Multi-Horizon Machine Learning Model for Home Energy Consumption Prediction Considering Appliances Usage

Home appliance energy consumption increases as technology penetration in home appliances increases. Predicting home energy consumption is a critical task in home energy management, including load scheduling, improving energy efficiency, and reducing carbon emissions. This paper proposes a multi-horizon prediction of home appliance power consumption using machine learning techniques. The proposed approach exposes both individual appliance and total power consumptions, and the impact on improving the prediction accuracy. Using appliance-level data collected from an IoT-based SCADA monitoring system, the tree-based models (Random Forest, Light Gradient Boosting Machine, and eXtreme Gradient Boosting) and deep learning models (Recurrent Neural Network, Long Short-Term Memory, and Gated Recurrent Unit) are evaluated across four prediction horizons: 1 min, 15 min, 1 hr, and 1 day. The evaluation results show that no single algorithm achieves the best performance across appliance types and time horizons. Thus, the appliance-specific, horizon-dependent model selection strategy is essential for optimizing prediction accuracy in home energy management systems. The proposed approach provides a valuable finding: incorporating all individual appliance power measurements, along with total power, as model inputs improves total power prediction performance.