Greenhouse digital twinning and equipment operation and maintenance system based on knowledge graph and data driving

Abstract

The invention discloses a greenhouse digital twinning and equipment operation and maintenance system based on a knowledge graph and data driving, which comprises a physical greenhouse, a greenhouse digital twinning platform and a greenhouse knowledge graph, and is characterized in that the physical greenhouse comprises an environment information acquisition system, a video acquisition system and an equipment state recognition model which are deployed in the physical greenhouse; the greenhouse digital twin platform is used for deploying the control platform, the virtual greenhouse scene model and the fault judgment module; the greenhouse knowledge graph performs information extraction on data acquired by a physical greenhouse, and converts acquired greenhouse environment information, greenhouse structure information, equipment association information and equipment state information into structured data information so as to facilitate identification and processing of a greenhouse digital twin platform. According to the invention, the state operation of the greenhouse equipment can be analyzed, and the fault diagnosis and processing efficiency of the greenhouse equipment can be improved, so that the overall management level of the greenhouse and the agricultural production efficiency are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of greenhouse equipment operation and maintenance, and relates to a greenhouse equipment operation and maintenance system that uses knowledge graphs to coordinate greenhouse data to build a greenhouse digital twin platform, and drives the system in real time through greenhouse environmental data and equipment operating status data detected by visual analysis. Background Art

[0002] With the growth of the global population and the increasing demand for food, traditional agriculture is facing unprecedented pressure. In order to improve agricultural production efficiency and product quality, smart agricultural technology has gradually become a research hotspot. As an important part of modern agriculture, greenhouse cultivation and its environmental control directly affect the growth and yield of crops. At present, the operation and maintenance of greenhouse equipment mostly rely on manual inspection, which is inefficient and cannot reflect the actual operating status of the equipment in real time, and cannot locate the fault link in time. Therefore, there is an urgent need for a system that can automatically and accurately detect the operating status of the equipment and perform fault maintenance to improve the level of intelligent greenhouse management.

[0003] The introduction of digital twin technology provides a new idea for the operation and maintenance of greenhouse equipment. Digital twin technology simulates and manages the physical system by constructing a virtual scene model of the physical greenhouse and keeping it synchronized with the physical greenhouse in real time. Through simulation and analysis, it optimizes the environmental control strategy and improves the operating efficiency and stability of the system. However, the current application of digital twin technology in greenhouses also faces some shortcomings. The virtual greenhouse scene in the digital twin mainly relies on manual construction, and the mutual mapping relationship between physical equipment and virtual equipment models is also mainly set manually. When the scale of the greenhouse and the number and type of equipment are large, the difficulty and cost of manual setting will increase significantly, which limits the efficient application of the technology. In this context, how to improve the efficiency of building virtual scene models in digital twins and realize the automatic completion of mutual mapping between physical equipment and virtual equipment has become an urgent problem to be solved. Summary of the invention

[0004] In view of the shortcomings in the prior art, the present invention provides a greenhouse digital twin and equipment operation and maintenance system based on knowledge graph and data-driven. Through the greenhouse knowledge graph, various data of the physical greenhouse are coordinated and a digital twin platform of the physical greenhouse is constructed, so that it can monitor and analyze the operating conditions of various equipment in the physical greenhouse in real time and accurately. The simulated operation results of the digital twin platform are used to compare and analyze the fault links, thereby improving the operational efficiency of the greenhouse and the safety of crop production.

[0005] The present invention achieves the above technical objectives through the following technical means.

[0006] A greenhouse digital twin and equipment operation and maintenance system based on knowledge graph and data-driven includes a physical greenhouse, a greenhouse digital twin platform and a greenhouse knowledge graph; the physical greenhouse includes an environmental information acquisition system, a video acquisition system and an equipment status recognition model deployed therein, the greenhouse digital twin platform is used to deploy a control platform, a virtual greenhouse scene model and a fault judgment module, the greenhouse knowledge graph extracts information from data acquired from the physical greenhouse, and converts the acquired greenhouse environmental information, greenhouse structure information, equipment association information and equipment status information into structured data information, so as to facilitate recognition and processing by the greenhouse digital twin platform.

