A building intelligent management and control method, device, equipment and medium

Abstract

The application relates to the technical field of intelligent buildings and city information model, and specifically discloses a building intelligent management and control method, device, equipment and medium. The application obtains multi-source heterogeneous data containing building information model data, geographic information system data, Internet of Things sensor data and video monitoring data, and fuses the data based on a unified coordinate system to construct a city information model three-dimensional digital bottom plate. Meanwhile, the application performs semantic understanding and intention recognition on natural language instructions to generate structured task instructions, thereby giving the system deep cognitive reasoning capability. Meanwhile, the application performs task decomposition and scheduling on the structured task instructions through a multi-agent collaborative engine, controls corresponding business agents to perform related operations, can effectively process complex linkage tasks across systems, realizes efficient decoupling and parallel processing of tasks, and thereby improves the intelligent level and response efficiency of building operation management.

Description

Technical Field

[0001] This application relates to the field of smart building and urban information modeling technology, specifically to a building intelligent management and control method, device, equipment and medium. Background Technology

[0002] Currently, City Information Modeling (CIM) platforms primarily serve as tools for 3D visualization of cities. Their information interaction and decision-making modes typically rely on a single, passive data presentation and manual interpretation. Specifically, users mainly use a mouse to click, drag, and zoom in a 3D scene to view static information such as building exteriors and spatial layouts, lacking intelligent analysis and proactive early warning capabilities for dynamic information such as building operational status, energy consumption trends, and equipment health.

[0003] Traditional building management systems are mostly decentralized and independent subsystems. The lack of unified data standards and interoperability between systems leads to data fragmentation and difficulties in collaboration. Operation and maintenance work still relies heavily on manual experience and repetitive operations, making it difficult to achieve intelligent closed-loop management of the entire process of planning, construction, and operation and maintenance.

[0004] Although existing research has attempted to combine artificial intelligence technology with CIM platforms, most of these efforts remain at the level of functional overlay and have not yet achieved deep integration of semantic understanding, intelligent reasoning, and multi-scenario collaboration, thus failing to effectively support natural interaction and proactive decision-making. Summary of the Invention

[0005] This application provides a method, device, equipment, and medium for intelligent building management and control to solve the above-mentioned technical problems.

[0006] Firstly, this application provides a building intelligent management and control method, including: Acquire first multi-source heterogeneous data, which includes building information model data, geographic information system data, Internet of Things sensor data, and video surveillance data; The first multi-source heterogeneous data is fused based on a unified coordinate system, and a three-dimensional digital base plate for the city information model is constructed based on the fused data. Obtain the user's natural language instructions, perform semantic understanding and intent recognition on the natural language instructions, and generate structured task instructions; The structured task instructions are decomposed and scheduled by a multi-agent collaborative engine to control the corresponding business agents to perform at least one operation related to building operation and management. The execution results of the business intelligence agent and the building operation status are mapped to the corresponding spatial locations in the three-dimensional digital base of the city information model and then visualized.

[0007] In one embodiment of this application, after acquiring the first multi-source heterogeneous data inside and outside the building, the method further includes: On the cloud platform, based on the building's historical operation data, equipment fault prediction models and energy consumption optimization models are trained, and the trained equipment fault prediction models and energy consumption optimization models are distributed to edge computing nodes. The edge computing node performs equipment health assessments and energy-saving strategy adjustments based on real-time building operation data, equipment fault prediction models, and energy consumption optimization models, and reports the equipment health assessment results and energy-saving strategy adjustment results to the cloud platform.

[0008] In one embodiment of this application, semantic understanding and intent recognition are performed on the natural language instructions to generate structured task instructions, including: Obtain a corpus of architectural terminology, and use the corpus to incrementally train a pre-trained large model to obtain a domain-specific large model. The natural language instructions are parsed based on the domain-specific large model and the pre-built dynamic knowledge graph of the construction domain to identify user intent; A structured task instruction is generated based on the user's intent. The structured task instruction includes the task type, target device, target spatial location, execution parameters, and feedback method.

[0009] In one embodiment of this application, a method for constructing a dynamic knowledge graph in the field of architecture includes: Acquire second multi-source heterogeneous data; Entities in the building domain are extracted from the second multi-source heterogeneous data. The entities include at least physical components, spatial units, equipment operating status, event records, fault descriptions, maintenance actions, personnel attributes, equipment parameters, and operational constraints. Define semantic relationships between the entities to form a network knowledge structure representing semantic connections between entities; wherein, the semantic relationships include at least spatial location relationships, topological connection relationships, control association relationships, monitoring correspondence relationships, and maintenance operation relationships; The entities and semantic relationships are stored in a graph database, and a unique identifier mapping is established between each entity and the spatial entity in the three-dimensional digital base of the city information model.

[0010] In one embodiment of this application, the multi-agent collaborative engine adopts a hierarchical distributed architecture of a master intelligent agent and business intelligent agents; the multi-agent collaborative engine decomposes and schedules the structured task instructions to control the corresponding business intelligent agents to execute at least one operation related to building operation management, including: The structured task instructions are input into the main control agent; The main control agent decomposes the task based on task association information, breaking down the structured task instructions into multiple sub-tasks. The multiple subtasks are scheduled to the corresponding business intelligent agents, and the business intelligent agents include at least one of the following: equipment control intelligent agent, energy and carbon management intelligent agent, intelligent operation and maintenance intelligent agent, and smart security intelligent agent; The scheduled subtasks are executed in parallel by the multiple business intelligence agents, and the execution status is fed back to the master control intelligence agent.

[0011] In one embodiment of this application, the execution result of the business intelligence agent and the building operation status are mapped to the corresponding spatial location in the three-dimensional digital base of the city information model and visualized, including: At least one of the equipment's operating parameters, energy consumption data, security events, and work order status is mapped and rendered to the corresponding spatial location in the three-dimensional digital base plate of the city information model. Different colors, icons, or animated codes are used to represent device status, energy consumption level, or event level; It receives natural language interactive query commands input into the three-dimensional digital base of the city information model, and outputs and displays the real-time status and historical trends of the area or device specified by the natural language interactive query command.

