Application of neuro inverse controller to power system stabilizer for a multimachine power system

This paper presents a neuro inverse controller for Power System Stabilizer application in a multimachine power system. The inverse mapping of the local controlled plant is approximated by an artificial neural network named Inverse Dynamics Neural Network (IDNN), which is used as a local inverse controller. There exists a difference between the inverse of a local plant and that of whole system including the interactions between generators. The difference is defined as Inverse Dynamics Error (IDE) and is assumed to be piecewisely continuous and smooth in the time domain. Under this assumption, the IDE is modeled as a time polynomial and calculated in a network called Error Reduction Network (ERN), which consists of IDNN's in the linear combination form. Therefore, the training of ANN is needed for IDNN only in the proposed controller. Two simulations are given to show the effectiveness of the proposed controller. One simulation in a single machine infinite bus power system that has no interaction shows the effectiveness of IDNN in case of an unstable and nonminimum phase plant, the other in a three machine system that has interactions shows the effectiveness of ERN.