[0007] In the above technical solution, the greenhouse knowledge graph adopts the triple format, including two categories: entity-relationship-entity and entity-attribute-attribute value, and is stored in the Neo4j graph database, where "relationship" is the spatial relationship and subordinate relationship between entities in the greenhouse scene described in the knowledge graph, and "attribute" includes the geometric attributes, position attributes and functional attributes of the entities in the greenhouse scene.

[0008] In the above technical solution, the greenhouse environment information is obtained by the environmental parameters acquired by the environmental information acquisition system as entities, and the corresponding triple expressions include: structured information extracted based on sensor data information and representation of the mutual influence between environmental parameters.

[0009] In the above technical solution, the greenhouse structure information is obtained by extracting information from standards / books related to greenhouse structures, and then defining entities, relationships and attributes for the overall structure of the greenhouse. Starting from the structural type of the greenhouse, each type of greenhouse structure is physically divided according to greenhouse enclosure structures, greenhouse equipment and greenhouse crops.

[0010] In the above technical solution, the device association information is used to determine the position of the device to be detected in the virtual greenhouse scene, and then establish a mapping relationship between the equipment in the physical greenhouse and the equipment model in the virtual greenhouse scene. The Neo4j graph database is accessed through Unity3D software, and the "relationship" and "attributes" of the physical camera in the knowledge graph are matched to the virtual camera in the corresponding virtual greenhouse scene, so as to realize the matching of the virtual greenhouse equipment model with the real equipment under the physical camera.

[0011] In the above technical solution, in the Unity3D scene, the physical camera position information stored in the Neo4j graph database is queried, and the position information and viewing angle of the physical camera are obtained through Cypher query; in the Unity3D software, a corresponding virtual camera is created for each physical camera; through the C# script, the position information of the physical camera in the Neo4j graph database is mapped to the virtual camera of the Unity3D software to ensure that the coordinates and direction of the virtual camera are consistent with the physical camera; then the view calculation method in the Unity3D software is used to determine the visible device model to be detected under the perspective of the virtual camera; the device to be detected under the field of view of the physical camera is determined through the relationship and attributes of the device model to be detected, thereby realizing the matching mapping of the device model in the virtual greenhouse model and the equipment in the physical greenhouse.

[0012] In the above technical solution, the equipment status information is mainly obtained by the equipment status recognition model to obtain the current status of the equipment to be detected in the physical greenhouse. The structured data obtained through information extraction is used in the fault diagnosis module to compare with the virtual greenhouse operation results to determine the fault.

[0013] In the above technical solution, the virtual greenhouse scene model is constructed in the following way:

[0014] Conduct a survey on the physical greenhouse site, extract the size, structure and material of the greenhouse enclosure, greenhouse crops and greenhouse equipment, build and render a three-dimensional model, and form a virtual greenhouse model library of various greenhouse elements;

[0015] Read the description of the virtual greenhouse scene generation logic from the knowledge graph in the Neo4j graph database, that is, the greenhouse structure information, build the corresponding scene generation script in the Unity3D software, and then set the corresponding geometric dimensions and spatial positions for each virtual device and structure based on the "relationships" and "attributes" of each greenhouse entity constructed in the knowledge graph, and adjust the models in the virtual greenhouse model library according to the positions and sizes obtained by the script, so that the layout of the virtual greenhouse scene is the same as the actual one.

[0016] In the above technical solution, the fault judgment module specifically includes:

[0017] Digital twin greenhouse operation: The control platform in the greenhouse digital twin platform collects and analyzes various sensor data in real time and makes decisions based on these data. The control platform sends instructions to the control cabinet of the physical greenhouse, requiring it to perform corresponding operations. While sending instructions, the corresponding operations are synchronously performed in the virtual greenhouse scene model.

