Denoising Transformer Partial Discharge Signals Using Wavelet Analysis

In this course project, I implemented wavelet analysis to process transformer partial discharge signals with excellent results. Specifically, wavelet analysis is a signal processing technique that decomposes signals into different frequency sub-signals, enabling better understanding and processing of signal components. The implementation involved applying discrete wavelet transform (DWT) to decompose the original signal into approximation and detail coefficients across multiple resolution levels. Using MATLAB's wavelet toolbox, I employed functions like wavedec for decomposition and wden for threshold-based denoising. The key algorithm involved selecting appropriate wavelet families (Daubechies wavelets were tested) and determining optimal thresholding rules (soft thresholding with level-dependent thresholds) to remove noise while preserving critical discharge pulse characteristics. After decomposition, I applied denoising techniques to the sub-signals by thresholding the detail coefficients at each level, effectively reducing noise interference. The denoised signal was then reconstructed using waverec function, resulting in a clearer output compared to the original signal. The processed signal demonstrated enhanced readability and interpretability with improved signal-to-noise ratio. Overall, wavelet analysis proves to be an effective methodology for processing partial discharge signals, providing valuable support for transformer monitoring and maintenance operations. The implementation successfully separated noise components from actual discharge patterns, making it particularly suitable for real-world industrial applications where signal clarity is crucial for accurate fault diagnosis.