[0012] In one embodiment of this application, the multi-source heterogeneous data is fused based on a unified coordinate system to construct a three-dimensional digital base plate for the city information model, including: Based on the data change rate of each data source corresponding to the first multi-source heterogeneous data, the time window corresponding to the data source is determined, and within the time window, the observed data is weighted and averaged based on the time distance between the observed data and the target time to obtain the time-aligned data. Using the coordinate system of the building information model data as a reference, the coordinate transformation matrix of each data source is determined by control point matching; Based on the coordinate transformation matrix, homogeneous coordinate transformation is used to convert the spatial coordinates of each data source to the unified coordinate system in order to construct the three-dimensional digital base of the city information model.

[0013] Secondly, this application provides a building intelligent control device, comprising: The data acquisition module is used to acquire first multi-source heterogeneous data, which includes building information model data, geographic information system data, Internet of Things sensor data, and video surveillance data. The data fusion module is used to fuse the first multi-source heterogeneous data based on a unified coordinate system, and to construct a three-dimensional digital base plate for the city information model based on the fused data. an instruction generation module, configured to acquire a user's natural language instruction, perform semantic understanding and intent recognition on the natural language instruction, and generate a structured task instruction; a task execution control module, configured to perform task decomposition and scheduling on the structured task instruction through a multi-agent collaboration engine, so as to control a corresponding business agent to execute at least one operation related to building operation management; a mapping control module, configured to map the execution result of the business agent and the building operation state to the corresponding spatial position in the three-dimensional digital base map of the city information model, and perform visual display.

[0014] In a third aspect, the present application provides an electronic device, comprising: one or more processors; and a memory, configured to store one or more programs, which when executed by the one or more processors, cause the processors to implement the above-mentioned building intelligent management and control method.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which when executed by a processor of a computer, causes the computer to execute the building intelligent management and control method.

[0016] Beneficial technical effects: By acquiring multi-source heterogeneous data including building information model data, geographic information system data, IoT sensor data and video surveillance data, and performing fusion processing thereon based on a unified coordinate system, the present application constructs a three-dimensional digital base map of a city information model, realizing the precise alignment of sensing data in the physical space; at the same time, the present application performs semantic understanding and intent recognition on natural language instructions to generate structured task instructions, thereby endowing the system with deep cognitive reasoning capability; in addition, the present application performs task decomposition and scheduling on structured task instructions through a multi-agent collaboration engine, and controls corresponding business agents to execute related operations, which can effectively process cross-system complex linkage tasks, realize efficient task decoupling and parallel processing, ensure an automated closed loop from perception to control, and improve the intelligence level and response efficiency of building operation management.

[0017] It should be understood that the foregoing general description and the following detailed description are merely exemplary and explanatory, and do not limit the present application. Brief Description of the Drawings

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0019] In the attached diagram: Figure 1 This is a schematic diagram of the architecture topology of a building intelligent management and control system provided in one embodiment of this application; Figure 2 This is a flowchart illustrating a building intelligent management and control method according to an embodiment of this application; Figure 3 A flowchart illustrating how to perform semantic understanding and intent recognition on natural language instructions to generate structured task instructions, according to an embodiment of this application; Figure 4 A flowchart illustrating the construction of a dynamic knowledge graph in the field of architecture, according to an embodiment of this application; Figure 5 This is a flowchart illustrating the construction of a three-dimensional digital base plate for a city information model according to an embodiment of this application; Figure 6 This is a schematic diagram of the principle of a building intelligent control device according to an embodiment of this application; Figure 7 A schematic diagram of a computer system suitable for implementing the memory of the embodiments of this application is shown. Detailed Implementation

[0020] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0021] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. The drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the shape, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0022] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of this application. However, it will be apparent to those skilled in the art that embodiments of this application may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of this application.

[0023] For ease of understanding, the following explains some key terms used in this application: City Information Model (CIM) refers to a three-dimensional digital space that integrates urban information based on technologies such as Building Information Modeling (BIM), Geographic Information System (GIS), and the Internet of Things (IoT), combining multi-scale spatial data (above and below ground, indoor and outdoor, historical and current status) with urban perception data.

[0024] Building Information Modeling (BIM) data refers to building model data built on the basis of various relevant information data of a building project. It includes geometric information, material properties, spatial relationships and construction information of building components.

[0025] Geographic Information System (GIS) data refers to data that describes the macro-geographic environment, such as geographic spatial location and topography.

[0026] A city information model's 3D digital baseboard refers to a high-precision 3D digital spatial carrier that can dynamically attach data based on internal and external conditions. Its function is to provide precise physical spatial anchor points for building operation data and to provide a basic scene for visualization and spatial query. It includes high-precision 3D terrain, building models, underground pipelines, transportation facilities, and other spatial objects, each with a globally unique spatial identifier.

[0027] Domain-specific large models refer to large language models that are incrementally trained using a corpus of architectural professional knowledge, based on general pre-trained large models. These models possess the ability to understand and reason about architectural professional knowledge. Their function is to perform semantic understanding and intent recognition on input natural language instructions and output structured task instructions.

[0028] A multi-agent collaborative engine refers to a task scheduling and execution framework consisting of a master intelligent agent and multiple business intelligent agents. Its function is to receive structured task instructions, break them down into several sub-tasks, schedule them to the corresponding business intelligent agents for execution, and aggregate the execution status feedback of each business intelligent agent to form a closed-loop control.

[0029] Retrieval-Augmented Generation (RAG) refers to a technique that retrieves relevant knowledge from an external knowledge base before a large model generates a response or instruction, and then provides the retrieval results as contextual hints to the large model to improve the accuracy and professionalism of the model's output.

[0030] Please see Figure 1 , Figure 1 This is a schematic diagram of the architectural topology of a building intelligent management and control system provided in one embodiment of this application. Figure 1 As shown, the system as a whole can adopt a layered architecture of cloud, edge, and terminal collaboration, comprising, from bottom to top, a data acquisition and sensing layer 110, an edge computing and collaboration layer 120, a cloud-based intelligent analysis and service layer 130, and a multimodal interactive application layer 140. The data acquisition and sensing layer 110 consists of various IoT sensors, high-definition video cameras, building information model data sources, and geographic information system data interfaces deployed inside and outside the building, responsible for real-time or periodic collection of multi-source heterogeneous data on the building's structure, environment, equipment operation, and personnel activities. The edge computing and collaboration layer 120 consists of edge servers or intelligent gateways deployed locally in the building, responsible for preliminary processing, filtering, lightweight AI analysis of raw data, and issuing low-latency localized control commands to the building automation system. The cloud-based intelligent analysis and service layer 130 includes a data fusion engine, a domain-specific large-scale model training and inference platform, a dynamic knowledge graph construction and management center, a multi-agent collaborative scheduling center, and a business model library. It receives data and events uploaded from the edge computing and collaboration layer 120, performing in-depth analysis, model training, global optimization, and complex decision generation. The multimodal interactive application layer 140 provides various interactive terminals for different roles such as administrators, maintenance personnel, and ordinary users, including a web-based management backend, mobile applications, mini-programs, and voice assistants, enabling natural language dialogue with the system, 3D scene roaming, data visualization, and task management.

