Support vector regression based S-transform for prediction of distribution network failure

Many of the electrical systems throughout the world are experiencing problems with aging insulation. When an insulation system fails, the results are usually catastrophic. Insulation failure can cause sustained interruption which can cause substantial financial loses due to lost production and damage to expensive equipment. These losses can amount to thousands of ringgit (RM) per hour. With the ability to predict when a possible insulation failure will occur, power utility's engineer will be able to reduce customers lost profit opportunities. In this paper a new technique to predict the occurrences of a network failure is proposed. This new technique, which comprise of the S-transform and support vector regression (SVR) will analyze a set of power quality measurement data and predict the potential occurrences of possible insulation failure in the supply systems. Several studies were performed to evaluate the performance of the new technique. Overall, the results of the studies showed that the new technique is able to predict the occurrences of incipient fault with an accuracy of 100%.