统计与数据科学学院知识库

习近平总书记从提出绿水青山就是金山银山在到要在2060年实现“双碳”目标，彰显大国担当。作为世界上能源最大的能源消费和生产国，我国碳市场已经取得了跨越式进展，2010年到2021年，从首次提出建设碳市场到全国碳市场开市交易，意味着我国低碳发展已经取得阶段性胜利，为实现新能源转型取得了更为广阔的市场，基于此，深入挖掘研究碳排放权交易价格（以下简称碳交易价格）的影响因素以及内在机理对预测碳交易价格帮助建立稳定有效的碳定价机制具有战略意义和实践价值。

首先，通过梳理碳交易价格影响因素及碳价预测等相关文献，深入挖掘碳排放权及碳交易价格的理论特征及属性。同时从经济形势、金融市场、国际碳市场、能源价格、气候环境、互联网大数据等6个方面构建影响碳交易价格的指标体系，运用线性插值将数据补全，最后进行描述性统计分析。

其次，立足于2021年7月16日-2022年9月1日的23个变量的时间序列数据，以正则稀疏化模型为基准点，分析6个一级指标对碳交易价格的影响。通过实证分析验证了宏观经济因素与金融市场对碳交易价格的影响较为显著，互联网大数据对碳交易价格的影响较小但有影响，其中化石能源价格对碳交易价格的影响较为显著，天然气价格与碳交易价格呈负相关关系，煤炭价格与碳交易价格呈负相关关系。国际碳市场以及大气环境对碳交易价格的影响并不显著。

最后，在以下2个方面进行碳交易价格的预测，一、直接将被解释变量用LSTM模型进行单因子预测，然后再将所有被解释变量以及解释变量纳入LSTM模型进行多因子预测。二、在SVR和LSTM模型的基础上引入Lasso，弹性网、弹性MCP等8个正则化模型进行变量选择，通过构建变量组合模型对碳交易价格进行预测分析。研究发现，对预测结果精度进行对比分析，得出组合模型的预测效果更好，评估指标更低，并且预测得出碳交易价格在2022年9月1日至9月5日这5天呈平稳上升发展趋势。本文通过实证结果得出弹性MCP-LSTM组合模型可以更有效预测碳交易价格。

基于以上研究分析，本文从推进扩大全国碳交易市场的参与规模、优化排放核算标准及配额分配方案、发展碳金融衍生产品等方面提出建议，为全国碳交易市场的稳步发展以及减少温室气体的排放作出贡献。

General Secretary Xi JinPing has proposed that green mountains are golden mountains and silver mountains to achieve the goal of " double carbon " by 2060, highlighting the responsibility of a big country. As the world 's largest energy consumer and producer, China 's carbon market has made great strides. From 2010 to 2021, from the first proposal to build a carbon market to the opening of a national carbon market, it means that China 's low-carbon development has achieved a phased victory. In order to achieve a broader market for new energy transformation, based on this, it is of strategic significance and practical value to deeply explore the influencing factors and internal mechanisms of carbon emission trading prices ( hereinafter referred to as carbon trading prices ) to predict carbon trading prices and help establish a stable and effective carbon pricing mechanism.

Firstly, by combing the relevant literature on the influencing factors of carbon trading price and carbon price prediction, the theoretical characteristics and attributes of carbon emission rights and carbon trading price are deeply explored. At the same time, the index system affecting carbon trading price is constructed from six aspects : economic situation, financial market, international carbon market, energy price, climate environment and Internet big data. Linear interpolation is used to incomplete the data, and finally descriptive statistical analysis is carried out.

Secondly, based on the time series data of 23 variables from July 16,2021 to September 1,2022, the influence of six first-level indicators on carbon trading prices is analyzed with the regularized sparse model as the reference point. Through empirical analysis, it is verified that macroeconomic factors and financial markets have a significant impact on carbon trading prices, while the international carbon market and atmospheric environment have no significant impact on carbon trading prices. Internet big data has a small but influential impact on carbon trading prices. Among them, fossil energy prices have a significant impact on carbon trading prices. Natural gas prices are negatively correlated with carbon trading prices, and coal prices are negatively correlated with carbon trading prices.

Finally, the prediction of carbon trading price is carried out in the following two aspects. First, the explanatory variables are directly predicted by the LSTM model for single factor prediction, and then all the explanatory variables and explanatory variables are included in the LSTM model for multi-factor prediction. Based on the SVR and LSTM models, eight regularization models such as Lasso, elastic net and elastic MCP are introduced to select variables, and the carbon trading price is predicted and analyzed by constructing a variable combination model. The study found that by comparing and analyzing the accuracy of the prediction results, it is concluded that the combined model has better prediction effect and lower evaluation index, and it is predicted that the carbon trading price will rise steadily from September 1 to September 5,2022. Through empirical results, this paper concludes that the flexible MCP-LSTM combination model can predict carbon trading prices more effectively.

Based on the above research and analysis, this paper puts forward suggestions from the aspects of promoting the expansion of the participation scale of the national carbon trading market, optimizing the emission accounting standards and quota allocation schemes, and developing carbon financial derivatives, so as to contribute to the steady development of the national carbon trading market and reduce greenhouse gas emissions.