Paper Conference

Proceedings of Building Simulation 2017: 15th Conference of IBPSA


Application of Support Vector Machines for Predicting the Performance of Air-Source Domestic Hot Water Heat Pump Systems

Romana Markovic1, Caroline Lorz1, Je´roˆme Frisch1, Christoph van Treeck1
1E3D - Institute of Energy Efficiency and Sustainable Building,RWTH Aachen University, Aachen, Germany*Corresponding author

Abstract: A compact air-to-water heat pump (AWHP) is a novel technology that combines a domestic hot water (DHW) system with an air conditioning system. Its main application focuses on residential houses and apartments. However, the performance of the DHW AWHP is widely unpredictable. Additionally, occupant’s behavior is one of the main reasons for the widely variable system’s coefficient of performance (COP). For that purpose, different support vector machines (SVM) are trained to recognize the patterns of the hot water consumption and to predict the resulting COP. The predictions are made in a discrete and independent time domain. In the following step, the predicted DHW consumption in sub-hourly intervals will be integrated per week, and the resulting weekly COP will be computed. As the result, operation of AWHP system over 51 weeks can be evaluated, and the predicted COP will be compared to the measured performance. The results pointed out comparable performance as detailed calibrated thermal simulations, while the computational costs of the evaluation process remained lower in comparison to the physical system’s modeling. Since no AWHP system specific information are required, this method is applicable for a wide range of DHW AWHP systems.
Pages: 596 - 605