Digital twin simulation test method and system applied to building system

Abstract

The invention provides a digital twinborn simulation test method and system applied to a building system, and the method comprises the steps: firstly, carrying out the construction processing of a knowledge graph of physical parameters of the building system, generating a physical parameter knowledge graph comprising structure, equipment, environment parameter nodes and a plurality of correlation edges, and then constructing a digital twinborn graph model based on the knowledge graph, the method comprises the following steps of: generating a graph test scene configuration sequence according to a model, including a node parameter disturbance rule and the like, executing a graph coupling simulation test by inputting the sequence into the model, and generating a graph test response data set including a node response sequence and the like; and finally, outputting a digital twin simulation test evaluation report based on the data set, wherein the report covers node response characteristics and the like. According to the method, the building system is comprehensively and accurately simulated, and the test efficiency, accuracy and comprehensiveness are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital twinning, in particular to a digital twinning simulation test method and system applied to a building system. BACKGROUND

[0002] In the field of building systems, as the scale of buildings expands and the functions become more complex, higher requirements are placed on performance monitoring, fault prediction and optimized management of building systems. Traditional building system test methods mainly rely on the construction of actual physical environments and on-site testing. The above-mentioned methods not only have high costs, require a large amount of manpower, material resources and time for equipment installation, debugging and operation, but also have long test periods, making it difficult to comprehensively test a variety of complex scenarios in a short period of time.

[0003] At the same time, the traditional test method is difficult to accurately grasp the complex relationship between the various parameters in the building system, and cannot effectively simulate the impact of changes in different parameters on the entire system, resulting in test results that are not accurate and comprehensive, making it difficult to meet the development needs of modern building system intelligent and refined management. SUMMARY

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present application, the embodiments of the present application provide a digital twinning simulation test method applied to a building system, the method comprising:

[0005] performing a knowledge graph construction process of physical parameters of a building system to generate a physical parameter knowledge graph;

[0006] performing a digital twinning graph model generation process based on the physical parameter knowledge graph to construct a digital twinning graph model of the building system;

[0007] generating a graph test scene configuration sequence according to the inter-layer association edge set and the node interaction path of the digital twinning graph model;

[0008] inputting the graph test scene configuration sequence into the digital twinning graph model to perform graph coupling simulation testing and generate a graph test response data set;

[0009] performing graph structure test result analysis based on the graph test response data set and outputting a digital twinning simulation test evaluation report of the building system, the digital twinning simulation test evaluation report containing graph node response characteristics, associated edge influence characteristics and abnormal path graph risk characteristics.

[0010] In still another aspect, the embodiment of the present application also provides a digital twin simulation test system applied to a building system, comprising a processor, a machine readable storage medium, the machine readable storage medium is connected with the processor, the machine readable storage medium is used for storing programs, instructions or codes, and the processor is used for executing the programs, instructions or codes in the machine readable storage medium to realize the method described above.

[0011] Based on the above aspects, the embodiment of the present application realizes parameter conduction and logical reasoning between different parameter layers by constructing a physical parameter knowledge graph containing structure parameter nodes, device parameter nodes, environment parameter nodes and various associated edges, and a digital twin graph model generated based on the physical parameter knowledge graph, and more realistically simulates the actual operation of the building system by using a hierarchical structure and an interlayer associated edge set. According to a graph test scene configuration sequence generated by the model, various test scenes can be flexibly set to enhance the diversity and pertinence of the test. The configuration sequence is input into the model to perform graph coupling simulation test, and comprehensive graph test response data set can be quickly obtained. Finally, a digital twin simulation test evaluation report is output based on the data set, the performance of the building system is evaluated from multiple dimensions such as node response, associated edge influence and abnormal path risk, and the efficiency, accuracy and comprehensiveness of the building system test are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 is a flowchart of the digital twin simulation test method applied to a building system provided by an embodiment of the present application.

[0013] Figure 2 is a schematic diagram of exemplary hardware and software components of the digital twin simulation test system applied to a building system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0014] The present application will be described in detail below with reference to the accompanying drawings of the specification, Figure 1 is a flowchart of the digital twin simulation test method applied to a building system provided by an embodiment of the present application, and the digital twin simulation test method applied to a building system will be described in detail below.

[0015] Step S110: Perform a knowledge graph construction process of building system physical parameters to generate a physical parameter knowledge graph, the physical parameter knowledge graph contains structure parameter nodes, device parameter nodes, environment parameter nodes and associated edges between nodes, and the associated edges between nodes contain location associated edges, influence associated edges and interaction associated edges.

[0016] In this embodiment, a commercial office building is taken as a unified application scenario, and the physical parameter knowledge graph can show the internal relationship between the building structure, equipment and environment. First of all, the structure parameter node represents various structural elements of the building, such as walls, floors, beams and columns; the equipment parameter node covers various equipment in the building, such as elevators, air conditioners, lighting systems, etc.; the environment parameter node reflects the environmental conditions of the building, such as indoor temperature, humidity, light intensity, etc. The association edge between nodes is the link connecting these nodes. The location association edge reflects the spatial relationship between the structure and the equipment, such as the elevator is installed at a specific floor and shaft position; the influence association edge describes the influence of the equipment on the environment or the environment on the equipment, for example, the operation of the air conditioner will change the indoor temperature; the interaction association edge represents the interaction between the environment and the structure, such as the humidity outside may affect the material performance of the wall.

[0017] Step S111: Collecting physical parameter data of the building system, the physical parameter data including structure class parameter data, equipment class parameter data and environment class parameter data.

[0018] In this step, the collection of physical parameter data of the commercial office building is a multi-channel and multi-mode process. For structure class parameter data, it can be obtained from the design drawings and construction documents of the building. These documents record detailed information such as wall thickness, floor height, beam size, etc. For example, the design drawings will clearly mark the height of each floor and the material and thickness of the wall. For equipment class parameter data, on the one hand, the basic parameters of the equipment can be obtained from the technical specifications of the equipment, such as the lifting capacity and running speed of the elevator, the refrigerating capacity and power of the air conditioner, etc.; on the other hand, the running data of the equipment can be collected in real time through the monitoring system of the equipment, such as the running time of the elevator and the actual power consumption of the air conditioner, etc.

[0019] For environment class parameter data, various sensors need to be installed in the building. Temperature sensors, humidity sensors and light sensors are installed at different floors and areas to obtain real-time data such as indoor temperature, humidity and light intensity. At the same time, outdoor weather data such as outdoor temperature, wind speed and rainfall can be obtained from the meteorological department.

[0020] Step S112: Performing graph node extraction processing on the structure class parameter data to extract structure parameter entities and entity attributes, and generating structure parameter nodes, the structure parameter nodes including node identification, parameter type and attribute description information.

[0021] Step S1121: Performing data cleaning processing on the collected structure class parameter data to remove duplicate data records and invalid data fields, and retaining valid data containing structure feature description.

[0022] In the scenario of a commercial office building, the collected structural parameter data can come from multiple different channels, such as design drawings, construction records, and post-maintenance documents. These data can have repetitions and invalidities. For example, the thickness of a wall can be recorded in both the design drawings and the construction records, but due to different recording times and methods, there can be inconsistencies in the data format or slight differences in the numerical values. At the same time, some data fields can be invalid due to recording errors or incompleteness, such as a record that only records the length of a wall without width and thickness. To remove these repeated and invalid data, the key information of the data, such as the location and size of the wall, can be compared to identify repeated data records, and the most accurate and complete one can be selected to keep. For invalid data fields, they can be judged and corrected according to the context of the data and other related records, and if they cannot be corrected, they will be deleted directly. After data cleaning, the valid data containing structural feature descriptions are retained, which will be used for subsequent entity recognition and attribute extraction.

[0023] Step S1122: performing entity recognition processing on the cleaned structural parameter data, identifying the key words representing the structural entities in the data based on a preset structural parameter entity dictionary, the structural entities including wall entities, floor entities, beam column entities, and space region entities.

[0024] The cleaned structural parameter data needs to be subjected to entity recognition to find out the key words representing the structural entities. The preset structural parameter entity dictionary is a set of vocabulary containing various structural entity names. In the scenario of a commercial office building, when "second floor wall" appears in the data, by comparing with the structural parameter entity dictionary, the corresponding floor entity of "second floor" and the wall entity of "wall" can be identified. The dictionary will define the name of each entity and related synonyms and near-synonyms to improve the accuracy of entity recognition. For the two words "column" and "beam column", it will be clearly defined in the dictionary that they both belong to the beam column entity. By traversing the cleaned structural parameter data, the words in the data are matched with the key words in the dictionary, so that the structural entities contained in the data are identified.

[0025] Step S1123: performing attribute extraction processing on the identified structural entities, extracting attribute information of each structural entity from the data, the attribute information including entity size attributes, material attributes, location attributes, and connection attributes.

[0026] After identifying the structural entities, attribute information of each structural entity needs to be extracted from the data. For the wall entity in the commercial office building, its attribute information contains multiple aspects. Entity size attributes can be obtained from design drawings, such as the length, width and thickness of the wall. Material attributes can be known through construction documents or material inspection reports, such as whether the wall is built with bricks, concrete or other materials. Location attributes can be determined according to the floor and specific coordinate information, such as which position on which floor the wall is located. Connection attributes describe the connection mode of the wall with other structural entities, whether it is directly connected with other walls or connected through beams and columns. For the floor entity, its attribute information can include the height, area of the floor and the connection relationship with the upper and lower floors, etc. Through analysis of the data, these attribute information of each structural entity is extracted.

[0027] Step S1124: Assign a unique node identifier to each structural entity, which consists of an entity type prefix and a serial number.

[0028] In order to accurately identify each structural entity in the knowledge graph, it is necessary to assign a unique node identifier to them. In the construction of structural parameter nodes of the commercial office building, the node identifier consists of an entity type prefix and a serial number. For the wall entity, the prefix of its node identifier is “wall-”, followed by a serial number, such as “wall-01”, “wall-02”, etc. The above identification method can clearly indicate the entity type represented by the node, and distinguish different entities of the same type through the serial number. For the floor entity, the node identifier can be “floor-01”, “floor-02”, etc. Through the above method, a unique node identifier is given to each identified structural entity, which facilitates the management and query of these nodes in the knowledge graph.

[0029] Step S1125: According to the type of the structural entity, the node is divided into parameter types, including wall parameter type, floor parameter type, beam column parameter type and space region parameter type.

