融合深度学习和信号分解的碳价预测方法研究

Research on the Method of Carbon Price Prediction by Integrating Deep Learning and Signal Decomposition

碳排放权交易市场有效降低碳排放量,促进经济绿色转型,碳价格预测研究有助于完善碳配额分配制度和构建合理定价机制。针对碳价数据非线性、高噪声、强波动的特点,本文提出了一种将深度学习与信号分解相结合的碳价预测方法。该方法以门控循环单元(GRU)作为预测模型基础,融合自适应噪声完备集合经验模态分解算法(CEEMDAN)以提取碳价序列的多尺度时频特征,降低原序列噪声,提高模型对碳价数据的预测能力,同时引入麻雀搜索算法(SSA)选取最优模型结构参数,增强模型结构稳定性。本文采用湖北省和北京市两地碳交易市场价格数据进行预测实验,实验结果证明:本文提出的CEEMDAN-SSA-GRU模型在碳价预测方面具备显著优势,该模型可以准确预测不同区域和不同时间尺度的碳价数据,同时保持显著的预测稳定性。

The carbon emission trading market can effectively reduce carbon emissions and promote the green transformation of the economy.The research on carbon price forecasting helps to improve the carbon quota allocation system and build a reasonable pricing mechanism.Considering the characteristics of nonlinearness,high noise and strong fluctuation of carbon price data,this paper proposes a prediction method combining deep learning and signal decomposition.In this method,the Gated Recurrent Unit(GRU)is used as the basis of the prediction model,and the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)is integrated to extract the multi-scale time-frequency characteristics of the carbon price sequence,reduce the noise of the original sequence,and improve the predictive ability of the model.The Sparrow Search Algorithm(SSA)is introduced to select the model structural parameters and enhance the stability of the model structure.In this paper,the price data of carbon trading market in Hubei Province and Beijing are used for prediction experiments.The experiment results show that the CEEMDAN-SSA-GRU model has significant advantages in carbon price prediction,and can accurately predict carbon price data in different regions and different time scales,while maintaining significant prediction stability.