Abstract: The practical mathematical model for gravity field recovery using satellite average accelerations derived from SST-HL mode is discussed.As the numerical differentiation will rapidly amplify the high-frequency noise in orbit-derived accelerations,the filtering method and specific data processing scheme for suppressing the high-frequency noise of satellite average accelerations are presented.As a test,an Earth’s gravity field model WHU-GOCE-SST01S（RE） up to degree and order 100 is recovered with Kaula regularization constrain from 61-day GOCE kinematic orbits and accelerometer data.The result shows that the total accuracy of WHU-GOCE-SST01S（RE） is more accurate than EIGEN-CHAMP03S model and GOCONSGCF2TIMR1 model,which validates the effectiveness of the proposed data processing method.