Power Data-Carbon Emission Prediction Model Based On Stacking Ensemble And Hyperparameter Optimization With Cross-Validation Method

Abstract

At present, the ‘‘greenhouse effect’’ caused by energy and environmental pollution makes energy carbon emission become the focus of the society, and accurate carbon emission prediction for high emission enterprises is the premise for the realization of emission reduction targets. This paper presents a power data-carbon emission prediction model based on a stacking ensemble and its hyperparameter optimization with a Cross-Validation method. Firstly, on the basis of obtaining power data and corresponding carbon emission data samples, a feature selection method of Emission Factor-Grey Correlation analysis is proposed for data specification.Then, the first layer sub-model is constructed separately, and the optimization method based on cross validation is combined to train respectively.Finally, the results of multiple single models are integrated by Stacking. The simulation results show that the proposed Cross-Validation optimization method can effectively improve the generalization ability of the model, and the carbon emission prediction model can reduce the maximum prediction error and improve the average prediction accuracy, which is better than the prediction of a single model. In addition, the model takes electricity consumption as the only input of the prediction model involving enterprise production data, which solves the time lag problem of enterprise carbon emission data mainly relying on carbon verification work and the difficulty of obtaining industrial fossil energy consumption data.