The relevant factors causing the steady-state error of capacitor voltage transformers（CVTs） are analyzed. A parametric model is developed according to the structure and parameters of CVT. The effectiveness of the model is verified by comparing the error simulation results with experimental data of the accuracy test. For the insulation on-line monitoring， a new method based on this model is developed to correct the steady-state error of CVT， in which the voltage signals of the measuring winding and current signals of the secondary winding with load are collected and the nonlinear least-squares fitting algorithm based on Levenberg-Marquardt algorithm is adopted to obtain the fundamental components. The equations of the model are utilized to determine the primary voltage of CVT. The analysis results indicate that the errors in ratio and phase due to the changes in the frequency， size of load and power factor can be markedly reduced， providing secondary monitoring facilities with high precision voltage signals.