Hybridizing UFO with Other ML Tools to Locate Faults by Just Knowing Relay Operating Times

Abstract

Universal functions originator (UFO) is a new machine learning (ML) tool that can find relationships between responses and predictors and then automatically formulate them as mathematical equations using the required number of analytic functions and arithmetic operators. Since it was introduced in the literature there is still an urgent question about whether it is worthwhile to hybridize it with other ML tools, such as linear regression (LR), nonlinear regression (NLR), support vector machine (SVM), and artificial neural network (ANN). This study is the first attempt to hybridize UFO, as a universal transformation unit (UTU), with the preceding ML tools. The goal here is to let UTU take care of the non-linearity issue of the dataset before being sent to other ML tools. These new hybrid computing systems are applied to locate three-phase (3ϕ) faults in an electric power network by utilizing the operating times measured from the two-end numerical directional overcurrent relays (DOCRs) of a faulty line. The results show that the hybrid approaches are viable where their estimations are much better than those obtained by the classical ML tools. This study proves that the strong side of UFO can be integrated with others to have superior computing systems.