The air conditioning system is one of the most power-intensive pieces of electrical equipment in an office building. In some large and medium-sized cities, the air conditioning load has accounted for 30% to 40% of peak load in summer. On the premise of guaranteeing the necessary comfort of people in the environment, it is of practical significance to study the optimal energy-saving control of frequency conversion air-conditioning frequency for promoting energy-saving and emission reduction and building a conservation-oriented society. Using DIgSILENT simulation software, a frequency conversion air conditioning model based on an electromagnetic transient is built. This includes the frequency converter, motor, compressor and heat exchange models. Due to the complexity of the algorithm, there is low precision and large error in previous frequency control methods of frequency conversion air conditioning. Therefore, starting from the compressor frequency controller, a frequency controller based on Extended State Observer (ESO) nonlinear optimal control is designed under the traditional PI control mode. The simulation results show that frequency control based on ESO and nonlinear optimal control can improve temperature control precision and reduce energy loss. This work is supported by National Natural Science Foundation of China (No. 60971077) and Science and Technology Project of the Headquarter of State Grid Corporation of China (Techniques and Application Research of Load Modeling Considering High-proportion Electrical and Electronic Equipment).