Optimizing Elevator Energy Use with Machine Learning-Based Real-Time Prediction

With the rapid development of modern buildings and the increasing demand for energy management, elevator energy consumption has gradually become a focus of attention. This study constructs a real-time prediction system for elevator energy consumption based on machine learning. The aim is to optimize elevator energy management strategies by analyzing historical operational data and environmental data. A random forest model is employed for energy consumption prediction. Through feature importance analysis, SHAP value interpretation, and partial dependence plot analysis, key factors influencing energy consumption, such as hourly operation frequency, elevator door openings, floor usage, building energy consumption, and load, are identified. Model evaluation results show that the random forest model performs excellently in terms of prediction accuracy. By establishing a performance monitoring system, the prediction performance of the model is continuously tracked and optimized, achieving real-time prediction and dynamic management of elevator energy consumption. After adjustments and optimizations, the final model’s prediction error is significantly reduced, accurately capturing the energy consumption trends. This study not only enhances the energy efficiency management of elevators but also provides a reference for energy optimization of other complex systems.