Improving the Reliability of Pumped-Storage Power Plants in the Operational Phases using Data Mining Algorithms

Abstract

The sudden breakdowns of the power generation facilities caused a serious concern about the stable operation of the power system. However, since electricity system in South Korea is not connected to neighboring country, it is necessary to devise methods to counteract emergency such as unexpected failure of a power plant or sudden increase of electric power demand. In response, South Korea is operating seven pumped-storage power plants whose capacity is equal to that of about five nuclear power plants, accounts for 5.9% of the total power capacity of South Korea. The pumped turbine is a machine that operates through forward and reverse rotations to store energy and produce electric power. It requires high reliability for rapidly generating electricity on load. The losses caused by the shutdown of the power plants due to turbine failures, and the costs of their recovery, are enormous. Thus, improving the reliability of a pumped-storage power plant will have a significant influence on the national power system. In this regard, this research seeks to increase the reliability of the pumped storage power plant by identifying the problematic factors. This study proposes models of Case Based Reasoning (CBR) for eliminating possible problematic factors in pumped storage power plants before operation phase through predicting and checking failure. The development of the model involves analyzing operational data from 36 pumped-storage power plants; conducting decision tree analysis on 19 problems in order to select significant problems; and developing CBR models which can predict and check possible problems to improve the reliability of the pumped-storage power plant.