[0030] After assigning a node identifier to the structural entity, the node needs to be divided into parameter types according to the type of the structural entity. In the commercial office building, different types of structural entities have different parameter characteristics. The nodes of the wall parameter type mainly focus on the size, material, location and connection attributes of the wall; the nodes of the floor parameter type focus on the height, area and load-bearing capacity of the floor; the nodes of the beam column parameter type focus on the size, strength and material of the beam column; the nodes of the space region parameter type involve the size, purpose and lighting of the space. Through parameter type division of the node, these structural parameter nodes can be more clearly organized and managed, and it is also convenient for targeted query and analysis in the knowledge graph.

[0031] Step S1126: integrate the node identification, parameter type and extracted attribute information to generate a structure parameter node, and the attribute description information of the structure parameter node includes entity size attribute value, material attribute value, position attribute value and connection attribute value.

[0032] After the node identification allocation and parameter type division are completed, the node identification, parameter type and extracted attribute information are integrated to generate a structure parameter node. In the scenario of a commercial office building, for a wall node, the node identification is "wall-01", the parameter type is a wall parameter type, and the attribute description information includes the size attribute value (such as length, width, thickness) of the wall, the material attribute value (such as bricks), the position attribute value (such as located at a certain position on the second floor) and the connection attribute value (such as connected to other walls through concrete). By integrating these information together, a complete structure parameter node is formed, which can comprehensively describe the characteristics and attributes of the wall.

[0033] Step S1127: perform attribute integrity check on the generated structure parameter node to ensure that each node contains at least one core attribute, perform attribute completion processing on the node missing the core attribute, and check again after completion until all structure parameter nodes pass the attribute integrity check.

[0034] The generated structure parameter node needs to be checked for attribute integrity to ensure that each node contains at least one core attribute. In the commercial office building, for a wall node, the core attributes may be thickness and material. If a wall node only contains the position attribute and lacks the thickness and material attributes, the wall node does not meet the attribute integrity requirement. For the node missing the core attribute, attribute completion processing needs to be performed. The missing attribute information can be obtained by consulting relevant design documents, construction records or conducting on-site measurement, etc. After completing the attributes, the node is checked again to check whether there is still a missing core attribute. Repeat the process until all structure parameter nodes pass the attribute integrity check, ensuring the quality and accuracy of the structure parameter nodes.

[0035] Step S113: perform graph node extraction processing on the device class parameter data to extract device parameter entities and entity attributes, and generate device parameter nodes, which include node identification, parameter type and attribute description information.

[0036] In this step, the device class parameter data of the commercial office building is subjected to graph node extraction processing. First, the device class parameter data also needs to be cleaned to remove duplicate and invalid data. The device class parameter data can come from the specifications of the equipment, monitoring systems, and maintenance records, etc. There can be duplicate equipment operation data or incorrect parameter records in these data. After cleaning, entity recognition is performed on the device class parameter data, and based on the preset device parameter entity dictionary, keywords representing device entities such as elevators, air conditioners, lighting systems, etc. are identified. Then, the identified device entities are subjected to attribute extraction, and their attribute information is extracted, including the model, power, and operating parameters of the equipment. Each device entity is assigned a unique node identifier, such as "elevator-01", "air conditioner-02", etc., and the nodes are classified by parameter type according to the type of the equipment, such as elevator parameter type, air conditioner parameter type, etc. The node identifier, parameter type, and extracted attribute information are integrated to generate a device parameter node, and finally the attribute integrity of the device parameter node is verified to ensure that each node contains at least one core attribute, such as the core attributes of the elevator node, which can be the load capacity and running speed.

[0037] Step S114: performing graph node extraction processing on the environment class parameter data, extracting environment parameter entities and entity attributes, and generating environment parameter nodes, the environment parameter nodes containing node identifiers, parameter types, and attribute description information.

[0038] For the environment class parameter data of the commercial office building, graph node extraction processing is performed in this step. First, the environment class parameter data is cleaned to remove possible abnormal data and duplicate records. These data come from sensors within the building and the meteorological department, and may produce abnormal values due to sensor failure or data transmission problems. After cleaning, entity recognition is performed on the data based on the preset environment parameter entity dictionary, and environment entities such as indoor temperature, humidity, and light intensity are identified. Then, the identified environment entities are subjected to attribute extraction, and their attribute information is extracted, such as the measurement time of temperature and the variation range of humidity. Each environment entity is assigned a unique node identifier, such as "indoor temperature-01", "humidity-02", etc., and the nodes are classified by parameter type according to the type of the environment entity, such as temperature parameter type, humidity parameter type, etc. The node identifier, parameter type, and extracted attribute information are integrated to generate an environment parameter node, and then the attribute integrity of the environment parameter node is verified to ensure that each node contains at least one core attribute, such as the core attributes of the temperature node, which can be the measurement value and the measurement time.

[0039] Step S115: performing edge relationship extraction processing on the physical association relationship between the structure parameter nodes and the device parameter nodes, generating location association edges, the location association edges containing edge identifiers, association types, and association strength descriptions.

[0040] In the commercial office building, there is a physical association between the structure parameter node and the device parameter node, and the location association edge is generated through the edge relationship extraction processing. For example, there is a location association between the elevator device and the floor structure of the building. By analyzing the installation location of the device and the spatial layout of the structure, the above association is determined. The edge identifier is used to uniquely identify this location association edge, such as "structure-elevator location association-01". The association type clearly indicates that the above association is a location association. The association strength description can be determined according to the closeness of the device and the structure. The elevator is closely connected with the shaft structure of the floor where it is located, and the association strength is high; while the association strength of some small lighting devices and wall structures is relatively low. By analyzing the physical association between the structure parameter node and the device parameter node, the information of the location association edge is accurately extracted.

[0041] Step S116: Perform edge relationship extraction processing on the physical association between the device parameter node and the environment parameter node to generate the influence association edge, which contains edge identifier, association type and association strength description.

[0042] In the scenario of a commercial office building, there is a physical association between the device parameter node and the environment parameter node, and the influence association edge is generated through the edge relationship extraction processing. The operation of the air conditioning device will have an impact on the indoor temperature and humidity and other environmental parameters. By analyzing the running parameters of the device and the changes of the environmental parameters, the above influence association is determined. The edge identifier is used to uniquely identify this influence association edge, such as "device-environment influence association-01". The association type clearly indicates that the above association is the influence of the device on the environment. The association strength description can be determined according to the influence degree of the device on the environmental parameters. Large air conditioning devices have a greater impact on indoor temperature, and the association strength is high; while small ventilation devices have a relatively small impact on indoor air quality, and the association strength is low. By in-depth study of the physical association between the device parameter node and the environment parameter node, the information of the influence association edge is accurately extracted.

[0043] Step S117: Perform edge relationship extraction processing on the physical association between the environment parameter node and the structure parameter node to generate the interaction association edge, which contains edge identifier, association type and association strength description.

[0044] In the commercial office building, there is a physical association relationship between the environmental parameter nodes and the structural parameter nodes, and the interactive association edges are generated through edge relationship extraction processing. The outdoor humidity and temperature and other environmental parameters will affect the material performance of the wall, and the structure of the wall will also affect the indoor environmental parameters. By analyzing the interaction between environmental parameters and structural parameters, the above interactive association relationship is determined. The edge identifier is used to uniquely identify this interactive association edge, such as "environment-structure interactive association-01". The association type clearly indicates that the above association is the interaction between the environment and the structure. The association strength can be determined according to the degree of influence of the environment on the structure and the degree of feedback of the structure on the environment. Long-term high-humidity environment has a greater impact on wooden structural walls, and its association strength is higher. While the influence of ordinary wall structure on indoor light intensity is relatively small, its association strength is lower. Through comprehensive analysis of the physical association relationship between environmental parameter nodes and structural parameter nodes, the information of interactive association edges is accurately extracted.

[0045] Step S118: integrate the structural parameter nodes, device parameter nodes, environmental parameter nodes, and location association edges, influence association edges, and interactive association edges, verify the node attribute integrity and edge relationship rationality through the graph consistency verification tool, and generate a physical parameter knowledge graph.

[0046] After the generation of structural parameter nodes, device parameter nodes, environmental parameter nodes, and location association edges, influence association edges, and interactive association edges, they are integrated together. In the construction of the knowledge graph of the commercial office building, all structures, devices, and environmental nodes and the association edges between them are combined into a whole. Then, the graph consistency verification tool is used to verify the integrated result. The tool checks the integrity of the node attributes to ensure that the attribute information of each node is accurate and complete. At the same time, it can verify the rationality of the edge relationship to check whether the association type and association strength of the association edge are consistent with the actual situation. For example, it checks whether the location association edge accurately reflects the actual location relationship of the structure and the device, and whether the influence association edge reasonably describes the degree of influence of the device on the environment. If it is found that the node attributes are not complete or the edge relationship is not reasonable, the corresponding nodes and edges need to be corrected. After several checks and corrections, until all node attributes are complete and edge relationships are reasonable, the physical parameter knowledge graph is finally generated. The physical parameter knowledge graph can comprehensively and accurately describe the physical parameters of the commercial office building and their mutual relationships.

[0047] Step S120: based on the physical parameter knowledge graph, perform digital twin graph model generation processing to construct a digital twin graph model of the building system, the digital twin graph model includes a structural parameter graph node layer, a device parameter graph node layer, an environmental parameter graph node layer, and a set of inter-layer association edges, the set of inter-layer association edges is used to realize parameter transmission and logical reasoning between different graph node layers.

[0048] After the physical parameter knowledge graph of the commercial office building is generated, a digital twin graph model generation process is performed based on the knowledge graph. The digital twin graph model is a digital mapping of the physical system of the commercial office building, which is composed of a structure parameter graph node layer, a device parameter graph node layer, an environment parameter graph node layer, and an interlayer association edge set. The structure parameter graph node layer organizes the structure parameter nodes in the knowledge graph according to certain rules, the device parameter graph node layer processes the device parameter nodes in a similar manner, and the environment parameter graph node layer integrates the environment parameter nodes. The interlayer association edge set connects different graph node layers to realize parameter transmission and logical reasoning between different graph node layers. For example, when the bearing capacity of a floor node in the structure parameter graph node layer changes, the change information can be transmitted to the device parameter graph node layer through the interlayer association edge set, affecting the operation of the devices on the floor; at the same time, according to the preset logical reasoning rules, the influence of the above change on the indoor environment parameters in the environment parameter graph node layer can be inferred.

