Substation automation using ART 1 based neural network

In this paper we describe the design, analysis and implementation of online security assessment in substations on an alternative computational paradigm which is inherently parallel and distributed. The massively parallel and distributed nature of artificial neural network (ANN) is well suited in this respect. A site survey was conducted in a 110 KV substation of Kerala State Electricity Board, Palakkad, Kerala. The basic system problems had been studied and tabulated. The different fault conditions had been tabulated and coded. Coding was done using 7 bit pattern. These codes were used to train the adaptive resonance theory 1 based online training and recognition system developed in the Centre for Artificial intelligence and Neural Network (CANN), NSS College of Engineering, Palakkad. The weight files were compressed by the technique developed by the authors and stored. The ART 1 based package uses 70 input nodes and 125 recognition layer nodes. The recognition layer nodes is a function of pattern set corresponding to the system problems