Datasets Comparison

5-ECK consists of five types of energy sources: 1) electricity, 2) gas, 3) water, 4) hot water, and 5) heating. However, because gas was not collected for the full period, we used the remaining four data types. This dataset includes a total of 47,117 records observed from Jan. 2020 to Mar. 2021 with a 15-minutes time period. We transformed the time period of the energy consumption from 15 minutes to an hour consistently. Here, missing values are filled based on linear interpolation, and min-max scaling is used for normalization.

Time Series Analysis, Energy Consumption

Creative Commons Attribution 4.0 International

Four types of Energy Consumption in Korea (4-ECK-2021) Dataset This dataset includes three types of energy sources: 1) electricity, 2) water, 3) hot water, and heating. It has a total of 47,117 records observed at 15-minute intervals from Jan. 2020 to Mar. 2021. To define another dataset that contains only correlated energy sources where the Pearson correlation coefficient is more than 0.2, we redefine 4-ECK-2021 as 3-ECK-2021 by excluding a non-correlated energy source. For data preprocessing, we transformed the period from 1 or 15 minutes to an hour consistently and missing values were filled based on linear interpolation. Min-max scaling was used for normalization. Five types of Energy Consumption in Korea (5-ECK-2022) Dataset This dataset includes five types of energy sources: 1) electricity, 2) gas, 3) water, 4) hot water, and 5) heating. It has a total of 13,584 records observed at hourly intervals from Aug. 2020 to Feb. 2022. We redefine 5-ECK-2022 as 3-ECK-2022 and 4-ECK-2022 by excluding non-correlated energy sources. Specifically, a heating energy source, which is the weakest correlated with the other energy sources, is excluded from 4-ECK-2022; heating and hot water energy sources are excluded from 3-ECK-2022. Missing values were replaced with one week ago records and min-max scaling was applied for normalization.

Time Series Analysis, Energy Consumption

Creative Commons Attribution 4.0 International