An approach to secondary voltage control: a solution with local ANN controllers

The paper presents a completely decentralized approach to secondary voltage control. Local secondary voltage controllers (SVQC) based on the artificial neural networks (ANN) are attached to each generating unit's excitation controller. For operation, the SVQC requires only local bus voltage and reactive power measurements obtained on the lines connected to that bus. The ANN SVQCs are trained on a large set of power system states. The optimal power flows are used as performance target in the ANN training process. The proposed secondary voltage control concept employing an artificial neural network was derived for the generators of the modified New England test system. The tests showed that voltage control kept the voltage profile close to the required one. Furthermore, it reduced the active and reactive power system losses, thus resembling the optimal power flow solution.