Transmission Line Fault Classification and Location Using Wavelets and Machine Learning

When a fault occurs on a high-voltage transmission line, the protection and the maintenance crew need to know two things quickly: which phases are involved (fault type) and where the fault is (distance). This project finds both from the three phase voltages and currents recorded at ONE end of a 400 kV, 200 km line: 10 fault types (AG, BG, CG, AB, BC, CA, ABG, BCG, CAG, ABC) plus no fault, and the fault distance in km. ALL DATA ARE SIMULATED: a three-phase line model written for the project (20 coupled pi-sections between two Thevenin sources with load flow, solved exactly, recorder anti-aliasing filter and measurement noise) generates 5,000 fault records, and is cross-checked with pandapower (IEC 60909 short-circuit currents, 0.0006 %) and an ngspice netlist (0.0006-0.0015 % RMS). The classical relay methods - phase-selection rules and the one-ended reactance and Takagi locators - are compared with db4 wavelet (DWT) energy features and Random Forest / SVM / gradient boosting / MLP, a 1-D CNN on the raw post-fault cycle and a hybrid locator (Takagi estimate + machine-learning correction), on a random split and on held-out fault resistances and inception angles. A GNU Octave cross-check (run), a MATLAB version (written, not run) and a web application with five databases are included. Everything runs on a student laptop CPU in about 7-10 minutes.

Category: Electrical and Electronics Engineering (Power Systems Protection, Transmission Lines, Fault Analysis, Wavelet Transform, Signal Processing, Machine Learning, Deep Learning, pandapower, ngspice, GNU Octave, MATLAB, Python)

Electrical & Electronics Engineering Project EE08 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.