Load forecasting for home energy management using light gradient boosting machine learning algorithm

Le résumé fourni par la source

The deployment of smart metres has revolutionised the smart grid ecosystem by enabling advanced monitoring and load prediction techniques, which benefit both electricity providers and consumers. These techniques utilise historical consumption data for short-term energy demand forecasting, supporting efficient energy management. Electricity providers can optimise supply strategies for enhanced reliability and cost-effectiveness, while consumers can reduce energy usage and costs through informed planning. This paper explores the application of various machine learning models, including LightGBM, K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Linear Regression, Decision Trees, Random Forest, XGBoost, CatBoost, and Extremely Randomized Trees (ERT), for short-term load forecasting using a publicly available dataset of London households. LightGBM emerged as the most effective model, achieving a lowest Mean Absolute Percent Error (MAPE) of 0.18% and an R2 Score of 0.8629. The data was preprocessed using Google Colaboratory, and performance evaluation highlights the transformative potential of smart metering and machine learning, particularly LightGBM, in optimising energy usage and grid efficiency.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.