Due to the problem that the abnormal telemetry data of meteorological satellites in the threshold rang cannot be detected, this paper proposes the Dynamic Kernel Principal Component Analysis (DKPCA) method to detect the abnormity of satellite telemetry. The telemetry data during the normal operation are collected, and the DKPCA model is established to obtain the control limit of the variable. Then the current telemetry data are detected to determine whether there is a fault. The method solves the problem of sequence correlation between observed data by the dynamic representation of initial data. The introduced DKPCA can transform the nonlinear problem of complex telemetry data into linear problems. The effect is verified by the data of the meteorological satellite FY3C in orbit. The results show that the observed data under normal conditions can be used to detect abnormal satellite telemetry, and the fault omission can be avoided effectively.