[0031] Please see Figure 2 , Figure 2 This is a flowchart illustrating a building intelligent management and control method according to an embodiment of this application. This building intelligent management and control method is applied to the system architecture described in the above embodiment. Figure 2 As shown, intelligent building management and control methods include: Step S210: Obtain the first multi-source heterogeneous data, which includes building information model data, geographic information system data, Internet of Things sensor data, and video surveillance data. Specifically, in practical engineering applications, Building Information Modeling (BIM) data can come from the design model database during the architectural design phase and the as-built model database during the completion phase. It provides precise component-level geometric and attribute information of the building's interior, such as the three-dimensional shape, size, material, and installation location of walls, beams, columns, doors, windows, electromechanical pipelines, and equipment on each floor. Geographic Information System (GIS) data can be obtained through GIS service interfaces, which provides macro-geographic environment and topographic information of the building's exterior, such as the topographic elevation of the site where the building is located and the spatial distribution of surrounding roads and municipal facilities. Internet of Things (IoT) sensor data is generated by real-time monitoring of sensors deployed throughout the building, including dynamic time-series information such as temperature, humidity, energy consumption, and equipment operating status. Video surveillance data is generated by cameras in the video surveillance system, providing intuitive visual contextual information for various areas of the building.

[0032] Step S220: The first multi-source heterogeneous data is fused based on a unified coordinate system, and a three-dimensional digital base plate of the city information model is constructed based on the fused data. Specifically, based on a unified spatial coordinate benchmark, the building information model (BIM) data undergoes lightweight processing and is precisely spatially matched and fused with terrain data provided by the geographic information system (GIS) to construct a high-precision 3D digital baseboard covering both the interior and exterior of buildings, both above and below ground. Building upon this, time-series data collected by IoT sensors, event data generated from structured video surveillance data, and business data such as work orders and asset information are further linked and bound to corresponding spatial entities in the BIM 3D digital baseboard. For example, the real-time reading of a temperature sensor is bound to the spatial entity node of the room where it is installed, and the operating parameters of an air conditioning unit are bound to the corresponding equipment component node of that unit in the 3D digital baseboard.

[0033] Step S230: Obtain the user's natural language instructions, perform semantic understanding and intent recognition on the natural language instructions, and generate structured task instructions; Specifically, when a user inputs unstructured natural language commands via voice, text, or other means, such as "view the current temperature and energy consumption curve of the conference room in the East Wing on the third floor," "book a conference room for 9:00 AM next Monday and turn on the air conditioning in advance," or "which device has had the most failures in the last three months," the command is sent to a domain-specific large-scale model deployed in the cloud-based intelligent analysis and service layer. This domain-specific large-scale model performs deep semantic analysis on the commands, accurately identifying whether the user's intent falls under the categories of query, control, booking, or analysis. The intent is then transformed into structured, machine-executable task instructions, which specify the task type, target spatial location, involved equipment, time parameters, and expected feedback method.

[0034] Step S240: The structured task instructions are decomposed and scheduled through the multi-agent collaborative engine to control the corresponding business agent to perform at least one operation related to building operation and management. Specifically, in the actual operation of smart buildings, user-issued instructions often involve the linkage of multiple underlying subsystems. For example, the instruction "prepare a meeting room" actually includes multiple sub-tasks such as environmental equipment adjustment, reservation status locking, and notification information dissemination, involving multiple subsystems such as air conditioning, lighting, and access control. After receiving the structured task instruction, the multi-agent collaborative engine decouples the instruction into multiple clearly defined sub-tasks and distributes each sub-task to a business intelligent agent with specific professional capabilities. Each business intelligent agent executes the sub-task in parallel after receiving it, driving the underlying equipment to complete the corresponding actions. Operations related to building operation and management can include at least one of equipment control, energy consumption management, intelligent operation and maintenance, or security management.

[0035] Step S250: Map the execution results of the business intelligence agent and the building operation status to the corresponding spatial locations in the three-dimensional digital base of the city information model, and then visualize them.

[0036] Specifically, the system renders the execution results of underlying equipment status changes, energy consumption data fluctuations, security event locations, and maintenance work order progress, along with building operation status information, in real time using intuitive visual elements on the corresponding 3D coordinates of the city information model's 3D digital base. For example, the status change of an malfunctioning air conditioning unit is presented as a color change on the city information model's 3D digital base, and the energy consumption distribution of a floor is overlaid on the floor's 3D space in the form of a heat map. Users can gain a comprehensive overview of the entire situation in a unified and intuitive 3D interface.

[0037] This application acquires multi-source heterogeneous data, including building information modeling data, geographic information system data, IoT sensor data, and video surveillance data, and fuses them based on a unified coordinate system to construct a three-dimensional digital base for urban information models. This breaks down the data silos caused by the independent construction of each subsystem in traditional building management systems and achieves precise alignment of sensing data in physical space.

[0038] Furthermore, this application performs semantic understanding and intent recognition on natural language instructions to generate structured task instructions, thereby endowing the system with deep cognitive reasoning capabilities, enabling it to accurately parse human natural language and complex business logic, and reducing the barriers to human-computer interaction.

[0039] Meanwhile, this application uses a multi-agent collaborative engine to decompose and schedule structured task instructions, and controls the corresponding business agents to perform related operations. This can effectively handle complex cross-system linkage tasks, achieve efficient decoupling and parallel processing of tasks, ensure an automated closed loop from perception to control, and improve the intelligence level and response efficiency of building operation and management.

[0040] Furthermore, this application maps the execution results of the business intelligence agent and the building's operating status to the corresponding spatial location in the three-dimensional digital base of the city information model and displays it visually, enabling managers to intuitively obtain the overall dynamic operating status. This not only improves the efficiency of situational awareness but also further improves the closed-loop link of perception, cognition, control, and display, thereby significantly enhancing the intelligent fusion of building lifecycle data and proactive decision-making capabilities.

