Abstract: There is a great difference between control costs of primary frequency regulation and secondary frequency regulation of thermal units, so it is necessary to respectively evaluate their control performance to provide a basis for reasonable compensation or payment. Primary and secondary frequency regulation response power are the basic parameters for the control performance evaluation of the units, but they are interwoven with the active power of the unit and cannot be used directly, resulting in the lack of the evaluation basis. Based on the time-frequency difference of the primary and secondary frequency regulation power time series obtained by the numerical analysis method, an iterative separation algorithm is proposed by the least square fitting of inertia proportional and delay parameter and Hodrick-Prescott trend feature extraction function. The response power of the system frequency response model and the secondary frequency regulation instruction are used as the initial iteration values of the primary and secondary frequency regulation power, respectively, which can further improve the data separation effect. The effectiveness and feasibility of the proposed method are verified by a simulation case.