RECURRENT NEURAL NETWORKS FOR PHASOR DETECTION AND ADAPTIVE IDENTIFICATION IN POWER-SYSTEM CONTROL AND PROTECTION

A multi-input multi-output (MIMO) recurrent neural network (RNN) is us ed as a versatile tool for the high-speed phasor detection and the ada ptive identification of control and protection signals in power system s, For the application as a phasor detector, a fast pseudo-gradient tr aining is performed off-line to estimate the time-invariant weights of the RNN, This network is then operated in real-time, in recall mode o nly, to behave as a nonlinear fixed-coefficient filter, For the applic ation as an adaptive identifier of nonlinear components, training is p erformed off-line for initializing the connection weights, but subsequ ently, they are continuously updated in real time, This results in an adaptive identifier suitable for detecting abrupt changes in complex n onlinear systems, Following an initial evaluation on synthetic signals , these two proposed RNN's are then validated using realistic waveform s generated from a series-compensated power system model.