[0018] The communication information received by the control cabinet in the virtual greenhouse scene and the operating status of the equipment model are compared with the corresponding control cabinet and equipment in the physical greenhouse; if the operating status of the equipment model in the virtual greenhouse scene is inconsistent with the operating status of the equipment in the physical greenhouse, it is considered that the equipment has sent a fault, and the specific link of the fault is further confirmed;

[0019] Check whether the control cabinet in the physical greenhouse receives the communication information of the control platform. If the control cabinet in the physical greenhouse does not receive the information, there may be a communication loss failure, and the fault link is located between the control platform and the control cabinet in the physical greenhouse. Otherwise, determine whether the control instructions issued by the control cabinet in the virtual greenhouse scene are consistent with the instructions issued by the control cabinet in the physical greenhouse. If they are inconsistent, the control cabinet in the physical greenhouse fails, and the fault link is located in the control cabinet in the physical greenhouse. Conversely, if the instructions issued by the control cabinet are consistent, but the equipment in the physical greenhouse is not running or the running status is opposite to the instructions, there is an equipment failure or the control instruction fails to be executed successfully, and the fault link is located in the equipment in the physical greenhouse.

[0020] In the above technical solution, the device state recognition model is constructed in the following way: the operation video of the device to be detected is imported into the video segmentation software, and a picture library is generated by extracting video frames; then, the devices to be detected in the picture are classified according to type and operation status, and various states are marked by box selection using a labeling tool to form a data set; after the labeling is completed, the data set is divided into a training set, a validation set and a test set, the data set is used to train the target detection model, and the model is optimized according to the training results; finally, the optimized target detection model is used as the device state recognition model.

[0021] The beneficial effects of the present invention are:

[0022] (1) The present invention coordinates the physical greenhouse information by constructing a greenhouse knowledge graph, and performs a structured representation of the environmental information, structural information, equipment-related information, and status information of the physical greenhouse, thereby realizing the construction of a greenhouse digital twin platform, and real-time response to the operating status of the physical greenhouse equipment through data-driven. The knowledge graph technology processes the data in a structured manner, so that the model in the virtual greenhouse scene can automatically build a greenhouse virtual scene model according to the correspondence between relationships and attributes. At the same time, the mapping relationship between devices is automatically identified and updated, making the mapping process more efficient and easy to manage, reducing the difficulty and cost caused by the increase in the number and complexity of equipment.

[0023] (2) The present invention creates cameras in the corresponding virtual greenhouse scene based on the relationship and attributes of cameras in the physical greenhouse in the knowledge graph, uses the cone detection of the virtual camera to determine the device model in the virtual scene, and then determines the corresponding device to be detected in the field of view of the camera in the physical greenhouse based on the relationship and attributes of the device model, thereby realizing the automatic matching and mapping of the device model in the virtual greenhouse model and the equipment in the physical greenhouse, reducing the difficulty and cost brought by manual matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the greenhouse digital twin and equipment operation and maintenance system based on knowledge graph and data drive according to the present invention;

[0025] Figure 2 It is a schematic diagram of the mapping relationship between the physical greenhouse equipment of the present invention and the virtual greenhouse equipment model in the greenhouse digital platform;

[0026] Figure 3 This is a flow chart of greenhouse equipment fault detection based on the greenhouse digital twin platform described in the present invention. DETAILED DESCRIPTION

[0027] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.

[0028] like Figure 1 As shown, a greenhouse digital twin and equipment operation and maintenance system based on knowledge graph and data-driven includes three parts: a physical greenhouse, a greenhouse digital twin platform and a greenhouse knowledge graph. Among them, the physical greenhouse includes an environmental information acquisition system, a video acquisition system and an equipment status recognition model deployed therein; the greenhouse digital twin platform is used to deploy a control platform, a virtual greenhouse scene model and a fault judgment module, and drives the virtual greenhouse scene model through greenhouse twin data to realize condition detection and fault judgment of equipment in the physical greenhouse; the greenhouse knowledge graph is used to extract information from the data obtained from the physical greenhouse, and convert the obtained greenhouse environmental information, structural information, equipment association information, and equipment status information into structured data information to facilitate recognition and processing by the greenhouse digital twin platform. The details are as follows:

[0029] 1. Physical Greenhouse

[0030] (1) Construction of environmental information collection system

[0031] The environmental information collection system mainly includes various sensors deployed inside the greenhouse, including temperature sensors, humidity sensors, light intensity sensors, carbon dioxide concentration sensors and soil moisture sensors, etc., to ensure that greenhouse environmental data can be collected in real time.

[0032] (2) Video acquisition system construction

[0033] The video acquisition system uses Hikvision cameras. After selecting the equipment to be tested in the physical greenhouse, the camera is set up above the physical greenhouse. The angle of the camera is adjusted to be able to capture the equipment to be tested and record the operation video of the equipment to be tested in real time.

[0034] (3) Construction of equipment status recognition model

[0035] After obtaining the operation video of the device to be detected, the operation video is manually segmented, the collected operation video is imported into the video segmentation software, and a picture library is generated by extracting video frames; then, the devices to be detected in the picture are classified according to type and operation status (on and off), and various states are marked by box selection using annotation tools to form a data set; after the annotation is completed, the data set is divided into a training set, a validation set, and a test set. By selecting a target detection algorithm suitable for video analysis requirements, the target detection model is trained using the data set, and the model is optimized based on the training results; finally, the optimized target detection model is used as the device status recognition model to ensure that it can recognize the device status in real time.

[0036] 2. Greenhouse Knowledge Graph

[0037] The greenhouse knowledge graph is mainly composed of structured data converted from greenhouse environment information, greenhouse structure information, equipment association information, and equipment status information. According to the construction process of the knowledge graph, a top-down construction method is selected. The overall knowledge graph adopts a triple format, including entity-relationship-entity and entity-attribute-attribute value, and is stored in the Neo4j graph database. The "relationship" is mainly the spatial relationship attribute and subordinate relationship attribute between the entities in the greenhouse scene described in the knowledge graph, and the "attribute" includes the geometric attributes, position attributes and functional attributes of the entity.

[0038] (1) Greenhouse environmental information is mainly composed of environmental parameters (including temperature, humidity, light intensity, etc.) obtained by the environmental information collection system. Taking temperature as an example, the corresponding triples are mainly expressed as follows: greenhouse environmental information-including-temperature, temperature-value-23℃, etc., which are structured information extracted from sensor data information. It also includes representations of the mutual influence between entities such as temperature, humidity, and air pressure, such as temperature-inversely proportional to-humidity, etc., which are based on physical theorems or information extracted from relevant standards and books.

[0039] (2) The greenhouse structure information is extracted from the standards and books related to greenhouse structure, and then the entities, relationships and attributes of the overall structure of the greenhouse are defined; starting from the structural type of the greenhouse, each type of greenhouse structure is divided into entities according to the greenhouse enclosure structure, greenhouse equipment and greenhouse crops; the subordinate relationship and spatial relationship of each part of the greenhouse are structured to determine the relative orientation and combination of each entity model when generating the virtual greenhouse scene; at the same time, the structured representation of the corresponding entity attributes (including geometric attributes, position attributes and functional attributes) is combined to determine the actual size and absolute position of each entity model, so as to realize the rapid generation of the virtual greenhouse scene in the visualization software Uniyt3D. Taking the greenhouse skylight as an example, the main triple expression forms are: subordinate relationship expression: skylight-belongs to-top of greenhouse; spatial relationship expression: skylight-below-external shading; attribute-related expressions such as: skylight-number-5, skylight-coordinate-(x, y, z).

