Research on Carbon Emission Prediction Method Based on Deep Learning: A Case Study of Shandong Province

Abstract

Accurate prediction of carbon emissions is of great significance for achieving the goal of carbon peak and carbon neutralization. The traditional carbon emission prediction models are usually designed for the fixed mathematical model based on energy, population, economy, policy and other factors. However, due to numerous influencing factors and difficulties in selection, it is difficult to get accurate prediction results from the model. In order to solve these problems, this thesis uses the method of deep learning to predict carbon emissions, and designs LSTM-Attention model which can accurately predict carbon emissions. Taking Shandong Province, China as an example, based on the carbon emissions data obtained by CEADS, potential influencing factors of car- bon emissions are selected from various fields of Shandong Statistical Yearbook, and three carbon emissions prediction experiments with differ- ent numbers of influencing factors are conducted. The results show that, compared with other prediction methods, the LSTM-Attention model de- signed in this thesis has achieved a better prediction effect. Even under the influence of strong correlation between the variables used to predict carbon emissions, such as population and economy, our model can still predict carbon emissions more stably and accurately.