Enhancing Microgrid Efficiency: A Comparative Analysis of Forecasting Techniques for Load Demand Prediction

Abstract

This study compared three machine learning techniques - Ridge Regression, ARIMA, and Random Forest Regression - to forecast short-term electricity demand in a Nigerian university microgrid. The goal was to identify the most accurate method for predicting the university's power needs 24 hours ahead. Six years of historical load data and weather information were used to train and evaluate the models. Random Forest Regression (RFR) emerged as the clear winner, achieving a significant improvement in accuracy compared to both Ridge Regression (RG) and ARIMA. Notably, RFR offered a 45% reduction in Root Mean Squared Error (RMSE) and a 33% decrease in Mean Absolute Percentage Error (MAPE) compared to RG. These results suggest that RFR provides the most precise predictions for university electricity demand in this scenario