Research On Carbon Dioxide Emission Prediction Based On SOA-LSTM And Influencing Factors

In September 2020,General Secretary Xi Jinping made action instructions for my country’s carbon dioxide emissions,pointing out that the carbon peak should be achieved in2030 and carbon neutrality should be achieved in 2060.This will have a great impact on the development of many industries in my country.Before the arrival of 2060,how to complete the "double carbon" commitment on time will become a key work.Through the prediction of carbon emissions and the analysis of influencing factors,it can be used as a reference for the measures that need to be implemented to achieve this goal.In order to accurately grasp the relevant situation of carbon emissions,appropriate indicators can be selected and a prediction model can be established for analysis and prediction.This paper takes 1997-2017 as the research time range,takes carbon emission-related data as the research object,and selects the population size,per capita GDP,thermal power generation ratio,and secondary industry ratio in combination with the relevant literature and the characteristics of carbon emissions.passenger and freight turnover and urbanization rate as the factors affecting carbon emissions.In the analysis of the influencing factors of carbon emissions,the improved STIRPAT model and ridge regression are used to analyze the influencing factors,and the influence direction and intensity of different factors on carbon emissions are obtained.For the prediction of carbon emissions,the deep learning method is used to build the SOALSTM model based on the Long short-term memory(LSTM)neural network and combined with the Seagull optimization algorithm(SOA).The advantages of fast convergence speed,good stability,and strong robustness solve the problems of LSTM that needs to adjust many parameters,consumes time and energy,and the prediction effect is not ideal.Taking the above six influencing factors as input,carbon emissions as output for training,and comparing with other neural network models.The analysis results of influencing factors show that population size,per capita GDP,proportion of thermal power generation,the proportion of secondary industry and passenger and freight turnover play a positive role in carbon emissions.will increase by 1.919%,0.248%,0.862%,1.104% and 0.332%;the urbanization rate will have a reverse effect,and every 1%increase in the urbanization rate will lead to a 0.361% decrease in carbon emissions.In the study of the prediction model of carbon emissions,the results show that compared with other neural network models,the prediction ability of the SOA-LSTM model optimized by the seagull optimization algorithm has been greatly improved,in which the absolute mean error(MAE)is140.1,the mean square The root error(RMSE)is 167.7,and the mean absolute value of error(MAPE)is 0.014.Compared with LSTM,SOA-BP,and GA-LSTM,they have advantages in each evaluation index,and can achieve more effective prediction results.It is effective to optimize LSTMs that require a lot of parameter tuning.