The future of electrical power system operation and control is towards increased use of intelligent devices thus making possible the smart grid. Transient stability analysis is currently performed using offline simulations to determine critical clearing times etcetera. There is no established method of monitoring the load angle of a generator hence there is no capability to validate offline simulation results on the physical system. The research work presented herein shows results obtained by using a neural network to estimate the load angle of a generator. The method is tested on a generator simulated on the Real-Time Digital Simulator (RTDS) and CT and VT signals are sent to a physical PMU whose outputs are sent to a neural network in MATLAB for load angle estimation. It will be shown that an adequately trained neural network can estimate the load angle of a generator.
Muzothule Kubheka, John Van Coller • University of the Witwatersrand