[0040] (3) The device association information is used to determine the position of the device to be detected in the virtual greenhouse scene, and then establish a mapping relationship between the device model in the physical greenhouse and the device model in the virtual greenhouse scene, laying the foundation for subsequent state identification and fault diagnosis. The structure is as follows: Figure 2 The main process is to access the Neo4j graph database through Unity3D software, and match the "relationship" and "attribute" of the physical camera in the knowledge graph to the virtual camera in the corresponding virtual greenhouse scene, so as to further match the virtual greenhouse equipment model with the real equipment under the physical camera. The specific steps are as follows:

[0041] First, in the Unity scene, query the physical camera location information stored in the Neo4j graph database, and obtain the physical camera location information and viewing angle (such as x, y, z coordinates and orientation information) through Cypher query; in Unity3D software, create a corresponding virtual camera for each physical camera; through C# script, map the location information of the physical camera in the Neo4j graph database to the virtual camera of Unity3D software to ensure that the coordinates and direction of the virtual camera are consistent with the physical camera; then use the view calculation methods such as cone detection and ray detection in Unity3D software to determine the visible device model to be detected under the perspective of the virtual camera; determine the device to be detected under the field of view of the physical camera through the relationship and attributes of the device model to be detected, thereby realizing the matching mapping between the device model in the virtual greenhouse model and the device in the physical greenhouse. Taking the fan as an example, the triple expression form is such as: fan model 1-mapping-fan 1.

[0042] (4) Equipment status information is mainly the status of the equipment to be detected in the current physical greenhouse obtained through the equipment status recognition model. The structured data obtained through information extraction is used in the fault diagnosis module to compare with the operation results of the virtual greenhouse to determine the fault. Taking the fan as an example, the triple expression form is as follows: fan-operating status-open class.

[0043] 3. Greenhouse Digital Twin Platform

[0044] (1) Construction of virtual greenhouse scene model

[0045] Collect various elements of the greenhouse, build its virtual model to form a virtual greenhouse model library, and generate virtual greenhouse scenes in Unity3D software through the description of greenhouse structure and equipment related information in the knowledge graph; the various elements of the greenhouse include: greenhouse enclosure model, mainly greenhouse roof, walls, columns and other enclosure frames; greenhouse equipment 3D model, mainly greenhouse fans, greenhouse skylights, shading nets and other greenhouse working equipment; greenhouse crop model, mainly the 3D model of greenhouse crops and the attached planting media (soil and flower pots, etc.). The details are as follows:

[0046] ① Establish a virtual greenhouse model library; conduct a survey on the physical greenhouse site, extract the size, structure and material of the greenhouse enclosure structure, greenhouse crops and greenhouse equipment, use Solidworks 3D modeling software to build a 3D model, and then render the model to improve the similarity between the 3D model and the real scene, forming a virtual greenhouse model library of various greenhouse elements.

[0047] ②Generation of virtual greenhouse scene model: Install and configure the Neo4j plug-in in the Unity3D software so that it can connect to the Neo4j graph database, read the description of the virtual greenhouse scene generation logic (i.e., greenhouse structure information) in the knowledge graph in the graph database, and build the corresponding scene generation script in the Unity3D software. Then, based on the "relationships" and "attributes" of the greenhouse entities constructed in the knowledge graph, set the corresponding geometric dimensions and spatial positions for each virtual device and structure, and adjust the models in the 3D model library according to the positions and sizes obtained by the script, so that the layout of the virtual greenhouse scene is the same as the actual one.

[0048] (2) Control platform construction

[0049] The main function of the control platform is to be used for decision analysis. The collected environmental data is input into the virtual greenhouse scene model. According to the collected environmental parameters, analysis and decision-making are performed through the decision model in the control platform (specifically the existing technology), the operation requirements of the greenhouse equipment model are judged and corresponding control instructions are issued, and the instructions are sent to the control cabinets in the physical greenhouse and virtual greenhouse scene models.

[0050] (3) Fault diagnosis module

[0051] ① Operation of digital twin greenhouse

[0052] a. Control platform data analysis and decision-making: The control platform in the greenhouse digital twin platform collects and analyzes various sensor data in real time and makes decisions based on these data; for example, if the temperature sensor in the greenhouse detects high temperature, the control platform will decide to start the cooling equipment;

[0053] b. Implement communication and control physical equipment: The control platform sends instructions to the control cabinet of the physical greenhouse, requiring it to perform corresponding operations, such as starting the cooling equipment;

[0054] c. Virtual greenhouse synchronous simulation: When sending instructions, the corresponding operations will be executed synchronously in the virtual greenhouse scene model to simulate the operation of the cooling equipment.

