<p>Capacitive voltage transformer (CVT) is an essential power measurement equipment in the power grid, which generates errors in long-term operation. Therefore, it is necessary to quantify the measurement performance of CVT and predict its measurement deterioration trend. This study proposes a measurement performance index to characterize the ratio error of CVT and a trend prediction method for CVT measurement deterioration based on double Gaussian model-KAN fusion. First, approximation and detail coefficients are formed after multilayer wavelet transform on the secondary side voltage of CVT. The maximum approximate coefficient is selected from the approximate coefficients, and the State of Performance (SOP) representation ratio error of CVT is calculated using the maximum approximate coefficient. Then, a Variational Modal Decomposition Mean Difference (VMD-MD) method is proposed to decompose the SOP sequence of CVT in multiple layers. The residual decomposed from the SOP sequence is used to characterize the deterioration trend of SOP, and the double Gaussian model is used to model and predict it. The Intrinsic Mode Functions (IMFs) decomposed from the SOP sequence are used to characterize the deterioration fluctuation of SOP, and the KAN algorithm is used to predict it. Finally, all the predicted results are added to represent the deterioration trend of CVT. Using CVTs and SWCVT-3 CVT online test system of China Electric Power Research Institute, the three-phase voltage data with increasing ratio error are collected, and the proposed double Gaussian model-KAN fusion method is tested. During the experiment, the ratio error of CVT was characterized effectively by SOP, and the proposed double Gaussian model-KAN fusion method could accurately predict the CVT SOP deterioration trend.</p>