The use of renewable energy in buildings is gaining traction recently, while there is growing interest in photovoltaic thermal (PVT) systems that utilize photovoltaics and solar thermal energy at the same time. However, there are limitations in introducing PVT systems to actual buildings, as many architects have concerns about how much space they would occupy in buildings and how they would look with buildings. In this regard, this study developed the Curtain Wall-Type Liquid-type Photovoltaic Thermal System (CW-PVT), which can be integrated as part of a building and used on the wall of a building. Furthermore, to analyze the system''s performance, this study built a real-scale experimental plant and conducted an experiment under actual outdoor conditions. The experiment was conducted for the performance of single and serial layout CW.L-PVTs, and measurement instruments were used to collect inlet and outlet water temperature, module surface temperature, electricity production, and weather data. In addition, to control hot water and electricity produced from CW.L-PVTs, this study suggested the energy production prediction algorithm and optimal control algorithm by using the multiple linear regression based on the measured data. This study found the following results: When inlet water temperature was 10℃, single CW.L-PVTs showed 0.43 in the coefficient of thermal efficiency, 2.71 in the coefficient of thermal loss, 0.1 in the coefficient of electricity efficiency, and 2.56 in the coefficient of electricity loss. Meanwhile, CW.L-PVTs with a serial layout out 3 modules showed 0.51 in the coefficient of thermal efficiency, -5.6 in the coefficient of thermal loss, 0.12 in the coefficient of electricity efficiency, and 0.15 in the coefficient of electricity loss. Based on the measured data, this study suggested the thermal and electrical energy production prediction algorithms. The root-mean-square deviation (RMSE) of the prediction models was 0.414 and 6.793, respectively. Moreover, this study proposed and developed the optimal control algorithm, which predicts inlet water temperature depending on weather data and outlet water temperature, and the RMSE of the model was calculated as 0.599, indicating that both the prediction and optical control models had great Accuracy.