Electric Power Carbon Emission Prediction based on Stacking Ensemble Model with K-fold Cross Validation

• Main Authors: Shen, Yu, Zeng, Zhensong, Lin, Weiwei, Que, Dingfei, Huang, Zhenyu
• Format: Conference Proceeding
• Language: English
• Subjects: Carbon dioxide; Data models; electric power carbon emission prediction; k-fold cross validation; Low-carbon economy; pearson correlation coefficient; Power system management; Predictive models; Regulation; Stacking; Stacking ensemble model
• ISSN: 2688-0938
• DOI: 10.1109/CAC57257.2022.10055874

With the increasing global awareness of climate change, countries and regions worldwide have made ambitious plans, regulations and supports, for the carbon-neural aspirations. For China, decarbonizing its power sector, which accounts for 40% of the total emission, is particularly important to reach its carbon-neutral objective. Research shows that an accurate projection of electricity consumption, if available, can significantly contribute to power management tasks, leading to better power efficiency and directly contributing to the decarbonization task for the power sector. To solve this problem, this paper proposes a novel electric power carbon emission prediction approach based on the stacking ensemble model. It constructs a stacking ensemble model with k nearest neighbor (KNN), decision tree (DT), AdaBoost and XGBoost as the base learners and gradient boosting decision tree (GBDT) as the meta learner to improve the final prediction accuracy. The experimental results show that the stacking ensemble model constructed in this paper performs better than the traditional model in terms of RMSE, MAE and R 2 .