Open AccessForecasting the carbon price sequence in the Hubei emissions exchange using a hybrid model based on ensemble empirical mode decomposition
Author(s) -
Wu Qunli,
Liu Ziting
Publication year - 2020
Publication title -
energy science and engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.638
H-Index - 29
ISSN - 2050-0505
DOI - 10.1002/ese3.703
Subject(s) - hilbert–huang transform , mode (computer interface) , carbon price , component (thermodynamics) , ensemble forecasting , econometrics , series (stratigraphy) , sequence (biology) , support vector machine , time series , computer science , statistics , artificial intelligence , mathematics , greenhouse gas , machine learning , energy (signal processing) , ecology , paleontology , physics , thermodynamics , genetics , biology , operating system
The prediction of carbon price is exceedingly essential for the regulators, investors, and participants of the carbon trading market. It is the basis for formulating market policies and improving risk management capabilities. China's carbon price series are nonlinear and nonstationary, so it is difficult to predict accurately with traditional models. This paper proposes a multiscale ensemble prediction model based on ensemble empirical mode decomposition (EEMD‐ADD) to improve the prediction accuracy of carbon price. Firstly, EEMD is used to decompose the carbon price sequence into several intrinsic mode functions (IMFs), and these IMFs are divided into high‐frequency component, low‐frequency component and the trend component. Then, LSSVM, PSO‐LSSVM, and BA‐LSSVM are used to predict the three components respectively after comparative analysis. Finally, the results are combined to obtain the final prediction value. In the empirical analysis of the Hubei Emissions Exchange, the proposed model outperforms other comparative models. The RMSE, MAE, and MAPE values of the EEMD‐ADD model are 0.6180, 0.4726, and 1.6342, and the DS, CP, and CD values are 94.36, 92.16, and 96.48. In addition, the model performed best in other time periods. The results suggest that the proposed model is effective and could predict carbon prices more accurately.