Harmonic Detection in Power Electronics Converters Using Machine Learning

Ensuring precise classification of a Power Quality Disturbance (PQD) signal is crucial for maintaining the safety of power system networks. Non-linear loads introduce harmonics into the system, causing distortion in voltage and current signals. This study introduces a method for identifying power quality issues by decomposing the voltage waveform in the frequency domain and applying Kernel Support Vector Machines (SVM) to the preprocessed voltage data. The research compares the performance of AI-based classification of power quality events using time-domain data and data preprocessed with the Fourier transform, followed by machine learning techniques on an optimized model to evaluate its accuracy. Simulation results indicate that Kernel-based SVM outperforms traditional probabilistic algorithms in detecting harmonics in PQD signals.