Performance Prediction of Ground Source Heat Pump System With Data-Driven Modelling: Application of Machine Learning Techniques

Abstract

Ground source heat pumps (GSHP) offer a clean and sustainable energy solution by using ground heat as a heat source. Various machine learning techniques are used to estimate coefficient of performance (COP), one of the key metrics for evaluating ground source heat pump systems. In this study, three different machine learning methods, Artificial Neural Network (ANN), Support Vector Machines (SVM) and Gaussian Process Regression (GPR), were used to estimate the performance of the system, using measurements made on an experimental GSHP system established in Sivas province of Turkey. Five different combinations are recommended for each method. Training and testing data were used to evaluate the effectiveness of each method in estimating the COP in the GSHP system. In the first stage, ANN, SVM and GPR models were trained on training data, and in the second stage, validation was carried out with test data. The statistical results obtained show that all three machine learning methods are successful in estimating COP in the GSHP system. R2 values of the model MIP1 in which all variables are used are determined as 0.9726 for ANN, 0.9543 for SVM, and 0.9650 for GPR. This study has demonstrated that different machine learning models can be used as an effective method for predicting the performance of GSHP systems. Additionally, by comparing the applicability and performance of these models, it provides information on determining the most appropriate model that can be used to increase the energy efficiency of GSHP systems and optimize costs.