[0041] In one embodiment, after acquiring the first multi-source heterogeneous data inside and outside the building, the method further includes: on a cloud platform, training an equipment fault prediction model and an energy consumption optimization model based on the building's historical operating data, and distributing the trained equipment fault prediction model and energy consumption optimization model to edge computing nodes; on the edge computing nodes, performing equipment health assessments and energy-saving strategy adjustments based on the building's real-time operating data, the equipment fault prediction model, and the energy consumption optimization model, respectively, and reporting the equipment health assessment results and energy-saving strategy adjustment results to the cloud platform.

[0042] Specifically, on the cloud platform side of the cloud-based intelligent analysis and service layer, due to the large volume and long period of historical building operation data, the cloud platform utilizes its abundant computing resources to deeply mine and analyze the historical operation data, training equipment failure prediction models and energy consumption optimization models. The equipment failure prediction model is used to predict the probability of equipment failure based on the historical operating parameter variation patterns, while the energy consumption optimization model is used to calculate energy-saving operation strategies based on historical energy consumption data and operating conditions. After training, the cloud platform distributes these two types of models to the edge computing and collaboration layer edge computing nodes deployed locally on the building.

[0043] On the edge computing node side, the edge computing node performs real-time analysis of video surveillance data, such as real-time artificial intelligence analysis of video streams to extract structured event information such as personnel intrusion, area intrusion, flames and smoke; at the same time, the edge computing node filters and compresses IoT sensor data, such as denoising, packaging and initially filtering high-frequency sampled time-series data such as temperature, humidity and energy consumption, and then only uploads the processed data or abnormal events to the cloud platform.

[0044] Furthermore, edge computing nodes perform low-latency control for equipment health assessments and energy-saving strategy adjustments based on real-time building operation data. Because edge computing nodes are deployed locally within the building, close to the data source and control terminal, they can directly access equipment fault prediction models and energy consumption optimization models sent from the cloud. This allows for rapid responses to emergencies or real-time environmental changes, directly issuing control commands to the building automation system, such as localized low-latency control of lighting, air conditioning, and access control equipment. After execution, the edge computing nodes report the control results to the cloud platform, forming a closed loop of two-way cloud-edge collaboration.

[0045] Through the above approach, the cloud is responsible for model training and global optimization based on historical trend data, while the edge is responsible for lightweight inference and rapid control based on real-time data. This establishes a two-way data flow mechanism for model distribution and anomaly reporting, which not only ensures low latency in the control of building-level equipment, but also reduces cloud communication bandwidth and concurrent processing pressure, thereby jointly improving the overall operational efficiency of the building intelligent management and control system.

[0046] Please see Figure 3 , Figure 3 This is a flowchart illustrating how to perform semantic understanding and intent recognition on natural language instructions to generate structured task instructions, according to one embodiment of this application. Figure 3 In this process, semantic understanding and intent recognition are performed on natural language instructions to generate structured task instructions, including: Step S310: Obtain the architectural professional corpus and use the architectural professional corpus to incrementally train the pre-trained large model to obtain the domain-specific large model. Specifically, a construction-related corpus can include various professional texts such as building design codes, equipment operation and maintenance manuals, operation and maintenance records, and industry standards. It can also cover corpora from multiple related fields, including construction, energy management, and equipment management. To ensure the quality and coverage of the corpus, data cleaning, terminology alignment, and domain annotation are required during its construction to create a structured training dataset.

[0047] Pre-trained large-scale models can employ general-purpose language models with the Transformer architecture (such as LLaMA, GPT series, or domestically developed open-source basic models). These models already possess strong language understanding and generation capabilities on massive general-purpose corpora. However, general-purpose models often lack sufficient understanding of specialized terminology, contextual logic, and implicit constraints in architectural scenarios. Therefore, incremental training (or domain-adaptive fine-tuning) is employed, using an architectural corpus as training data to further pre-train or fine-tune the pre-trained large-scale model. This allows the model to master architectural domain-specific vocabulary, equipment names, spatial relationship expression habits, and business logic, thereby obtaining a domain-specific large-scale model. Specifically, self-supervised tasks such as Masked Language Modeling (MLM) and Next Sentence Prediction (NSP) can be used, combined with domain-specific entity recognition and relation extraction auxiliary loss functions, to shift the model parameters towards the architectural domain space, thus obtaining a domain-specific large-scale model. After incremental training, the model can accurately distinguish professional components such as "shear wall", "frame column" and "post-cast strip", understand the constraint relationship between parameters such as "anchorage length", "concrete grade" and "seismic grade", and can parse information such as construction sequence, safety inspection and acceptance conditions implied in natural language.

[0048] Step S320: Parse natural language instructions based on the domain-specific large model and the pre-built dynamic knowledge graph of the construction domain to identify user intent; In this context, a dynamic knowledge graph refers to a graph-structured knowledge base built using entities in the construction field as nodes and semantic relationships between entities as edges. This knowledge graph can be incrementally updated based on data changes; for example, when construction specifications are changed or on-site measurement data is transmitted back, the attribute values ​​and relationship weights in the graph can be dynamically adjusted.

[0049] During the parsing process, when a user inputs a natural language command, the initial semantic parsing is first performed by a domain-specific large model deployed in the cloud. Key entities and intent frameworks are extracted from the user's command, and these are used as indexes to trigger retrieval enhancement of the dynamic knowledge graph (RAG) to obtain the latest engineering parameters, spatial coordinates and historical records related to the current command. Finally, the command is deeply semantically disambiguated and structured intent is output.

[0050] Step S330: Generate structured task instructions based on user intent. The structured task instructions include task type, target device, target spatial location, execution parameters, and feedback method.

[0051] After the user's intent is identified, it is converted into a structured task instruction that can be directly parsed and executed by downstream assembly and commissioning units, equipment control systems, or construction management platforms.

