School of Statistics, Southwestern University of Finance and Economics, Chengdu, China.
School of Economics and Management, University of Chinese Academy of Sciences, Beijing, China.
J Environ Manage. 2024 Jun;362:121253. doi: 10.1016/j.jenvman.2024.121253. Epub 2024 Jun 1.
Carbon trading is one of the pivotal means of carbon emission reduction. Accurate prediction of carbon prices can stabilize the carbon market, mitigate investment risks, and promote green development. In this study, firstly, the IVMD and ICEEMDAN are used to decompose carbon price quadratically; secondly, the Dispersion entropy is used to identify the sequence frequency, and then the SOA-LSSVM model and TCN model are used to predict the high-frequency and low-frequency sequences, respectively; finally, the prediction results are integrated by SOA-GRU. As a result, the hybrid IVMD-ICEEMDAN-SOALSSVM/TCN-SOAGRU model was constructed. This framework consistently performs best under two carbon markets, the CEEX Guangzhou and the EU ETS, compared with 21 comparative models, with MAPEs of 0.42% and 0.83%, respectively. The main contributions are as follows: (1) A novel IVMD-ICEEMDAN secondary decomposition method is proposed, which improves the problem of poorly determining the value of the decomposition modal number K in the traditional VMD method and improves the efficiency of the carbon price sequence decomposition. (2) A hybrid forecasting model of LSSVM and TCN is proposed, effectively capturing the features of different sequences. (3) Optimization for LSSVM and GRU using SOA improves the stability and adaptability of the model. The article provides governments, enterprises, and investors with novel and effective carbon price forecasting tool.
碳交易是减少碳排放的主要手段之一。准确预测碳价可以稳定碳市场，降低投资风险，促进绿色发展。在这项研究中，首先使用 IVMD 和 ICEEMDAN 对碳价进行二次分解；其次，使用分散熵来识别序列频率，然后分别使用 SOA-LSSVM 模型和 TCN 模型预测高频和低频序列；最后，通过 SOA-GRU 对预测结果进行集成。结果构建了混合 IVMD-ICEEMDAN-SOALSSVM/TCN-SOAGRU 模型。与 21 种对比模型相比，该框架在两个碳市场（广州碳交易所和欧盟 ETS）中的表现始终最佳，其平均绝对百分比误差（MAPE）分别为 0.42%和 0.83%。主要贡献如下：（1）提出了一种新的 IVMD-ICEEMDAN 二次分解方法，解决了传统 VMD 方法中模态数 K 值难以确定的问题，提高了碳价序列分解的效率。（2）提出了一种 LSSVM 和 TCN 的混合预测模型，有效地捕捉了不同序列的特征。（3）使用 SOA 对 LSSVM 和 GRU 进行优化，提高了模型的稳定性和适应性。本文为政府、企业和投资者提供了一种新颖有效的碳价预测工具。
