In order to solve the problem that the efficiency of high temperature adiabatic compressed air energy storage system is difficult to predict and the discharge capacity of the system is uncertain because the system often operates under off-design conditions, an intelligent control system for high temperature adiabatic compressed air energy storage system based on machine learning prediction is designed. According to the input power grid load command, predict the outlet pressure of the gas storage salt cavern of the system, and adjust the internal parameters such as the molten salt flow of the salt-gas heat exchanger, the water flow of the water-gas heat exchanger and the air flow of the system, so as to maximize the system efficiency of the unit. The results show that the system can effectively improve the power grid's capacity to absorb renewable energy and meet the current demand for energy saving and emission reduction.