Noise Reduction and Confidence Level Analysis in MMG-based Transient Fault Location

A multi-resolution morphological gradient (MMG) filtering technique is applied successfully into the transient fault locator. The performance of MMG will inevitably deteriorate when various disturbances are imposed on the transient signals. These disturbances can be considered as noise. In this paper, median filter is applied to fault location for noise reduction in the transient signals and to improve the performance of accurate fault location using MMG. The confidence analysis is introduced to discuss the reliabilities of fault location under different signal-noise-ratios in noisy environments. Since the accuracy of fault location relies on the accurate extraction of the maxima of MMG, a confidence level index (CLI) is defined based on the analysis of these maxima. A hypothesis testing is used to determine whether an assertion of maxima of time-tags is reasonable. By using the CLI, the reliability of fault location is able to be measured. The analysis of receiver operating characteristic (ROC) curves with respect to CLI has shown that the median filter can improve the discrimination of accurate and inaccurate locations and the performance of the MMG detection scheme has been improved therefore