Deep Learning Models in Irradiance Forecasting

Abstract

The slides present a high-level knowledge of the accurate prediction of solar irradiance power from a particular location using hybrid machine learning models viz Stacked Stateless/ Stateful GRU, LSTM and Autoencoders, which can be proved to be viable if applied to prior installation of solar photovoltaic cells in a particular area. The project tries to save the cost prior to the installation of solar panels by accurately predicting the appropriate location from where power can be elicited to meet the desired electric power required for running industries. The analysis of the hybrid machine learning models is done to determine which model is best suited for prediction by feeding them with data such as geometrical coordinates, solar parameter like GHI and weather parameters like temperature and wind speed etc.