ANN in local secondary voltage control concept

The paper presents an implementation of artificial neural networks (ANN) in a decentralized approach to secondary voltage control. Each generating unit is equipped with one local secondary voltage controller (SVQC), which are trained on a large set of power system states. Data, necessary for the ANN controller operation are acquired locally at the controlled generator node. The ANN has been trained using the optimal power flow results as the target values. The proposed ANN secondary voltage control concept was tested on the 30-bus New England test system. It maintained a suboptimal power system voltage profile while reducing the active and reactive power system losses. Each generator was controlled individually, so the time required for training was short. Fast response of the ANN controller makes it suitable for real-time implementation. The new control concept is a promising tool for generators to maintain a suboptimal voltage profile in a deregulated environment.