[0052] Specifically, the Task Type field identifies the macro-level category of the instruction, such as "Assembly Operation," "Lifting Operation," "Measurement Operation," "Calibration Operation," or "Quality Inspection Operation." This field determines the main logical branch of the subsequent processing flow. The Target Equipment field specifies one or more equipment identifiers to perform the task, and may include a list of alternative equipment or redundant configurations to improve fault tolerance. The Target Spatial Location indicates the precise 3D pose and area of ​​action defined in a unified coordinate system (i.e., CIM 3D digital baseboard) that the target equipment ultimately needs to reach or perform operations on. The Execution Parameters field dynamically organizes a specific set of parameters according to the task type. For assembly operations, this includes the desired pose (position coordinates, Euler angles), desired contact force threshold, feed rate, acceleration limit, force control compliance coefficient, etc.; for lifting operations, this includes the lifting weight, lifting point coordinates, path planning method, wind speed limit, sway suppression parameters, etc.; for measurement operations, this includes the measurement point number, measurement accuracy, sampling frequency, and allowable error range, etc. Execution parameters not only derive from explicit values ​​in user instructions but also incorporate default values ​​and constraints from dynamic knowledge graph completion (e.g., automatically completing standard installation gaps based on component models). The logical consistency between parameters is verified through a domain-specific large model. The feedback method field specifies the information feedback mechanism during and after execution, including the feedback data type (e.g., real-time force data, pose data, image, or point cloud data), feedback frequency (e.g., 100Hz, 1Hz, or event-triggered), anomaly reporting conditions (e.g., deviation exceeding limits, force overload, communication interruption), and feedback target address (e.g., local console, cloud monitoring platform, or mobile terminal). Furthermore, structured instructions can include timestamps, priorities, execution validity periods, and dependencies (e.g., preceding task IDs) to facilitate scheduling and conflict detection by the task scheduling system.

[0053] This application significantly improves the accuracy and robustness of natural language instruction understanding in architectural scenarios by integrating a domain-specific large-scale model and a dynamic knowledge graph as a dual parsing mechanism. On one hand, the large-scale model handles irregular language expressions, flexibly adapting to different wording habits and colloquial descriptions; on the other hand, the knowledge graph provides structured domain knowledge and real-time updated engineering data, ensuring that the parsing results conform to actual specifications and site conditions. The resulting structured task instructions not only fully preserve user intent but also supplement implicit engineering parameters and constraints, laying a solid foundation for subsequent precise control and adaptive adjustment, effectively reducing the error rate of manual instruction conversion, and improving the level of intelligence in construction and assembly operations.

[0054] Please see Figure 4 , Figure 4 This is a flowchart illustrating the construction of a dynamic knowledge graph in the field of architecture, according to an embodiment of this application. Figure 4Methods for constructing dynamic knowledge graphs in the field of architecture include: Step S410: Obtain the second multi-source heterogeneous data; wherein, the second multi-source heterogeneous data includes: building information model data, Internet of Things system data, operation and maintenance work order record data, and design specification document data; Specifically, maintenance work order records cover structured or semi-structured work orders generated in business processes such as property management, equipment inspection, repair and maintenance, and cleaning and maintenance. The content includes fault descriptions, maintenance actions, personnel involved, working hours, and spare parts replacement information. This type of data is usually stored in relational databases or Excel spreadsheets and is an important basis for extracting entity information such as fault events, maintenance actions, and personnel attributes.

[0055] Design specification documents include national standards, industry standards, enterprise-level construction process standards, safety operating procedures, and design specification documents. This type of data is mainly unstructured or semi-structured text, containing a large number of operational constraints (such as allowable deviation ranges, material strength requirements, and construction sequence), parameter thresholds, and equipment selection rules. It is the main source of constraining knowledge and reasoning rules in the knowledge graph.

[0056] When acquiring the above data, different extraction interfaces need to be designed according to the characteristics of each data source: for BIM, the component tree and attribute table can be extracted through the IFC parser or API interface; for IoT data, real-time subscription can be performed through message queues (such as MQTT) or time-series database interfaces; for maintenance work orders, batch extraction can be performed through database connections or ETL tools; for specification documents, OCR combined with natural language processing technology is used for content extraction and structured transformation. After acquisition, all data undergoes preprocessing steps such as deduplication, timestamp alignment, coordinate system unification, and encoding mapping to provide high-quality input for subsequent entity extraction.

[0057] Step S420: Extract entities from the building domain from the second multi-source heterogeneous data. The entities include at least physical components, spatial units, equipment operating status, event records, fault descriptions, maintenance actions, personnel attributes, equipment parameters, and operating constraints. Specifically, domain-specific Named Entity Recognition (NER) models and rule-matching methods can be used to extract various entities from the aforementioned second multi-source heterogeneous data, and then perform type labeling and attribute filling. The specific entity categories and their extraction methods are as follows: Physical components refer to actual physical objects in a building, such as columns, beams, walls, slabs, doors and windows, pipes, cables, and equipment units. Component instances and their static attributes, such as geometric dimensions, materials, strength grades, and production batches, are extracted from BIM data. Simultaneously, the current status of the components (intact, damaged, under repair, etc.) is extracted from maintenance records.

[0058] A spatial unit refers to an area within a building that has spatial division significance, such as floors, rooms, fire compartments, shafts, work surfaces, and construction sections. Spatial unit information mainly comes from BIM, supplemented by construction layout drawings, and its attributes include area, volume, elevation, and purpose.

[0059] Equipment operating status refers to the real-time operating parameters of various electromechanical equipment (such as tower cranes, construction elevators, pumping equipment, ventilation fans, distribution boxes, etc.), including current, voltage, speed, load, temperature, running time, and switch status. These entities are driven by IoT data streams, and their attribute values ​​are dynamically updated over time, carrying timestamps and confidence levels.

[0060] Event logs refer to recordable events that occur during building operation and maintenance or construction, such as "concrete pouring completed," "tower crane slewing over-limit alarm," and "materials arriving and passing inspection." Event entities include attributes such as event type, occurrence time, trigger source, associated components or equipment, and processing status, and are mainly extracted from work order records and IoT alarm logs.

[0061] A fault description refers to an abnormal phenomenon that occurs in equipment or components, such as "pipeline leakage," "motor overheating," or "weld cracking." The fault description entity is extracted from the fault description field of the work order, inspection records, and sensor anomaly data, and includes attributes such as fault code, severity level, location of occurrence, and possible causes.

[0062] Maintenance actions refer to human operations performed on building components or equipment, such as "tightening bolts," "replacing seals," "calibrating sensors," and "spraying fire-retardant coating." Maintenance action entities are extracted from the maintenance measures field of the work order and associated with the tools used, spare parts, time spent, and personnel performing the actions.

[0063] Personnel attributes refer to the information of personnel involved in construction or operation and maintenance work, including name, job type, skill level, work team, qualification certificate (such as welder's certificate, tower crane operator's certificate), and current task assignment.

[0064] Equipment parameters refer to the inherent technical specifications and performance indicators of the equipment, such as rated lifting capacity, rated power, accuracy class, stroke range, and service life. These are usually extracted from the equipment nameplate, manufacturer's manual, or design specification documents and do not change or change very little over time.

