Smart and effective shading blind Control can cope with dynamic daylight environment changes and continuously create a visually comfortable visual daylight environment. Existing control systems have insufficient dynamic prediction and smart decision-making capabilities for indoor daylight environments due to the unclear influence mechanism of indoor daylight environment visual comfort indicators. This research combines machine learning techniques and the dynamic programming algorithm in reinforcement learning to construct an indoor daylight environment dynamic prediction model and an intelligent agent, so as to solve the bottlenecks of insufficient nonlinear fitting and insufficient Control accuracy in indoor daylight environment dynamic prediction. It realizes the effective Control of intelligent shading blinds in terms of comfort performance.