[0049] Step S121: Extracting structure parameter nodes from the physical parameter knowledge graph, layering and aggregating the structure parameter nodes according to parameter types to generate a structure parameter graph node layer, the structure parameter graph node layer containing multiple hierarchical sub-nodes, the hierarchical sub-nodes connected by internal association edges.

[0050] In the scenario of a commercial office building, after extracting structure parameter nodes from the physical parameter knowledge graph, layering and aggregating the structure parameter nodes according to parameter types to generate a structure parameter graph node layer. According to the types of structure parameter nodes, such as wall parameter type, floor parameter type, beam column parameter type, and space region parameter type, they are aggregated into different levels. For floor parameter type nodes, nodes of different floors are aggregated together to form a level. In each level, there can be multiple hierarchical sub-nodes. In the floor level, each specific floor node is a hierarchical sub-node. The hierarchical sub-nodes are connected by internal association edges, which can represent the upper and lower level relationships, adjacent relationships, etc. between nodes. The first floor node and the second floor node are connected by internal association edges, indicating that they are adjacent floor relationships. Through the above layering and aggregation and connection of internal association edges, a clear hierarchical and clear structure structure parameter graph node layer is formed.

[0051] Step S122: Extracting device parameter nodes from the physical parameter knowledge graph, layering and aggregating the device parameter nodes according to parameter types to generate a device parameter graph node layer, the device parameter graph node layer containing multiple hierarchical sub-nodes, the hierarchical sub-nodes connected by internal association edges.

[0052] In the commercial office building, after extracting the equipment parameter nodes from the physical parameter knowledge graph, the equipment parameter graph node layer is generated by hierarchical aggregation according to the parameter type. According to the types of the equipment parameter nodes, such as the elevator parameter type, the air conditioner parameter type, and the lighting parameter type, they are aggregated into different levels respectively. For the nodes of the elevator parameter type, the nodes of different elevators are aggregated together to form a level. In each level, there are also multiple level sub-nodes, and each elevator node is a level sub-node. The level sub-nodes are connected by internal association edges, which can represent the cooperative working relationship and the position relationship between the devices. The multiple elevators on the same floor are connected by internal association edges, which represent that they serve the personnel transportation together on the floor. Through the above hierarchical aggregation and the connection of the internal association edges, an ordered equipment parameter graph node layer is formed.

[0053] Step S123: Extracting the environment parameter nodes from the physical parameter knowledge graph, generating the environment parameter graph node layer by hierarchical aggregation according to the parameter type, and the environment parameter graph node layer contains multiple level sub-nodes, which are connected by internal association edges.

[0054] In the construction of the environment parameter graph node layer of the commercial office building, after extracting the environment parameter nodes from the physical parameter knowledge graph, hierarchical aggregation is performed according to the parameter type. According to the types of the environment parameter nodes, such as the temperature parameter type, the humidity parameter type, and the light intensity parameter type, they are aggregated into different levels respectively. For the nodes of the temperature parameter type, the temperature nodes of different positions and times are aggregated together to form a level. In each level, there can be multiple level sub-nodes, and each specific temperature measurement node is a level sub-node. The level sub-nodes are connected by internal association edges, which can represent the mutual influence relationship and the spatial position relationship between the environment parameters. The temperature nodes of adjacent areas on the same floor are connected by internal association edges, which represent that they are adjacent in space and the temperature can influence each other. Through the above hierarchical aggregation and the connection of the internal association edges, a reasonable environment parameter graph node layer is formed.

[0055] Step S124: Extracting the position association edges from the physical parameter knowledge graph, binding the level sub-nodes of the structure parameter graph node layer and the level sub-nodes of the equipment parameter graph node layer through the position association edges, and generating a subset of structure-equipment inter-layer association edges.

[0056] In the construction of the digital twin graph model of the commercial office building, after extracting the location correlation edges from the physical parameter knowledge graph, the hierarchical child nodes of the structural parameter graph node layer and the hierarchical child nodes of the equipment parameter graph node layer are bound through the location correlation edges. The elevator equipment node and the floor node of the floor where the elevator is located are bound through the location correlation edges to form a subset of structural equipment interlayer correlation edges. These edge subsets record the correlation information of the structure and the equipment in the spatial position, and through them, the parameter conduction between the structural parameter graph node layer and the equipment parameter graph node layer can be realized. When the structure of the floor node changes, through the subset of structural equipment interlayer correlation edges, the change information can be transmitted to the corresponding elevator equipment node, affecting the operation safety evaluation of the elevator, etc.

[0057] Step S125: Extracting the influence correlation edges from the physical parameter knowledge graph, binding the hierarchical child nodes of the equipment parameter graph node layer and the hierarchical child nodes of the environment parameter graph node layer through the influence correlation edges, and generating a subset of equipment environment interlayer correlation edges.

[0058] In the commercial office building, after extracting the influence correlation edges from the physical parameter knowledge graph, the hierarchical child nodes of the equipment parameter graph node layer and the hierarchical child nodes of the environment parameter graph node layer are bound through the influence correlation edges. The air conditioning equipment node and the environment parameter nodes such as indoor temperature and humidity are bound through the influence correlation edges to generate a subset of equipment environment interlayer correlation edges. These edge subsets record the influence information of the equipment on the environment, and through them, the parameter conduction between the equipment parameter graph node layer and the environment parameter graph node layer can be realized. When the operation parameters of the air conditioning equipment node change, through the subset of equipment environment interlayer correlation edges, the change information can be transmitted to the environment parameter graph node layer, affecting the environment parameters such as indoor temperature and humidity.

[0059] Step S126: Extracting the interaction correlation edges from the physical parameter knowledge graph, binding the hierarchical child nodes of the environment parameter graph node layer and the hierarchical child nodes of the structural parameter graph node layer through the interaction correlation edges, and generating a subset of environment structure interlayer correlation edges.

[0060] In the construction of the digital twin graph model of the commercial office building, after extracting the interaction correlation edges from the physical parameter knowledge graph, the hierarchical child nodes of the environment parameter graph node layer and the hierarchical child nodes of the structural parameter graph node layer are bound through the interaction correlation edges. The outdoor humidity environment parameter node and the wall structure node are bound through the interaction correlation edges to generate a subset of environment structure interlayer correlation edges. These edge subsets record the interaction information between the environment and the structure, and through them, the parameter conduction between the environment parameter graph node layer and the structural parameter graph node layer can be realized. When the humidity value of the outdoor humidity environment parameter node changes, through the subset of environment structure interlayer correlation edges, the change information can be transmitted to the structural parameter graph node layer, affecting the material performance of the wall structure, etc.

[0061] Step S127: integrate the structure-device interlayer association edge subset, the device-environment interlayer association edge subset, and the environment-structure interlayer association edge subset to form an interlayer association edge set, and the interlayer association edge set includes an edge weight attribute for representing parameter conduction strength.

[0062] In the digital twin graph model of the commercial office building, the structure-device interlayer association edge subset, the device-environment interlayer association edge subset, and the environment-structure interlayer association edge subset are integrated to form an interlayer association edge set. The interlayer association edge set includes an edge weight attribute for representing parameter conduction strength. For the edges in the structure-device interlayer association edge subset, if the elevator is closely associated with the structure of the floor, the edge weight attribute value is higher, indicating that the parameter conduction strength is greater; while for some small devices and structure association edges, the edge weight attribute value is relatively low. Through the edge weight attribute, the strength of parameter conduction between different graph node layers can be accurately described, making the parameter conduction and logical reasoning more accurate and reasonable.

[0063] Step S128: build a graph reasoning engine, which realizes parameter conduction calculation and logical relationship reasoning based on the interlayer association edge set, integrates the structure parameter graph node layer, the device parameter graph node layer, the environment parameter graph node layer, the interlayer association edge set, and the graph reasoning engine to generate a digital twin graph model.

[0064] Step S1281: design a parameter conduction calculation module of the graph reasoning engine, which receives the parameter value of the starting graph node and the edge weight attribute of the interlayer association edge, and calculates the input parameter value of the intermediate graph node through a weighted summation algorithm, the input of the weighted summation algorithm being the product of the starting graph node parameter value and the edge weight attribute, and the output being the input parameter value of the intermediate graph node.

[0065] In the graph reasoning engine of the commercial office building, the parameter conduction calculation module is responsible for calculating the input parameter value of the intermediate graph node. When the parameter value of the starting graph node changes, the parameter conduction calculation module receives the parameter value and the edge weight attribute of the interlayer association edge connecting the starting graph node and the intermediate graph node. Through the weighted summation algorithm, the starting graph node parameter value is multiplied by the edge weight attribute to obtain the input parameter value of the intermediate graph node. If the starting graph node is a floor node, and the carrying capacity parameter value changes, the interlayer association edge connecting the floor node and the elevator device node on the floor has a certain edge weight attribute. Multiplying the carrying capacity parameter value of the floor node by the edge weight attribute can obtain the input parameter value of the elevator device node, which reflects the degree of influence of the change of the floor structure on the elevator device.

[0066] Step S1282: design a logical relationship reasoning module of the graph reasoning engine, the logical relationship reasoning module calls a preset logical reasoning rule library based on the association type of the inter-layer association edge, and performs reasoning on the parameter relationship between different graph node layers, and the logical reasoning rule library includes a position association reasoning rule, an influence association reasoning rule, and an interaction association reasoning rule.

[0067] In the graph reasoning engine of the commercial office building, the logical relationship reasoning module calls a preset logical reasoning rule library based on the association type of the inter-layer association edge to perform reasoning. For the position association reasoning rule, if the position association of a certain device node and a structure node does not meet the preset safety distance requirement, the logical relationship reasoning module can infer that the installation of the device may have safety hazards. For the influence association reasoning rule, when the running power of the air conditioning device node increases, according to the influence association reasoning rule, it can be inferred that the indoor temperature will decrease, and the humidity may also change. For the interaction association reasoning rule, when the humidity value of the outdoor humidity environment parameter node increases, according to the interaction association reasoning rule, it can be inferred that the material performance of the wall structure node may be affected. Through these logical reasoning rules, the logical relationship reasoning module can accurately reason the parameter relationship between different graph node layers.

[0068] Step S1283: design a node state updating module of the graph reasoning engine, the node state updating module receives an input parameter value of an intermediate graph node, and calculates an output parameter value of the node by a state transition function in combination with attribute description information of the node, and the state transition function is determined based on a parameter type of the node.

