A Novel Approach to Predict Unplanned Interruptions in Medium-Voltage Power Distribution Grids using Data Mining in Tabriz City

In this research, an approach based on data mining is employed to investigate big data in order to characterize hidden, meaningful patterns and rules among data. Due to availability of historic interruption recording system in Tabriz City, Iran, and thus access to large amounts of historical data of the Eastern Azerbaijan Province, Iran, it is tried in this paper to predict the future interruptions in Eastern Azerbaijan Distribution System (EADS) via a novel data mining-based approach. According to the collected data, future fault occurrences are forecasted monthly in one year time horizon in terms of transient and unknown faults, insulator faults, jumper faults, fallen object faults (i.e., trees) on power lines, etc. Finally, predictions are made for upcoming six months considering real values of faults. The obtained results of this paper clearly substantiate that the data mining can be used effectively to predict distribution network interruptions with a high probability of 85% in long term period.