The Evolutional Prediction Model of Carbon Emissions in China Based on BP Neural Network

This paper analyzes and processes the time sequence of carbon emissions from the perspective of the BP neural network, revealing the structure and the law of all kinds of index systems of carbon emissions and describing the dynamic characteristics of carbon emissions system. Based on the synthesis of the energy structure factors, energy efficiency factors, economic development factors, population factors,per capita consumption factors and population urbanization rate factors as the main influencing factors, the author tries to establish the neural network prediction model of carbon emissions in China, simulating it with MATLAB, and forecast carbon emissions of the next decade with three exponential smoothing methods .The simulation results show that the selected input index plays an important role in the carbon emissions. This article adopts the improving BP neural network LM algorithm to predict the carbon emissions in China, with an interesting finding that the predicted carbon emissions and the actual ones are so close that the relative error can be controlled within three percent, which shows that this neural network of carbon emissions forecast can be adopted by the macro economic department as a reliable basis for decision-making. The above result is achieved in the existing policies and measures conditions of energy conservation and emission reduction, but to achieve the carbon emissions proposed by 2050 in Copenhagen world climate summit, the control policy and the related factors of internal and external model need to be analyzed. This paper analyzes the energy conservation and emission reduction concretely by 2050 and designs three scenarios, corresponding to the low, medium and high energy conservation and emission reduction policy.They are forecasted respectively with the forecasting results that in the situation of low energy conservation and emission reduction, that is if any new climate change countermeasure is not taken, the prediction of carbon emissions is 84.7598 billion tons; In the situation of medium energy conservation and emission reduction, that is, in the premise of considering the requirements of the sustainable development, energy security, domestic environment and low carbon conditions in the economic society, the prediction values of carbon emissions is 68.6690 billion tons; In the situation of high energy conservation and emission reduction, that is, with the consideration of the domestic economy and social development and the demand of environment development, the technology of non-fossil energy and emission reduction makes a key breakthrough. It forecasts the carbon emissions of 46.4767 billion tons. Simulation shows that using the improved model of neural network method to predict carbon emissions in 2050, with the analysis of the main factors of carbon emissions, the results can be more accurate and comprehensive.