Improving Performance of Wilcoxon Criterion under Solving Problems with Typical Singularities in Measurements

Abstract

The paper discusses the approach that supports measurement data treatment on significant events from electrical consumption load profile. The problem of detecting significant deviations in the accounting of electricity consumption is under consideration. The data available from power meters and intelligent sensors is used to classify the events in power supply process. This information is of high importance and can be utilized to predict violations. It leads to rapid decision-making and allows avoiding the penalties on behalf of the power distribution utilities. The data received from smart meters is usually represented in the form of so-called “load profiles”. The article introduces an approach for improving the solvability of the data discrimination problem through the non-parametric Wilcoxon sign ranks criterion under the cases with singularities in compared data samples of the type “vertical offset” and “scale”. The modified Wilcoxon criterion is considered as a tool for identifying deviations in the conduct of the technological process, which do not exceed the threshold values, but have a significant impact on the technical and economic indicators. The proposed method of identifying critical events, which uses the modified Wilcoxon criterion, makes it possible to determine both the fact of the presence of an event and to provide their identification. It is proved that the power of the proposed modified Wilcoxon test does not degrade comparing with the traditional Wilcoxon test. The executed example of processing the power consumption data of the petrochemical industrial enterprise allowed testing the effect of using the modified Wilcoxon criterion as a part of the computational procedure for the developed algorithm.