Renewable energy generation prediction using machine learning models

Abstract

Renewable energy utilization is now widespread because of its friendliness to the environment, its sustainability which allows it to grow a society sustainably. Rwanda plans to increase its electricity generation by increasing the number of hydroelectric and solar power plants in the country. However, due to the uncertainty in the availability of renewable energy sources, it’s challenging to successfully integrate them into the electric grid. Conducting accurate forecasts of these energy sources can help address this challenge. In a bid to address this challenge, this study aims to use Machine Learning which has been proven to be faster and more efficient in making forecast of Time Series data. This work aims to present the performance of a Machine Learning model, the SARIMA model while forecasting solar and hydroelectric energy. The SARIMA model is selected because it can make good predictions with Time Series data which has a seasonal component, and hydroelectric and solar energy are affected by seasonal factors like temperature and relative humidity. The ANN and SVM models have also been proven to perform efficiently in the prediction of hydroelectric and solar energy. Due to the unavailability of adequate data, forecasting could be only done by creating SARIMA models for separate plants. The results indicated that the forecasts produced for the plants could become more precise. This is because of the stationary nature of the data of most of the plants. This research can be continued with the provision of adequate data for analyzing other Machine Learning methods