摘要
对碳价格的准确预测,有助于碳中和方案的制定和碳排放权交易市场的稳定发展。目前的碳价格预测模型,没有考虑碳价格数据中隐含的未来数据信息,也没有针对不同地区碳价格进行自适应的参数优化。深度学习通过引入非线性因素,可以对非线性的碳价格数据进行有效预测。为此在深度学习的基础上,充分利用双向长短期记忆模型双向处理上下文信息以及变分模态分解算法对于非线性和非平稳数据适用性良好等特点,对变分模态分解算法进行参数优化,提出一种基于信号分解算法的碳价格混合预测模型。首先,使用变分模态分解算法作为分解算法,以最小化平均近似熵为目标,使用遗传算法对变分模态分解算法的主要参数进行优化;接着,使用优化参数后的分解算法,对原始碳价格进行分解;之后,使用双向长短期记忆模型对分解结果分别进行预测,再将各个预测结果聚合,得到最终预测价格。基于深圳市历史碳价格数据的实验结果表明,该模型可以有效预测碳价格,并且相较于其他模型,该模型预测准确率更高。
Accurate prediction of carbon price can not only promote effective carbon neutralization solutions,but also contribute to the development of carbon trading market.However,the current carbon price prediction model does not consider the future data information implied in the carbon price data,nor does it carry out adaptive parameter optimization for carbon prices in different regions.By introducing nonlinear factors,deep learning can effectively predict the nonlinear carbon price data.Based on deep learning,taking full advantage of Bidirectional Long Short-Term Memory(BiLSTM)in processing contextual information and the good applicability of Variational Mode Decomposition(VMD)to nonlinear and non-stationary data,this paper optimized the parameters of VMD and proposed a carbon price hybrid prediction model based on signal decomposition algorithms.Firstly,VMD is used as the decomposition algorithm,and then Genetic Algorithm(GA)is used to optimize the parameters of VMD for minimizing the Approximate Entropy(ApEn).Secondly,the original carbon price is decomposed by VMD with the optimized parameters.Thirdly,BiLSTM neural network is used to predict the decomposed results respectively,and then the prediction results are aggregated to obtain the final predicted price.The experimental results based on the historical carbon price data of Shenzhen show that this model can predict the carbon price effectively.Compared with other models,the prediction accuracy of this model is significantly improved.
作者
任冠宇
栾皓轮
万剑雄
李雷孝
王晓磊
REN Guanyu;LUAN Haolun;WAN Jianxiong;LI Leixiao;WANG Xiaolei(School of Data Science and Application,Inner Mongolia University of Technology,Hohhot 010051,China;School of Economics and Management,Inner Mongolia University of Technology,Hohhot 010051,China)
出处
《内蒙古工业大学学报（自然科学版）》
2023年第4期355-362,共8页
Journal of Inner Mongolia University of Technology：Natural Science Edition
基金
内蒙古自治区重点研发与成果转化计划项目(2021CG0033,2022YFSJ0013)
内蒙古自治区高等学校“青年科技英才”支持计划项目(NJYT22084)。
关键词
碳价格预测
变分模态分解
双向长短期记忆模型
carbon price prediction
variational modal decomposition
bidirectional long short-term memory
