Bayesian inference-based deep learning approach for performance prediction and uncertainty quantification of solar photovoltaic-driven heat pump systems

Solar photovoltaic-driven heat pump systems hold significant potential for greenhouse heating, cooling, and waste heat recovery applications. However, their performance is influenced by multiple interacting factors, including environmental conditions, operational characteristics, and load demand properties. These factors exhibit complex nonlinear relationships, which often cause the system to operate in an unstable state over time. Traditional physical modeling approaches face notable limitations in handling such nonlinearities, making it challenging to accurately predict system dynamics and to quantify performance uncertainty. With the rapid advancement of artificial intelligence technologies, deep learning has emerged as an effective tool to address these challenges. This study proposes a deep learning-based Bayesian inference method for performance prediction and uncertainty quantification in photovoltaic heat pump systems. The approach combines deep learning models—MLP, CNN, and ResNet—for predicting key performance parameters, followed by Bayesian inference to estimate their posterior distributions. Input features include ambient temperature, water temperature, and the inlet and outlet pressures and temperatures of the compressor, while output parameters cover power consumption, cooling capacity, heating capacity, and COP. The results demonstrate that the ResNet-MCMC model can deliver accurate predictions even with a dataset size of 1,000, while effectively quantifying the uncertainties associated with cooling capacity, heating capacity, and power consumption. The corresponding R2 values reach 0.997, 0.992, and 0.981, respectively. Moreover, the ResNet-MCMC model achieves reductions of up to 36 % in nRMSE and 40.4 % in RMSE compared to the ResNet model. Overall, his study offers a reliable reference for performance prediction and uncertainty quantification of PV-driven heat pump systems.