Research on dynamic operation strategy of heating, power generation, and air purification-type BIPVT system combined with machine learning

The building-integrated photovoltaic/thermal (BIPVT) system with air purification has great application potential in the post-pandemic era. However, existing systems cannot adjust dynamically, leading to unmet demands and higher energy consumption. Machine learning is an effective tool for optimizing building systems but is rarely applied to air-purifying BIPVT systems. This study proposes a dynamically adjustable BIPVT system that integrates heating, power generation, and air purification. A performance prediction model combined with machine learning is developed and validated. An optimized operation strategy is designed with multi-objective optimization algorithm and compared to traditional strategies. The impact of occupant density on strategy selection is analyzed. Key findings include: (1) The developed neural network model achieved prediction errors of 0.96 % for room temperature and 4.31 % for power generation. (2) The optimized strategy maintained indoor bacteria concentrations within safe limits throughout the day, providing 5.51 h of thermal comfort and generating 0.59 MJ of net power. (3) Compared to traditional strategies, the optimized strategy achieves a 208.5 % improvement in energy saving or extends the indoor air safety time by 7.16 h (4) Increased occupant density heightens the demand for air purification, resulting in longer air purification mode times, extended thermal comfort duration, and reduced power generation.