A carbon price prediction model based on secondary decomposition algorithm and optimized back propagation neural network

Abstract

Abstract Carbon trading is one of the important mechanisms used to reduce carbon dioxide emissions. The increasing interest in the carbon trading market has heightened the need to decrease the prediction error of the carbon price. In this paper, a new hybrid model for carbon price forecasting is proposed, and the secondary decomposition algorithm is innovatively introduced into carbon price forecasting. First, time series data were decomposed into several intrinsic modal functions by empirical mode decomposition (EMD). Second, the first intrinsic mode function (IMF1) was further decomposed by variational mode decomposition (VMD). Then, the model input was determined by partial autocorrelation analysis (PACF). Finally, the back propagation (BP) neural network model optimized by genetic algorithm (GA) was utilized for prediction. In the empirical analysis of the Hubei market, the proposed model outperforms other comparative models. The mean absolute percentage error (MAPE), goodness of fit (R2) and root mean square error (RMSE) of the model are 1.7577%, 0.9929 and 0.5441, respectively. In the complementary cases of the Beijing and Shanghai carbon markets, the model also performs best. The results suggest that the proposed model is effective and robust and could predict carbon prices more accurately.