A residual life prediction method of a specific satellite attitude control system is explored on the basis of reconfigurable analysis. In order to investigate the influence caused by multi-state failures of the system, the concept of degraded state or partial failure is introduced and different states are defined according to the severity of the failure. There are transitions between states, and their conditional probability can be calculated through Kaplan-Meier estimator and then fitted as parametric Weibull distribution via the maximum likelihood estimation (MLE) approach. The states and the transitions constitute the Petri net model of the satellite attitude control system, on which the residual life can be predicted through a large sample of simulations both off-line and on-line. The simulation results verify the effectiveness of the proposed method.