Probabilistic Neural Network Procedure for Classification and Forecasting Voltage Sags on Distribution System

The paper describes a procedure for the automatic analysis of monitored voltage waveforms in distribution systems, the classification of possible voltage sags with identification of their origin and the forecasting of each type of expected occurrences in a given time period. A Probabilistic Neural Network (PNN) method is employed in order to classify sags according both to their origin and severity. Application of the procedure would allow industrial users to predict economic losses due to poor quality of the supply and utilities to anticipate possible customers complaints and take suitable counter-measures.