In order to overcome the difficulty to predict the system efficiency of high temperature adiabatic compressed air energy storage system and the uncertainty of discharge capacity of the system, an intelligent control system for the energy storage system based on machine learning prediction is designed. According to the input grid load command, the system predicts the outlet pressure of the gas storage salt cavern, and adjusts the internal parameters, including molten salt flow of the system salt gas heat exchanger, water flow and air flow of the water gas heat exchanger, so as to reach the maximum system efficiency of the unit. The device is compact and efficient, which can effectively improve the capacity of the power grid to absorb renewable energy and meet the current demand for energy conservation and carbon emission reduction.