Load forecasting is a critical component in modern power systems, ensuring efficient energy dispatch and grid stability. Traditional methods like ARIMA and and exponential smoothing often fall short in capturing nonlinear trends and responding to rapid fluctuations in demand and weather conditions. This is where machine learning offers a significant advantage, with its ability to model complex patterns from data. This paper proposes a short-term load forecasting framework using three machine learning models—Random Forest, Support Vector Regression (SVR), and Linear Regression. The system predicts electricity demand for the next 48 h using historical weather and load data and compares actual versus predicted load values. Among the three, Random Forest delivers the best performance based on RMSE and R2 metrics, highlighting the potential of data-driven methods for improved forecasting accuracy in renewable-based microgrids.