An Enhanced K-Means Clustering Algorithm for Pattern Discovery in Big Data Analysis of 3-Phase Electrical Quantities

Abstract

The Unila Internet of Things Research Group (UIRG) was developed online monitoring of power distribution system based on Internet of Things (IoT) technology on Department of Electrical Engineering University of Lampung (Unila), has been running for several months, this system monitored electrical quantities of 3-phase main distribution panel of H-building. The measurement system involve multiple sensors such current sensors and voltage sensors, the measurement data stored in to database server and shown the information in a real-time through a web-based application.Main objective of this research was to capture, analyze, and identified the knowledge pattern of electrical quantities data measurements, using Cross-Industry Standard Process for Data Mining (CRISP-DM) data mining framework, for helping the stake holders to continuous improvement of the quality of electricity services, the initial research limited to total 770847 electrical quantities recorded data that save on database system, since 1 September - 31 October 2018, the dataset consist of 21 attribute electrical quantities such as; voltage, current, power factor values, energy consumption, frequency, on H building 3-Phase main panel control.Rapidminer as leading application on knowledge discovery application was used to analyze the big data, K-Mean cluster algorithm implemented to identify the data pattern, the result indicated that 3-Phase load was unbalanced, and Phase-0 was the most utilized phase, based on from total 5 cluster analysis result.