An adaptive neuro fuzzy power system stabilizer for damping inter-area oscillations in power systems

Abstract

An adaptive neuro-fuzzy inference system (ANFIS) based PSS is proposed in this paper. The controller is essentially divided into two sub-systems, a recursive least square identifier for the generator and an adaptive neuro fuzzy PSS to damp the oscillations. The PSS is coupled to a single machine in every area and the parameters of this PSS are tuned online in order to minimize a cost function. The cost function consists of a summation of terms, in which each term is made up of the square of the difference in speed between the machine to which the PSS is connected and another machine in that same area (the number of terms equal the number of machines in that area excluding the machine installed with a PSS). The PSS is trained to reduce the speed difference between machines in every area while helping to reduce inter area oscillations. The proposed technique is illustrated on a 2 area 4-machine 13 bus system. This ANFIS PSS showed satisfactory performance under severe faulting conditions, where a three-phase fault applied to a line, was cleared after a extended period of time. The conventional PSS and the ANFIS using the original cost function (consisting of just the square of the speed difference of the generator installed with the PSS) failed to perform under such conditions.