Development of Artificial Intelligence Techniques for Solar PV Power Forecasting for Dehradun Region of India

Abstract

The need of reducing emission of carbon dioxide became possible with the increased penetration of solar photovoltaic (PV) power generation. The variable and intermittent nature of solar PV power generation affects the stability of the distribution grid. It comes to be necessary to forecast the generated PV power to avoid such type of uncertain conditions. This paper presents the empirical comparisons of six developed supervised learning algorithms to predict the solar power generation for the Dehradun region in India. The algorithms namely multiple linear regression (MLP), ridge regression, decision tree (DT), random forest (RF), support vector machine (SVM) and K-nearest neighbor (KNN) are modified in accordance with the higher prediction accuracy. The detailed empirical comparisons of results are discussed on the basis of mean absolute percentage error (MAPE) and root mean squared error (RMSE). It is found that the best performance of RF method with MAPE as 2.2790% and RMSE as 0.8792%. Tree based algorithms have shown the improved performance among all the methods while SVM and ridge techniques perform quietly low.