CN-115902508-B - Semantic segmentation and topology reconstruction-based rapid positioning method for heating of power equipment

Abstract

The invention relates to a quick positioning method for overheat faults of power equipment based on semantic segmentation and topology reconstruction. After marking the electric equipment by category, the in-phase identification of the electric equipment is realized by using the connected area analysis, whether the areas and the horizontal coordinates of the three-phase various electric equipment are all in an arithmetic array is calculated and judged, if the areas and the horizontal coordinates of the three-phase various electric equipment are all in the arithmetic array, various electric equipment in the in-phase is deduced according to the largest area phase, the smallest horizontal coordinate or the largest area in the various electric equipment, a local electric topology diagram is formed, the local electric topology diagram is compared with a transformer substation main wiring diagram, whether the in-phase connection relation of various electric equipment in the local electric topology diagram is correct is checked, and therefore the connection relation topology reconstruction of the electric equipment is realized, and fault positioning is realized based on the local electric topology diagram and the transformer substation main wiring diagram. The invention can rapidly realize the fault association and calculation of the in-phase equipment, provides a pre-diagnosis basis for fault diagnosis, and is convenient for rapid fault positioning.

Inventors

• LIU HE
• DING GUILI
• WANG ZONGYAO
• ZHAO TIANCHENG
• LI JIASHUAI
• YANG DAIYONG
• YU QUNYING
• CHEN JIEYUAN
• XU ZHIHAO
• SANG JINHAI
• KANG BING

Assignees

• 国网吉林省电力有限公司电力科学研究院
• 南昌工程学院

Dates

Publication Date

20260505

Application Date

20221031

Claims (4)

1. 1. The utility model provides a quick positioning method of power equipment heating based on semantic segmentation and topology reconstruction, which is characterized by comprising the following steps: Dividing power equipment in an infrared image by using a semantic dividing method, and marking different types of power equipment in the infrared image with different colors; step 2, performing binarization processing on the infrared image subjected to semantic segmentation, and respectively labeling the power equipment with different phase sequences according to the 1 st, 2 nd and 3 rd phases by using connected region analysis to realize in-phase identification of the power equipment; Step 3, solving the areas of the similar power equipment of each phase, and solving the arithmetic series of the areas of the adjacent two phases of the similar power equipment; step 4, solving the abscissa of the same kind of power equipment in each phase, and solving the arithmetic series of the abscissa of the adjacent two phases of the same kind of power equipment; Step 5, if the areas and the horizontal coordinates of all the three-phase power equipment are in an arithmetic series, deducing all the power equipment with the same phase according to the largest area phase and the smallest horizontal coordinate phase in all the power equipment to form a local electric topological graph, comparing the local electric topological graph with a transformer substation main wiring graph, checking whether the same phase connection relation of all the power equipment in the local electric topological graph is correct, if so, combining the local electric topological graph and the infrared image to form a power equipment topological association graph, if not, returning to the step 1 to analyze the next group of infrared images; And 6, marking the fault phase power equipment in the verified local electrical topological graph according to the fault diagnosis result of the infrared image, and marking and displaying the fault phase power equipment in the transformer substation main wiring graph according to the matching relation between the local electrical topological graph and the transformer substation main wiring graph, so that the fault phase power equipment can be rapidly positioned.
2. 2. The rapid positioning method of heating of power equipment based on semantic segmentation and topology reconstruction according to claim 1, wherein the connected region analysis uses a Two-Pass algorithm (Two-Pass algorithm), a first Pass sets a mark for all pixels of an infrared image, records to which connected region each mark belongs, and a second Pass marks each pixel as the connected region to which each mark belongs.
3. 3. The method for quickly positioning heating of power equipment based on semantic segmentation and topology reconstruction according to claim 1 is characterized in that in step 3, BFS breadth first search algorithm is used for obtaining areas of 1 st, 2 nd and 3 rd phase power equipment, and whether the areas of the corresponding power equipment are in an arithmetic series is calculated according to an arithmetic series judgment formula respectively, namely whether the following formulas are satisfied simultaneously: |S n1 -S n2 |＝|S n2 -S n3 | Wherein n represents the type of power equipment, S n1 、S n2 、S n3 is the area of the power equipment of the 1 st, 2 nd and 3 rd phases, The average area of the three-phase n-type power equipment is σ n , and the standard deviation of the area of the n-type power equipment is shown.
4. 4. The quick positioning method for heating of power equipment based on semantic segmentation and topology reconstruction according to claim 1, wherein in step 4, gray centroid method is used to obtain the abscissa of the 1 st, 2 nd and 3 rd phase power equipment, and whether the abscissa of the power equipment is in an arithmetic progression is calculated according to an arithmetic progression judgment formula, namely, whether the following formulas are satisfied at the same time: |x n1 -x n2 |＝|x n2 -x n3 | Wherein x n1 、x n2 、x n3 is the abscissa of the power equipment of the 1 st, 2 nd and 3 rd phases, The average abscissa of the three-phase n-th type power equipment is gamma n , and the abscissa standard deviation of the n-th type power equipment is gamma n .