[0069] In the graph reasoning engine of the commercial office building, the node state updating module is responsible for updating the state of the node. When the intermediate graph node receives an input parameter value, the node state updating module calculates the output parameter value of the node by a state transition function in combination with the attribute description information of the node. For the elevator device node, the attribute description information thereof includes load capacity, running speed, etc., and when an input parameter value transmitted due to a change in the floor structure is received, a new running state parameter value of the elevator device node is calculated according to the state transition function determined based on the parameter type of the elevator, such as the running speed that may need to be adjusted to adapt to the change in the floor structure. Different types of nodes have different state transition functions to ensure accurate calculation of the output parameter value of the node.

[0070] Step S1284: design a path search module of the graph reasoning engine, the path search module adopts a depth-first search algorithm or a breadth-first search algorithm to find all possible transmission paths between any two graph nodes based on a set of inter-layer association edges, and outputs the graph node identifiers and the inter-layer association edge identifiers contained in the paths.

[0071] In the graph reasoning engine of the commercial office building, the path search module is used to find all possible conduction paths between any two graph nodes. Based on the inter-floor association edge set, the module can use a depth-first search algorithm or a breadth-first search algorithm for path search. When a conduction path between a floor node and a lighting device node located on the floor needs to be found, the path search module starts from the floor node and searches along the inter-floor association edge. If a depth-first search algorithm is used, the search can be as deep as possible along a path until it cannot continue, and then backtracking to the last node to continue searching other paths; if a breadth-first search algorithm is used, the adjacent nodes can be searched layer by layer until the target node is found or all possible paths are traversed. Finally, the path search module outputs the graph node identifiers and inter-floor association edge identifiers contained in the path.

[0072] Step S1285: The parameter conduction calculation module, the logical relationship reasoning module, the node state updating module, and the path search module are integrated to build the core processing unit of the graph reasoning engine, and the data interaction interface and the calling sequence between the modules are defined.

[0073] In the construction of the graph reasoning engine of the commercial office building, the parameter conduction calculation module, the logical relationship reasoning module, the node state updating module, and the path search module are integrated to build the core processing unit. In order to ensure the cooperative work between the modules, the data interaction interface and the calling sequence between the modules need to be defined. When the parameter value of the starting graph node changes, the parameter conduction calculation module first calculates the input parameter value of the intermediate graph node according to the edge weight attribute of the inter-floor association edge; then the input parameter value is passed to the node state updating module, which updates the state of the node in combination with the node attribute information; at the same time, the logical relationship reasoning module infers the parameter relationship according to the association type of the inter-floor association edge and the preset logical reasoning rule; the path search module can find the conduction path between any two graph nodes when needed. By defining the data interaction interface and the calling sequence between the modules, the graph reasoning engine can run efficiently and accurately.

[0074] Step S1286: The parameter conduction calculation accuracy, the logical relationship reasoning correctness, and the node state updating real-time of the graph reasoning engine are verified by inputting preset test parameter values and known conduction paths, and the algorithm parameters and reasoning rules in the modules are adjusted until the graph reasoning engine meets the preset performance indicators.

[0075] After the graph reasoning engine of the commercial office building is built, it needs to be verified and adjusted. By inputting preset test parameter values and known conduction paths, the parameter conduction calculation accuracy, logical relationship reasoning correctness and node state update real-time of the graph reasoning engine are verified. Input a known floor node carrying capacity change parameter value and the corresponding conduction path, observe whether the input parameter value of the elevator equipment node calculated by the graph reasoning engine is accurate; at the same time, check whether the influence on other related nodes inferred by the logical relationship reasoning module is reasonable; also check the real-time of the node state update module to update the node state. If it is found that the performance of the graph reasoning engine does not meet the preset performance indicators, the algorithm parameters and reasoning rules in the module need to be adjusted. Adjust some coefficients in the weighted summation algorithm of the parameter conduction calculation module, or modify the logical reasoning rules of the logical relationship reasoning module. After several verifications and adjustments, until the graph reasoning engine meets the preset performance indicators, it can accurately and efficiently complete the tasks of parameter conduction calculation, logical relationship reasoning and node state update, etc.

[0076] Step S130: generating a graph test scene configuration sequence according to the inter-layer associated edge set and node interaction path of the digital twin graph model, the graph test scene configuration sequence including graph node parameter perturbation rules, associated edge trigger conditions and test path control logic.

[0077] After the digital twin graph model of the commercial office building is built, a graph test scene configuration sequence is generated according to the inter-layer associated edge set and node interaction path. The configuration sequence is an important basis for graph coupling simulation testing. The graph node parameter perturbation rules are used to control the parameter change of the starting graph node, the associated edge trigger conditions determine when the inter-layer associated edge is triggered, and the test path control logic guides the progress of the test process. By reasonably designing the graph test scene configuration sequence, the performance and accuracy of the digital twin graph model can be comprehensively tested.

[0078] Step S131: analyzing the inter-layer associated edge set of the digital twin graph model, identifying all possible interaction paths between the structure parameter graph node layer, the equipment parameter graph node layer and the environment parameter graph node layer, each interaction path including a starting graph node, an intermediate graph node, a terminal graph node and an inter-layer associated edge connecting each node.

[0079] In the scenario of a commercial office building, the inter-layer association edge set of the digital twin graph model is parsed to identify all possible interaction paths between different graph node layers. Starting from a floor node in the structural parameter graph node layer, an interaction path is formed by connecting to an elevator device node in the device parameter graph node layer through an inter-layer association edge, and then connecting to an indoor temperature node in the environment parameter graph node layer through another inter-layer association edge. Each interaction path includes a starting graph node (such as a floor node), an intermediate graph node (such as an elevator device node), a terminating graph node (such as an indoor temperature node), and inter-layer association edges connecting the nodes. By parsing the inter-layer association edge set in detail, all possible interaction paths are found, providing a basis for subsequent test scenario configuration.

[0080] Step S132: Based on the node types and inter-layer association edge attributes of the interaction paths, test scenario types are classified, each test scenario type corresponding to the cooperative test requirements of one or more interaction paths, and the scenario test objectives are determined for each test scenario type.

[0081] In the digital twin graph model test of a commercial office building, test scenario types are classified based on the node types and inter-layer association edge attributes of the interaction paths. According to the types of the starting graph node, intermediate graph node, and terminating graph node in the interaction path, as well as the association type and edge weight attributes of the inter-layer association edge, the test scenarios are classified into different types. If the interaction path mainly involves the location association between structure and device, it is classified as a structure-device location association test scenario type; if the interaction path mainly reflects the influence of the device on the environment, it is classified as a device-environment influence test scenario type. Each test scenario type corresponds to the cooperative test requirements of one or more interaction paths, and the scenario test objectives are determined for each test scenario type. For the structure-device location association test scenario type, the scenario test objective may be to verify whether the influence of structural changes on device installation and operation meets safety requirements; for the device-environment influence test scenario type, the scenario test objective may be to evaluate whether the adjustment effect of device operation on environmental parameters meets expectations.

[0082] Step S133: For each test scenario type, according to the starting graph node parameter type of the corresponding interaction path and the scenario test objective, a graph node parameter perturbation rule is designed, which includes the initial parameter value setting of the starting graph node, the parameter continuous variation mode, and the parameter cooperative perturbation relationship.

[0083] In the digital twin graph model test of the commercial office building, the graph node parameter perturbation rules are designed for each test scenario type. For the structure-equipment location association test scenario type, if the starting graph node is a floor node, the graph node parameter perturbation rules are designed according to the parameter type (such as bearing capacity) and the scene test target (verify the impact of structural changes on equipment). The initial bearing capacity parameter value of the floor node is set, and the parameter continuous change mode is determined, whether it is linear change or nonlinear change. At the same time, considering the parameter coordinated perturbation relationship, if the bearing capacity of the floor node changes, it may have a coordinated perturbation relationship with the parameters of multiple equipment nodes on the floor, such as the load limit of the elevator equipment which may need to be adjusted accordingly. By reasonably designing the graph node parameter perturbation rules, different actual situations can be simulated, and the performance of the digital twin graph model can be comprehensively tested.

[0084] Step S134: Design the associated edge trigger condition according to the scene test target and the inter-layer associated edge attribute of the interaction path, the associated edge trigger condition includes the starting graph node parameter threshold condition, the intermediate graph node state condition and the inter-layer associated edge conduction strength condition.

[0085] In the digital twin graph model test of the commercial office building, the associated edge trigger condition is designed according to the scene test target and the inter-layer associated edge attribute of the interaction path. For the equipment-environment impact test scenario type, the scene test target is to evaluate the adjustment effect of air conditioning equipment on indoor temperature. According to the inter-layer associated edge attribute of the interaction path, such as the edge weight attribute reflecting the influence strength of the air conditioning equipment on the indoor temperature, the associated edge trigger condition is designed. The parameter threshold condition of the starting graph node (air conditioning equipment node) can be that the running power of the air conditioner reaches a certain value; the state condition of the intermediate graph node (such as the temperature sensor node) can be that the indoor temperature does not reach the preset target temperature; and the inter-layer associated edge conduction strength condition can be that the edge weight attribute reaches a certain range. When these conditions are met, the associated edge is triggered to start parameter conduction and logical reasoning.

[0086] Step S135: Design the test path control logic based on the graph node parameter perturbation rules and the associated edge trigger condition, the test path control logic includes the interaction path selection strategy, the node parameter injection sequence and the test process interruption mechanism.

[0087] In the digital twin graph model test of the commercial office building, the test path control logic is designed based on the graph node parameter perturbation rule and the associated edge trigger condition. The interaction path selection strategy determines which paths to choose among the many possible interaction paths for testing. According to the focus of the test scenario and resource constraints, the most representative interaction paths can be selected. The node parameter injection sequence determines when the parameters of the starting graph node are injected to ensure the rationality of the test process. If there are multiple starting graph nodes, the parameters need to be injected in a certain order to avoid parameter conflicts. The test process interruption mechanism specifies when to interrupt the test process. When the test result is abnormal or the preset test time is reached, the test is interrupted to ensure the safety and effectiveness of the test.

[0088] Step S136: The graph node parameter perturbation rule, associated edge trigger condition and test path control logic are associated and integrated according to the test scenario type to generate a graph test scenario configuration sequence.

