Research On Price Forecast Of National Carbon Emission Trading Market Based On Combination Forecast Model

The significant amounts of greenhouse gases represented by carbon dioxide,have a profound impact on the escalating severity of global warming.Establishing carbon trading markets is effective measures to effectively reduce carbon emissions and slow down global warming.In September 2020,China,recognized as the foremost emitter of carbon,publicly declared its intention to achieve the objective of "peak carbon dioxide emissions" before 2030,and additionally,to attain "carbon neutrality" before 2060.In addition,subsequent to the establishment of eight pilot markets such as Beijing and Tianjin,China formally instituted a consolidated national carbon emission trading market on July 16,2021.Accurate and stable forecasting of carbon price plays an important role in facilitating energy preservation and emission reduction,and achieving the goals of carbon peak and carbon neutrality as soon as possible in China.Consequently,this paper selected the closing rates of Chinese Emission Allowances(CEA)in the national carbon emission trading market from July 16,2021 to December 30,2022,and selected eight influencing factors from three aspects of macroeconomic,energy price and climate.Since the carbon price series is non-stationary and non-linear,firstly,the CEEMDAN algorithm is initially employed to decompose the price series into eight eigenmode functions and residual term of different frequencies,subsequently reconstituted into the high-frequency group,the low-frequency group and the trend term.The analysis shows that the high-frequency group has high fluctuation frequency,which reflects the short-term random fluctuation of carbon price.Therefore,the model of ARIMA-GARCH is used to fit and prediction.Moreover,the low-frequency group changes slowly,which reflects the long-term fluctuation of carbon price,and it is highly correlated with Brent crude oil,EUA price and industrial index,so multivariate.LSTM model is used for prediction.In addition,the trend item is an increasing sequence,which reflects the trend of carbon price and is highly correlated with CSI 300 index and industrial index,so the multivariate LSTM model is also used for prediction.Finally,the carbon price prediction results are the sum of the predicted values of each component.By comparing the selected prediction metrics including MSE,RMSE,MAE and MAPE values with other single or combined models such as ARIMA,LSTM and EEMD-ARMALSTM,etc,it can be seen that the CEEMDAN-ARIMA-GARCH-LSTM model constructed in this paper has the highest prediction accuracy the best predictive effect.Consequently,the combined prediction model proposed in this paper has advantages in carbon price prediction and can provide new ideas for improving the prediction accuracy of carbon price and other financial asset prices.