Evaluation Of Rooftop Solar PV Potential Based On Multi-Source High-Resolution Remote Sensing Image

Fossil energy is dominant in the current energy consumption structure.According to2017 BP Statistical Review of World Energy,petroleum,which is the most important fuel,accounted for one-third of global primary energy consumption.Coal and natural gas accounted for 28%and 25%,respectively.With the rapid development of global economy,the demand of global energy will continue to grow.However,huge fossil energy consumption is gradually leading to the depletion of non-renewable fossil energy resources.In addition,the development and utilization of fossil energy is a key factor that causes global environmental change and pollution,which brings about serious ecological and environmental problems such as the greenhouse effect,the global warming and so on.Changing the current energy structure which is dominated by fossil energy and developing new energy forms are the strategic choices for most countries,especially for China.In the 13th Five-Year Plan of China,new energy industries and the related industries area were listed as key development and strategic emerging industries,and the development of these industries has received strong support from the Chinese government.Solar energy resources were widely promoted because of its advantages such as widely distribution,enrichment and pollution-free,etc.China has a vast territory and most of the country lies in the middle and low latitudes where the solar radiation is strong and solar resources are abundant,making China an appropriate country to develop and utilize solar energy resources.Roof photovoltaic?PV?power generation is the first priority of solar energy resources development because it is not limited by the geographical position and is tightly integrated with buildings without occupying additional land resources,which not only saves land resources,but also improves the utilization rate of land resources.National Energy Administration announced that the government aimed to reach the capacity of 100 GW of solar photovoltaic by 2020 in the 13th Five-Year Plan.The National Development and Reform Commission also issued a document stating that the scale of photovoltaic poverty alleviation during the“Thirteenth Five-Year Plan”period would achieve 15GW.In China,rooftop resource is rich because there are more than 40billion m² of existing residential buildings in China,and 2 billion m² of new buildings are built every year.However,the installed capacity of PV was just 6 million KW in 2015.Compared to other countries,the rooftop PV of China developed slowly,and the speed of development was far less than the expectations.The main reasons were the lack of building data,unclear PV rooftop,and the lack of evaluation methods of PV rooftop resources.At present,researches on solar energy resources in China are focused on the assessment of the richness and exploitable value of solar energy resources.Few scholars have carried out studies on the evaluation of rooftop solar PV potential.However,the high-precision solar rooftop PV potential map of the roof is the foundation of solar PV development,and is the key point to promote the large-scale development of rooftop solar PV.Therefore,in order to realize the rapid development of China's rooftop PV industry,it is urgent to develop and establish a scientific and feasible method for assessing the potential of distributed solar PVs on the rooftops of urban buildings,so as to promote the large-scale development and utilization of rooftop solar PV and promote the green,low-carbon and clean development of China's energy system.Based on the above requirements,this paper proposed a method to estimate the potential of rooftop solar PV through retrieving the rooftop features from multi-source high-resolution remote sensing satellite images.The key technologies involved in this method was optimized and innovated to realize the high-precision building rooftop solar energy PV potential evaluation.The main research contents are as follows:1.Extraction of rooftop outlines from high-resolution remote sensing imagesThe plane area of buildings rooftops is the basis for calculating the available area of rooftop PV.Therefore,this thesis was focused on the research of the object-oriented building extraction method.1)For the building extraction from high-resolution remote sensing images,an optimization approach for multi-resolution segmentation was proposed,which was integrated with height information and vegetation information.The optimized approach was used to realize high-precision high-resolution remote sensing image segmentation and improve the accuracy of image segmentation.2)After image segmentation,an optimal feature set for building identification were extracted.Based on the optimal feature set,the BP neural network was used to extract the building,and the outlines of buildings were generalized using the morphological method to calculate the high-precision plane area of the building rooftops.2.Rooftop PV potential assessment based on rooftop's feature parametersCombining the extraction result of rooftops and digital surface model?DSM?retrieved from stereo images,rooftop PV potential assessment based on feature parameters of rooftop was studied.1)Considering the different resolution of remote sensing images,different methods were proposed to extract the rooftop parameters of different types of buildings,which enlarged the application range of the rooftop PV potential evaluation method.2)According to the features of different types of rooftops,the parameters of rooftop PV system were estimated.Combining the PV system parameters and solar radiation of rooftops,the distribution of high precision and the building-scale rooftop PV potential were obtained.Based on the above research content,the three key technologies involved in this method was optimized and innovated,specifically as follows:1.Segmentation of high resolution remote sensing image with multiple