Research On Carbon Emission Prediction Of Power Industry Based On Improved LSSVM

The electric power sector,which accounts for the highest proportion of carbon dioxide emissions in China,has become the focus of promoting carbon emission reduction.Reasonable and accurate analysis and prediction of carbon emissions in the electric power industry is conducive to the early realization of China’s carbon peak goal.The improved machine learning algorithm and scenario analysis method are combined to analyze and forecast the power demand in Hunan Province.On this basis,the carbon emissions of power in Hunan Province are predicted with other multi-source data,and policy suggestions are put forward for the carbon emission reduction of power.In this paper,from the perspective of power demand prediction,aiming at the problem of inaccurate subdivision of power demand influencing factors,the power demand is divided into household electricity consumption and industry electricity consumption,and the corresponding influencing factor sets are constructed respectively.Then,the random forest algorithm is used to screen each influencing factor.In order to solve the problems of slow convergence speed and weak searching ability of traditional optimization algorithms,the simulated annealing algorithm and particle swarm optimization algorithm are improved respectively,and the two improved algorithms are combined with the least square support vector machine regression algorithm to build the power demand prediction model.The results show that the prediction accuracy of the improved algorithm is much higher than other prediction algorithms,and the prediction results can provide data reference for the power demand development planning of Hunan Province.Next,aiming at the low accuracy of existing carbon emission prediction,on the basis of the boundary definition of carbon emissions in the power industry,the calculation of carbon emissions and the power demand forecast,a multi-source data combining the night light data and economy and energy data of carbon emissions prediction model of the electric power industry in Hunan province is proposed.The results show that the proposed model has lower error and better fitting effect.The prediction results are helpful to analyze the spatio-temporal evolution pattern of power carbon emissions,the relationship between power carbon emissions and social economy,and provide a reference for planning and decision-making of regional power carbon peak work.Finally,scenario analysis method was used to make reasonable assumptions on the future change rate of the influencing factors in Hunan Province,such as per capita GDP,population urbanization rate and the proportion of new energy power generation.Eight future scenario development modes were designed.The results showed that the power carbon emission of Hunan province will peak at 78,879,200 tons in 2032 under the baseline scenario,and 77,738,300 tons and 77,119,700 tons respectively in 2030 under the two ideal scenarios.The power carbon emission of Hunan Province will peak after 2035 under other scenarios.Based on the research results,it can be concluded that the future focus of Hunan province’s power carbon emission reduction is to further adjust the power generation structure,and accelerate the proportion of new energy generation and installed new energy.Then the spatial and temporal distribution characteristics of power carbon emission in key areas and coordinate regional carbon emission reduction should be paid attention to.Finally,the implementation of supporting policies can be considered when formulating carbon emission reduction policies in the future.