[0089] In the digital twin graph model test of the commercial office building, the graph node parameter perturbation rule, associated edge trigger condition and test path control logic are associated and integrated according to the test scenario type to generate a graph test scenario configuration sequence. For each test scenario type, the corresponding graph node parameter perturbation rule, associated edge trigger condition and test path control logic are combined together. For the structure-equipment location association test scenario type, the graph node parameter perturbation rule, associated edge trigger condition and test path control logic under this scenario are integrated into a configuration unit; for the equipment-environment influence test scenario type, the same integration is performed. Finally, all the configuration units of the test scenario types are combined together to form a complete graph test scenario configuration sequence.

[0090] Step S140: The graph test scenario configuration sequence is input into the digital twin graph model to perform graph coupling simulation test, generating a graph test response data set containing a graph node response sequence, associated edge conduction relationship data and abnormal path graph data.

[0091] In the digital twin graph model test of the commercial office building, the graph test scenario configuration sequence is input into the digital twin graph model to perform graph coupling simulation test. The graph node parameter perturbation rule in the graph test scenario configuration sequence will change the parameters of the starting graph node, the associated edge trigger condition determines when the inter-layer associated edge conducts the parameters, and the test path control logic guides the progress of the test process. During the test process, the graph reasoning engine of the digital twin graph model will conduct parameter conduction and logical reasoning according to the inter-layer associated edge set, record the response of the graph node, the conduction relationship of the associated edge and the possible abnormal path. Finally, a graph test response data set is generated, which contains a graph node response sequence, associated edge conduction relationship data and abnormal path graph data.

[0092] Step S141: Analyzing the graph node parameter perturbation rules in the graph test scene configuration sequence, generating a graph node parameter injection sequence to be injected into the digital twin graph model, which contains the parameter values of the starting graph node in the interaction path and the parameter injection time points.

[0093] In the digital twin graph model test of a commercial office building, the graph node parameter injection sequence is generated by analyzing the graph node parameter perturbation rules in the graph test scene configuration sequence. For the graph node parameter perturbation rules of the structure-equipment location association test scene type, the initial bearing capacity parameter value of the floor node and the parameter continuous change mode are set. According to these rules, the parameter value of the starting graph node (floor node) in the interaction path and the parameter injection time point are determined. If the parameter continuous change mode is linear change, the floor node bearing capacity parameter values at different time points are calculated at certain time intervals to form the graph node parameter injection sequence. This graph node parameter injection sequence will be injected into the digital twin graph model to simulate the impact of hierarchical structure changes on equipment in actual situations.

[0094] Step S142: Monitoring the graph node state of the digital twin graph model according to the associated edge trigger conditions in the graph test scene configuration sequence, and when the trigger conditions are met, injecting the graph node parameter injection sequence into the corresponding starting graph node in order of parameter injection time points.

[0095] In the digital twin graph model test of a commercial office building, the graph node state of the digital twin graph model is monitored according to the associated edge trigger conditions in the graph test scene configuration sequence. For the associated edge trigger conditions of the equipment-environment influence test scene type, it includes the starting graph node (air conditioning equipment node) parameter threshold condition, the intermediate graph node (temperature sensor node) state condition, and the interlayer associated edge conduction intensity condition. These conditions are continuously monitored, and when all conditions are met, the graph node parameter injection sequence is injected into the corresponding starting graph node in order of parameter injection time points. If the operating power of the air conditioning equipment node reaches the parameter threshold condition, the state of the temperature sensor node does not reach the preset target temperature, and the interlayer associated edge conduction intensity meets the condition, the parameter injection sequence of the air conditioning equipment node is injected in order of time points, and the influence of the air conditioning equipment on the indoor environment is simulated.

[0096] Step S143: The graph reasoning engine of the digital twin graph model performs parameter conduction processing based on the interlayer associated edge set. After the starting graph node receives the parameter injection value, it is transmitted to the intermediate graph node through the interlayer associated edge. After the intermediate graph node updates its state, it continues to transmit to the terminal graph node, forming a graph coupling conduction link.

[0097] In the digital twin graph model test of the commercial office building, when the starting graph node receives the parameter injection value, the graph inference engine of the digital twin graph model performs parameter conduction processing based on the inter-layer association edge set. After the starting graph node (such as a floor node) receives the parameter injection value, it is stored in the current state attribute of the starting graph node. The path search module of the graph inference engine finds all directly connected inter-layer association edges starting from the starting graph node based on the inter-layer association edge set, and determines the intermediate graph nodes (such as elevator equipment nodes on the floor) directly connected to the starting graph node. The parameter conduction calculation module calls the edge weight attribute of the inter-layer association edge between the starting graph node and the intermediate graph node, multiplies the parameter injection value of the starting graph node by the edge weight attribute to obtain the input parameter value of the intermediate graph node. The intermediate graph node receives the input parameter value, and the node state update module calls the state transition function corresponding to the parameter type of the intermediate graph node to fuse and calculate the updated state attribute value of the input parameter value and the current state attribute value of the intermediate graph node, and completes the state update of the intermediate graph node. The path search module continues to find all directly connected inter-layer association edges starting from the updated state of the intermediate graph node, and determines the next level intermediate graph node or the terminal graph node (such as an indoor temperature node) directly connected to the intermediate graph node. If it is connected to the next level intermediate graph node, the weighted calculation of the parameter conduction calculation module and the state update of the node state update module are repeated until the conduction reaches the terminal graph node; if it is connected to the terminal graph node, the parameter conduction calculation module calculates the input parameter value of the terminal graph node, and the terminal graph node updates the state attribute value after receiving the input parameter value, and completes a parameter conduction. Record all graph node identifiers, inter-layer association edge identifiers, parameter values and conduction time points involved in the conduction process from the starting graph node to the terminal graph node to form a graph coupling conduction link.

[0098] Step S144: During the operation of the graph coupling conduction link, the test path control logic collects the structure node response sequence of the structure parameter graph node layer, the device node response sequence of the device parameter graph node layer, and the environment node response sequence of the environment parameter graph node layer, and integrates to generate a graph node response sequence.

[0099] In the digital twin graph model test of the commercial office building, the node response sequence of each graph node layer is collected by the test path control logic during the operation of the graph coupling conduction link. The test path control logic specifies the time interval and node range of the collection. According to the test path control logic, the response sequence of the structural nodes such as floor nodes and wall nodes in the structural parameter graph node layer is collected, and the parameter values of the structural nodes are recorded to reflect the changes over time; the response sequence of the equipment nodes such as elevator equipment nodes and air conditioning equipment nodes in the equipment parameter graph node layer is collected; the response sequence of the environmental nodes such as indoor temperature nodes and humidity nodes in the environmental parameter graph node layer is collected. The structural node response sequence, the equipment node response sequence and the environmental node response sequence are integrated to generate a graph node response sequence, which reflects the response of the nodes in each graph node layer to the parameter changes during the graph coupling conduction process.

[0100] Step S145: Correlation analysis is performed on the structural node response sequence, the equipment node response sequence and the environmental node response sequence in the graph node response sequence, the parameter value change relationship between different graph node layers through the inter-layer correlation edges is extracted, and correlation edge conduction relationship data is generated.

[0101] In the digital twin graph model test of the commercial office building, the structural node response sequence, the equipment node response sequence and the environmental node response sequence in the graph node response sequence are subjected to correlation analysis. By comparing the node response sequences of different graph node layers, the parameter value change relationship between different graph node layers through the inter-layer correlation edges is extracted. If the bearing capacity parameter value of the floor node changes, the change of the operating parameters (such as the load limit) of the elevator equipment node on the floor is observed, and the parameter value change relationship between the two is analyzed; at the same time, the influence of the operation of the elevator equipment node on the environmental nodes such as indoor temperature is analyzed. Through the above correlation analysis, the correlation edge conduction relationship data is generated, which can help to deeply understand the parameter conduction mechanism between different graph node layers.

[0102] Step S146: The state values of the graph nodes are monitored during the operation of the graph coupling conduction link, and when the state value of a graph node exceeds the normal range, the graph reasoning engine is used to track the path from the starting abnormal graph node to the final affected graph node through the intermediate conduction graph nodes, record the graph node identifiers, inter-layer correlation edge identifiers and conduction order in the abnormal path, and generate abnormal path graph data.

[0103] In the digital twin graph model test of the commercial office building, the state values of each graph node are monitored during the operation of the graph coupling conduction link. If the state value (such as the running speed) of a certain graph node (such as an elevator equipment node) exceeds the normal range, the graph reasoning engine traces the conduction path of the abnormal state value. The path search module of the graph reasoning engine finds all possible conduction paths starting from the starting abnormal graph node (the elevator equipment node) based on the inter-layer association edge set, determines the intermediate conduction graph nodes (such as the structure nodes of the floor) and the final impact graph nodes (such as the indoor environment nodes of the floor). The graph node identifiers, inter-layer association edge identifiers and conduction sequences in the abnormal path are recorded to generate abnormal path graph data. These abnormal path graph data can help analyze the causes and impact range of abnormal conditions.

[0104] Step S147: Integrate the graph node response sequence, association edge conduction relationship data and abnormal path graph data to generate a graph test response data set.

[0105] In the digital twin graph model test of the commercial office building, the graph node response sequence, association edge conduction relationship data and abnormal path graph data are integrated to generate a graph test response data set. The graph node response sequence reflects the response of the nodes in each graph node layer to parameter changes; the association edge conduction relationship data embodies the parameter conduction mechanism between different graph node layers; and the abnormal path graph data records the conduction path and impact range of abnormal conditions. Integrating these data together forms a complete graph test response data set.

[0106] Step S150: Perform graph structure test result analysis based on the graph test response data set, and output a digital twin simulation test evaluation report of the building system, which includes graph node response characteristics, association edge impact characteristics and abnormal path graph risk characteristics.

[0107] After the digital twin graph model test of the commercial office building is completed, perform graph structure test result analysis based on the graph test response data set, and output a digital twin simulation test evaluation report. Through analysis of the graph test response data set, the performance of the digital twin graph model and the actual situation of the building system can be comprehensively understood. The graph node response characteristics reflect the response law of the nodes in each graph node layer to parameter changes; the association edge impact characteristics embody the parameter conduction impact between different graph node layers; and the abnormal path graph risk characteristics evaluate the risk degree and impact range of abnormal conditions. Comparing these characteristics with the preset graph structure test evaluation benchmark, the compliance degree of each characteristic and the graph nodes or association edges that need to be optimized are determined, and finally the digital twin simulation test evaluation report is output in the preset report format.

