Accurate electricity demand forecasting is crucial for effective energy management, particularly during peak periods. It supports grid stability, minimizes waste, and enhances the integration of renewable energy. However, forecasting is challenging because demand varies with weather, seasonal patterns, and daily fluctuations. Hybrid models, such as Prophet combined with XGBoost, can improve accuracy by capturing complex demand patterns, contributing to a more resilient energy infrastructure. This study uses four years and three months of time-series data from the Ceylon Electricity Board (CEB) in Sri Lanka. It compares statistical models Exponential Smoothing (ES), Seasonal Autoregressive Integrated Moving Average (SARIMA), and Prophet with a hybrid Prophet + XGBoost model. Forecasting accuracy is measured using Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Square Error (RMSE). The hybrid model shows superior performance, reducing MAE by 19.45, RMSE by 18.73, and MAPE by 1.32% compared to Prophet alone. These findings indicate that combining statistical and machine learning methods can substantially enhance electricity demand forecasting, which is crucial for smart grid advancement, renewable energy integration, and sustainable energy planning. Although developed with Sri Lankan data, the methodology is adaptable to other regions, offering a scalable solution for global energy forecasting challenges.