Accurate load forecasting is essential for effective planning and operation of modern power systems. In this study, we present a hybrid forecasting framework that blends multiple approaches to improve monthly electricity load predictions. First, we employ a SARIMAX model to capture seasonal and trend patterns, then incorporate ensemble-based machine learning methods—XGBoost and Random Forest—to handle complex nonlinear relationships, and finally integrate a deep learning method, the LSTM network, to leverage temporal dynamics more effectively. By combining these complementary techniques and refining the outputs through stacking and weighted averaging, our framework capitalizes on their individual strengths. Applied to real-world data from the National Power Control Center (NPCC) in Pakistan, the proposed model outperforms standalone techniques, achieving up to a 99% reduction in forecasting error compared to the best single-model baseline. These results confirm that the integrated approach offers a more robust and accurate solution for monthly load forecasting.