Hao Zhe Tan0000-0002-4082-3322Yun Seng Lim0000-0002-7211-9524Jianhui Wong0000-0003-4889-4907Lee Cheun Hau2026-10-142026-10-142026-03-06979833156119210.1109/CPEEE69412.2026.11521293https://dspace-cris.utar.edu.my/handle/123456789/11729The current energy infrastructure of Malaysia is overly centralized that focuses on supply-side management, which consists of large and peaking power plants. The current approach is not economically nor environmentally sustainable due to the infrastructural upgrade cost as well as the high carbon emissions rate. Thus, battery energy storage system (BESS) technology is proposed to overcome this issue. To utilize the BESS, this paper proposed a maximum demand reduction strategy that leverages a deep learning model as the core of a short-Term forecasting algorithm. The proposed algorithm analyses the load demand data to estimate the peak demand threshold for the BESS to coordinate energy supply when the load demand exceeds that threshold. With a 60-minute lookback window, it produces the most optimal results. When tested with multiple experimental sites' dataset, the robust algorithm can effectively reduce their maximum demand as long as it is initially trained with the site's data. Furthermore, it can achieve a maximum demand reduction factor of at least 43% among four experimental sites. © 2026 IEEE.en-USBattery energy storage system (BESS)Deep learningMaximum demand reductionsShort-Term forecastingBattery storageElectric utilitiesForecastingLearning systemsSecondary batteriesBattery energy storage systemBattery energy storage systemsCurrent energyDemand reductionForecasting algorithmLoad demandMaximum demandMaximum demand reductionShort-term forecastingDigital storagetext::conference output::conference proceedings::conference paper