[0108] Step S151: Extract the graph node response sequence from the graph test response dataset, analyze the response law of the structure node response sequence of the structure parameter graph node layer with the parameter injection value, the response law of the device node response sequence of the device parameter graph node layer with the parameter injection value, and the response law of the environment node response sequence of the environment parameter graph node layer with the parameter injection value, and integrate to generate the graph node response characteristics.

[0109] In the analysis of the test results of the digital twin graph model of the commercial office building, the graph node response sequence is extracted from the graph test response dataset. The structure node response sequence of the structure parameter graph node layer is analyzed to observe the change of the load capacity parameter value of the floor node with the parameter injection value of the starting graph node (such as the upper structure node), and the response law is summarized, whether it is linear response or nonlinear response. For the device node response sequence of the device parameter graph node layer, the response law of the running parameters (such as load limit, running speed) of the elevator device node with the change of the load capacity parameter of the floor node is analyzed. For the environment node response sequence of the environment parameter graph node layer, the response law of the temperature value of the indoor temperature node with the running parameter of the air conditioning device node is analyzed. The node response laws of these different graph node layers are integrated to generate the graph node response characteristics, which can help understand the sensitivity and response mode of the nodes of each graph node layer to parameter changes.

[0110] Step S152: Extract the associated edge conduction relationship data from the graph test response dataset, analyze the parameter influence law between the structure parameter graph node layer and the device parameter graph node layer through the structure-device interlayer associated edge conduction, the parameter influence law between the device parameter graph node layer and the environment parameter graph node layer through the device-environment interlayer associated edge conduction, and the parameter influence law between the environment parameter graph node layer and the structure parameter graph node layer through the environment-structure interlayer associated edge conduction, and integrate to generate the associated edge influence characteristics.

[0111] In the analysis of the test results of the digital twin graph model of the commercial office building, the associated edge conduction relationship data is extracted from the graph test response dataset. The parameter influence law between the structure parameter graph node layer and the device parameter graph node layer through the structure-device interlayer associated edge conduction is analyzed, and the change of the load capacity of the floor node how to affect the running parameters of the elevator device on the floor. For the parameter influence law between the device parameter graph node layer and the environment parameter graph node layer through the device-environment interlayer associated edge conduction, the influence of the running of the air conditioning device on the indoor temperature and humidity and other environmental parameters is analyzed. For the parameter influence law between the environment parameter graph node layer and the structure parameter graph node layer through the environment-structure interlayer associated edge conduction, the influence of outdoor humidity on the performance of wall structure materials is analyzed. The parameter influence laws between these different graph node layers are integrated to generate the associated edge influence characteristics, which can help understand the role and influence degree of the interlayer associated edge in the parameter conduction process.

[0112] Step S153: Extract abnormal path graph data from the graph test response dataset, analyze the path rule of abnormal state values from the starting abnormal graph node to the final impact graph node through the intermediate conduction graph node, the range rule involving the graph node layer, and the speed rule through the interlayer association edge conduction, and integrate the abnormal path graph risk characteristics by combining the importance rule of the abnormal graph node in the digital twin graph model.

[0113] In the digital twin graph model test result analysis of the commercial office building, abnormal path graph data is extracted from the graph test response dataset. The path rule of abnormal state values is analyzed to determine the type order of graph nodes through which abnormal state values are transmitted, whether it is from structure nodes to equipment nodes and then to environment nodes, or other orders. The number of graph node layers involved in each abnormal path graph data is counted to determine the range of graph node layers covered by abnormal state values, whether it is a local range or a global range. The conduction time difference between adjacent graph nodes of abnormal state values is calculated according to the conduction time point sequence to analyze the speed rule of conduction through the interlayer association edge, whether it is fast conduction or slow conduction. The graph node importance annotation table of the digital twin graph model is queried to obtain the importance annotation corresponding to the starting abnormal graph node identifier, whether it is a core node or a general node. The path rule, range rule, speed rule, and importance annotation are analyzed in association to determine that the global range mode of the core node annotation combined with the fast conduction mode corresponds to a specific risk characteristic, the local range mode of the general node annotation combined with the slow conduction mode corresponds to another specific risk characteristic, and other combinations correspond to corresponding risk characteristics. The risk characteristic descriptions corresponding to different combinations are integrated to generate abnormal path graph risk characteristics, which can help to evaluate the risk degree and impact range of abnormal conditions.

[0114] For example, step S1531: Extract all abnormal path graph data from the graph test response dataset, each abnormal path graph data containing a starting abnormal graph node identifier, a sequence of intermediate conduction graph node identifiers, a final impact graph node identifier, a sequence of interlayer association edge identifiers, and a sequence of conduction time points.

[0115] In the analysis of the test results of the digital twin graph model of the commercial office building, all abnormal path graph data are extracted from the graph test response data set. Each abnormal path graph data records in detail the conduction process of the abnormal situation. The starting abnormal graph node identification clearly identifies the starting point of the abnormal situation, the intermediate conduction graph node identification sequence records the intermediate nodes passed by the abnormal state value, the final impact graph node identification points out the final impact of the abnormal situation, the interlayer association edge identification sequence records the interlayer association edges passed by the abnormal state value during conduction, and the conduction time point sequence records the conduction time between nodes and edges. For example, in the digital twin graph model test of the commercial office building, abnormal state value conduction of the elevator equipment node may occur. The abnormal path graph data will clearly record the identification of the starting abnormal graph node as the elevator equipment node, and the identification of the intermediate structure node may be passed through the intermediate conduction graph node identification sequence, and finally affect the indoor environment node of the floor, whose identification is the final impact graph node identification. At the same time, the interlayer association edge identification sequence of the abnormal state value from the elevator equipment node to the structure node through a specific structure equipment interlayer association edge, and then to the indoor environment node through other association edges, and the specific time points of each conduction, forming the conduction time point sequence, can be recorded.

[0116] Step S1532: For each abnormal path graph data, analyze the types of starting abnormal graph node identification, intermediate conduction graph node identification sequence and final impact graph node identification, determine the graph node type order of abnormal state value transmission, and summarize the path rule, which includes structure device environment path mode, device environment structure path mode and environment structure device path mode.

[0117] In this step, for each abnormal path graph data extracted, the types of starting abnormal graph node identification, intermediate conduction graph node identification sequence and final impact graph node identification are analyzed in detail. In the scenario of a commercial office building, the starting abnormal graph node can be of device type (such as elevator equipment node), structure type (such as floor structure node) or environment type (such as indoor temperature abnormal environment node). The node types contained in the intermediate conduction graph node identification sequence are also different, which can pass through structure nodes, device nodes, etc. in turn. The final impact graph node can also be of different types. By analyzing the order of these node types, the graph node type order of abnormal state value transmission can be determined. If the starting abnormal graph node is of device type, passes through structure nodes in the middle, and finally affects environment nodes, a device-structure-environment path mode is formed. According to the analysis results of a large number of abnormal path graph data, common path rules are summarized, such as structure device environment path mode, device environment structure path mode and environment structure device path mode, etc.

[0118] Step S1533: Count the number of graph node layers involved in each abnormal path graph data, determine the range of graph node layers covered by the abnormal state value, summarize the range law of the graph node layers involved, which includes the local range mode covering two graph node layers and the global range mode covering three graph node layers.

[0119] In the analysis of the test results of the digital twin graph model of the commercial office building, for each abnormal path graph data, the number of graph node layers involved needs to be counted. The graph node layers include the structure parameter graph node layer, the equipment parameter graph node layer and the environment parameter graph node layer. If an abnormal path graph data only involves the nodes of the structure parameter graph node layer and the equipment parameter graph node layer, the number of graph node layers covered by the abnormal state value is two, belonging to the local range mode. For example, the abnormal state value of the elevator equipment node is only conducted to the structure node of the floor, and does not affect the nodes of the environment parameter graph node layer. If the abnormal path graph data involves the nodes of the structure parameter graph node layer, the equipment parameter graph node layer and the environment parameter graph node layer, the number of graph node layers covered by the abnormal state value is three, belonging to the global range mode. For example, the abnormal state value of the elevator equipment node is first conducted to the structure node of the floor, and then affects the environment parameter nodes such as temperature and humidity in the room. Through the statistics and analysis of a large number of abnormal path graph data, the range law of the graph node layers involved is summarized.

[0120] Step S1534: Calculate the conduction time difference between adjacent graph nodes according to the conduction time point sequence, analyze the speed law of conduction through the interlayer association edge, which includes the fast conduction mode with shorter conduction time difference and the slow conduction mode with longer conduction time difference.

[0121] In the analysis of the abnormal path of the digital twin graph model of the commercial office building, the conduction time difference between adjacent graph nodes is calculated according to the conduction time point sequence in the abnormal path graph data. The conduction time point sequence records the specific time when the abnormal state value reaches each graph node. By subtracting the conduction time of the previous adjacent graph node from the conduction time of the next graph node, the conduction time difference between the two adjacent graph nodes can be obtained. If a certain abnormal state value is conducted from the elevator equipment node to the structure node of the floor, the conduction time point sequence shows that the time of reaching the elevator equipment node is a specific time, and the time of reaching the structure node is another time later. Subtracting the two times gives the conduction time difference. When the conduction time difference is short, it means that the abnormal state value conducts faster between the two adjacent graph nodes, belonging to the fast conduction mode; otherwise, when the conduction time difference is long, it belongs to the slow conduction mode.

[0122] Step S1535: Query the graph node importance annotation table of the digital twin graph model to obtain the importance annotation corresponding to the starting abnormal graph node identifier. The importance annotation includes core node annotation and general node annotation.

[0123] In the digital twin graph model of a commercial office building, there is a graph node importance annotation table. This table annotates the importance of each graph node, and the annotation types include core node annotation and general node annotation. Core nodes are usually nodes that play a key role in the normal operation of the entire building system, such as the main control elevator node in the elevator equipment node and the key beam column node in the structure parameter graph node layer. General nodes are nodes that have relatively small impact on system operation, such as some auxiliary lighting equipment nodes. When analyzing abnormal path graph data, the graph node importance annotation table needs to be queried to obtain the importance annotation corresponding to the starting abnormal graph node identifier. If the starting abnormal graph node is the main control elevator node, the importance annotation can be obtained through the table, which is the core node. If it is a certain ordinary lighting equipment node, its importance annotation is a general node.

[0124] Step S1536: Correlation analysis of path rules, range rules, speed rules, and importance annotations. The combination of global range mode and fast conduction mode of core node annotation corresponds to a specific risk characteristic, the combination of local range mode and slow conduction mode of general node annotation corresponds to another specific risk characteristic, and other combinations correspond to corresponding risk characteristics.

