Robustness Of Photovoltaic Electric Field Information Extraction Based On Highresolution Image And Its Thermal Environment Effect

As one of representative of new energy, photovoltaic electric field is extending rapidly in Chi na, especially in desert area of the western China. It’s necessary to obtain photovoltaic electric field automatically, quickly and accurately and to analysis its thermal environment effect on the surrounding desert based on remote sensing image. It will not only help the government to grasp the dynamic information of photovoltaic electric field timely but also be of great significance to land cover, engineering construction, energy utilization and environmental protection. In this paper, combining with spectral, geometric and texture features, the expert knowledge was established based on object-based image analysis method through the segmentation parameter optimization.It would be used for extracting photovoltaic electric field in the form of rule sets based on the middle and high resolution remote sensing image. Then we discussed the impact of Phase, Sensor, Spatial resolution and Region differences on the robustness of the rule sets and gave advices when the rule sets were used in different remote sensing images. Finally, the influence of the photovoltaic electric field located in Ningxia Desert Photovoltaic Industry Park, Zhongwei City on the thermal environment of the surrounding desert was analyzed. The main conclusions are as follows:1. The optimal segmentation parameters of the photovoltaic electric field in the spatial resolution of 0.5-30 m can be obtained by the ED2 segmentation parameter optimization algorithm. The optimal segmentation scale(y) and spatial resolution(x) fit the function relationship(y = 300.98x-0.571) with high shape factor. It has great reference significance for the optimization segmentation parameters of the photovoltaic electric field.2. In the support of object-based image analysis method, two rule sets were established to extract photovoltaic electric field by using GF-1 and Landsat 8 OLI fusion image respectively. For the remote sensing images with spatial resolution higher than 2 m, the photovoltaic matrix could be extracted by using RRI, LSW, GLCM-Dissimilarity and brightness features. Then using the context and geometric features, we extract the internal roads and gaps. Finally, the extracted results were optimized by using mathematical morphology, and the comprehensive extraction of photovoltaic electric field was completed. The rule set had the advantages of few features, clear structure and high readability. The overall accuracy of the extraction was higher than 93%. For Landsat 8 OLI remote sensing image, most of the photovoltaic electric field can be extracted by using RRI, NDBI, NDVI, Brightness and standard deviation of the Blue band. The overall accuracy was slightly lower than that of high-resolution remote sensing image, between 85% and 90%.3. The influence of Phase, Sensor, Spatial resolution and Region factors on the robustness of the two rule sets were quantitatively analyzed. For the rule set 1, in the case of obtaining the optimal segmentation, the robustness of the rule set in different remote sensing images was phase>sensor>spatial resolution>region.(1). In the same region, for the remote sensing images that using the same sensor and spatial resolution, the rule set 1 had the highest robustness and could be reused directly, which was conducive to large-scale production;(2). In the same region, for the remote sensing images with same spatial resolution but different sensor, the rule set 1 had a better robustness. As long as the threshold value of the spectral features were adjusted properly, it could work smoothly.(3). For the remote sensing images with the same sensor and spatial resolution while in different regions, the rule set 1 had a better robustness. The threshold value of texture features should be adjusted properly before reuse.(4). Spatial resolution had the greatest influence on the robustness of rule set 1, reflecting in aspects such as texture, geometry and context. For rule set 2, the robustness of the rule set in different remote sensing images was Phase > Spatial resolution > Region. Compared with the rule set 1, the 3 factors had a great influence on the robustness of rule set 2. The rule set based on the high-resolution remote sensing image was more suitable for the automatic or semi-automatic extraction of the photovoltaic electric field.4. In this studied area, the average temperature of the photovoltaic electric field was 2.37℃lower than that of desert. The construction of photovoltaic electric field was advantageous to reduce the local temperature in the desert area. The area, perimeter and shape index of the photovoltaic electric field had a low correlation with temperature. Photovoltaic panels and vegetation were the factors that affect the cooling effect of the photovoltaic electric field in the desert region. The two factors were mutual influence and mutual gain, and the photovoltaic panel was in dominant. By analyzing the MODIS temperature day and night in the area of the Desert Photovoltaic Industrial Park, it was found that the cooling effect of the photovoltaic electric field on the desert was mainly concentrated in the daytime and easy to form a "cold island" center. While at night, the thermal environment of the desert area was less affected.