区域碳排放趋势预测与驱动因素实证研究

Empirical Analysis of Regional Carbon Emission Trend Forecasting and Its Driving Factors

摘要: 在全球变暖问题如此严峻的背景下，碳排放问题已经成为全球关注的焦点。中国作为世界上最大的发展中国家之一，提出了“双碳”的目标。为了实现这一目标，区域碳排放的预测及其影响因素分析显得尤为重要。本文首先进行文献调研并收集了上海市的碳排放及经济社会相关指标的时序数据，并进行了数据的预处理。然后，利用自回归移动平均模型(ARIMA)、BP神经网络模型、最小二乘支持向量机(LSSVM)单一模型对碳排放进行预测，研究发现，在单一模型中预测效果最好的为ARIMA模型。在此基础上，将ARIMA模型与其他两种模型通过优化赋权构建不同的两种组合预测模型：ARIMA-BP组合预测模型以及ARIMA-LSSVM组合预测模型，进一步缩小预测误差选取最优模型。最后，从经济、社会、能源等方面选取与碳排放数据相关的指标，基于这些影响碳排放量的指标，运用相关性分析、灰色关联、回归分析等方法进行碳排放影响因素分析。研究结果表明，ARIMA-BP组合模型在区域碳排放预测中表现出显著优势，能够有效提高预测精度。通过影响因素分析，识别出人均可支配收入、人均GDP以及能源消耗总量等是影响区域碳排放的关键因素。本文旨在通过数据驱动的方法，构建区域碳排放的预测模型并分析其关键影响因素，以期为区域碳排放政策的制定提供一定的科学依据。

Abstract: Against the backdrop of increasingly severe global warming, carbon emissions have become a focal point of international concern. As one of the world’s largest developing countries, China has put forward the “Dual Carbon” targets. To achieve this goal, forecasting regional carbon emissions and analyzing their influencing factors are particularly important. This study begins with a comprehensive literature review, followed by the collection and preprocessing of time-series data related to carbon emissions and socio-economic indicators in Shanghai. Subsequently, three single forecasting models—Autoregressive Integrated Moving Average (ARIMA), Back Propagation (BP) neural network, and Least Squares Support Vector Machine (LSSVM)—were employed. The results indicate that the ARIMA model outperforms the other two in terms of prediction accuracy. Based on this, two hybrid models—ARIMA-BP and ARIMA-LSSVM—were developed using optimized weight allocation methods, aiming to further reduce forecasting errors. The hybrid ARIMA-BP model was ultimately identified as the optimal model for regional carbon emission forecasting. Furthermore, from economic, social, and energy consumption perspectives, relevant indicators were selected for correlation analysis, grey relational analysis, and regression analysis to explore the key drivers of carbon emissions. The findings reveal that per capita disposable income, per capita GDP, and total energy consumption are the most significant factors influencing regional carbon emissions. By adopting a data-driven approach, this research constructs an effective forecasting framework for regional carbon emissions and identifies their key determinants, thereby providing a scientific basis for formulating targeted carbon reduction policies.