최민희(Min-Hee Choi) ; 정남준(Nam-Joon Jung) ; 이규철(Kyu-Chul Lee) ; 정재성(Jae-Sung Jeong) ; 서인용(In-Young Seo)
Damage to the power system caused by natural disasters, including typhoons, is gradually increasing. The amount of the power outage caused by major typhoons shows 1.25 million households by “Rusa” in 2002, 1.44 million by “Maemi” in 2003, 1.68 million by “Kompasu” in 2010, 1.93 million by “Bolaven” in 2012 and 0.25 million by “Chaba” in 2016. Power companies are striving to establish an integrated system and simulators to predict power facility damage by natural disasters in advance and to establish a rapid response system in case of damage. In this paper, we developed the power facility damage prediction algorithm applied artificial neural network (ANN) for 6 kinds of natural disasters such as typhoon, strong wind, heavy rain, heavy snow, cold wave and heat wave. The algorithm consists of three phases: ① the establishment of big data by extracting meteorological data from the Automatic Weather System from 2007 to 2018, ② the analysis of the correlation between the power failures and the weather conditions(such as wind speed, rainfall, etc.) and ③ the evaluation of damage prediction algorithms using the ANN. In particular, comparisons and analyses with the Linear Regression(REG) algorithm were performed to assess the accuracy of the ANN algorithm.
This algorithm was applied to Typhoon “Chaba” in 2016 to predict the failure of electric wires and Cut Out Switch (COS) in Seogwipo. The prediction error(MAE) of the ANN is 0.127, which is better than the performance of the REG.