Leveraging AI tools along the heat pump value chain from component design to operation supervision
Artificial Intelligence (AI) provides engineers with powerful tools. We showcase applications of AI tools along the heat pump value chain. The development of efficient heat pump circuits with minimized refrigerant charges requires complex trade-offs between component selection and overall system performance. By using Machine Learning (ML) models trained on a large dataset from a heat pump development project, a design optimization tool to assist engineers in identifying the most optimal component combinations is developed and evaluated. The tool’s main goal is to recommend component combinations that achieve the highest possible efficiency (COP: Coefficient of Performance) with the lowest refrigerant charge for a given heat capacity and to reduce development times. On building energy systems (BES) design level, reliable annual simulations of variable speed heat pumps require a detailed performance map (> 100 operation points) of the heat pump. In early stages of heat pump and BES design, available component or equipment data is often limited. By applying ML models including deep learning tools to predict heat pump performance out of a minimal dataset, the number of required measurement points can be drastically reduced. We explore, evaluate, and show how to leverage these tools for an efficient BES design process. At the operational stage of the heat pump lifecycle, large amounts of sensor data are generated continuously by building automation systems. However, this data is often unstructured, noisy, and lacks explicit labels, posing significant challenges for direct ML application. We address this by developing a modular data-driven monitoring framework using task-specific ML models for fault detection, operational state classification, and performance evaluation. A central element is the integration of expert feedback to refine model outputs and validate anomalies. This human-in-the-loop strategy enhances model transparency, which is essential in the BES context, where explainability and operational relevance are critical for deployment.