Description

Semantic segmentation and topology reconstruction-based rapid positioning method for heating of power equipment Technical Field The invention relates to a quick positioning method for heating of power equipment based on semantic segmentation and topology reconstruction, and belongs to the field of image processing of power equipment of a transformer substation. Background In recent years, many monitoring technologies have been actively popularized and applied, and among them, infrared thermal imaging technologies are favored because of their advantages of no power failure, no sampling, no disassembly, and the like. However, the infrared image detection has limited analysis and judgment capabilities on defects, so that the state level of the intelligent detection equipment is limited. In order to improve the state level of the intelligent detection equipment, a more intelligent image processing method is adopted to divide the infrared image, and a required power equipment area is extracted so as to facilitate subsequent fault positioning and early warning. The color information in the infrared image corresponds to the temperature information of the equipment, so that the infrared image can be used as an important basis for early warning of the power equipment faults. In the existing power equipment fault analysis processing, fault diagnosis is realized through shape recognition mainly based on comparison analysis of fault images and normal images. The relevance among the devices is not considered, in fact, the device clusters are mutually affected, the relevance among the devices is found, and the relevance among the devices is mined, so that the rapid and accurate fault analysis is facilitated. Disclosure of Invention In view of the technical problems, the invention provides a rapid positioning method for heating of power equipment based on semantic segmentation and topology reconstruction. The technical scheme adopted by the invention is that the method for rapidly positioning the heating of the power equipment based on semantic segmentation and topology reconstruction comprises the following steps: Dividing power equipment in an infrared image by using a semantic dividing method, and marking different types of power equipment in the infrared image with different colors; step 2, performing binarization processing on the infrared image subjected to semantic segmentation, and respectively labeling the power equipment with different phase sequences according to the 1 st, 2 nd and 3 rd phases by using connected region analysis to realize in-phase identification of the power equipment; Step 3, solving the areas of the similar power equipment of each phase, and solving the arithmetic series of the areas of the adjacent two phases of the similar power equipment; step 4, solving the abscissa of the same kind of power equipment in each phase, and solving the arithmetic series of the abscissa of the adjacent two phases of the same kind of power equipment; Step 5, if the areas and the horizontal coordinates of all the three-phase power equipment are in an arithmetic series, deducing all the power equipment with the same phase according to the largest area phase, the smallest horizontal coordinate or the largest area phase in all the power equipment to form a local electric topological graph, comparing the local electric topological graph with a transformer substation main wiring graph, checking whether the same phase connection relation of all the power equipment in the local electric topological graph is correct, if so, combining the local electric topological graph and the infrared image to form a power equipment topological association graph, if not, returning to the step 1 to perform the next group of infrared image analysis; And 6, marking the fault phase power equipment in the verified local electrical topological graph according to the fault diagnosis result of the infrared image, and marking and displaying the fault phase power equipment in the transformer substation main wiring graph according to the matching relation between the local electrical topological graph and the transformer substation main wiring graph, so that the fault phase power equipment can be rapidly positioned. Further preferably, the connected region analysis uses a Two-Pass algorithm (Two-Pass algorithm), where the first Pass sets a flag for all pixels of the infrared image, records to which connected region each flag belongs, and the second Pass marks each pixel as belonging to the connected region. Further preferably, in step 3, the areas of the 1 st, 2 nd and 3 rd phase power devices are obtained by using a BFS breadth first search algorithm, and whether the areas of the corresponding power devices are in an arithmetic series or not is calculated according to an arithmetic series judgment formula, namely, whether the following formulas are simultaneously satisfied: Wherein n represents the type of power equipment, S n1、Sn2、Sn3 is the a