Analysis of the effects of load model on a single phase fault locator based on support vector machines

This paper presents an analysis of the performance of a fault locator based on support vector machines (SVM) for power distribution system, using different load models. Four load models are considered. SVM is training four times using one load model each time and was validated for all of the proposed load models. The proposed methodology is tested in a standard IEEE 34-bus power distribution system, which is subdivided into 11 zones, considering a database of 66420 of single-phase different conditions. This analysis shows that load model variation in validation, greatly affects the fault locator performance.