Volume 38: Energy Transitions toward Carbon Neutrality: Part I

Development of FDD Model for a Real-life Case using Transfer Learning with Synthetic Data Han-Gyeong Chu, Seongkwon Cho, Cheol-Soo Park

Abstract

The Air Handling Unit (AHU) system is influenced by various types of errors, which can cause thermal discomfort of occupants and energy waste in building. Therefore, an early and accurate Fault Detection and Diagnosis (FDD) is important for optimal control of building heating/cooling systems and increasing occupant productivity. The data-driven FDD is promising because it is convenient compared to the first principles-based rule set that demands in-depth expertise. However, in order to realize the data-driven FDD for real-life cases, the data imbalance problem in FDD must be solved. In this study, the authors suggest a novel approach that generates synthetic data from an entire building system simulation tool, HVACsim+ and then use them as a source model for applying transfer learning to a target AHU system. For the transfer learning, only the normal operational data from the existing target system was used. It is found that the transfer learning approach is satisfactory, confirming that the proposed method will be effective in mitigating the data imbalance issue in developing the data-driven FDD.

Keywords building energy, transfer learning, synthetic data, automated fault detection and diagnostics