Xie Cherng Miow0000-0002-4082-3322Yun Seng Lim0000-0003-4889-4907Lee Cheun HauWai Meng Chin0000-0002-7211-9524Jianhui Wong2025-09-122025-09-122025-01-0110.1088/1742-6596/2943/1/012006https://dspace-cris.utar.edu.my/handle/123456789/11349Flexible loads are common building demands that have gathered attention in demand-side management. However, further study is still required in applying machine learning-based methods on flexible loads to provide energy savings in practical studies. Thus, this article introduces an Artificial Neural Network (ANN)-based energy management controller for flexible loads, particularly VRV X AC and refrigerators. The controller balances energy savings and indoor comfort by controlling temperature set points for VRV X AC and duty cycles for refrigerators. Experimental data from the Daikin Research & Development building in Malaysia shows that the ANN-based controller can achieve energy savings of up to 27% while maintaining indoor comfort at an average temperature of 26.76°C with a Desired Energy Saving Coefficient (DESC) of 0.95. It is important to note that DESC represents the trade-off between energy savings and indoor comfort, with higher DESC values offering more significant savings at the potential cost of reduced comfort. Although maximum DESC settings have demonstrated up to 36% energy savings on-site, they have led to some compromise in comfort, evident from the higher average temperature of 27°C. This article highlights the fine balance between energy savings and indoor comfort during high building power demand. © 2025 Institute of Physics Publishing. All rights reserved.enBalance energiesDemand-sideEnergyEnergy savingsEnergy-savingsFlexible loadsIndoor comfortsManagement controllerNetwork-basedNeural-networksDemand side managementtext::conference output::conference proceedings::conference paper