FAULT CLASSIFICATION AND LOCATION IN MVDC SHIPBOARD POWER SYSTEMS USING EXTREME LEARNING MACHINE

Abstract

In this article, a new fault classification and location method for medium voltage direct current (MVDC) shipboard power systems is presented. Smooth and uninterrupted power supply in MVDC shipboard power systems requires an efficient fault classification and location method. A parametric technique called autoregressive (AR) signal modeling is used to extract features from the current signal for full cycle duration at the point of measurement. The AR coefficients of the modeled current signal are used as input for the fault classifier and fault locators. The proposed fault classifier and fault locators are designed using machine intelligent technique based on extreme learning machine (ELM). The proposed fault classifier and fault locator has been tested with 8640 and 6480 cases respectively with wide variation in system parameters. Test results indicate that the proposed method is simple, reliable, effective, and accurate than the existing method in fault classification and location for MVDC shipboard power systems.