Applying Recurrent Neural Networks to Static VAR Compensator

The purpose of this paper is to use recurrent neural networks (RNN) to control and adjust the switching of thyristor in a Static VAR Compensator (SVC) to adjust the voltage. In the new control scheme, instead of just using a feedback loop, same as neural network several feedback loop conventional recurrent are employed. In the proposed controller model RNN provides a sample of the connected system, and its output provides part of input for the RNN controller, then sends the control signals to SVC system. Three types of non-linear modes were selected for testing new control system operation for voltage regulation in IEEE Std 519-1992. The test consists of three-phase power system fault that opens one of the transmission lines in a transitional two-track system and suddenly changes in load demand. The results show that the proposed control system is able to adjust voltage in desirable range.