Osaka, Japan – As organizations intensify their efforts to reduce energy consumption and curb carbon emissions, one critical aspect that demands optimization is indoor heating and cooling systems. Heating, Ventilation, and Air Conditioning (HVAC) systems, on average, account for approximately 40% of a building’s total energy usage. Implementing methods that conserve electricity while still ensuring a comfortable indoor environment for occupants can play a pivotal role in the battle against climate change.
Researchers from Osaka University have now demonstrated substantial energy savings through the application of a novel AI-driven algorithm designed for HVAC system control. Notably, this approach doesn’t rely on intricate physics models or exhaustive prior knowledge about the specific building.
During cold weather, conventional sensor-based systems often struggle to determine when heating should be deactivated. Factors like thermal interference from lighting, equipment, or the heat generated by occupants can complicate this decision, resulting in unnecessary HVAC operation and energy wastage.
To address these challenges, the research team employed a control algorithm that predicted the thermodynamic response of the building based on collected data. This predictive approach proved more effective than attempting to calculate the intricate influences on temperature, such as insulation and heat generation. Hence, data-driven methodologies can often outperform complex models when sufficient data is available. In this case, the HVAC control system was trained to “learn” the symbolic relationships among variables, including power consumption, using a substantial dataset.
The algorithm successfully achieved energy savings while maintaining occupant comfort. Lead author Dafang Zhao noted, “Our autonomous system delivered significant energy savings, surpassing 30% for office buildings, by harnessing the predictive capabilities of machine learning to optimize HVAC operation times. Importantly, the rooms remained comfortably warm, even during winter.”
The algorithm’s objectives included minimizing total energy consumption, the discrepancy between actual and desired room temperatures, and changes in power output during peak demand. Senior author Ittetsu Taniguchi highlighted the adaptability of their system, stating, “Our system can be easily customized to prioritize energy conservation or temperature accuracy, depending on the specific requirements of the situation.”
As we collectively strive towards a carbon-neutral economy, it is increasingly likely that corporations will need to lead the charge in innovation. The researchers anticipate that their approach may witness rapid adoption, especially in times of rising energy costs, benefiting both the environment and corporate sustainability.
- Press release
