Faulty operation in buildings is known to waste 15% to 30% of total HVAC energy on average - yet fault diagnostic tools have not been widely adopted by the industry due to the current lack of low-cost, accurate solutions.
We have developed a technology for detecting and diagnosing faults, that that combines high-level data-driven analytics and component-level diagnostics using a Bayesian network. This combination provides holistic building diagnostics that can be readily conveyed to a building operator using a probabilistic output. This diagnostic technology was designed specifically to be able to automatically learn and adapt to the wide variety of components and custom systems typically found in medium and large commercial buildings. This automatic customization reduces the upfront cost of implementation and also increases the diagnostic accuracy (and reduces false alarms), addressing the key obstacles to widespread adoption of fault diagnostic tools.
This software is presently being demonstrated in four commercial buildings, in which we have performed over 300 unique fault experiments spanning two years of research. The diagnostic accuracy of this novel technology is over 95% accurate, with a false alarm rate of less than 1%. The demonstration of this technology is being expanded to an additional 6 buildings over the next month, and the software is being converted from research-grade into a commercially viable product. We are seeking funding to assist with this conversion, and to assist with the design and implementation of the front-end/UX on our path to commercialization.