Short-term Wind and PV Generation Forecasting of time-series using ANN

The main sources of energy from renewable energy sources (RES) that have penetrated into the generation mix depend on the climate and cannot be shipped. The data of Independent Ontario Electricity System Operator (IESO), Canada and meteorological parameters available near the Oak Ridge Laboratory, USA are considered for the short-term forecast of generation from wind and photovoltaic (PV) sources, according to some hypotheses. Hourly data for two years 2017 and 2018 is taken and training validation and testing of the artificial neural network (ANN) is performed. Three models for the short-term prognosis of time-series are taken for the study, namely, non-linear automatic regression with exogenous input (NARX) Automatic nonlinear regression (NAR) and Input-Output model. The Bayesian regularization method (BR) is used for training. NARX shows the minimum mean square error (MSE) and regression during training, validation and testing. When implementing the prediction by this ANN, it is expected that the generation from these RES can be more dependable to meet the load contributing a further reduction in the cost of the conventional generation.