Machine Learning Using High-Precision Data for Fault Location

Fault location is a high priority for both researchers and utilities in the transmission industry. However, traditional methods tend to either require expensive equipment or systems that are far too slow to be effective at finding and clearing faults quickly. Therefore, this paper is written with the purpose of using machine learning, which is far more efficient once properly trained, to identify the location of faults in a grid using voltage measurements. This is done using a variety of classification and neural network algorithms, with the intention of determining the most accurate, or most efficient, algorithm available. Training data is gathered using Real Time Digital Simulation (RTDS) hardware, which is specially designed to simulate power flows with high precision, and in this paper is verified with an unmodified WECC 9 bus grid. After said training data is gathered, they are used to test the accuracy and speed of each algorithm using Monte Carlo simulations, resulting in a better picture of how useful and scalable each algorithm is in fault detection.