Purpose: This study aims to evaluate the energy reduction performance of HVAC systems by applying optimal supply air temperature control based on an integrated machine learning approach. In conventional HVAC operation, fixed supply air temperature settings or reactive control strategies are commonly used, limiting the ability to respond effectively to varying load conditions and resulting in suboptimal energy performance. To address this limitation, a predictive and adaptive control strategy is proposed to dynamically adjust supply air temperature according to system conditions. Method: An integrated machine learning algorithm (IMLA) was developed by combining two machine learning models: an artificial neural network for system condition estimation and an optimization-based learning algorithm for determining control variables. The proposed method determines the optimal control variable at each time step based on predicted system conditions, enabling adaptive operation under varying load profiles. The performance of the method was evaluated using a central HVAC system under typical cooling season conditions. Result: The results show that total energy consumption decreased from 85,381.8kWh to 80,307.5kWh, corresponding to a reduction of 5.9% (5,074.3kWh) compared to baseline operation. Component-wise analysis indicates that chiller energy decreased by 7.2% (79,492.0kWh to 73,748.5kWh) and pump energy decreased by 9.1% (2,732.6kWh to 2,483.7kWh), while fan energy increased by 29.1% (3,157.2kWh to 4,075.3kWh). Monthly results further show higher energy reduction during partial-load periods, with savings of 12.9% in June and 10.3% in September. Despite the increase in fan energy, the overall reduction was achieved due to the dominant decrease in chiller energy.
Seong et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: