A multi-source heterogeneous identification algorithm for intelligent management of power plants

Abstract

The application relates to an identification algorithm, in particular to a multi-source heterogeneous identification algorithm for intelligent management of power stations, and belongs to the technical field of intelligent management of power stations.The intelligent management system is connected with an external server, the intelligent management system manages photovoltaic components through regional analysis and geographic coding, the intelligent management system comprises an information acquisition module, an image detection module, a regional control module and a user data platform, the information acquisition module is divided into an image acquisition module and a video acquisition module, is used for acquiring image and video information of the photovoltaic components, the information acquisition module is electrically connected with a photovoltaic component database through the image acquisition module and the video acquisition module, and the photovoltaic component database is used for storing data information of the photovoltaic components; the application can complete the inspection of a target region with extremely high efficiency, does not need a large amount of manpower and time to specially process detection and identification of such repetitive and monotonous labor, and improves the detection processing efficiency.

Description

TECHNICAL FIELD

[0001] The application relates to an identification algorithm, in particular a multi-source heterogeneous identification algorithm for intelligent management of power stations, and belongs to the technical field of intelligent management of power stations. BACKGROUND

[0002] With the continuous development of digitalization, technologies such as Internet of Things, big data and cloud computing have been integrated into various industries, promoting the development of cities towards smart cities. In the development process of smart cities, the photovoltaic components in the city need to be detected and managed. Most of the time, artificial neural networks and deep learning technologies are used to extract target features and learn according to the constructed data set, and feature extraction and inference are performed in newly photographed photovoltaic component images to screen and locate the photovoltaic component targets to be detected. At present, the publication number (CN110989685A) discloses an unmanned aerial vehicle cruising detection system and a cruising detection method thereof, which provides an unmanned aerial vehicle cruising detection method. In this method, the unmanned aerial vehicle flies in the substation based on a GPS-free navigation system. The method comprises the following steps: the unmanned aerial vehicle flies in the substation according to a preset route; the unmanned aerial vehicle is connected with a base station arranged in the substation to realize positioning; a detection device is carried on the unmanned aerial vehicle and moves with the unmanned aerial vehicle to obtain data at detection points in the substation; finally, the unmanned aerial vehicle or the detection device transmits the detected temperature or image data to a processing device in the substation, and the processing device processes the temperature or image data and makes a corresponding response. Compared with the conventional human patrol mode, the unmanned aerial vehicle inspection improves the inspection efficiency and reduces the inspection cost.

[0003] The publication number (CN105718866A) discloses a visual target detection and identification method, which comprises the following steps: extracting basic visual features from training samples to train a cascade classifier, obtaining a preliminary detection model of the target; extracting strong visual features from the training samples to train a strong classifier, obtaining a secondary discrimination verification model of the target; using a sliding window strategy to scan the image, obtaining a candidate image region; using the secondary discrimination verification model to perform secondary discrimination verification on the target candidate region, obtaining the final detection and identification result. The application uses multiple basic visual features in the preliminary detection process, enhances the robustness to complex application scenarios, and uses a cascade classifier to quickly filter out most of the non-target regions, greatly improving the detection speed. Then, the local features with stronger description ability and the strong classifier with better classification performance are used to perform secondary discrimination verification on the candidate region, further removing the non-target regions that are difficult to distinguish, effectively improving the accuracy of detection and identification.

[0004] However, the above two technical solutions can only analyze the properties of the target from one dimension, but when collecting and detecting photovoltaic module images, the different spectral characteristic properties of the photovoltaic module itself have high similarity, and it is difficult to locate the specific problem module through visual comparison, and a large amount of manpower is required to process and judge whether there is a target to be detected in the image and video shot by the unmanned aerial vehicle, the detection efficiency is not high, and there are certain disadvantages in the use process.

[0005] Therefore, a multi-source heterogeneous identification algorithm for intelligent management of power stations is provided. SUMMARY

[0006] The purpose of the present application is to provide a multi-source heterogeneous identification algorithm for intelligent management of power stations, which can complete the inspection of the target area with high efficiency, and can no longer need a large amount of manpower and time to specially process detection and identification of such repetitive and monotonous labor, effectively improving the efficiency of photovoltaic module detection and processing.

[0007] The present application realizes the above-mentioned purpose through the following technical scheme, a multi-source heterogeneous identification algorithm for intelligent management of power stations, comprising an intelligent management system, the intelligent management system is connected with an external server, the intelligent management system manages photovoltaic modules through region analysis and geocoding, the intelligent management system comprises an information collection module, an image detection module, a region control module and a user data platform, the information collection module is divided into an image collection module and a video collection module, and is used for collecting image and video information of the photovoltaic module, the information collection module is electrically connected with a photovoltaic module database through the image collection module and the video collection module;

[0008] The photovoltaic module database is used for storing data information of the photovoltaic module, and the photovoltaic module is connected with the image detection module, the image detection module is used for detecting and managing real-time module information collected by the information collection module and historical data in the photovoltaic module database, the photovoltaic module database is fused with an AI visual algorithm, which is used for evaluating the module from multiple dimensions throughout the entire life cycle;

[0009] The photovoltaic module database is connected with a data statistical module, the data statistical module is used for analyzing real-time data information and historical data information of the photovoltaic module database, the data statistical module is electrically connected with the region control module and the user data platform, the region control module is used for controlling the flight position of the unmanned aerial vehicle according to the module data statistical information, and the user data platform is used for displaying the monitored module information.

[0010] Further, in order to establish a diversified database through the component parameters and the array parameters, more accurate comparison between real-time data and historical data can be achieved, the photovoltaic component database includes component parameters and array parameters, the component parameters are used to obtain component manufacturer, component type, component model and component power information, and the array parameters are used to obtain factory environment, component installation mode, inverter type and component string power generation information.

[0011] Further, in order to organically integrate the AI visual algorithm with the photovoltaic component database, the photovoltaic component can be evaluated throughout the whole life cycle from multiple dimensions, and corresponding suggestions can be given combined with business needs, the component parameters and the array parameters are integrated with the AI visual algorithm, and are used to manage component surface dirt, hidden crack defects and hot spot defects.

[0012] Further, in order to detect the facility state through the component facility data quantity statistical unit and form an analysis report, the quantity information of the photovoltaic component can be more intuitively obtained, the photovoltaic component can be conveniently managed as a whole, and when the historical data and real-time data are displayed in the form of charts through the component data analysis display unit and the real-time data analysis display unit, the real-time data and the historical data can be conveniently compared, the change of the photovoltaic component can be conveniently monitored, the data statistical module includes a component facility quantity statistical unit, a component data analysis display unit and a real-time data analysis display unit, the component facility quantity statistical unit is used for state detection of the infrastructure, forms a data statistical analysis report, and displays data summary in the form of charts, the component data analysis display unit is used for statistical analysis of historical data through the photovoltaic component database, forms a regional data index, and the real-time data analysis display unit is used for real-time data statistics of the component facility, and displays the data in the form of charts, so that the dynamic of the photovoltaic component can be mastered and understood.

[0013] Further, in order to collect the image information of the photovoltaic component by the image detection module through the control unit, the abnormal information can be alarmed, the abnormal information can be excluded to make the image detection module collect the image information of the photovoltaic component again, the historical data information of the photovoltaic component corresponding to the real-time data can be more accurately collected, the image detection module collects and collects the image information of the photovoltaic component through the control unit, processes and outputs the image information of the photovoltaic component, and alarms the abnormal information.

[0014] Further, in order to enable the image detection module to collect images of photovoltaic modules with different characteristics and classifications through the control unit, to be flexible in use, to expand the detection range, and to control the region control module through the control unit to enable the unmanned aerial vehicle to fly to the corresponding region to collect photovoltaic module information, the control unit collects images according to the characteristics and classifications of the photovoltaic modules, packages real-time images and historical images of the photovoltaic modules, and enables the photovoltaic module image data to be transmitted to the region control module and the user data platform.

[0015] Further, in order to enable the image detection module to collect images of photovoltaic modules with different characteristics and classifications through the control unit, to be flexible in use, to expand the detection range, and to control the region control module through the control unit to enable the unmanned aerial vehicle to fly to the corresponding region to collect photovoltaic module information, the control unit collects images according to the characteristics and classifications of the photovoltaic modules, packages real-time images and historical images of the photovoltaic modules, and enables the photovoltaic module image data to be transmitted to the region control module and the user data platform.

[0016] Further, in order to enable the image detection module to collect images of photovoltaic modules with different characteristics and classifications through the control unit, to be flexible in use, to expand the detection range, and to control the region control module through the control unit to enable the unmanned aerial vehicle to fly to the corresponding region to collect photovoltaic module information, the control unit collects images according to the characteristics and classifications of the photovoltaic modules, packages real-time images and historical images of the photovoltaic modules, and enables the photovoltaic module image data to be transmitted to the region control module and the user data platform.

[0017] Technical effects and advantages of the present application: by adding component parameters and array parameters on the basis of the original visual detection algorithm, a diversified photovoltaic module database can be formed, the AI visual algorithm is organically combined with the photovoltaic module database, the monitoring and management of the photovoltaic module in multiple dimensions can be realized, the real-time data collected by the information collection module and the historical data in the photovoltaic module database can be automatically compared and detected through the image detection module, and the flight position of the unmanned aerial vehicle can be controlled through the region control module, the target area can be patrolled with extremely high efficiency, a large amount of manpower and time is no longer needed to specially process detection and identification of such repetitive and monotonous labor, and the efficiency of photovoltaic module detection and processing is effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is the overall framework diagram of the intelligent management system of the present application;

[0019] Figure 2 This is an overall framework diagram of the photovoltaic module database of this invention;

[0020] Figure 3 This is an overall framework diagram of the data statistics module of the present invention;

[0021] Figure 4 This is an overall framework diagram of the image detection module of the present invention;

[0022] Figure 5 This is an overall framework diagram of the area control module of the present invention;

[0023] Figure 6 This is a flowchart illustrating the usage of the intelligent management system of the present invention;

[0024] Figure 7 This is a flowchart of the infrared hot spot screening algorithm of the present invention;

[0025] Figure 8 This is a flowchart of the string identification and positioning algorithm of the present invention;

[0026] Figure 9 These are important formula diagrams related to the string identification and positioning algorithm of this invention; Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Please see Figures 1-6 As shown, a multi-source heterogeneous identification algorithm for intelligent power plant management includes an intelligent management system connected to an external server. The intelligent management system manages photovoltaic modules through regional analysis and geocoding. During use, the system allows for flexible selection of regions, making it applicable to photovoltaic module detection in different areas and effectively expanding its scope of application. The intelligent management system includes an information acquisition module, an image detection module, a regional control module, and a user data platform. The information acquisition module is divided into an image acquisition module and a video acquisition module, used to collect image and video information from the photovoltaic modules. During use, the operator can adjust the modules as needed, improving flexibility. The information acquisition module is electrically connected to a photovoltaic module database through the image and video acquisition modules. The collected photovoltaic module image and video information can be saved in the photovoltaic module database for easy comparison with historical data and can also be used as a comparison value for subsequent inspections.

[0029] The photovoltaic module database is used for storing data information of photovoltaic modules, and the photovoltaic modules are connected with the image detection module, which is used for detecting and managing the real-time module information collected by the information collection module and the historical data in the photovoltaic module database. In use, the image detection module collects the historical data in the photovoltaic module database through the control unit, which facilitates the comparison of real-time data and historical data while ensuring the accuracy of the compared data. The collection of historical data in the photovoltaic module database can be selected according to the characteristics and classification of photovoltaic modules, improving the data collection rate. The photovoltaic module database integrates AI vision algorithm, which is used to evaluate the components throughout the entire life cycle from multiple dimensions. Since each photovoltaic component defect feature has different degrees of relevance to the type, batch, power, and other characteristics of the component, and is closely related to environmental factors such as factory type and installation method, in use, multi-dimensional data such as component parameters and array parameters are added, and the corresponding database is established, which greatly improves the accuracy of the detection algorithm and reduces the possibility of false detection and missed detection.

[0030] The photovoltaic module database is connected with a data statistical module, which is used for analyzing the real-time data information and historical data information of the components in the photovoltaic module database. The data statistical module is electrically connected with the regional control module and the user data platform. The regional control module is used for controlling the flight position of the unmanned aerial vehicle according to the component data statistical information, and the user data platform is used for displaying the monitored component information. In use, the photovoltaic module image and video information are statistically analyzed by the data statistical module, and the image and video information data with discrepancies in the comparison results in the photovoltaic module database are transmitted to the user data platform, which facilitates the operator to locate the target photovoltaic module. The data meeting the requirements is transmitted to the regional control module, so that the regional control module detects other photovoltaic modules by the unmanned aerial vehicle, effectively improving the detection efficiency.

[0031] As shown in Figure 2 The photovoltaic module database includes component parameters and array parameters. The component parameters are used to obtain the component manufacturer, component type, component model, and component power information. The array parameters are used to obtain the factory environment, component installation method, inverter type, and component string power generation information. The component parameters and array parameters are integrated with the AI vision algorithm, which is used to manage the component surface dirt, hidden crack defects, and hot spot defects. In use, the photovoltaic module database formed by the component manufacturer, component type, component model, component power, factory environment, assembly installation method, inverter type, and component string power generation information is integrated with the AI vision, which effectively solves the problem that the traditional database only stores and manages the original data and cannot be analyzed in conjunction with business needs, and still requires manual analysis of a large amount of original data.

[0032] As shown inFigure 3 As shown, the data statistics module includes a component setting quantity statistics unit, a component data analysis display unit, and a real-time data analysis display unit. The component setting quantity statistics unit is used for state detection of the infrastructure, forms a data statistics analysis report, and displays data summary in a chart form. The component data analysis display unit is used for statistical analysis of historical data through the photovoltaic component database, forms a regional data index. The real-time data analysis display unit is used for real-time data statistics of the component infrastructure and displays in a chart form. The dynamic of the photovoltaic component is grasped and understood. The state of the photovoltaic component infrastructure, the real-time data analysis result, and the regional historical data information are displayed on the user data platform in a chart form. In use, the detection data of the photovoltaic component is more intuitively embodied, and automatic comparison of real-time data and historical data is facilitated, and the efficiency of automatic detection is improved.

[0033] As shown in Figure 4 The image detection module identifies and collects photovoltaic component image information through the control unit, processes and outputs the photovoltaic component image information, and alarms abnormal information. The control unit collects images according to the characteristics and classification of the photovoltaic component, packs real-time images and historical images of the photovoltaic component, and transmits the photovoltaic component image data to the regional control module and the user data platform. In use, the image information collection rate and accuracy of the photovoltaic component are improved by collecting according to the characteristics and classification of the photovoltaic component. The situation that the historical data collected in the photovoltaic component database does not correspond to the real-time data collected by the information collection module is effectively prevented, and the detection error is effectively prevented. The use effect is good.

[0034] As shown in Figure 5 The regional control module includes an airborne control unit and a ground control unit. The airborne control unit and the ground control unit are wirelessly connected through sensors. The airborne control unit and the ground control unit are used for controlling flight deviation of the unmanned aerial vehicle and monitoring flight state. The unmanned aerial vehicle is used as a carrier. The carrier has a universal quick-mount interface, can quickly replace the mounted equipment according to the task type, uses a visible light or multi-spectrum holder as a task load, is used for shooting images, finding a target to be detected, and adjusting the flight deviation of the unmanned aerial vehicle and adjusting the flight state of the unmanned aerial vehicle during the shooting of the images, thereby ensuring the accuracy of the shooting of the images and improving the image collection effect.

[0035] As shown in Figure 6As shown, the intelligent management system uses the process, which includes: first, starting the intelligent recognition software, starting the server to provide push and reporting services; second, starting the unmanned aerial vehicle, opening the live streaming function, and making the unmanned aerial vehicle start to perform the flight task; then making the server receive the unmanned aerial vehicle push stream and extracting pictures and videos at a certain frequency for detection and recognition; finally, after the server detects the target, the information is packaged and pushed to the user data platform. In use, it effectively solves the problem that the traditional unmanned aerial vehicle camera cannot automatically detect the captured picture, and a large amount of manpower is needed to process and judge whether there is a target to be detected. The traditional database only stores and manages the original data, and cannot be linked with business demand for analysis. Manual analysis of a large amount of original data is still needed. Image detection technology can only analyze the properties of the target from one dimension, but the different spectral characteristics of photovoltaic components have high similarity, and it is difficult to locate the specific problem of the component by visual comparison only.

[0036] As shown in Figure 7 The infrared hot spot screening algorithm needs to input an infrared thermal imaging picture with a complete photovoltaic panel as a sample. The photovoltaic panel in the sample picture needs to represent all the photovoltaic panels in the pictures to be screened, that is, the sample photovoltaic panel needs to have the average imaging style of all photovoltaic panels. Then, the rotation angle of the photovoltaic panel in the sample is automatically detected by means of a straight line detection algorithm, and the image is rotated by the corresponding angle to make the photovoltaic panel appear vertically or horizontally in the image. At this time, the four vertices of the photovoltaic panel are marked on the software by manual marking to accurately cut the photovoltaic panel sample. All pictures to be detected are read in turn, and the rotation angle of the photovoltaic panel is obtained by means of a straight line detection algorithm, and the corresponding rotation transformation is made, so that the photovoltaic panel in the detected image has invariance in rotation with the sample. Next, a template matching algorithm is used to extract the possible photovoltaic panel candidate box in the image, and a suitable threshold is selected by previewing the matching effect to filter the matched candidate box, so that the accuracy of the photovoltaic panel candidate box is the highest. After completing the above matching steps, only the photovoltaic panel type and the defect type to be identified need to be selected to start detection. The defect detection algorithm will extract the defect features according to the photovoltaic panel imaging and output the results.

[0037] As shown in Figure 8As shown, the string recognition positioning algorithm is trained by a large amount of data annotation on the AI semantic segmentation model, so that the AI model learns the semantic features of the photovoltaic panel, and then the photographed picture is input into the model during prediction to obtain the predicted photovoltaic panel area. However, the predicted photovoltaic panel area is relatively rough, and the gap between different panel blocks is easy to cause the adhesion between photovoltaic panels, so the prediction result cannot be directly used as the photovoltaic panel segmentation area. The predicted photovoltaic panel segmentation area is subjected to a graphics inflation operation, and the reliable area is fully expanded. Then, the advantages of the fine extraction features of the traditional image algorithm are used to extract the photovoltaic panel area, and the intersection of the photovoltaic panel area and the AI model prediction area after processing is obtained, so that a relatively fine photovoltaic panel area can be obtained under the condition of filtering out background interference.

[0038] Since only the GPS position of the center position of the picture is recorded during picture storage, the accurate photovoltaic panel center position cannot be obtained, so the string positioning algorithm is used to calculate the specific position information of the string. After the photovoltaic panel is extracted, the photovoltaic panel center pixel position can be located, and the pixel offset between the photovoltaic panel center position and the picture center position can be obtained. At this time, the camera parameters, the state of the unmanned aerial vehicle during shooting, and the camera projection model can be used to calculate the physical distance offset of the photovoltaic panel and the center, and the transformation formula is as shown in the formula. Figure 9

[0039] In use, first, according to the business requirements, the intelligent management system performs regional analysis and geographic coding search to determine the detection range, starts the unmanned aerial vehicle to open the push stream function, and makes the unmanned aerial vehicle start to perform the flight task through the area control module, so that the information acquisition module acquires the photovoltaic module image and video information, and the photovoltaic module real-time data acquisition information is transmitted to the photovoltaic module data. Through the image detection module, the data information corresponding to the real-time information collected by the information acquisition module in the photovoltaic module database is collected, and the photovoltaic module database and the AI vision algorithm are organically integrated. From multiple dimensions, the photovoltaic module is automatically compared, and the corresponding suggestions are given according to the business requirements. The use process no longer needs a large amount of manpower and time to specially process detection and identification of such repetitive and monotonous labor. At the same time, the unmanned aerial vehicle performs the shooting task, so that the target area is inspected with extremely high efficiency, the accuracy of the detection algorithm is greatly improved, and the possibility of false detection and missed detection is reduced.

[0040] ​It will be obvious to a person skilled in the art that the application is not limited to the details of the foregoing exemplary embodiments and can be implemented in other concrete forms without departing from the spirit or essential characteristics of the application. The embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. No reference signs in the claims should be considered as limiting the scope of the claims to the identity of the reference signs therein.

[0041] Furthermore, it should be understood that although the description is made on the basis of the embodiments, not every embodiment contains only one independent technical solution, and the description of the specification is only for the sake of clarity, and those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that those skilled in the art can understand.

Claims

1. A multi-source heterogeneous identification algorithm for intelligent management of power plants, comprising an intelligent management system, characterized in that: The intelligent management system is connected with an external server, the intelligent management system manages photovoltaic components through regional analysis and geocoding, the intelligent management system comprises an information acquisition module, an image detection module, a regional control module and a user data platform, the information acquisition module is divided into an image acquisition module and a video acquisition module, and is used for acquiring image and video information of the photovoltaic components, and the information acquisition module is electrically connected with a photovoltaic component database through the image acquisition module and the video acquisition module; The photovoltaic component database is used for storing data information of the photovoltaic components, and the photovoltaic components are connected with the image detection module, the image detection module is used for detecting and managing real-time component information collected by the information acquisition module and historical data in the photovoltaic component database, and the photovoltaic component database is fused with an AI vision algorithm, which is used for evaluating components throughout the entire life cycle from multiple dimensions; The photovoltaic component database is connected with a data statistical module, the data statistical module is used for analyzing real-time data information and historical data information of the components in the photovoltaic component database, the data statistical module is electrically connected with the regional control module and the user data platform, the regional control module is used for controlling the flight position of the unmanned aerial vehicle according to the component data statistical information, and the user data platform is used for displaying the monitored component information; the photovoltaic component database comprises component parameters and array parameters, the component parameters are used for acquiring component manufacturer, component type, component model and component power information, and the array parameters are used for acquiring factory environment, component installation mode, inverter type and component string power generation information. 2.The multi-source heterogeneous identification algorithm for intelligent management of power plants according to claim 1, wherein: The component parameters and the array parameters are fused with the AI vision algorithm, and are used for managing component surface dirt, hidden crack defects and hot spot defects. 3.The multi-source heterogeneous identification algorithm for intelligent management of power plants according to claim 1, wherein: The data statistical module comprises a component parameter statistical unit, a component data analysis display unit and a real-time data analysis display unit, the component parameter statistical unit is used for state detection of infrastructure, forms a data statistical analysis report, and displays data summary in a chart form, the component data analysis display unit is used for statistical analysis of historical data by connecting the photovoltaic component database, forms a regional data index, and the real-time data analysis display unit is used for real-time data statistics of component facilities, and displays data in a chart form, so that the dynamic of the photovoltaic components can be mastered and understood. 4.The multi-source heterogeneous identification algorithm for intelligent management of power plants according to claim 1, wherein: The image detection module identifies and acquires photovoltaic component image information through a control unit, processes and outputs the photovoltaic component image information, and alarms abnormal information. 5.The multi-source heterogeneous identification algorithm for intelligent management of power plants according to claim 4, characterized in that: The control unit identifies and acquires images according to characteristics and classification of the photovoltaic components, packs real-time images and historical images of the photovoltaic components, and sends photovoltaic component image data to the regional control module and the user data platform. 6.The multi-source heterogeneous identification algorithm for intelligent management of power plants according to claim 1, wherein: The area control module comprises an airborne control unit and a ground control unit, the airborne control unit and the ground control unit are wirelessly connected through sensors, and the airborne control unit and the ground control unit are used for controlling flight deviation of the unmanned aerial vehicle and monitoring flight states. 7.The multi-source heterogeneous identification algorithm for intelligent management of power plants of claim 1, wherein: The intelligent management system uses a process, which comprises the following steps: firstly, starting intelligent identification software, starting a server to provide push flow and reporting services; secondly, starting the unmanned aerial vehicle, opening the live streaming function, and making the unmanned aerial vehicle start to execute a flight task; then making the server receive the unmanned aerial vehicle push flow and detecting and identifying pictures and videos at a certain frequency; finally, after the server detects the required target, the information is packaged and pushed to a user data platform.

Images