Scholarlink

HVAC-driven load shifting is a demand-side management strategy that can help prevent grid congestion and the economic penalization of the user, caused by the peak demand surges due to the charging load of electric vehicles. In this context, the present work investigates the impact of a set of HVAC-driven interventions on the load profile of a warehouse (including both storage and office zones) case study, aimed at balancing the charging load of electric trucks upon their arrival. Therefore, a set of interventions is simulated using EnergyPlus and its Python API, which include overcooling of the indoor spaces, maintaining a reduced setpoint for a certain period, followed by a subsequent setpoint relaxation interval. Simulations are conducted for the entire cooling season, and the resulting data is used to train machine learning(ML) pipelines designed to predict the duration of overcooling and setpoint relaxation phases. The performance of various ML algorithms is assessed and compared while employing the sliding training window scheme. Lastly, the selected pipeline is employed to determine the suitable schedule for the aforementioned interventions, and the corresponding impact (through performing re-simulations) on balancing the charging load of electric trucks is assessed.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Implementation of a Machine Learning–Based Precooling-Assisted HVAC Load Shifting Procedure to Balance Electric Vehicle Charging Load in a Warehouse

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Farzad Dadras Javan,
  • Arya Assadian,
  • Fabio Rinaldi,
  • Sara Perotti,
  • Behzad Najafi

摘要

HVAC-driven load shifting is a demand-side management strategy that can help prevent grid congestion and the economic penalization of the user, caused by the peak demand surges due to the charging load of electric vehicles. In this context, the present work investigates the impact of a set of HVAC-driven interventions on the load profile of a warehouse (including both storage and office zones) case study, aimed at balancing the charging load of electric trucks upon their arrival. Therefore, a set of interventions is simulated using EnergyPlus and its Python API, which include overcooling of the indoor spaces, maintaining a reduced setpoint for a certain period, followed by a subsequent setpoint relaxation interval. Simulations are conducted for the entire cooling season, and the resulting data is used to train machine learning(ML) pipelines designed to predict the duration of overcooling and setpoint relaxation phases. The performance of various ML algorithms is assessed and compared while employing the sliding training window scheme. Lastly, the selected pipeline is employed to determine the suitable schedule for the aforementioned interventions, and the corresponding impact (through performing re-simulations) on balancing the charging load of electric trucks is assessed.