A study on the applicability of PV/T-EAHE in different regions based on machine learning prediction algorithms

• Establish an XGBoost model to evaluate PVT/EAHE thermoelectric performance. • In the selected regions, annual overheating discomfort hours reduce by 61.5%-100%. • In the selected regions, total annual power generation increased by 1.24%-4.26%. • Winter H CD rises 0.41%–8.63% in severe-cold/cold regions; elsewhere, minimal gains. • In the selected regions, annual energy output increased by 0.13%-47.21%. To investigate the regulatory, energy-saving, and applicability aspects of the thermal and electrical performance of the solar Photovoltaic/Thermal and Earth-Air Heat Exchanger (PV/T-EAHE) in different regions, this study established a database comprising 897600 sets of combined experimental test data and system simulation data. Six machine learning models were employed to train the thermal and electrical performance of the PV/T-EAHE, and the system operation effects in seven cities under different thermal zones were predicted. The thermal and electrical performance with and without PV/T-EAHE was compared. The results indicate that Extreme Gradient Boosting (XGBoost) demonstrated the best performance on the test set, with coefficients of determination (R²) reaching 99.89% for indoor temperature and 99.99% for instantaneous power generation predictions. The PV/T-EAHE system exhibited considerable potential for addressing overheating and cooling regulation challenges across the annual cycle in the selected cities, reducing annual hours of overheating discomfort by 61.5%–100% and increasing total annual power generation by 1.24%–4.26%. In severe cold and cold regions, the system increases H CD by 0.41% to 8.63% in winter. However, the system exhibited weaker cold regulation capabilities for winter indoor temperatures in regions with hot summer and cold winter (Chengdu), hot summer and warm winter (Guangzhou), and mild regions (Kunming). Additionally, the annual energy output of the system across all thermal zones has declined. Specifically, it decreased by 15.87% in Hami, 47.21% in Shenyang, 18.46% in Beijing, 18.66% in Xi'an, 11.89% in Chengdu, 0.13% in Kunming, and 17.79% in Guangzhou.