[0125] In the abnormal risk analysis of the digital twin graph model of a commercial office building, the path rules, range rules, speed rules, and importance annotations obtained in the previous analysis need to be correlated. When the importance annotation of the starting abnormal graph node is a core node, and the conduction of the abnormal state value conforms to the global range mode and belongs to the fast conduction mode, the above combination corresponds to a specific high-risk characteristic. The abnormality of the core node will quickly spread to the structure, equipment, and environment three graph node layers, which may cause serious problems in the operation of the entire building system, such as the abnormality of the main control elevator node quickly affecting the floor structure and indoor environment, which may endanger personnel safety and normal operation of equipment. When the importance annotation of the starting abnormal graph node is a general node, and the conduction of the abnormal state value conforms to the local range mode and belongs to the slow conduction mode, it corresponds to a relatively low risk characteristic. The abnormality of the ordinary lighting equipment node slowly affects the local environmental parameters, and has little impact on the entire building system. For other combinations of different path rules, range rules, speed rules, and importance annotations, there are corresponding risk characteristics. Through the above correlation analysis, the risk degree of the abnormal situation can be more accurately evaluated.

[0126] Step S1537: Integrate the risk characteristic descriptions corresponding to different combinations to generate the abnormal path graph risk characteristic.

[0127] After completing the correlation analysis of the path law, range law, speed law and importance annotation, the risk characteristic descriptions corresponding to different combinations are integrated. In the digital twin graph model of the commercial office building, for the combination of global range mode and fast conduction mode annotated by the core node, the risk characteristic description may be a serious threat to the operation of the entire building system, which may cause personnel casualties and major equipment damage; for the combination of local range mode and slow conduction mode annotated by the general node, the risk characteristic description may have a slight impact on the local environment and does not affect the normal operation of the overall system. The risk characteristic descriptions of these different combinations are integrated according to certain logic to form a comprehensive abnormal path graph risk characteristic. This characteristic can comprehensively reflect the propagation risk of abnormal state values in the digital twin graph model.

[0128] Step S154: Compare the graph node response characteristic, associated edge influence characteristic and abnormal path graph risk characteristic with the preset graph structure test evaluation benchmark to determine the compliance degree of each characteristic and the graph node or associated edge that needs to be optimized.

[0129] In the test result analysis of the digital twin graph model of the commercial office building, the obtained graph node response characteristic, associated edge influence characteristic and abnormal path graph risk characteristic are compared and analyzed with the preset graph structure test evaluation benchmark. The preset graph structure test evaluation benchmark is formulated according to the design requirements and safety standards of the building system, which specifies the reasonable range of graph node response, the normal degree of associated edge influence and the acceptable level of abnormal path graph risk. For the graph node response characteristic, compare whether the node response law of each graph node layer is consistent with the response law specified in the benchmark. If the response of the floor node to parameter changes is too sensitive or lagging, which exceeds the benchmark range, then it means that the response characteristic of the floor node does not meet the requirements. For the associated edge influence characteristic, check whether the parameter influence law between different graph node layers is consistent with the benchmark. If the influence degree of the air conditioning equipment on the indoor temperature is significantly different from the influence degree specified in the benchmark, then the corresponding device environment interlayer associated edge may need to be optimized. For the abnormal path graph risk characteristic, compare the risk degree with the acceptable risk level in the benchmark. If the risk characteristic of the abnormal situation exceeds the benchmark range, then find out the starting abnormal graph node or related interlayer associated edge that causes high risk for optimization. Through the above comparison and analysis, the compliance degree of each characteristic is determined, and the graph node or associated edge that needs to be optimized is found out.

[0130] Step S155: Integrate the graph node response characteristic, associated edge influence characteristic, abnormal path graph risk characteristic and comparison and analysis results according to the preset report format, and output a digital twin simulation test evaluation report.

[0131] After the digital twin graph model testing of the commercial office building is completed, relevant information is integrated and a digital twin simulation test evaluation report is output according to a preset report format. The preset report format typically includes a title, table of contents, main body, and conclusion. The main body first describes the response characteristics of the graph nodes in detail, including the node response patterns of each graph node layer, using specific analysis results and examples to illustrate the node's response to parameter changes. Next, it explains the influence characteristics of associated edges, introducing the parameter influence patterns transmitted between different graph node layers through inter-layer associated edges, and the impact of these patterns on the building system operation. Then, it describes the risk characteristics of the abnormal path graph, including the path patterns, range patterns, and speed patterns of abnormal state values, as well as the corresponding risk levels. Next, it presents a detailed comparative analysis of the graph node response characteristics, associated edge influence characteristics, and abnormal path graph risk characteristics with the preset graph structure test evaluation benchmark, indicating the degree of conformity of each characteristic and the graph nodes or associated edges that need optimization. Finally, the conclusion summarizes the main findings and recommendations of the test, providing a reference for the optimization and improvement of the building system. By integrating data according to a preset report format, the output digital twin simulation test evaluation report can clearly and comprehensively present the test results, providing a basis for decision-making for relevant personnel.

[0132] Figure 2 The illustration shows exemplary hardware and software components of a digital twin simulation test system 100 for building systems, which can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the digital twin simulation test system 100 for building systems and to perform the functions in this application.

[0133] For example, a digital twin simulation test system 100 applied to a building system may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the digital twin simulation test system 100 applied to a building system may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The digital twin simulation test system 100 applied to a building system also includes an I / O interface 150 between the computer and other input / output devices.

[0134] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned digital twin simulation test method applied to building systems is implemented.

[0135] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A digital twin simulation testing method for building systems, characterized in that, The method includes: Perform knowledge graph construction and processing of building system physical parameters to generate a physical parameter knowledge graph; Based on the physical parameter knowledge graph, a digital twin graph model generation process is performed to construct a digital twin graph model of the building system; A test scenario configuration sequence is generated based on the set of inter-layer related edges and node interaction paths of the digital twin graph model. The graph test scenario configuration sequence is input into the digital twin graph model to perform graph coupling simulation testing, generating a graph test response dataset; Based on the graph test response dataset, the graph structure test results are analyzed, and a digital twin simulation test evaluation report of the building system is output. The digital twin simulation test evaluation report includes the response characteristics of graph nodes, the influence characteristics of associated edges, and the risk characteristics of abnormal path graphs.

2. The digital twin simulation testing method for building systems according to claim 1, characterized in that, The process of constructing a knowledge graph of the physical parameters of the building system generates a physical parameter knowledge graph, including: Collect physical parameter data of building systems, including structural parameter data, equipment parameter data, and environmental parameter data; The structural parameter data is subjected to graph node extraction processing to extract structural parameter entities and entity attributes, and structural parameter nodes are generated. The structural parameter nodes contain node identifiers, parameter types and attribute descriptions. The device parameter data is subjected to graph node extraction processing to extract device parameter entities and entity attributes, and generate device parameter nodes. The device parameter nodes include node identifier, parameter type and attribute description information. The environmental parameter data is subjected to graph node extraction processing to extract environmental parameter entities and entity attributes, and generate environmental parameter nodes. The environmental parameter nodes include node identifier, parameter type and attribute description information. Perform edge relationship extraction processing on the physical association relationship between the structural parameter node and the device parameter node to generate position association edges. The position association edges include edge identifier, association type and association strength description. Perform edge relationship extraction processing on the physical association relationship between the device parameter node and the environmental parameter node to generate influential association edges, which include edge identifier, association type and association strength description; Perform edge relationship extraction processing on the physical association relationship between the environmental parameter node and the structural parameter node to generate interactive association edges. The interactive association edges include edge identifier, association type and association strength description. The structural parameter nodes, equipment parameter nodes, environmental parameter nodes, and location-related edges, influence-related edges, and interaction-related edges are integrated. The integrity of node attributes and the rationality of edge relationships are verified through a graph consistency verification tool to generate a physical parameter knowledge graph.

3. The digital twin simulation testing method for building systems according to claim 1, characterized in that, The process of generating a digital twin model based on the physical parameter knowledge graph to construct a digital twin model of the building system includes: Structural parameter nodes are extracted from the physical parameter knowledge graph and aggregated hierarchically according to parameter type to generate a structural parameter graph node layer. The structural parameter graph node layer contains multiple hierarchical child nodes, and the hierarchical child nodes are connected by internal association edges. Equipment parameter nodes are extracted from the physical parameter knowledge graph and aggregated hierarchically according to parameter type to generate a device parameter graph node layer. The device parameter graph node layer contains multiple hierarchical child nodes, and the hierarchical child nodes are connected by internal association edges. Environmental parameter nodes are extracted from the physical parameter knowledge graph and aggregated hierarchically according to parameter type to generate an environmental parameter graph node layer. The environmental parameter graph node layer contains multiple hierarchical child nodes, and the hierarchical child nodes are connected by internal association edges. Position-related edges are extracted from the physical parameter knowledge graph, and the hierarchical child nodes of the structural parameter graph node layer and the hierarchical child nodes of the equipment parameter graph node layer are bound by position-related edges to generate a subset of inter-structure and inter-equipment layer related edges. Influence-related edges are extracted from the physical parameter knowledge graph, and the hierarchical child nodes of the device parameter graph node layer and the hierarchical child nodes of the environmental parameter graph node layer are bound by influence-related edges to generate a subset of inter-device and inter-environmental related edges. Interactive association edges are extracted from the physical parameter knowledge graph, and the hierarchical child nodes of the environmental parameter graph node layer and the hierarchical child nodes of the structural parameter graph node layer are bound by interactive association edges to generate a subset of inter-environmental and structural layer association edges. The inter-layer related edge subsets of the structure equipment, the inter-layer related edge subsets of the equipment environment, and the inter-layer related edge subsets of the environmental structure are integrated to form an inter-layer related edge set, which includes an edge weight attribute to represent the parameter transmission strength. A graph reasoning engine is constructed, which realizes parameter transmission calculation and logical relationship reasoning based on the set of inter-layer related edges. The structural parameter graph node layer, equipment parameter graph node layer, environmental parameter graph node layer, and the set of inter-layer related edges are integrated with the graph reasoning engine to generate a digital twin graph model.

