Predicting Daily Incoming Solar Energy Output using Weather Data

Abstract

This project ‘Prediction of Solar energy using weather data’ is built using machine learning algorithms. Solar energy is most useful renewable source of energy so accurately predicting solar energy is required as their is a need to make up deficits from traditional fuel plants to match the total power generation with the instantaneous power consumption. Project main focus is based on prediction of Solar energy output recorded by solar photovoltaic panels using weather metrics like Temperature ,Cloud coverage , humidity ,visibility etc on a particular day. Type of problem of this project is regression based machine learning problem .Various famous machine learning algorithms like Random Forest ,Lasso and Ridge and K-nearest neighbours are used in order to accurately forecast solar PV output .Dataset was gathered from National Oceanic and Atmospheric Administration’s (NOAA) Global Ensemble Forecast System (GEFS) which generally forecasts weathers. Every machine learning algorithm is evaluated based on error evaluation metrics like mean squared error, mean absolute error and R-squared score. Many visualizations plots are also depicted in order to have better understanding of model. Feature selection is done on the dataset in order to select the best predictor features from a range of features. At last results and model comparison is done in order to evaluate which model among the three is best fitted for Solar PV output forcast.