基于时变高阶矩NAGARCHSK-LSTM模型的中国碳排放权价格预测  

Forecast of Carbon Price in China Based on the NAGARCHSK-LSTM Model with Time-varying High-order Moments

作　　者：云坡 唐文之 黄荷暑[2] YUN Po;TANG Wenzhi;HUANG Heshu(School of Economics and Management, Hefei University, Hefei 230601, China;School of Business, Anhui University, Hefei 230601, China)

机构地区：[1]合肥学院经济与管理学院,安徽合肥230601 [2]安徽大学商学院,安徽合肥230601

出　　处：《安徽农业大学学报（社会科学版）》2021年第5期48-57,共10页Journal of Anhui Agricultural University:SOC.SCI.

基　　金：教育部人文社会科学研究青年基金项目“基于双向多层循环神经网络时变高阶矩传染的碳金融资产定价研究”(21YJC790152);国家自然科学基金青年项目“多种低碳融资方式下供应链减排运作与契约协调机制研究”(71904041)。

摘　　要：有效的碳价预测有助于碳市场以较低成本解决环境问题,实现碳排放减少的目标。现有研究忽略市场不对称和极端因素对碳价的时变高阶矩冲击关系,预测准确性存在质疑。基于碳价非对称性、极端冲击敏感性强以及时变波动等专属特征,构建新的机器学习碳价预测模型NAGARCHSK-LSTM。研究显示,NAGARCHSK-LSTM模型能有效捕捉碳价时变高阶矩特征,碳价预测精度和鲁棒性均优于其他基准模型,特别是模型长期预测优势得到验证。研究为投资者研判市场行情、开展价格分析提供技术手段。Effective carbon price prediction can help the carbon market to solve environmental problems at a lower cost and achieve less carbon emissions.Existing studies ignore the impact of market asymmetry and extreme factors on carbon price with time-varying higher-order moments,so the accuracy of prediction is questionable.Based on specific characteristics such as carbon price asymmetry,high sensitivity to extreme shock and time-varying volatility,a new machine learning carbon price prediction model NAGARCHSK-LSTM is constructed.The results show that the proposed model can effectively capture the time-varying higher-order moment characteristics of carbon price.In terms of accuracy of carbon price prediction and robustness,the proposed model is better than other benchmark models,and the long-term prediction advantage of the model has been verified in particular.To sum up,our research provides a technical means of judging market conditions and carrying out price analysis for investors.

关 键 词：碳价预测 时变高阶矩 NAGARCHSK模型 LSTM模型 

分 类 号：F832.5[经济管理—金融学]