4. The digital twin simulation testing method for building systems according to claim 1, characterized in that, The step of generating a graph test scenario configuration sequence based on the set of inter-layer related edges and node interaction paths of the digital twin graph model includes: The set of inter-layer association edges of the digital twin graph model is analyzed to identify all possible interaction paths between the structural parameter graph node layer, the device parameter graph node layer, and the environmental parameter graph node layer. Each interaction path includes a starting graph node, an intermediate graph node, a ending graph node, and inter-layer association edges connecting each node. Test scenario types are divided based on the node types and inter-layer association edge attributes of the interaction path. Each test scenario type corresponds to the collaborative testing requirements of one or more interaction paths, and the scenario testing objectives are determined for each test scenario type. For each test scenario type, graph node parameter perturbation rules are designed based on the parameter type of the starting graph node in the corresponding interaction path and the scenario test objective. The graph node parameter perturbation rules include the initial parameter value setting of the starting graph node, the continuous parameter change mode, and the parameter cooperative perturbation relationship. Based on the inter-layer association edge attributes of the scenario test target and interaction path, design association edge triggering conditions. The association edge triggering conditions include the threshold conditions of the starting graph node parameters, the state conditions of the intermediate graph nodes, and the inter-layer association edge transmission strength conditions. The test path control logic is designed based on the graph node parameter perturbation rules and associated edge triggering conditions. The test path control logic includes the interactive path selection strategy, the node parameter injection order and the test process interruption mechanism. The graph node parameter perturbation rules, associated edge triggering conditions, and test path control logic are associated and integrated according to the test scenario type to generate a graph test scenario configuration sequence.

5. The digital twin simulation testing method for building systems according to claim 1, characterized in that, The step of inputting the graph test scenario configuration sequence into the digital twin graph model to perform graph coupling simulation testing and generate a graph test response dataset includes: The graph node parameter perturbation rules in the graph test scenario configuration sequence are analyzed to generate a graph node parameter injection sequence for the digital twin graph model to be injected. The graph node parameter injection sequence includes the parameter values ​​of the starting graph node in the interaction path and the parameter injection time point. According to the associated edge triggering conditions in the graph test scenario configuration sequence, the graph node status of the digital twin graph model is monitored. When the triggering conditions are met, the graph node parameter injection sequence is injected into the corresponding starting graph node in the order of parameter injection time. The graph reasoning engine of the digital twin graph model performs parameter propagation processing based on the set of inter-layer associated edges. After the starting graph node receives the parameter injection value, it propagates it to the intermediate graph node through the inter-layer associated edges. After the intermediate graph node updates its own state, it continues to propagate to the terminating graph node, forming a graph coupling propagation link. During the operation of the graph coupling transmission link, the structural node response sequence of the structural parameter graph node layer, the device node response sequence of the equipment parameter graph node layer, and the environmental node response sequence of the environmental parameter graph node layer are collected according to the test path control logic and integrated to generate the graph node response sequence. Correlation analysis is performed on the structural node response sequence, device node response sequence, and environmental node response sequence in the graph node response sequence to extract the parameter value change relationship between different graph node layers through inter-layer correlation edges, and generate correlation edge transmission relationship data; During the operation of the graph coupling transmission link, the status values ​​of each graph node are monitored. When the status value of a graph node exceeds the normal range, the graph inference engine tracks the path of the abnormal status value from the initial abnormal graph node through the intermediate transmission graph nodes to the final affected graph node. The graph node identifiers, inter-layer association edge identifiers and transmission order in the abnormal path are recorded to generate abnormal path graph data. By integrating the graph node response sequences, associated edge propagation relationship data, and abnormal path graph data, a graph test response dataset is generated.

6. The digital twin simulation testing method for building systems according to claim 1, characterized in that, The process involves analyzing the graph structure test results based on the graph test response dataset and outputting a digital twin simulation test evaluation report for the building system, including: Graph node response sequences are extracted from the graph test response dataset. The response patterns of structural node response sequences in the structural parameter graph node layer, equipment node response sequences in the equipment parameter graph node layer, and environmental node response sequences in the environmental parameter graph node layer are analyzed to generate graph node response characteristics. Extract the associated edge transmission relationship data from the graph test response dataset, analyze the parameter influence law between the structural parameter graph node layer and the device parameter graph node layer through the associated edge between the structural and device layers, the parameter influence law between the device parameter graph node layer and the environmental parameter graph node layer through the associated edge between the device and environmental layers, and the parameter influence law between the environmental parameter graph node layer and the structural parameter graph node layer through the associated edge between the environmental and structural layers, and integrate them to generate associated edge influence characteristics; Extract abnormal path graph data from the graph test response dataset, analyze the path pattern of abnormal state values ​​from the initial abnormal graph node through intermediate transmission graph nodes to the final affected graph node, the range pattern of graph node layers involved, and the speed pattern of transmission through inter-layer associated edges, and combine the importance pattern of abnormal graph nodes in the digital twin graph model to integrate and generate the risk characteristics of abnormal path graphs; The response characteristics of the graph nodes, the influence characteristics of the associated edges, and the risk characteristics of the abnormal path graph are compared and analyzed with the preset graph structure test and evaluation benchmarks to determine the degree of compliance of each characteristic and the graph nodes or associated edges that need to be optimized. The report integrates the graph node response characteristics, associated edge impact characteristics, abnormal path graph risk characteristics, and comparative analysis results according to the preset report format, and outputs a digital twin simulation test evaluation report.

7. The digital twin simulation testing method for building systems according to claim 2, characterized in that, The step of performing graph node extraction processing on the structural parameter data to extract structural parameter entities and entity attributes, and generating structural parameter nodes, includes: Perform data cleaning on the collected structural parameter data to remove duplicate data records and invalid data fields, and retain valid data containing descriptions of structural features; Entity recognition processing is performed on the cleaned structural parameter data. Based on a preset structural parameter entity dictionary, keywords representing structural entities in the data are identified. The structural entities include wall entities, floor entities, beam and column entities, and spatial area entities. The identified structural entities are subjected to attribute extraction processing to extract attribute information for each structural entity from the data. The attribute information includes entity size attributes, material attributes, location attributes, and connection attributes. Each structural entity is assigned a unique node identifier, which consists of an entity type prefix and a sequence number; The nodes are classified into parameter types according to the type of structural entity. The parameter types include wall parameter types, floor parameter types, beam and column parameter types, and spatial area parameter types. The node identifier, parameter type, and extracted attribute information are integrated to generate a structural parameter node. The attribute description information of the structural parameter node includes entity size attribute value, material attribute value, location attribute value, and connection attribute value. Perform attribute integrity checks on the generated structural parameter nodes to ensure that each node contains at least one core attribute. Perform attribute completion processing on nodes that are missing core attributes, and check again after completion, until all structural parameter nodes pass the attribute integrity check.

8. The digital twin simulation testing method for building systems according to claim 3, characterized in that, The graph reasoning engine includes: The parameter transmission calculation module of the design graph inference engine receives the parameter values ​​of the starting graph nodes and the edge weight attributes of the inter-layer related edges, and calculates the input parameter values ​​of the intermediate graph nodes through a weighted summation algorithm. The input of the weighted summation algorithm is the product of the parameter values ​​of the starting graph nodes and the edge weight attributes, and the output is the input parameter values ​​of the intermediate graph nodes. The design graph reasoning engine has a logical relationship reasoning module. The logical relationship reasoning module calls a preset logical reasoning rule library based on the association type of the inter-layer association edges to reason about the parameter relationships between different graph node layers. The logical reasoning rule library includes position association reasoning rules, influence association reasoning rules, and interaction association reasoning rules. The design graph inference engine includes a node state update module. This module receives input parameter values ​​from intermediate graph nodes and calculates the output parameter values ​​of the nodes using a state transition function, which is determined based on the node's parameter type. The design of the path search module of the graph reasoning engine uses a depth-first search algorithm or a breadth-first search algorithm based on the set of inter-layer associated edges to find all possible propagation paths between any two graph nodes, and outputs the graph node identifiers and inter-layer associated edge identifiers contained in the path. The core processing unit of the graph reasoning engine is constructed by integrating the parameter transmission calculation module, logical relationship reasoning module, node state update module and path search module, and the data interaction interface and calling order between each module are defined. By inputting preset test parameter values ​​and known transmission paths, the accuracy of parameter transmission calculation, the correctness of logical relationship reasoning, and the real-time update of node states of the graph inference engine are verified. The algorithm parameters and inference rules in the module are adjusted until the graph inference engine meets the preset performance indicators.

9. The digital twin simulation testing method for building systems according to claim 5, characterized in that, The graph inference engine of the digital twin graph model performs parameter propagation processing based on the set of inter-layer interconnected edges. After receiving the parameter injection value, the starting graph node propagates it to the intermediate graph node through the inter-layer interconnected edges. The intermediate graph node updates its own state and continues to propagate to the terminating graph node, forming a graph coupling propagation link, including: The starting graph node receives the parameter injection value from the graph node parameter injection sequence and stores the parameter injection value in its current state attribute; The path search module of the graph reasoning engine searches for all directly connected inter-layer edges originating from the starting graph node based on the set of inter-layer related edges, and determines the intermediate graph nodes that are directly connected to the starting graph node. The parameter propagation calculation module calls the edge weight attribute of the inter-layer association edge between the starting graph node and the intermediate graph node, and multiplies the parameter injection value of the starting graph node with the edge weight attribute to obtain the input parameter value of the intermediate graph node; The intermediate graph node receives the input parameter value. The node state update module calls the state transition function corresponding to the parameter type of the intermediate graph node, merges the input parameter value with the current state attribute value of the intermediate graph node to calculate and update the state attribute value, and completes its own state update. The path search module continues to search for all directly connected inter-layer association edges originating from the intermediate graph node after the update status, and determines the next level intermediate graph node or the terminal graph node that is directly connected to the intermediate graph node. If the connection is to the next level intermediate graph node, the weighted calculation of the parameter propagation calculation module and the state update steps of the node state update module are repeated until the propagation reaches the terminal graph node. If the connection is to a terminal graph node, the parameter transmission calculation module calculates the input parameter value of the terminal graph node. After receiving the input parameter value, the terminal graph node updates its own state attribute value, thus completing one parameter transmission. Record all graph node identifiers, inter-layer associated edge identifiers, parameter values, and transmission time points involved in the transmission process from the starting graph node to the ending graph node, forming a transmission log of the graph coupling transmission link.

10. A digital twin simulation testing system for building systems, characterized in that, The system includes a processor and a memory, the memory being connected to the processor. The memory is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the memory to implement the digital twin simulation testing method for building systems as described in any one of claims 1-9.

Images