Power Quality Disturbance Classification Using Signal Processing and Deep Learning

Voltage sags, swells, interruptions, harmonics, flicker, transients, notches and spikes damage equipment and stop production. This project classifies 16 power-quality (PQ) classes - normal, 8 single IEEE 1159 disturbances and 7 combinations (sag + harmonics, swell + harmonics, interruption + harmonics, flicker + harmonics, flicker + sag, flicker + swell, sag + transient) - from 10 cycles (200 ms) of one voltage waveform. THE TRAINING AND TEST DATA ARE SYNTHETIC: the parametric IEEE 1159 disturbance equations of the PQ literature generate 16,000 training and 8,000 test signals (3.2 kHz, random parameters, fixed seeds), and every test signal is evaluated noiseless and at 50, 40, 30 and 20 dB SNR. A classical rule-based RMS / THD detector with the IEEE 1159 thresholds is the baseline; db4 wavelet (DWT), S-transform and time-domain PQ-index features with SVM and Random Forest and a small 1-D CNN on the raw waveform are compared with it. The deployed 1-D CNN reaches 98.38 % noiseless and 91.40 % at 20 dB (rules 82.53 / 64.45 %). A GNU Octave cross-check (run), a MATLAB version and a Scilab version (written, not run) and a web application with five databases are included. Everything runs on a student laptop CPU in about 15-20 minutes.

Category: Electrical and Electronics Engineering (Power Quality, IEEE 1159, Signal Processing, Discrete Wavelet Transform, S-Transform, Machine Learning, Deep Learning, 1-D CNN, GNU Octave, MATLAB, Scilab, Python)

Electrical & Electronics Engineering Project EE09 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.