Abstract:
Heat pumps have emerged as key technology to pave the way for a sustainable heat supply in buildings. However, the ongoing shortage of trained experts counteracts the current heat pump trend. Increasing the capacity of experts, services like Fault Detection and Diagnosis (FDD) can support the identification of malfunctions and integration of methods for predictive maintenance. The primary objective of FDD is to detect faults, diagnose their causes, and possibly enable correction to prevent efficiency losses as well as system damage or downtime. This involves a comparison between a fault-free reference case and the real system. In research, machine-learning methods like Artificial Neural Networks (ANN) show the capability to learn the behavior of fault-free systems. In practice, however, the implementation of ANN is limited due to missing data in operation for training. Therefore, it is common to utilize physically simulated data for pre-training. One way to achieve high efficiency in heat pumps is to maximize heat transfer in the evaporator. Fouling within this component therefore leads to significant performance degradation and reduced system lifespan. As a result, this work introduces and evaluates an extendable FDD method for evaporator fouling in air-to-water heat pumps. To detect evaporator fouling during operation using an ANN, a transient model of a refrigerant cycle provides the training data. Based on literature, the fouling effect is emulated afterwards, serving the data for the reference system considering faulty operation. Applying the present concept, we reveal a reduction in COP due to evaporator fouling of approximately 3 % over a whole year, while our fault detection methodology detects 55.65 % of the faults within the given heat pump model. Overall, this study provides insights into the performance of FDD methods for evaporator fouling in air-to-water heat pumps, which can help to improve the efficiency and reliability within the system lifespan. The results of this study demonstrate that the concept of FDD offers the potential to be applied in practice, and proposes recommendations for future perspectives about ANN within FDD in heat pump systems. Β© (2023) by ECOS 2023 All rights reserved.
