This study aimed to develop a building-integrated photovoltaic (BIPV) device and optimal control methods that increase the photovoltaic (PV) efficiency and visual comfort of the indoor space. A louver-type PV-integrated shading device was suggested and an artificial neural networks (ANN) model was developed to predict PV electricity output, work plane illuminance, and daylight glare index (DGI). The slat tilt angle of the shading device was controlled to maximize PV electricity output based on three different strategies: one without visual comfort constraints, and the other two with visual comfort constraints: work plane illuminance and DGI. Optimal tilt angle was calculated using predictions of the ANN. Experiments were conducted to verify the system modeling and to evaluate the performance of the shading device. Experiment results revealed that the ANN model successfully predicted the PV output, work plane illuminance, and DGI. The PV-integrated shading device was more efficient in producing electricity than the conventional wall-mount PV systems, the control method without visual comfort constraints was most efficient in generating electricity than the other two with such constraints, and excluding the constraints resulted in less comfortable visual environment and reduced energy benefit. From the results analysis, it can be concluded that based on the accurate predictions, the PV-integrated shading device controlled using the proposed methods produced more electricity compared to the wall-mount counterpart.