Open AccessThe use of artificial neural networks (ANN) in the prediction of energy consumption of air-source heat pump in retrofit residential housing
Author(s) -
Keng Ye,
Gulsun Demirezen,
Alan S. Fung,
E.G.O.N. Janssen
Publication year - 2020
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/463/1/012165
Subject(s) - artificial neural network , wind speed , energy consumption , linear regression , heat pump , air source heat pumps , air temperature , engineering , automotive engineering , environmental science , machine learning , computer science , meteorology , mechanical engineering , heat exchanger , electrical engineering , geography
Machine learning algorithms using Artificial Neural Network (ANN) were developed to predict the performance of heat pump systems in retrofit residential housing. The study attempts to address the research gap in the application of machine learning algorithms to real-life field measurements as a case study. Rowhouse units with electric resistance baseboard heating were retrofitted with Ductless Air Source Heat Pumps (DASHPs). Sensors were installed to collect the energy consumption data during the baseboard and DASHP monitoring periods. Linear and quadratic regression methods following the International Performance Measurement and Verification Protocol (IPMVP) were applied to predict energy consumption based on outdoor temperature and heating degree days. These predictions were compared against results from ANN models based on Levenberg-Marquardt algorithms using the hour of the day, day of the week, outdoor temperature, wind speed and direction, relative humidity, condition and indoor temperature as inputs. Preliminary results indicate that predictions from ANN models produced higher correlation of determination than those from IPMVP regression analysis.