[0065] Operational constraints refer to the conditions and restrictions that must be followed during construction, installation, or operation and maintenance, such as "lifting is prohibited when the wind speed is greater than level 6", "concrete curing time shall not be less than 7 days", and "the bolt tightening torque range is 200 to 220 N·m". Operational constraint entities are extracted from design specifications and process procedures and include attributes such as constraint conditions, applicable objects, upper / lower thresholds, and consequences of violation.

[0066] Step S430: Define semantic relationships between entities to form a network knowledge structure representing semantic connections between entities based on semantic relationships; wherein, semantic relationships include at least spatial location relationships, topological connection relationships, control association relationships, monitoring correspondence relationships, and maintenance operation relationships; Specifically, based on the business logic and professional knowledge in the construction field, the following five types of core semantic relationships are defined: Spatial location relationships describe the relative positions of entities in physical space, including "located in" (e.g., "Tower Crane No. 2 is located on the east working face"), "adjacent to" (e.g., "The meeting room is adjacent to the corridor"), "containing" (e.g., "A floor contains rooms"), "passing through" (e.g., "A pipe passes through a wall"), and "supporting" (e.g., "A column supports a beam"). These relationships are mainly extracted from the geometric layout and spatial topology of BIM and calibrated in conjunction with the overall construction plan.

[0067] Topological connections describe the connection status of components or equipment in a structural or piping system, including "rigid connections" (such as rigid connections between beams and columns), "hinged connections," "welded connections," "flange connections," "electrical connections" (such as a distribution box connected to a lighting circuit), and "pipe connections" (such as a water supply pipe connected to a valve). These relationships are extracted from the connection logic of the design drawings and the component associations in the BIM, and are used to support path analysis and fault propagation tracing.

[0068] Control relationships describe the command and feedback relationships between the control system and the controlled equipment, including "control" (e.g., a central controller controlling elevator operation), "feedback" (e.g., sensors feeding back data to the controller), "setting" (e.g., an operator setting PID parameters), and "linkage" (e.g., a fire alarm linking to the start of a smoke exhaust fan). These relationships are extracted from the control system configuration table and logic diagram, and dynamically updated in conjunction with the control command flow of the IoT system.

[0069] Monitoring correspondences describe the measurement relationships between sensors or monitoring points and the objects being measured. These include "monitoring" (e.g., strain gauges monitoring stress in steel beams), "acquisition" (e.g., data acquisition devices collecting temperature and humidity values), "alarms" (e.g., level switches alarming for exceeding limits), and "verification" (e.g., comparing manually measured values ​​with sensor values ​​for verification). These relationships are extracted from IoT device layout diagrams and monitoring solution documents to ensure that each monitored entity can be associated with a specific physical object and attribute.

[0070] Maintenance work relationships describe the execution and dependencies between maintenance tasks and related entities, including "repair" (e.g., a repair work order for a water pump), "replacement" (e.g., replacing a filter), "inspection" (e.g., an inspection route covering a power distribution room), and "maintenance" (e.g., quarterly maintenance of a specific elevator model). These relationships are extracted from maintenance work order templates and preventive maintenance plans, and also include time frequency and responsible party associations.

[0071] Step S440: Store entities and semantic relationships in a graph database, and establish a unique identifier mapping for each entity to spatial entities in the 3D digital base of the city information model.

[0072] Graph databases are database systems specifically designed for storing and querying graph-structured data (such as Neo4j). They use nodes (entities), edges (semantic relationships), and attributes (key-value pairs) as their core data model, enabling efficient execution of complex graph traversal and path query operations. This step first imports or writes in real-time the entity nodes and their attributes, and the relationship edges and their attributes obtained in steps S420 and S430, according to the definition of a graph database.

[0073] Simultaneously, a unique identifier mapping is established between each entity in the knowledge graph and the spatial entity in the 3D digital base of the city information model. This unique identifier mapping includes spatial coordinates and a globally unique identifier for the component. In other words, each entity node in the graph database not only carries semantic information at the text level, but is also forcibly bound to its precise geometric position and unique identity code in the 3D digital base, giving each text node in the knowledge graph a definite physical coordinate anchor point in 3D space.

[0074] In one embodiment, the multi-agent collaborative engine adopts a hierarchical distributed architecture of a master intelligent agent and business intelligent agents; the multi-agent collaborative engine decomposes and schedules structured task instructions to control the corresponding business intelligent agents to execute at least one operation related to building operation and management, including: The structured task instructions are input to the master control agent; the master control agent decomposes the task based on task association information, breaking down the structured task instructions into multiple sub-tasks; the multiple sub-tasks are scheduled to the corresponding business agents, which include at least one of the following: equipment control agent, energy and carbon management agent, intelligent operation and maintenance agent, and smart security agent; the scheduled sub-tasks are executed in parallel by the multiple business agents, and the execution status is fed back to the master control agent.

[0075] Specifically, a multi-agent collaborative engine deployed in the cloud-based intelligent analysis and service layer inputs structured task instructions generated by a domain-specific large model into the master intelligent agent. The master intelligent agent acts as the scheduling hub of the entire execution system. Instead of blindly distributing tasks, it utilizes task association information in a dynamic knowledge graph to decompose tasks and locate resources. For example, by querying the knowledge graph, it obtains the specific identifiers, physical topology connections, and control logic dependencies of the target conference room's air conditioning, lighting, and access control equipment, thereby decomposing the macroscopic structured task instructions into several sub-tasks with clear boundaries and well-defined objectives. The master intelligent agent then schedules each sub-task to the corresponding business intelligent agent. Business intelligent agents are specialized execution units in specific domains, including at least one of the following: equipment control intelligent agent, energy and carbon management intelligent agent, intelligent operation and maintenance intelligent agent, and smart security intelligent agent. The system comprises three intelligent agents: a device control agent, a smart agent, and an intelligent energy management agent. The device control agent sends commands to the edge control layer to control the start / stop and parameter adjustment of equipment such as air conditioners, lighting, and curtains. The energy and carbon management agent calls the energy consumption analysis model to calculate and recommend energy-saving operation strategies. The intelligent operation and maintenance agent generates inspection work orders or early warning information based on the equipment fault prediction model. The smart security agent handles security incidents and links video surveillance and access control systems. Each business agent executes its sub-tasks in parallel without blocking each other, thus shortening the overall task execution time.

