Solar Irradiance Prediction from Historical Trends Using Deep Neural Networks

Accurate prediction of solar irradiance is difficult because of the uncertainties in weather parameters. In this paper we present a forecasting model for solar irradiance based on historical data analysis and meteorological features, with the help of deep neural networks. A unique feature of the method is the use of historical trends of solar irradiance, which significantly improves the accuracy of our prediction model. We have validated the performance of our model using the data obtained from National Solar Radiation Database (NSRDB). As an application, the forecasted irradiance values are used to predict the power output of a photovoltaic (PV) plant set up in Kharagpur, India, within a low margin of error.