Solar Power Generation Prediction

Abstract

Predicting sun irradiance has been a crucial subject in the production of renewable energy. Prediction enhances solar system development and operation and provides several financial benefits to power companies. Statistical techniques like artificial neural networks (ANN), support vector machines (SVM), or autoregressive moving average can be used to forecast the irradiance (ARMA). However, because to their scalability or the fact that they are unable to be employed with huge data, they either lack accuracy due to their inability to capture long-term reliance. Thus, in this paper the XGBoost algorithm is implemented for prediction and Optuna Algorithm for Hyper parameter tuning and optimizing the results. Aside from predicting the solar irradiance. It is crucial to create a tool that will estimate the entire amount of energy that can be produced by a solar power plant, array, or household solar setup based on the expected solar radiation and the site's particular solar panel or array parameters. In this work methodology designed and developed a system that will not only predict the solar irradiance for next 15 days based on real time forecast but it will also predict the power generation in units for your solar power panel or array. This system is currently implemented in a webapp that can be accessed through any browser.