[0076] Each business intelligence agent feeds back its execution status to the master intelligence agent in real time during and after execution. The execution status may include information such as execution success, device offline, and parameter exceeding limits, thus forming a closed-loop control link. If the master intelligence agent receives failure feedback, it can replan an alternative execution plan based on a dynamic knowledge graph or trigger an alarm to prompt manual intervention.

[0077] In one embodiment, the execution results of the business intelligence agent and the building's operating status are mapped to corresponding spatial locations in the three-dimensional digital base of the city information model and then visualized, including: Map and render at least one of the equipment's operating parameters, energy consumption data, security events, and work order status to the corresponding spatial location in the 3D digital base of the city information model; represent equipment status, energy consumption level, or event level through different colors, icons, or animation codes; receive natural language interactive query commands input in the 3D digital base of the city information model, and output and display the real-time status and historical trends of the area or equipment specified by the natural language interactive query command.

[0078] Specifically, the multimodal interactive application layer maps and renders multi-dimensional dynamic business data, including equipment operating parameters, regional energy consumption data, sudden security incidents, and work order status of maintenance personnel, onto corresponding spatial locations in the 3D digital base of the city information model. For example, the real-time air supply temperature of an air conditioning unit can be directly displayed near the geometric location of that unit in the 3D model, and the work order processing progress of a certain floor can be associated and displayed in the corresponding 3D area of ​​that floor.

[0079] To enhance the intuitiveness of situational awareness, visual coding can be used to represent device status, energy consumption levels, or event levels using different colors, icons, or animations. For example, normally functioning devices can be marked in green, floor areas with excessive energy consumption can be rendered as red heatmaps on a 3D baseboard, and access control nodes experiencing security intrusion events can display alarms as flashing icons.

[0080] Please see Figure 5 , Figure 5 This is a flowchart illustrating the construction of a three-dimensional digital base plate for a city information model, according to an embodiment of this application. Figure 5 In this process, a three-dimensional digital base for the city information model is constructed, including: Step S510: Determine the time window corresponding to each data source based on the data change rate of each data source corresponding to the first multi-source heterogeneous data, and perform a weighted average of the observation data based on the time distance between the observation data and the target time within the time window to obtain the time-aligned data. Specifically, the time window length for each data source is determined based on its rate of change. For example, for rapidly changing temperature sensor data, the variance of the historical data can be calculated. A larger variance indicates a faster rate of data change, thus allocating a shorter time window to that data source to ensure the temporal resolution of rapidly changing data. For data sources with slow changes, a longer time window is allocated to reduce computational overhead while maintaining data representativeness. After determining the time window length, within that time window, the observed data are weighted and averaged according to their temporal distance from the target time. Observations closer to the target time are assigned higher weights, resulting in time-aligned data.

[0081] Step S520: Using the coordinate system of the building information model data as a reference, determine the coordinate transformation matrix of each data source through control point matching; Specifically, the coordinate transformation matrix of each data source relative to the reference coordinate system of the building information model is determined by control point matching. Control point matching can select several spatial control points with clear correspondence in the coordinate system of the building information model and the coordinate system of the data source to be transformed, respectively. The rotation and translation parameters of the transformation matrix are calculated by matching the control point pairs.

[0082] Step S530: Based on the coordinate transformation matrix, homogeneous coordinate transformation is used to convert the spatial coordinates of each data source to a unified coordinate system in order to construct a three-dimensional digital base plate for the city information model.

[0083] Specifically, a coordinate transformation matrix is ​​used as the operator for homogeneous coordinate transformation, which converts the spatial coordinates of various data sources to a unified coordinate system. The homogeneous coordinate transformation process involves combining the three components of the three-dimensional coordinates with a constant to form a four-dimensional homogeneous coordinate vector. A matrix multiplication operation is then performed between the coordinate transformation matrix (containing rotation and translation parameters) and this homogeneous coordinate vector. The first three components are extracted from the transformation result to obtain the three-dimensional coordinates under the unified coordinate system. Finally, based on data that is aligned in both time and space, a high-precision three-dimensional digital base plate for the urban information model is constructed.

[0084] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0085] Please see Figure 6 , Figure 6 This is a schematic diagram illustrating the principle of a building intelligent control device according to an embodiment of this application. Figure 6 In China, intelligent building management and control devices include: The data acquisition module 610 is used to acquire multi-source heterogeneous data inside and outside the building. The multi-source heterogeneous data includes building information model data, geographic information system data, Internet of Things sensor data and video surveillance data. The data fusion module 620 is used to fuse multi-source heterogeneous data based on a unified coordinate system to construct a three-dimensional digital base plate for the city information model. The instruction generation module 630 is used to acquire the user's natural language instructions, perform semantic understanding and intent recognition on the natural language instructions, and generate structured task instructions. The task execution control module 640 is used to decompose and schedule structured task instructions through a multi-agent collaborative engine, so as to control the corresponding business agent to execute at least one operation related to building operation and management. The mapping control module 650 is used to map the execution results of the business intelligence agent and the building operation status to the corresponding spatial positions in the three-dimensional digital base plate of the city information model and to display them visually.

[0086] It should be noted that the building intelligent management and control device provided in the above embodiments and the building intelligent management and control method provided in the above embodiments belong to the same concept. The specific ways in which each module and unit performs operations have been described in detail in the method embodiments, and will not be repeated here. In practical applications, the building intelligent management and control device provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above, and this is not a limitation.

[0087] Embodiments of this application also provide an electronic device, including: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the memory enables the building intelligent management and control method described above.

[0088] Embodiments of this application also provide one or more machine-readable media having instructions stored thereon that, when executed by one or more processors, cause the processors to perform the building intelligent management and control method described above.

[0089] Figure 7 A schematic diagram of a computer system suitable for implementing the memory of embodiments of this application is shown. It should be noted that... Figure 7 The computer system with the memory shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0090] like Figure 7 As shown, the computer system 700 includes a Central Processing Unit (CPU) 701, which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, based on a program stored in Read-Only Memory (ROM) 702 or a program loaded from storage into Random Access Memory (RAM) 703. The RAM also stores various programs and data required for system operation. The CPU 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0091] The following components are connected to I / O interface 705: an input section 706 including a keyboard, mouse, etc.; an output section 707 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to I / O interface 705 as needed. A removable medium 711, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 710 as needed so that computer programs read from it can be installed into storage section 708 as needed.

