Forecasting Carbon Emissions in Chinese Coastal Cities Based on a Gated Recurrent Unit Model

Abstract

Abstract Scientific forecasting of carbon emission trends is an important basis for understanding carbon emission levels and is a key reference for achieving dual carbon goals. Carbon emission prediction models based on backpropagation neural networks and multiple linear regression methods have yielded relatively favourable results. Compared with traditional methods, deep learning offers plasticity and effectiveness, and has been verified by application in multiple fields. In this study, a carbon emission forecasting model was built based on the gated recurrent unit model to empirically evaluate data from 16 coastal cities in China from 1997 to 2019. Eight scenarios were established to forecast carbon emissions in Chinese coastal cities from 2020 to 2029, and to investigate the combined effects of the factors driving carbon emissions, including economy size, industrial structure, and public revenue. Our results predict increasing carbon emissions from 2020 to 2029, with substantially slower growth compared with the change in carbon emissions from 1997 to 2019. Furthermore, the eight scenarios show that changes in driving factors, including public revenue, industrial structure, and size of the economy, strongly impact on carbon emissions. The size of the economy has the most significant effect on carbon reduction. Therefore, the marine economic growth model needs to be redesigned and further developed with a view towards carbon emission reduction and low-carbon development.