The Research On Simulation Of Urban Electrical Consumption In China Based On Satellite Light Data

Electric is the symbol of modern industrial,and electric planning is the core of energy planning and construction.City is the basic unit in making energy supply and demand policy in our country.And planning urban electric scientifically provides timely basis and data support for making policy and scheduling resources,which need to calculate the urban Electric Power Consumption(EPC)accurately and recognize the inherent relations between EPC and economic development before.Satellite remote sensing data can avoid the shortcomings of traditional statistics,such as missing data and inconsistent statistical scale,and its space-time characteristics are suitable for using in multi-scale and long time series studies.With the improvement and faster publish of satellite product,it has become a research hotspot to simulate city EPC with satellite nighttime light(NTL)data.The current research mainly has the following problems:(1)The two types of satellite night light images correction hypothesis lacks certain rationality,and they were discontinuous in 2013.(2)The reasonable selection of auxiliary data is particularly important for establishing the EPC simulation model and improving the accuracy.However,few studies discussed the basis and evaluation criteria in auxiliary data selection.(3)There are various classification methods for sample cities,but few researches involve machine learning algorithm and compare the simulation results of different clustering methods horizontally.This subject carried out the research work in aiming at the problems existing in the existing methods.First,it corrected the DMSP-OLS and NPP-VIIRS satellite remote sensing images of nighttime lights,and extracted the Nighttime Lights dataset of Chinese cities.Secondly,it constructed the simulation model includes NTL-EPC foundation model,single factor model and multi-factor model based on single year dataset,and found optimal model in unified evaluation standard.At the same time,discussed the influence of common socio-economic factors on EPC and speculated the urban development model.Finally,it proposed the K-Means city classification method,constructed NTL-EPC linear model after classified,and proved the feasibility of the classification method through comparing with the other models under traditional classification methods horizontally.The experimental results show that:1)Compared with DMSP-OLS satellite,NPP-VIIRS satellite nighttime light image was the most ideal data source in EPC simulation.NPP-VIIRS satellite NTL-EPC logarithm zone model was the optimal foundation model,the~2 of the model was 0.73,root mean square error(RMSE)was 71.75,the average relative error(?R??E?)was 44.51%,and there were 45.68%high-precision cities in sample cities.2)After adding social-economic factors,most simulation models became better.According to the simulation results,the overestimation of EPC in the loess plateau region and some eastern cities was alleviated,and the relative error of cities in northeast China was reduced significantly.The~2 of Optimal single factor model can reach 0.93,RMSE was 40.11,?R??E?was 25.09%,and there were 61.88%high-precision cities.Compared with the optimal foundation models,the relative error of 67.63%cities were reduced.The optimal multi-factor model is the zone model based on breadth-first search algorithm,the model of~2 was 0.96,RMSE was 21.36,?R??E?was 18.47%,and there were 69.78%high-precision cities.Compared with the optimal foundation models,the relative error of 67.27%cities were reduced.3)Combined with machine learning,this study proposed the K-Means city classification method based on light structure,and divided the sample cities into five types according to the urbanization level through the above classification method.Then established the NTL-EPC linear simulation model,the?R??E?of the simulation results was 32.02%,which decreased 25%than no classification.There were 53.99%high-precision cities,which increased 13.59%than no classification.And compared with no classification,the relative error of 57.79%cities were reduced.However,only 19 cities’relative error increased by more than 0.25 percentage point compared with no classification,which EPC were too small.Compared with the traditional classification method,the advantage of K-Means city classification method is that it can get similar or better simulation results only relying on the characteristics of the lighting itself,and can get rid of the constraints of statistical data.Compared with the traditional classification method,the advantage of this algorithm is that it can get similar or better simulation results with the previous methods only by relying on the characteristics of the lighting itself,and can get rid of the constraints of statistical data.4)While constructing the NTL-EPC model,this study discussed the influence of population,economy,industrial structure and other factors on EPC.The results showed:when NTL keep unchanged,the city’s EPC increased with the development of most social-economic factors,and with the accumulation of human activities and the increase of annual average temperature.Due to the special situation of central heating in China,the positive influence of annual average temperature on EPC is more obvious in non-heating cities.This study divided the urban development into three stages:(1)Initial development stage:population,economy and science technology all have positive influences on the growth of EPC,and the economic factors are the main influencing factors.(2)Talent development stage:population,economy and science technology all have positive influences on the growth of EPC,but the main influencing factors are changed to population factors.(3)Science technology development stage:the positive effect of population factors on EPC gradually decreases,and the main factors that have a positive effect on EPC in cities return to economic factors.The influence of science and technology factors on city EPC is negative.