[0055] ②Fault detection

[0056] The communication information received by the control cabinet in the virtual greenhouse scene and the operating status of the equipment model are compared with the corresponding control cabinet and equipment in the physical greenhouse; if the operating status of the equipment model in the virtual greenhouse scene is inconsistent with the operating status of the equipment in the physical greenhouse, the equipment is considered to have a fault, and the specific link of the fault is further confirmed, such as Figure 3 As shown, the specific process is:

[0057] Check whether the control cabinet in the physical greenhouse receives the communication information of the control platform. If the control cabinet in the physical greenhouse does not receive the information, there may be a communication loss failure, and the fault link is located between the control platform and the control cabinet in the physical greenhouse. Otherwise, determine whether the control instructions issued by the control cabinet in the virtual greenhouse scene are consistent with the instructions issued by the control cabinet in the physical greenhouse. If they are inconsistent, the control cabinet in the physical greenhouse fails, and the fault link is located in the control cabinet in the physical greenhouse. Conversely, if the instructions issued by the control cabinet are consistent, but the equipment in the physical greenhouse is not running or the running status is opposite to the instructions, there is an equipment failure or the control instruction fails to be executed successfully, and the fault link is located in the equipment in the physical greenhouse.

[0058] The embodiments are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention.

Claims

1. A greenhouse digital twin and equipment operation and maintenance system based on knowledge graph and data-driven, characterized in that: It includes a physical greenhouse, a greenhouse digital twin platform and a greenhouse knowledge graph; the physical greenhouse includes an environmental information acquisition system, a video acquisition system and an equipment status recognition model deployed therein; the greenhouse digital twin platform is used to deploy a control platform, a virtual greenhouse scene model and a fault judgment module; the greenhouse knowledge graph extracts information from data acquired from the physical greenhouse, and converts the acquired greenhouse environmental information, greenhouse structure information, equipment association information and equipment status information into structured data information to facilitate recognition and processing by the greenhouse digital twin platform.

2. The greenhouse digital twin and equipment operation and maintenance system according to claim 1, characterized in that: The greenhouse knowledge graph adopts the triple format, including entity-relationship-entity and entity-attribute-attribute value, and is stored in the Neo4j graph database, where "relationship" is the spatial relationship and subordinate relationship between entities in the greenhouse scene described in the knowledge graph, and "attribute" includes the geometric attributes, position attributes and functional attributes of the entities in the greenhouse scene.

3. The greenhouse digital twin and equipment operation and maintenance system according to claim 2 is characterized in that: The greenhouse environment information is an entity of environmental parameters acquired by the environmental information acquisition system, and the corresponding triplet expressions include: structured information extracted based on sensor data information and representation of the mutual influence between environmental parameters.

4. The greenhouse digital twin and equipment operation and maintenance system according to claim 2, characterized in that: The greenhouse structure information is obtained by extracting information from standards / books related to greenhouse structures, and then defining entities, relationships and attributes for the overall structure of the greenhouse. Starting from the structural type of the greenhouse, each type of greenhouse structure is physically divided according to greenhouse enclosure structures, greenhouse equipment and greenhouse crops.

5. The greenhouse digital twin and equipment operation and maintenance system according to claim 2, characterized in that: The device association information is used to determine the position of the device to be detected in the virtual greenhouse scene, and then establish a mapping relationship between the device in the physical greenhouse and the device model in the virtual greenhouse scene. The Neo4j graph database is accessed through the Unity3D software, and the "relationship" and "attribute" of the physical camera in the knowledge graph are matched to the virtual camera in the corresponding virtual greenhouse scene, so as to achieve the matching of the virtual greenhouse device model with the real device under the physical camera.

