【摘要】 准确的碳价格预测有助于监管部门观测碳交易市场运行状况及投资者进行科学决策，对实现碳达峰和碳中和具有重要作用。但碳价序列具有非线性、非平稳性和高噪声的特性，很难对其进行准确预测。将完全自适应噪声集合经验模态分解(CEEMDAN)方法与门控循环单元(GRU)相结合，构建一个碳排放交易价格预测模型。该模型基于分解、集成思想，利用CEEMDAN将原始碳价序列分解，获得不同频率的本征模函数(IMF)和残差序列，使用GRU神经网络分别为各子序列建立预测模型，最后集成预测结果得到碳价预测值。以湖北省碳交易市场的日度成交价为例进行实证分析，结果表明：相较于其他5种基准模型，CEEMDAN-GRU模型具有更小的预测误差和更高的拟合优度，在碳价格预测上具有一定的优势。更多还原

【Abstract】 Accurate carbon price forecasting helps regulators to observe the operation of the carbon trading market and investors to make scientific decisions, which plays an essential role in achieving the goals of carbon peaking and carbon neutrality. However, the non-linearity, non-smoothness and high noise characteristics of carbon price series make it difficult to forecast them precisely. A carbon price prediction model was constructed by combining the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN) method with a gate recurrent unit(GRU).The model was developed based on the idea of decomposition and integration, using CEEMDAN to decompose the original series to obtain the intrinsic mode functions(IMFs) and residual series at different frequencies, followed by using the GRU neural network to build prediction models for each sub-series separately, and finally integrating the prediction results to obtain the forecasting carbon price. Taking the daily transaction price of the Hubei carbon trading market as an example for empirical analysis, the results show that the CEEMDAN-GRU model has smaller prediction error and higher fitting effect than the other five benchmark models, which provides certain advantages in carbon price prediction.更多还原