The system-level fault prediction of a closed-loop single-ended primary inductor converter (SEPIC) under variable operating conditions is not only affected by the fault mode, but also affected by changes in operating conditions. Aiming at this problem, a new system-level fault characteristic parameter (S-FCP) for the performance degradation state of the whole converter under variable operating conditions is proposed. A based gaussian process regression (GPR) method is proposed to predict the degradation state of S-FCP. Firstly, the influence of key component degradation of closed-loop SEPIC converter on system-level parameters under different operating conditions is studied. Secondly, the system-level parameters that are sensitive to degradation of all key components and have regular degradation trends are selected as the S-FCP of the closed-loop SEPIC converter, and the S-FCP under rated conditions is obtained by multivariate least squares regression. Finally, based on GPR, the fault prediction of S-FCP is carried out to realize the fault prediction of closed-loop SEPIC converter under variable operating conditions. Experimental results demonstrate the feasibility and effectiveness of the S-FCP extraction method and prediction framework.