Machine-Learning-Based Prediction of HVAC-Driven Load Flexibility in Warehouses

This paper introduces a methodology for predicting warehouse’s reduced load while offering flexibility. Physics-based energy simulations are first performed to model flexibility events, which involve adjusting cooling setpoints with controlled temperature increases reduce the load. The warehouse building encompasses office and storage spaces, three scenarios implemented, i.e., exclusive area cooling, in both expand study’s potential applications. Next, simulation data utilized training machine learning (ML)-based pipelines, five subsequent hourly consumption values an hour before setpoint adjustments, providing time plan participation demand response programs or prepare charging electric vehicles. For each scenario, performance of Artificial Neural Network (ANN) tree-based ML algorithm compared. Moreover, expanding window scheme is utilized, gradually incorporating new emulating online learning. results indicate superior algorithm, average error less than 3.5% across all cases maximum 7%. achieved accuracy confirms method’s reliability even dynamic where integrated space offices needs be predicted.