Abstract: Resistive current is a key parameter to judge the working state of a metal oxide arrester (MOA). The resistive current of a MOA is very small when the power system is working normally, and the measured resistive current is prone to be affected by the interference of high-frequency noise, white noise and random pulses, leading to the false alarms of alarming devices. At present, none of the existing denoising methods can completely eliminate the influence of the above interference on the resistive current of MOA. Therefore, this paper proposes a new method for eliminating the interference of resistive current of MOA based on adaptive variational mode decomposition (A-VMD) and adaptive singular value decomposition (A-SVD). Firstly, by sequentially changing the secondary penalty factor and the decomposition layer, and with the energy and loss indicators to measure the effect of VMD decomposition, the optimal parameters of decomposition layers and secondary penalty factor are searched out. Secondly, the A-SVD is used to eliminate the residual white noise in the resistive current after preliminary denoising by A-VMD, which provides a reliable basis for judging the insulating state of a MOA. The effectiveness of the proposed method is verified by simulation and measured data, and the processing results meet the requirements of actual projects.