-

The state of the power distribution system must be estimated for determining control strategies and making decisions by a pattern recognizer system. In this system, each state is a class of measured data which describes system status in that moment of the time.
In this paper two classifier systems are designed and compared using Radial Basis Function (RBF) and Multi Layer Perception (MLP) neural networks. It is shown that MLP performs better than RBF in this’ application. Also the effect of different sorts of data distribution in classification space, scaling operations, preprocessing, normalization, and conformal mappings on the classification space, adding noise to system input, optimum selection of error function order ( its Lebesgue norm) and maximum error reduction on learning, generalization, interpolation and extrapolation of neural networks are studied. Finally, a practical example is presented including simulation results to test performance of designed system