6. The greenhouse digital twin and equipment operation and maintenance system according to claim 5, characterized in that: In the Unity3D scene, query the physical camera location information stored in the Neo4j graph database, and obtain the location information and viewing angle of the physical camera through Cypher query; in the Unity3D software, create a corresponding virtual camera for each physical camera; through the C# script, map the location information of the physical camera in the Neo4j graph database to the virtual camera of the Unity3D software to ensure that the coordinates and direction of the virtual camera are consistent with the physical camera; then use the view calculation method in the Unity3D software to determine the visible device model to be detected from the perspective of the virtual camera; determine the device to be detected in the field of view of the physical camera through the relationship and attributes of the device model to be detected, thereby realizing the matching mapping of the device model in the virtual greenhouse model with the equipment in the physical greenhouse.

7. The greenhouse digital twin and equipment operation and maintenance system according to claim 2, characterized in that: The equipment status information mainly obtains the current status of the equipment to be detected in the physical greenhouse through the equipment status recognition model. The structured data obtained through information extraction is used in the fault diagnosis module to compare with the virtual greenhouse operation results to determine the fault.

8. The greenhouse digital twin and equipment operation and maintenance system according to claim 2, characterized in that: The virtual greenhouse scene model is constructed in the following way: Conduct a survey on the physical greenhouse site, extract the size, structure and material of the greenhouse enclosure, greenhouse crops and greenhouse equipment, build and render a three-dimensional model, and form a virtual greenhouse model library of various greenhouse elements; Read the description of the virtual greenhouse scene generation logic in the knowledge graph in the Neo4j graph database, that is, the greenhouse structure information, build the corresponding scene generation script in the Unity3D software, and then set the corresponding geometric dimensions and spatial positions for each virtual device and structure based on the "relationships" and "attributes" of each greenhouse entity constructed in the knowledge graph, and adjust the models in the virtual greenhouse model library according to the positions and sizes obtained by the script, so that the layout of the virtual greenhouse scene is the same as the actual one.

9. The greenhouse digital twin and equipment operation and maintenance system according to claim 8, characterized in that: The fault judgment module specifically includes: Digital twin greenhouse operation: The control platform in the greenhouse digital twin platform collects and analyzes various sensor data in real time and makes decisions based on these data. The control platform sends instructions to the control cabinet of the physical greenhouse, requiring it to perform corresponding operations. While sending instructions, the corresponding operations are synchronously performed in the virtual greenhouse scene model. The communication information received by the control cabinet in the virtual greenhouse scene and the operating status of the equipment model are compared with the corresponding control cabinet and equipment in the physical greenhouse; if the operating status of the equipment model in the virtual greenhouse scene is inconsistent with the operating status of the equipment in the physical greenhouse, it is considered that the equipment has sent a fault, and the specific link of the fault is further confirmed; Check whether the control cabinet in the physical greenhouse receives the communication information of the control platform. If the control cabinet in the physical greenhouse does not receive the information, there may be a communication loss failure, and the fault link is located between the control platform and the control cabinet in the physical greenhouse. Otherwise, determine whether the control instructions issued by the control cabinet in the virtual greenhouse scene are consistent with the instructions issued by the control cabinet in the physical greenhouse. If they are inconsistent, the control cabinet in the physical greenhouse fails, and the fault link is located in the control cabinet in the physical greenhouse. Conversely, if the instructions issued by the control cabinet are consistent, but the equipment in the physical greenhouse is not running or the running status is opposite to the instructions, there is an equipment failure or the control instruction fails to be executed successfully, and the fault link is located in the equipment in the physical greenhouse.

10. The greenhouse digital twin and equipment operation and maintenance system according to claim 1, characterized in that: The device state recognition model is constructed in the following manner: the operation video of the device to be detected is imported into the video segmentation software, and a picture library is generated by extracting video frames; then, the devices to be detected in the picture are classified according to type and operation state, and various states are marked by box selection using a marking tool to form a data set; After labeling, the dataset is divided into training set, validation set and test set. The dataset is used to train the target detection model, and the model is optimized based on the training results. Finally, the optimized target detection model is used as the device status recognition model.

Images