Artificial neural network model for predicting CO2 heat pump behaviour in domestic hot water and space heating systems

Abstract

The natural refrigerant, CO2, possesses thermophysical properties that make it highly suitable for domestic hot water (DHW) production using heat pump technology. In this study, the development and validation of an artificial neural network (ANN) model that enables efficient design and control of a CO2 heat pump is presented. The study employs experimental data from a CO2 heat pump with a nominal heat capacity of 8 kW. The fully instrumented rig includes the heat pump and a pump rig designed to generate system temperatures representative of various space heat and DHW demands. A comprehensive dataset was generated through systematic variation of inlet temperatures and setpoints. The ANN provides predictions for outlet temperatures, heat production, and electricity consumption utilizing inlet flow rates, temperatures, and setpoints as inputs. These predictions are important for condition monitoring or in a smart operation management framework that determines optimal schedules for the machine.