In this paper, the expert system is introduced in

order to detect and classify commonly power quality

disturbances. This system is using learning vector quantization

artificial neural networks. Clustering method named fuzzy

c-mean is also utilized to initialize weight vector of first hidden

layer. It can mitigate the disadvantage of LVQ ANN. The

proposed system employs wavelet decomposition coefficients for

extracting of deviated signals features. The determined feature

vector is derived from Standard Deviation of 10-level

decomposition detail coefficients. For the purpose of having

efficient network, just 3 characteristic points among 10 points

have been used, that leads to make networks training much