Research On China’s Carbon Emission Right Price Prediction Based On Deep Learning Mode

The national carbon market was launched on July 16,2021.The national carbon market is still in its initial operation stage,covering only the power sector,while each carbon pilot is still the main venue for carbon emissions trading.The price of carbon emission rights is the embodiment of the value of carbon emission rights that reflects key information such as market supply and demand,and the cost of emission reduction of enterprises.With the standardization of China’s carbon market,there is an urgent need to analyze the factors influencing the price of carbon emission rights and to forecast the price of carbon emission rights,so as to provide an important guarantee for market players to improve their risk prevention ability and reduce the loss of carbon asset value,as well as to provide a reference basis for the government authorities to design and manage the carbon market,so as to effectively meet the challenges of climate change and contribute to "double carbon" goal.In this paper,we consider the regional heterogeneity of carbon emission price influencing factors to forecast the carbon emission price of different carbon pilot projects.Firstly,we analyze the factors influencing the price of carbon emission rights in China’s carbon market from six dimensions based on domestic and international studies on the factors influencing the price of carbon emission rights,and screen the factors influencing the price of carbon emission rights in different pilot regions based on the Lasso regression model;secondly,we introduce CNN network and attention mechanism based on the traditional GRU neural network model,and carry out the feature extraction and weighting of the price of carbon emission rights.Finally,the CNN-GRU model based on the attention mechanism was used to predict the carbon emission rights prices of eight carbon pilot sites in China and compared with the prediction results of various models.The results of the study found that there is regional heterogeneity in the factors influencing the price of carbon emission credits,with the S&P 500 index,Shanghai Interbank Offered Rate,US dollar exchange rate,Euro exchange rate,oil price,natural gas price and whether the national carbon market is launched having significant effects on the price of carbon emission credits in most carbon pilots,while coal price,SSE index,Baidu index and compliance expiration date have less effects on the price of carbon emission credits;At the same time,the CNN-GRU model based on attention mechanism proposed in this paper has higher prediction accuracy compared with other deep learning models,and the predicted value of carbon emission rights price calculated by this model has the smallest error with the real value when the data samples are sufficient,but the predicted value of carbon emission rights price calculated when the data samples are small has higher prediction accuracy.However,the predicted value of carbon emission rights price is more inaccurate than the true value when the data samples are small,and the prediction of small samples is lacking.The results of this paper can help carbon market traders to manage carbon assets and carbon financial risks,avoid economic losses caused by abnormal price fluctuations,and help enterprises to reasonably arrange emission reduction costs and develop appropriate emission reduction plans.This paper suggests that policy makers need to pay attention to the influencing factors affecting the price of carbon emission rights in the process of promoting the national carbon market and consider the unbalanced development of different regions;carbon market participants establish a carbon emission rights price prediction system,improve the carbon asset management system,and establish a carbon cost decision-making mechanism.