SHORT-TERM LOAD FORECASTING IN SMART GRIDS USING MACHINE LEARNING: A COMPARATIVE ANALYSIS OF REGRESSION-BASED MODELS
Keywords:
Short-Term Load Forecasting, Smart Grids, Machine Learning, Regression Models, Linear Regression, Decision Tree, Random Forest, Energy Forecasting, Predictive Modeling, Load PredictionAbstract
Precise short-term load forecasting (STLF) is vital for the better efficiency, reliability, and sustainability of contemporary smart grids. With the rising deployment of smart meters and advanced metering infrastructure, there are ample amounts of high-resolution electricity consumption data available, which has paved the way for the use of machine learning (ML) methods for enhanced demand forecasting. The paper offers a comparative study of three regression-driven ML algorithms, Linear Regression, Decision Tree Regressor and Random Forest. Used for predicting hourly electricity load. The models are implemented and tested on actual smart meter data with features such as historical load, temperature, time of day, and day type. Evaluation is done based on the primary statistical metrics such as Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE). Results show that ensemble learning, especially the Random Forest model, notably surpasses conventional linear methods in forecasting with an MAPE as low as 5.3%. The research identifies the possibility of data-driven methods in developing smart grid operations and advocates for the incorporation of ML-based forecasting systems into real-time energy management and planning.
