| Title |
A Machine Learning Based Algorithm for Short-Term Weekends Load Forecasting |
| Authors |
심상우(Sang Woo Shim) ; 이다한(Da Han Lee) ; 노재형(Jae Hyung Roh) ; 박종배(Jong-Bae Park) |
| DOI |
https://doi.org/10.5370/KIEE.2022.71.11.1578 |
| Keywords |
XGBoost; Feature Selection; Shapley Value; Correlation Coefficient; Weekends Load Forecasting |
| Abstract |
This paper presents a methodology for forecasting short-term weekends hourly loads using feature selection and hyperparameter. It starts with setting features necessary for forecasting loads and the peculiarity is that weather data divided by region are used as weather data across the country based on the number of people in each region. In order to improve the performance of forecasting, important variables are extracted through the SHAP(SHapley Additive exPlanations) method and Pearson Correlation Coefficient, and then optimized XGBoost parameters are found and applied through grid search. This paper tried to predict for every weekend of 2021, and the paper shows results for four weekends in September, or eight days. Errors are expressed through NMAE, MAPE and NRMSE to show the performance of the prediction model in various ways. Later studies will be conducted on forecasting algorithms for special days such as holidays as well as general weekends, and sensitivity analysis for each feature will also be considered. |