Development of a Machine Learning-Driven HVAC Load Shifting Strategy to Facilitate Balancing Electric Vehicle Charging Load in a Logistics Node

Introducing a fleet of electric vehicles (EVs) at logistics nodes and the resulting load of the corresponding charging stations can lead to local grid balancing issues, which (if not handled) can lead to potential blackouts and the economic penalization of these end-users due to the subsequent increase in their peak load. Therefore, proposing and employing HVAC-driven load-shifting strategies, which can mitigate the latter issues, is of increasing industrial interest. Accordingly in the present paper, a warehouse is considered as the case study. Next, a set of interventions is simulated, which includes slight overheating of the indoor spaces, followed by a period of preservation at an increased setpoint, and a subsequent setpoint relaxation interval (during which EVs can be charged). The simulations are performed using EnergyPlus software throughout the heating season and using the resulting generated data, a set of machine learning (ML)-based pipelines are developed. These pipelines, which predict the duration setpoint relaxation procedures, permit anticipated planning and execution of the interventions, being given the estimated arrival time of EVs. The performed simulations demonstrate that the mentioned interventions can effectively reduce the facility’s load to balance the charging load of EVs. The predictions of the developed ML-based pipelines reveal that the models estimate the ramp-down durations with acceptable accuracy. Therefore, it is shown that these interventions, while being scheduled using the proposed ML-based estimation pipelines, can be effectively deployed in practice, to handle the charging load of EVs, evading a surge in the facility’s load.