Research On My Country’s Energy Demand Forecast Based On Combination Model

Energy resources are an important material basis for human economic and social development.In the new stage of development,scientific and reasonable energy demand forecasting is not only the theoretical basis for the optimization and adjustment of the energy consumption structure in the process of achieving my country’s "dual-carbon" goal,but also an important basis for the government to formulate long-term energy development strategic plans.But in fact,energy demand is affected by many factors,and the amount of information contained in a traditional single model is very limited,which easily leads to a large deviation between the forecast result and the actual situation,thereby reducing the forecast accuracy.Therefore,this article comprehensively uses the combined model idea to construct a non-linear combined forecasting model of energy demand based on the Shapley value method and the inclusive test.The working steps and main conclusions of this paper are as follows:First of all,use qualitative analysis to analyze the current supply and demand situation of my country’s energy market and the factors affecting demand.On this basis,the core variables that have been sorted out are quantified,which lays the theoretical foundation for the construction of single and combined models in the following article.Secondly,by comparing and analyzing the advantages and disadvantages and applicable conditions of various energy demand forecasting methods,a single gray GM(1,1)model,multiple linear regression,BP neural network,partial least square regression,etc.,for medium and long-term forecasting of energy demand in my country is established.Model,and use the energy consumption from 1990 to 2019 as the observation value to test the validity of the single model.Thirdly,considering that the combination model has the advantages of integrating more effective information and improving the prediction accuracy,from the perspective of weight distribution;construct a combination model of equal weight and non-equal weight.In the non-equal weight combination model,the weight is calculated by Shapley value method and inclusive test.Finally,based on the combined model constructed in the previous article,forecasts of my country’s different types of energy demand from 2020 to 2030.The results show:(1)In the field of energy demand forecasting,BP neural network,partial least square regression,gray number and other dimensional recursive dynamic models are better than gray GM(1,1)model and multiple linear regression,with higher accuracy,The average relative error is less than 5%;(2)By comparing the data samples from 1990 to 2019,it is found that the non-equal weight combination model is better than the equal weight combination model in terms of forecasting accuracy,and can be used for medium and long-term forecasts of China’s energy demand;(3)By the end of2030,China’s total energy demand is expected to exceed 600,000 tons of standard coal.The total demand for traditional fossil energy such as coal,oil and natural gas will continue to increase in the short term.In the long term,the growth rate will be Will gradually slow down,and the consumption demand of renewable energy will show a significant exponential growth trend.