During the operation of the power system

uncertain factors such as high load conditions and strong electromagnetic interfe-rence seriously interfere with the effective collection of insulator leakage current signals

thereby affecting the accuracy of fault diagnosis.To address the above issues

it proposes and constructs a diagnostic system.This system takes the support vector machine(SVM) as its core and achieves functional improvement through the collaborative optimization of Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(ICEEMDAN) and sparrow search algorithm.This method first uses ICEEMDAN to perform adaptive decomposition processing on the leakage current signal

combines the correlation coefficient and kurtosis index to screen the key intrinsic mode function components

and realizes the fault feature extraction through signal reconstruction.Further

it adopts the spa-rrow search algorithm to globally optimize the core parameters of the SVM

constructs and a multi-classification diagnostic model with strong generalization ability

effectively improving the recognition accuracy of the insulator deterioration state under complex working conditions.The test results show that this model demonstrates excellent accuracy and robustness in the identification of insulator deterioration status.The optimal identification accuracy rate reaches 96.5%

and the average accuracy rate exceeds 93.5%

providing a reliable technical solution for the early detection and preventive maintenance of insulator deterioration in the power system.