Short-term Load Forecasting Using XGBoost and the Analysis of Hyperparameters

Abstract

Accurate load forecasting is getting vital with social and economic development to secure electricity supply and minimize redundant electricity generation. The load forecasting is also essential for efficient power system operation. As machine learning techniques become popular due to the breakthroughs in the application of intelligent systems such as speech or image recognition, variety of machine learning algorithms have also been applied to predict electricity demand. For load forecasting, this paper employs XGBoost algorithm that has recently been receiving attention. To yield the maximum performance of the XGBoost model, we performed grid search method to find optimal hyperparameters of XGBoost. The effects of the XGBoost model's hyperparameters on the model are assessed and visualized.