Multi-Stage Partial Discharge Signal Denoising Algorithm Based on SVD-ICEEMDAN

Partial Discharge (PD) detection serves as a crucial method for assessing the insulation performance of highvoltage electrical equipment. Accurate and timely detection of PD signals can help to predict potential insulation faults and prevent large-scale power outages. However, during the actual process of maintenance, the presence of one or more types of noise usually makes it hard for operators to clearly identify PD signals. Meanwhile, traditional denoising algorithms, such as Wavelet Transform (WT), Singular Value Decomposition (SVD), etc., are only applicable to suppress a specific noise, and have low robustness against superimposed multi-noise. Detecting PD signals is extremely challenging, requiring an algorithm that can effectively remove multiple superimposed noises. Therefore, we propose a multi-stage denoising algorithm based on SVD and Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN). SVD can suppress periodic narrow-band noise in the signal, preserving the key features of the original signal. ICEEMDAN achieves accurate signal decomposition and eliminates white noise interference by suppressing mode mixing through its Intrinsic Mode Function (IMF). Simulation results demonstrate that compared with traditional denoising algorithms, the multistage denoising algorithm based on SVD-ICEEMDAN can significantly suppress the superimposed interference in PD signals and improve the Signal-to-Noise Ratio (SNR) of PD signals.