Clustering Methodology for Defining a Short Test Sequence for Whole System Testing of Solar and Heat Pump Systems
Abstract
Dynamic whole system testing methods are currently applied to evaluate the performance of heating and cooling system. One major obstacle to the implementation of such dynamic test methods is the cost connected to the experimental phase. To define a short test sequence, an original implementation of “k-medoid” clustering has been implemented. The paper investigates on the length of the sequence and how the days should be described to achieve effective results. A family of solar assisted heat pump systems has been generated by varying the collector field area and storage tank volume and studied via numerical simulations. The results show that the days can be described as a function of average temperature and global horizontal irradiation. The simulation of different sequence lengths shows that a six-day sequence can be used to test the system with a possible deviation of results lower than 8% if compared to the annual simulation. The deviation is reducing with a higher number of clusters.