Abstract: In order to solve the problem of how to effectively participate in bidding declaration according to the reasonable price prediction interval, a price prediction method of frequency modulation auxiliary service market based on integrated learning is proposed, in order to increase the investment income of frequency modulation market participants. Firstly, the principles of machine learning methods, such as random forest, XGBoost and LightGBM are given, and the basic architecture of multi-type algorithms for integrated learning is introduced. Secondly, considering the coupling characteristics of power fluctuations, such as meteorology, new energy, power load, delivery and frequency modulation demand, through the analysis of the boundary data published by the day-ahead electric energy market, a frequency modulation price interval prediction model based on random forest-XGBoost-LightGBM multi-model ensemble learning is established, and an interval prediction evaluation index is established. Finally, the simulation calculation is carried out with a provincial power frequency modulation auxiliary service market as an example, and the effectiveness and economic feasibility of the proposed method in the market application of frequency modulation auxiliary service market entities are verified.