella automatic weather station (AWS) at Plymouth, UK. Numerical results indicate that a considerable improvement in forecasting process is achieved with 2-D predictive filtering compared to the conventional approaches.