Enhancing and Optimising Solar Power Forecasting in Dhar District of India using Machine Learning

Abstract

Power and energy systems around the world are expanding and evolving in tandem with technological advancement. In the current scenario, energy is a crucial requirement for the development for any country. Machine learning (ML) is used as a technology to address the requirement for quicker and more accurate analyses that would support the control and operation of modern power systems. In this paper, analysis is performed using Machine Learning and Deep Learning (DL) models to predict power estimation at a photovoltaic (PV) solar site with the capacity of 79.95 kW, installed in Dhar district, Madhya Pradesh (MP), India. The model's accuracy is evaluated using various statistical parameters, R2 score, Mean Square Error (MAE), Root Means Square Error (RMSE), Mean Square Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percent Error (MAPE) and variance. The proposed method Linear Regression (LR) algorithm shows a maximum R2 score of 0.99994, a small error metric of MAE 0.0091, and an RMSE of 0.121, which indicate the highest accuracy model as compared to other algorithms. Accurate prediction of solar power without irradiance, season-wise (five seasons in India) and month-wise, is also predicted with high accuracy using ten different models of machine learning and one deep learning method, with comparison of its results with the existing work.