[0092] Specifically, according to embodiments of this application, the processes described above with reference to the flowchart can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the building intelligent management and control method of the foregoing embodiments. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium 711. When the computer program is executed by the central processing unit (CPU) 701, it performs various functions defined in the system of this application.

[0093] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM) 703, read-only memory (ROM) 702, erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0094] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in the flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block or combination of blocks in the block diagram or flowchart may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0095] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0096] Another aspect of this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a computer's processor, causes the computer to perform the aforementioned building intelligent management and control method. This computer-readable storage medium may be included in the memory described in the above embodiments, or it may exist independently and not incorporated into that memory.

[0097] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the building intelligent management and control methods provided in the various embodiments described above.

[0098] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A building intelligent management and control method, characterized in that, include: Acquire first multi-source heterogeneous data, which includes building information model data, geographic information system data, Internet of Things sensor data, and video surveillance data; The first multi-source heterogeneous data is fused based on a unified coordinate system, and a three-dimensional digital base plate for the city information model is constructed based on the fused data. Obtain the user's natural language instructions, perform semantic understanding and intent recognition on the natural language instructions, and generate structured task instructions; The structured task instructions are decomposed and scheduled by a multi-agent collaborative engine to control the corresponding business agents to perform at least one operation related to building operation and management. The execution results of the business intelligence agent and the building operation status are mapped to the corresponding spatial locations in the three-dimensional digital base of the city information model and then visualized.

2. The building intelligent management and control method according to claim 1, characterized in that, After acquiring the first multi-source heterogeneous data, the process also includes: On the cloud platform, based on the building's historical operation data, equipment fault prediction models and energy consumption optimization models are trained, and the trained equipment fault prediction models and energy consumption optimization models are distributed to edge computing nodes. The edge computing node performs equipment health assessment and energy-saving strategy adjustment based on the building's real-time operation data, the equipment fault prediction model, and the energy consumption optimization model, and reports the equipment health assessment results and energy-saving strategy adjustment results to the cloud platform.

3. The building intelligent management and control method according to claim 1, characterized in that, The natural language instructions are semantically understood and intent-recognized to generate structured task instructions, including: Obtain a corpus of architectural terminology and use the corpus to incrementally train a pre-trained large model to obtain a domain-specific large model. The natural language instructions are parsed based on the domain-specific large model and the pre-built dynamic knowledge graph of the construction domain to identify user intent; A structured task instruction is generated based on the user's intent. The structured task instruction includes the task type, target device, target spatial location, execution parameters, and feedback method.

4. The building intelligent management and control method according to claim 3, characterized in that, Methods for constructing dynamic knowledge graphs in the field of architecture include: Acquire second multi-source heterogeneous data; Entities in the building domain are extracted from the second multi-source heterogeneous data. The entities include at least physical components, spatial units, equipment operating status, event records, fault descriptions, maintenance actions, personnel attributes, equipment parameters, and operational constraints. Define semantic relationships between the entities to form a network knowledge structure representing semantic connections between entities; wherein, the semantic relationships include at least spatial location relationships, topological connection relationships, control association relationships, monitoring correspondence relationships, and maintenance operation relationships; The entities and semantic relationships are stored in a graph database, and a unique identifier mapping is established between each entity and the spatial entity in the three-dimensional digital base of the city information model.

5. The building intelligent management and control method according to any one of claims 1 to 4, characterized in that, The multi-agent collaborative engine adopts a hierarchical distributed architecture of master intelligent agent and business intelligent agent; the multi-agent collaborative engine decomposes and schedules the structured task instructions to control the corresponding business intelligent agent to execute at least one operation related to building operation management, including: The structured task instructions are input into the main control agent; The main control agent decomposes the task based on task association information, breaking down the structured task instructions into multiple sub-tasks. The multiple subtasks are scheduled to the corresponding business intelligent agents, and the business intelligent agents include at least one of the following: equipment control intelligent agent, energy and carbon management intelligent agent, intelligent operation and maintenance intelligent agent, and smart security intelligent agent; The scheduled subtasks are executed in parallel by the multiple business intelligence agents, and the execution status is fed back to the master control intelligence agent.

6. The building intelligent management and control method according to claim 1, characterized in that, The execution results of the business intelligence agent and the building operation status are mapped to the corresponding spatial locations in the three-dimensional digital base of the city information model and then visualized, including: At least one of the equipment's operating parameters, energy consumption data, security events, and work order status is mapped and rendered to the corresponding spatial location in the three-dimensional digital base plate of the city information model. Different colors, icons, or animated codes are used to represent device status, energy consumption level, or event level; It receives natural language interactive query commands input into the three-dimensional digital base of the city information model, and outputs and displays the real-time status and historical trends of the area or device specified by the natural language interactive query command.

7. The building intelligent management and control method according to claim 1, characterized in that, The first multi-source heterogeneous data is fused based on a unified coordinate system, and a three-dimensional digital base plate for the city information model is constructed based on the fused data, including: Based on the data change rate of each data source corresponding to the first multi-source heterogeneous data, the time window corresponding to the data source is determined, and within the time window, the observed data is weighted and averaged based on the time distance between the observed data and the target time to obtain the time-aligned data. Using the coordinate system of the building information model data as a reference, the coordinate transformation matrix of each data source is determined by control point matching; Based on the coordinate transformation matrix, homogeneous coordinate transformation is used to convert the spatial coordinates of each data source to the unified coordinate system in order to construct the three-dimensional digital base of the city information model.

8. A building intelligent control device, characterized in that, include: The data acquisition module is used to acquire first multi-source heterogeneous data, which includes building information model data, geographic information system data, Internet of Things sensor data, and video surveillance data. The data fusion module is used to fuse the multi-source heterogeneous data based on a unified coordinate system to construct a three-dimensional digital base plate for the city information model. The instruction generation module is used to acquire the user's natural language instructions, perform semantic understanding and intent recognition on the natural language instructions, and generate structured task instructions. The task execution control module is used to decompose and schedule the structured task instructions through a multi-agent collaborative engine, so as to control the corresponding business agent to perform at least one operation related to building operation and management. The mapping control module is used to map the execution results of the business intelligence agent and the building operation status to the corresponding spatial positions in the three-dimensional digital base of the city information model and to display them visually.

9. An electronic device, characterized in that, include: One or more processors; and A memory for storing one or more programs, which, when executed by one or more processors, cause the processors to implement the building intelligent management and control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by the computer's processor, causes the computer to perform the building intelligent management and control method as described in any one of claims 1 to 7.

Images