RACA Journal I June 2026 www. refrigerationandaircon. co. za 29
FEATURE SERIES
AI
Another important application is load prediction. Traditional engineering approaches rely on detailed physical models of buildings – accounting for materials, geometry and thermal properties. While rigorous, these models are complex and often rely on assumptions that limit their accuracy in real-world operation.
Machine learning offers an alternative. By analysing historical data – such as weather patterns and system performance – AI can learn how a building behaves without requiring detailed knowledge of its physical structure. In many cases, this datadriven approach can produce more accurate predictions of cooling demand.
This has significant implications for both design and operation, particularly in retrofit environments where detailed building data may not be available.
PREDICTIVE MAINTENANCE PROTECTING PERFORMANCE Beyond optimisation, AI also enables predictive maintenance – an increasingly valuable tool in cooling systems. By learning the‘ normal’ performance of equipment, machine learning models can detect deviations that indicate deterioration or faults. This allows issues to be addressed early, before they lead to costly failures or efficiency losses.
For facility managers, this represents a shift from reactive maintenance to proactive asset management – reducing downtime and extending equipment life.
Despite its potential, the adoption of AI in cooling is not without challenges. One of the most significant barriers is trust. As Barraza notes in the podcast, transparency is a key concern among users. Many AI systems operate as‘ black boxes’, making decisions that are not easily understood by operators. This lack of visibility can lead to resistance, particularly in mission-critical environments.
Berger emphasises the importance of making AI systems transparent – showing users not only what decisions are made, but why. This includes providing visibility into alternative scenarios and allowing operators to override specific control actions when necessary.
Another major challenge is data quality. Unlike digital platforms that rely on vast, clean datasets, HVAC systems operate in environments where data is often incomplete, inconsistent or affected by sensor errors. Ensuring data reliability and filtering out faulty inputs is critical to the success of machine learning applications.
Finally, there is the risk of overengineering. Not every problem requires AI, and applying machine learning where simpler solutions exist can lead to unnecessary complexity and reduced reliability.
When applied effectively, AI in cooling delivers tangible benefits. The most obvious is energy savings. By continuously optimising system performance, machine learning can significantly reduce energy consumption – particularly in large, complex installations.
However, the benefits extend beyond energy:
• Reduced maintenance costs through early fault detection.
• Lower operational overhead through automation.
• Improved system reliability and performance.
As Berger notes, AI also frees up facility teams from routine troubleshooting, allowing them to focus on higher-value tasks.
AI MEETS RENEWABLE COOLING Looking ahead, the role of AI in cooling is set to expand further – particularly as systems become more complex with the integration of renewable energy. Hybrid systems combining solar thermal, storage and multiple chiller technologies introduce new levels of variability and interdependence. In these environments, traditional control strategies may no longer be sufficient.
AI offers a pathway to manage this complexity – optimising not just individual systems, but entire energy ecosystems. In this sense, AI is not simply an efficiency tool, it is an enabler of the next generation of sustainable cooling systems.
The insights from Berger and Barraza highlight an important reality: AI in cooling is neither a magic solution nor a distant future concept. For the HVAC & R sector, the challenge is not whether to adopt AI, but how to apply it effectively. As the industry moves toward more complex, data-driven systems, one thing is clear: The future of cooling will not just be efficient – it will be intelligent. RACA
AI is not replacing conventional control strategies but enhancing them.
Freepik. com artificial intelligence
RACA Journal I June 2026 www. refrigerationandaircon. co. za 29