Abstract:The powerhouse of large-scale hydropower station with long diversion is arranged underground, and its biggest feature is that it has large-scale cave groups, such as Baihetan Hydropower Station, wudongde hydropower station Hydropower Station and Xiluodu Hydropower Station, which undertake the transportation, emergency escape, ventilation, production and domestic water, power delivery and other channels of power station operation and maintenance management. The tunnels in the system are staggered and connected vertically and horizontally to form a complicated underground network, so the scientific management of cave groups is to ensure the safe and stable production of power station. Because there is no sunlight in the cavern group and it needs to run around the clock, the lighting management of the cavern group is extremely prominent. In order to ensure the orderly development of all work in the cavern group, and considering the rational distribution of resources and the non-waste of energy, strategies such as hierarchical and zoning control, intelligent prediction combined with reinforcement learning, multi-platform interconnection, and cloud platform data management and control are put forward to realize efficient and intelligent control of the cavern group lighting system of hydropower station, and solve the difficulties of management and control and energy waste of large-scale cavern group lighting system of hydropower station..