Discussion on “A Wavelet-neuro-fuzzy Combined Approach for Digital Relaying of Transmission Line Faults”

Abstract

The use of wavelet-neuro-fuzzy intelligence methodology for depicting digital relaying of transmission line faults is a timely approach, which should be elaborated more with future research along the same line. Power system simulation work by MATLAB software is a very useful tool, but without plausible reasoning, it may lead to mechanical achievements in terms of numerical analysis but might be short of sight when linguistic bases are concerned. It seems that the authors did not consider some basic conceptions underlying the fuzzy logic (FL) principle in their proposed model. They mentioned that usually membership functions (MFs) are chosen somewhat arbitrarily and are fixed in nature, which is not quite right. None of the MFs are fixed in nature, but they are flexible depending on the degree of expert view. They misused some points, especially concerning the MFs, which are fully contrary to the philosophy of FL. We would like to present the following important subjects.1) The authors made a terminological mistake in the their article by stating their fuzzy inference system (FIS) editor as the "Mamadani" approach, which should be the "Mamdani" approach, similar to what is available in the literature [1]. 2) On p. 1390 at the beginning of the last paragraph, it is stated that "Usually S a , S b and S c decreases as the fault distance increases from the source end, as represented in Figures 4(a), 4(b), 4(c) through triangular MF with a decreasing trend." The "decreasing trend" in this statement does not have any relation to the shape of the MF, as implied by the authors. Monotonically increasing and decreasing limbs of MFs are the requirement of normal fuzzy sets. 3) In FL philosophy, the most left and right MFs should have a membership degree equal to 1 [2, 3]. However, the rightmost MF, namely H for the input variable, S a has decreasing limbs toward "high" values, which is fully contrary to FL principle [3]. Therefore, MF H for S a is linguistically incorrect according to FL philosophy. Similar flaws could, of course, be noticed for the MFs of the input variables S b and S c and output variable D. In order to alleviate such situations, it is necessary to have either the rightmost limb of the H fuzzy set equal to 1, or there should be another MF to the right of H MF with the right limb equal to 1.