This study models the relationship between the number of power distribution facilities outages and weather factors by utilizing actual power outage data in Busan and Ulsan.
The actual number of power outages of 13 branches belonging to the Busan Uls...
This study models the relationship between the number of power distribution facilities outages and weather factors by utilizing actual power outage data in Busan and Ulsan.
The actual number of power outages of 13 branches belonging to the Busan Ulsan Regional Headquarters of Korea Electric Power Corporation(KEPCO) for 5 years between 2015 and 2019 was set as dependent variables. With reference to previous studies, wind speed, precipitation and snow accumulation 3 factors were set as highly affecting power distribution facilities outages. However, due to the characteristics of Busan and Ulsan, snow accumulation occurred only twice between 2015 and 2019(2018. 1. 10, 2019 1. 31). So, The lack of statistical observations resulted in no significant probability, which made statistical interpretation impossible. Therefore, 2 factors (wind speed and precipitation) excluding snow accumulation, were set as high-impact weather factors. Also, Through Pearson's correlation analysis, the maximum instantaneous wind speed and daily precipitation were set as the final independent variables, respectively
We performed multiple linear regression using IBM SPSS statistics 26, and multiple linear regression resulted in an under-fitting model. Thus, a polynomial non-linear regression analysis using the curve fitting toolbox of MATLAB R2021a could be performed to derive a predictive model of the optimal-fitting model.
First, based on data for 2015-2019, the power outage prediction model was verified by comparing the predicted number of power outages and the actual number of power outages.
Secondly, based on data for 2020, the power outage prediction model was verified by comparing the predicted number of power outages and the actual number of power outages
The data for 2020 showed that within ±4 of (actual number of power outages - predicted number of power outages) is 351 out of 366, accounting for 95.62%, which resulted in significantly higher performance of the model.
More than -4 of (actual number of power outages - predicted number of power outages) is 11 out of 366 cases, 3.00%, which is considered to be the case where the predicted number of power outages is higher than the actual number of power outages. This is the result of electricity operators proactively preparing for disaster situations by reflecting weather forecasts such as typhoons, and it is not a big problem because enough manpower and equipment are secured from the side of electricity operators to smoothly prepare for the actual power outage.
However, More than +4 of (actual number of power outages - predicted number of power outages) is 4 out of 366 cases, 1.09%, which is considered to be the case where the actual number of power outages is higher than the predicted number of power outages. This can cause major confusion in the field as manpower and equipment are not sufficiently secured when securing personnel and equipment based on the predicted number of power outages. In other words, if the (actual number of power outages - predicted number of power outages) is more than +4, it may be difficult for the electricity operators to respond appropriately to the actual power outages, and This is a case where the threshold of the derived power outage prediction model can be seen.
Future research projects are not limited to Busan and Ulsan, but I think it is necessary to identify regional weather factors nationwide and derive a prediction model for the number of power outages suitable for weather characteristics in each region.