Data mining-based cause identification of momentary outages in power distribution systems

Electric power distribution systems face outages that prevent them from serving customers. Short-term outages are known as momentary outages, and their causes are not usually recorded in the outage dataset. While, frequent occurrences of momentary outages may lead to a long-term permanent outage, which can significantly reduce system reliability. Unlike previous works which focused on permanent outage diagnosing and prediction, this paper proposes data-mining based approaches to identify the most probable momentary outages' causes. To achieve this goal, the outage dataset, sub-transmission substation load, and weather historical data are processed and integrated. Then, association rules that describe the antecedents leading to different permanent outages' and momentary outages' causes are derived by using the Apriori algorithm. The frequent itemsets of momentary outages are also obtained. Based on momentary outage rules and frequent itemsets, two procedures are proposed to find similarities between permanent and momentary outages to identify the most probable causes of momentary outages. Finding the cause of momentary outages, the operator can reduce the probability of permanent outage occurrences. Results of applying the proposed approaches on real data of a test distribution system show that expected energy not supplied of the distribution system can be decreased by more than 18%.

Automated summary

This paper proposes a data-mining based approach to identify the most probable causes of momentary outages in electric power distribution systems. By analyzing outage dataset, load, and weather historical data, association rules and frequent itemsets are derived to find similarities with permanent outages, thereby reducing the probability of permanent outages.