Bridging prediction and decision: An application-driven review of load forecasting in energy systems

Load forecasting plays a pivotal role in operational decision-making within modern electrical power systems, influencing grid resilience, energy efficiency, and demand flexibility. Despite substantial advances in forecasting techniques, their practical utility is often constrained by limited integration into real-world operations. This review highlights the application of load forecasting in energy system management, focusing on methods that directly inform operational decisions. It provides a comprehensive analysis of forecasting horizons, algorithmic frameworks – including machine learning and deep learning approaches – and preprocessing strategies that enhance prediction accuracy and adaptability. Special attention is given to probabilistic forecasting, which addresses uncertainties inherent in dynamic energy markets, alongside a consolidation of key performance metrics. Distinguishing itself from existing surveys, this study identifies critical limitations related to adaptive modeling, customization, and scalability. By emphasizing application-driven approaches, this work lays the foundation for developing interpretable, scalable, and context-aware forecasting models capable of supporting increasingly dynamic and data-intensive energy systems. • Reviews load forecasting approaches applied to real-world smart energy systems. • Classifies forecasting models by horizon, scale, and application domain. • Synthesizes advances in data preprocessing and feature engineering methods. • Identifies research gaps and outlines future directions for practical integration.