基于多尺度分析的碳市场价格区间预测研究

In recent years, the global carbon market has rapidly developed, which will continue to grow in the future as tighter actions will be taken to address the climate change. However, the high complex features of multiscale, uncertainty and dynamic interaction for carbon price have resulted in a serious loss of billions of euros for Chinese carbon assets. Along with the establishment of China’s carbon markets which will constantly deepen the links with the global carbon markets, the loss will continue to rise. Therefore, how to accurately forecast the carbon price has been emerged as a research hotspot. Facing with the key management scientific issues to be solved, from the perspective of multiscale analysis, this project will apply synergistically the multivariate empirical mode decomposition, multiple output least squares support vector machine, quantile regression averaging, sparse Bayesian learning and other approaches to carry out quantitative studies on interval forecasting for carbon price. Around the state and characteristics of carbon price forecasting, this study will establish the multiscale multi-step-ahead point forecasting, interval forecasting and density forecasting models for carbon price, for deeply quantitative researches of theories and methods system for multiscale interval forecasting for carbon price. This project will pay attention to combination of theory and practice, emphasize the multidisciplinary integration, and aim at promoting the development of carbon price forecasting to data-driven mode and multiscale interval forecasting direction, which will provide decision-makers with more information supports and decision-making references.

近年来，全球碳市场发展迅猛，受国际应对气候变化行动的影响，未来会继续发展。然而，碳价波动的多尺度性、不确定性和动态关联性等复杂特征，造成我国仅因价差碳资产流失高达数十亿欧元。随着我国逐步建成自己的碳市场并不断深化与全球碳市场链接，碳价波动不确定性将会增加，碳资产流失将会继续上升，点预测已无法满足需要，因此碳价区间预测问题凸显。本项目针对碳价区间预测面临的亟待解决的关键管理科学问题，基于多尺度预测框架，协同应用多变量经验模态分解、多输出最小二乘支持向量机、分位数回归平均等理论方法，紧密结合碳价预测特点与现状，建立碳价多尺度多步点预测模型、区间预测模型和密度预测模型，深入探讨定量研究碳价多尺度区间预测的理论方法体系。本项目注重理论方法研究与实践相结合，强调多学科的交叉融合，旨在推动碳价预测工作向数据驱动模式和多尺度区间预测方向发展，为决策者提供更多的信息支持和决策参考。

气候变暖是人类可持续发展面临的重大环境问题。全球碳减排实践证明，碳排放权交易市场是一种切实有效和低成本的碳减排政策，中国已经建立全国统一的碳市场。为促进碳市场稳定运行和有效减排，提升碳交易收益，本项目紧密围绕碳市场价格生成、波动机制，聚焦碳价点预测和区间预测等关键科学问题，建立碳市场的多尺度辨识模型、配额调整与分配模型，价格波动行为和风险特征识别与度量模型、多尺度自适应点预测模型和多尺度前端-中端和后端区间预测模型，深入开展碳价波动特征和预测方法研究。一方面，本项目系统全面的探索了碳市场的多尺度性、不确定性，异质性和动态关联性等复杂特征，揭示了碳市场复杂系统内部运行基本规律；另一方面，注重理论方法研究与实践相结合，强调多学科的交叉融合，协同应用博弈论、统计计量工具、金融分析方法、多变量经验模态分解、多输出最小二乘支持向量机、极限梯度提升机和稀疏贝叶斯学习等理论方法，提出碳配额动态调整与分配方案，开发了一系列数据驱动的多尺度高精度集成预测新技术，初步形成了碳价多尺度区间预测的方法体系。本研究进一步丰富和完善了碳市场理论体系与技术方案，为碳交易、碳市场健康稳定发展和“碳中和”中国方案提供信息支持和决策参考。