基于CEEMDAN和优化LSTM模型的碳价波动率预测研究  

Research on carbon price volatility prediction based on CEEMDAN and optimized

作　　者：段钧陶 杨晓忠[1] DUAN Juntao;YANG Xiaozhong(School of Mathematics and Physics,North China Electric Power University,Beijing 102206,China)

机构地区：[1]华北电力大学数理学院,北京102206

出　　处：《中国科技论文在线精品论文》2024年第2期283-293,共11页Highlights of Sciencepaper Online

基　　金：国家自然科学基金(11371135)。

摘　　要：本文以北京碳配额交易价格实际波动率为研究对象,构建以自适应噪声完备集合经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)和长短期记忆网络(long short-term memory networks,LSTM)为基础的混合预测模型,采用粒子群优化算法(particle swarm optimization,PSO)确定模型结构参数。实验结果证明:该模型具备提取多尺度复杂时间序列波动趋势和有效处理金融时间序列的优点,粒子群算法对预测模型结构参数的优化避免了因参数选取不当导致的拟合问题,该模型在碳价波动率预测方面具备较高的准确性和稳定性。This paper takes the realized volatility of Beijing carbon emission allowance trading prices as the research object,constructs a hybrid forecasting model based on complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and long short-term memory network(LSTM),and optimizes the model structural parameters through particle swarm optimization(PSO)algorithm.The experimental results demonstrate that the model has the advantages of extracting multi-scale complex time series volatility trends and effectively processing financial time series.The particle swarm optimization algorithm optimizes the structural parameters of the forecasting model to avoid fitting problems caused by improper parameter selection.The model has significant accuracy and stability in forecasting carbon price volatility.

关 键 词：应用统计数学 碳价波动率预测 CEEMDAN-PSO-LSTM模型 时间序列预测 

分 类 号：O213[理学—概率论与数理统计] TP183[理学—数学]