Influencing Factors And Price Prediction Models Of Carbon Emissions Prices In China Based On Informer Model

With the construction of China’s carbon emission market and the call for the "double carbon" initiative,China’s carbon emission market is playing an increasingly important role in the global carbon reduction and environmental protection initiatives.In this paper,we focus on two progressive themes,namely,the influencing factors and the price prediction model of Chinese carbon price,in order to identify the important factors influencing the Chinese carbon market and to construct a suitable carbon price prediction model for monitoring carbon prices and risk warning.The marginal contributions of this paper are mainly two: first,in the selection of influencing factors,this paper selects more domestic developmentrelated influencing factors based on the Chinese market,and finds that the domestic carbon market is still more influenced by overseas markets after incorporating local variables;second,in the selection of price prediction models,the more cutting-edge self-attentive neural network Informer model is selected,and through the evaluation metrics and DM test that proved the superiority of Informer model in carbon price predicting.In the exploration of the influencing factors,based on the demand-side perspective,this paper mainly selects 19 influencing factors from the six dimensions of fossil energy prices,economic development,international carbon market,clean energy development,commodity prices,and climate environment.Then this paper integrates two characteristic importance indicators,namely the integral gradient algorithm and the gradient Shapley value,and finds that the international market carbon emission rights prices,overseas commodity prices,RMB against developed countries exchange rates,and the carbon price of the developed countries are the most important factors.While the importance of influencing factors such as temperature,old and new energy generation situation,domestic commodity prices,and domestic energy prices are weaker.At the level of price prediction model selection,this paper introduces a more cutting-edge self-attentive neural network Informer model into the price prediction of China’s carbon market,and finds that the Informer model can better predict the carbon price in Hubei than traditional models.The evaluation metrics such as MAE,MSE,RMSE,MAPE,and MSPE all exceed the prediction performance of traditional deep neural network models such as long short-term memory neural network(LSTM)and back propagation neural network(BPNN)in the same sample.To ensure the significance of the conclusion,this paper also demonstrates through DM test that the prediction effect of Informer model is significantly better than that of BPNN and LSTM models.In addition,this paper also performs carbon price prediction for Fujian carbon price and national carbon emissions trading market,and finds that the prediction effect of Informer model is still better in these two markets,which tests the robustness of the prediction of Informer model.In this paper,we further expand the prediction steps from single step to multi-step,and find that the 5-step prediction of Informer model still has certain guiding effect that the prediction trend is basically consistent with the real carbon price;the 10-step prediction still maintains certain guiding effect on the trend,but the correctness rate has significantly declined,and the fineness is also far from the real carbon price;while the 20-step,60-step prediction basically fails and is not as good as the real carbon price.