featuresAccording to the characteristic of buildings,an image segmentation method integrated with height information and vegetation information was proposed to make full use of spectrum,shape and spatial information of images.Considering the sensitivity of height information and vegetation index for building,height information was used as a factor of heterogeneity and NDVI was used as an independent constraint conditions,to optimize the image segmentation.2.Buildings extraction based on the optimal feature set and the BP neural networkIn order to reduce the computational overhead of building extraction and to fully exploit the performance of the classifier,this paper proposed a feature selection solution of group corrected partial least squares generalized linear regression?G-PLSGLR?on the basis of partial least squares generalized linear regression?PLSGLR?.The objects with features of different ground type were selected as samples and the numbers of them were balanced by Bootstrap,then the features was grouped based on the Pearson correlation coefficient.The importance of Grouped features were ranked by PLSGLR regression coefficients and the insignificant features were removed.Finally,the optimal feature set which contained enough information for buildings extraction were selected by Bayesian information criterion?BIC?.Based on the optimal feature set,the building outlines were extracted from different resolution images using BP neural network.3.Rapid assessment of rooftop PV potential based on rooftop featuresDifferent resolutions of remote sensing images indicated direct impacts on the identification of rooftop types and the acquisition of parameters.For high spatial resolution remote sensing images,the probabilities of pitched and flat rooftop were calculated using the sampling method,and the rooftop features parameters of different rooftops were calculated.For very high spatial resolution remote sensing image,the slope and aspect were calculated based on DSM,the precise rooftop features parameters of different rooftops were calculated.The PV systems'parameters were estimated with the rooftop parameters.Finally,combining solar radiation parameters,a rapid evaluation of the rooftop PV potential was performedBy the above-mentioned research,the following conclusions were obtained:1.The optimized image segmentation method was implemented to 6 test scenes with different building densities and building types.The results obtained by contrastive analysis showed that the Area Fit Index?AFI?,Under-Segmentation?US?and Quality Rate?QR?of the optimized method were better than that before optimization.The results also showed that the image segmentation method integrated with multiple information could obtain more integral buildings which were always easily confused with other ground object?tree,road,etc.?,and the building edges were more obvious.This method improved the purity of objects and the segmentation accuracy.High precision segmentation results set foundation for the future precise building contour extraction.2.Based on the above segmentation results,the optimal feature set consist of GLCM_Mean,Density,DSM_Mean,Brightness and Ratio₁ were selected for building extraction using G-PLSGLR.The building outlines were extracted from different resolution images with the optimal feature set and BP neural network.Comparing the two building extraction results from Pleiades images using all features and optimal feature set,the two results showed little difference in accuracy.However,the time of features calculation and classification of the latter was less than the former,so the latter was more efficient.In addition,the building outlines were extracted from different resolution remote sensing images.The overall accuracies of the two results from Worldview-3 and Pleiades were 94%and 91%,respectively.Kappa coefficients reached 0.90 and 0.81,respectively.The two results both met the accuracy requirement.Building extraction with high efficiency and high precision provided a good foundation for rooftop photovoltaic potential assessment.3.The spatial distribution of building rooftop PV potential in the study area was estimated based on rooftop distribution,rooftop parameters,rooftop PV system parameters and solar irradiance.By comparing and analyzing the annual radiation of slope surface with different angles,the optimal installation angle were determined.For flat rooftops,the optimal installation angle of the PV array was 33°,the orientation was south.For high spatial resolution remote sensing images,the angle of the pitched rooftop was set as 20°and the orientation were set as south through sampling.The PV panels'areas of flat rooftop and pitched rooftop extracted from the Pleiades images were 2.4503km²and 1.0817 km²,respectively.The total annual solar radiation of study area was6813.67GWh,and the annual generating capacity was 1175.36GWh.For very high spatial resolution remote sensing image,the slope and aspect were calculated based on DSM,and they were used to identify rooftop types.The angle is the slope of pitched rooftop,and the orientation was same as that of the pitched rooftop.Up to 509 available rooftops with a total area of 853,910m² extracted from Worldview-3 images can be used to install photovoltaic system.The total annual solar radiation was 499.63GWh,and the annual generating capacity was 86.19GWh.The results presented that there would be great potential for the future development of the PV industry in the study area.In summary,for the problems of scattered rooftop distribution,inaccessible building data,and difficulty in accurate evaluation of rooftop PV potential,based on high-resolution remote sensing,a rapid assessment method for rooftop PV potential was proposed.The key technologies and methods were optimized to improve the accuracy of rooftop PV potential evaluation.With this method,a high-precision rooftop PV potential distribution map was obtained,which provided guidance for the future development of rooftop PV industry...