Embracing Artificial Intelligence in the HVAC Industry

Although still at its nascent stage, use of Artificial Intelligence (AI) in the HVAC industry will start growing rapidly from the next year. Before going further, let us have a brief idea on what AI is and how it does help us. Although there is a plethora of definitions of AI, perhaps the words used by IBM authors Cole Stryker and Eda Kavlakoglu constitute the most precise one.

According to them, “Artificial Intelligence (AI) is the technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity and autonomy.” They further elaborate that the applications and devices equipped with AI can see and identify objects. They can understand and respond to human language. They can learn from new information and experience. They can make detailed recommendations to users and experts. They can act independently, replacing the need for human intelligence or intervention.

Although a few companies have already started adopting small bites of AI in their systems, still lots of research and developmental works are going on across the world – to apply it in much wider and better way – targeting comfort enhancement, energy saving and air quality improvement. The purpose of this article is to present a few such R&D works and their findings so far.

AI’s impact on energy consumption: Schneider Electric

In a recent report, published through Schneider Electric, Rémi Paccou and Gauthier Roussilhe, Research Fellow and Doctoral Student at RMIT, have demonstrated how AI-powered HVAC systems can enhance energy efficiency and environmental conservation in buildings.

According to them, HVAC systems account for 35 – 65% of total building energy consumption. Their study has examined over 87 educational properties in Stockholm, Sweden, over an extended period under real-world conditions. Between 2019 and 2023, the study has observed a total carbon emission reduction of 65tCO₂e/y, roughly 60 times the actual embodied carbon footprint of the AI system deployed.

The research has revealed opportunities for even greater carbon reductions in environments with more demanding heating, cooling or air conditioning requirements. A comparative analysis between Stockholm and Boston has shown that implementing the same solution in Boston could yield carbon emission savings seven times higher than in Stockholm.

VRF system’s efficiency and thermal comfort: CEEE, UMD

According to the recent research findings published by the University of Maryland (UMD) Center for Environmental Energy Engineering (CEEE), a sophisticated type of Artificial Intelligence (AI), known as ‘deep learning’, could play an important role in reducing energy usage in the next generation of HVAC systems.

The researchers have explored AI’s impact on predicting power consumption in Variable Refrigerant Flow (VRF) technology – complex HVAC systems that have an outdoor unit and multiple indoor units – and presented their findings (paper) in the January 2025 issue of the ‘International Journal of Refrigeration’, now available online (https://www.sciencedirect.com/science/article/pii/S0140700724003591).

According to the paper, HVAC accounts for around 50% of a building’s total electricity consumption, so finding ways to slash HVAC energy consumption could significantly decrease a building’s overall energy usage. Energy consumption predictions are used to help optimise a VRF system’s efficiency and thermal comfort.

The researchers compared two types of AI models – a traditional machine learning model known as Artificial Neural Network (ANN) and a more recently developed deep learning model called Long-Short-Term Memory (LSTM). Both models use data to recognize patterns and produce insights and predictions – with LSTM requiring a larger data set. In this case, the models looked at data gathered from a VRF system installed at the university’s Glenn L. Martin Hall, taking into account variables like indoor and outdoor temperatures and humidity levels. Data was taken for a period of a year, so as to cover all four seasons.

As expected, the team found that the more data-intensive LSTM model had better accuracy in predicting power consumption. The big surprise, though, was that the LSTM model appeared to require less computing power and memory than the ANN model. The lead author of the paper, mechanical engineering graduate student Po-Ching Hsu, who had hypothesized that the more sophisticated LSTM model would require more computing power, commented, “That was counterintuitive.” He further explained, “What happens is that the ANN tries to improve its accuracy during the optimization process. So, it keeps increasing the parameters inside the structures in order to have a more complex model to predict the power consumption. But even with that, it still cannot achieve the same performance as LSTM.”

As per the paper, the optimised ANN model had over 13,500 trainable parameters, while the LSTM model had 1,809 trainable parameters. A higher number of trainable parameters in models is likely to result in increased memory usage on the computer and be computationally expensive.

According to Hsu, ‘deep learning’ could be a powerful tool in improving the energy efficiency of VRF systems, but the challenge is decreasing the time required to gather sufficient data. He said, “After the AC system is installed, it would need to collect data for one year to become optimized. We are trying to figure out if we can train this model with less data. Is there a way we can do this with data from a few days or a few weeks and still make very good predictions?”

The paper’s other co-authors are former CEEE researcher Lei Gao Ph.D. ’22, now on the R&D staff at Oak Ridge National Laboratory, and CEEE Co-Director Yunho Hwang, a mechanical engineering research professor.

AI helps heat pumps to operate more efficiently: Fraunhofer ISE

The Fraunhofer Institute for Solar Energy Systems ISE has been conducting research on a new generation of smart heat pumps that use artificial neural networks to adapt to environmental conditions and to learn as conditions change. This increases both the energy efficiency and user comfort. Extensive simulations have shown promising potential energy savings from 5 to 13% in addition to increased comfort. These results have been confirmed by measurements in an initial field test in a real building.

In the ‘AI4HP’ project, Fraunhofer ISE, together with the company Stiebel Eltron and the French research partners CEA List (Laboratory for Integration of Systems and Technologies) and LPNC (Laboratoire de Psychologie et NeuroCognition) as well as the industrial partner EDF R&D, has gathered important findings on new adaptive control methods for heat pumps based on neural networks.

They focused on the potential, flexibility and practical suitability of AI controls. According to them, up to now, heat pumps for residential heating purposes have mainly been controlled using static heating curves set once during installation. In most cases, the curves have not been optimised for the building, as this is only achievable through a time-consuming calibration. Furthermore, heating curves do not account for short or long-term changes, such as solar radiation, occupant usage or building renovation and aging. In this project, the specific building behaviour patterns, e.g., how it changes with varying solar radiation, is learned by Artificial Intelligence (AI) that continuously analyses recorded measured values.

The new AI heat pump controller was evaluated in extensive simulation tests, in which three buildings, each of a different construction year and refurbishment status, were simulated for the period of one heating season. The questions on self-calibration and the adaptability to new environmental conditions were both answered positively. Depending on the building, the resulting energy savings were shown to be 13% on average compared to the standard heating curve. These savings were due, in particular, to an improved matching of the reference room temperature and the setpoint temperature. Further energy savings can be expected if the controller is extended to include the efficiency characteristics of the heat pump.

By P. K. Chatterjee (PK)