中高分辨率卫星影像的装置级别工业热源识别

Method development for device-level industrial heat source identification using medium and high-resolution satellite images

针对精细化工业热源遥感监测困难与误差较大等问题,本文聚焦单个规模较大的工厂内部热场分布情况,提出一种基于中高分辨率卫星影像(Landsat 8)的装置级别工业热源的多层级识别方法。该方法基于双通道非线性劈窗算法反演的地表温度,利用多种空间统计分析方法分别对单期影像开展热源装置范围的识别检测,筛选出识别位置准确、识别能力稳定的方法,开展多时相热源装置识别分析,确定厂区范围内稳定出现的热源装置的位置,并结合高分辨率卫星影像进一步确定热源装置的边界范围。在某石油化工企业厂区的实验研究结果如下:5种方法均具有识别热源装置的能力。其中,温度1.5倍标准差分级与焦点统计分析方法适合捕捉厂区内最主要的热源装置,冷热点分析、温度1倍标准差分级方法以及聚类与异常值分析方法更适用于识别厂区内一般的热源装置。聚类与异常值分析方法识别结果正确检出率在80%以上,整体遗漏率低于10%,且各期识别结果受季节变化影响较小,更适用于识别装置级别(主要包括生产装置、循环水、锅炉装置、储存库、火炬装置以及罐区)的热源对象。识别的热源装置中,生产装置占比约55%,是厂区主要的热源装置。研究结果表明,中高分辨率卫星遥感影像数据源可以有效监测工业企业的精细尺度发热单元,为环境保护与管理、工业去产能监测提供技术支撑。

This study focuses on the internal thermal field distribution of a single large-scale factory to address the problems of insufficient refinement and large errors in the existing remote sensing monitoring of industrial heat sources. A device-level industrial heat source identification method based on medium and high-resolution satellite images is also proposed.The surface temperature is first obtained based on the dual-channel nonlinear split-window algorithm. Then, several spatial statistical analysis methods are used to identify candidate high-temperature areas, and the location of high-temperature devices in the plant is determined by multi-temporal superposition analysis of the identification results. By using high-resolution satellite images, the range of the high-temperature device is determined, the difference in the accuracy of the recognition results of several methods is compared and analyzed, and the method with the best recognition effect is ascertained.(1) Each method has the ability to identify heat source devices. The boundary of the standard deviation grading of temperature, that is,1.5×, and the focal statistical analysis method is clearer. The accuracy rate and fall rate of the cell point are more than 85%. However, the omission rate of the general heat source device is also high at more than 50%. Thus, it is suitable for capturing the most important heat source device in the factory area. The identification results of the cold and hot spot analysis method are relatively stable, but the range is larger than that of the actual heat source device, and the non-falling rate is 23%. The result is more misidentified pixels, but more heat source devices can be captured. The identification results of the temperature 1 standard deviation grading method and the clustering and outlier analysis method are closely aligned. The accuracy rate is about 82%, and the omission rate is within 10%, which is more suitable for identifying heat source devices in the plant area.(2) Comprehensive analysis of the identification effect of the five methods shows that cluster and outlier analysis methods have good identification ability, the correct detection rate of identification results is more than 80%, the overall omission rate is less than 10%, and the identification results of each period can be obtained. This method is less affected by seasonal changes and is more suitable for identifying general heat source devices in the factory.(3) Judging by the characteristics of high-resolution satellite images, the heat source device mainly includes production device, circulating water, boiler device, storage warehouse, flare device,and tank farm. Among them, the production device accounts for about 55%, which is the main heat source device in the factory. The difference between the average temperature of each type of device and the factory area shows that the production equipment is also the main high-temperature area.Medium and high-resolution satellite remote sensing image data sources can effectively monitor the fine-scale heating units of industrial enterprises. They can provide technical support for environmental protection and management and industrial overcapacity reduction monitoring.