Natural language-based two-dimensional configuration diagram and three-dimensional digital twin synchronous generation method

Abstract

The application discloses a natural language-based two-dimensional configuration graph and three-dimensional digital twin synchronous generation method, comprising the following steps: inputting a natural language instruction, a pre-constructed unified intermediate representation and a current unified intermediate representation into a large language model output logical topology as a prompt; performing structure checking and point position resolvable checking on the logical topology according to a pre-defined structure constraint to obtain a logical topology that passes the checking; a deterministic layout engine calculates geometric coordinates according to the logical topology that passes the checking and a pre-defined fixed layout rule to generate a complete unified intermediate representation; and inputting the complete unified intermediate representation into a two-dimensional vector rendering engine and a three-dimensional rendering engine respectively to synchronously generate a two-dimensional configuration graph and a three-dimensional digital twin scene. The application realizes synchronous generation of a two-dimensional configuration graph and a three-dimensional digital twin, replaces a large language model with a deterministic layout to avoid confusion, and introduces a checking closed loop to ensure that the configuration can be rendered.

Description

Technical Field

[0001] This application relates to the fields of building energy management and building automation technology, specifically to a method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language. Background Technology

[0002] HVAC / refrigeration systems in public buildings such as stadiums, hotels, office buildings, and hospitals are generally equipped with monitoring systems, requiring the creation of two types of visualization interfaces: two-dimensional configuration diagrams and three-dimensional digital twin scenes. Two-dimensional configuration diagrams (e.g., process flow diagrams / P&ID / SCADA mimic) use graphic elements (e.g., units, valves, pumps, headers, sensors, etc.) and pipelines to represent the system structure and water / medium flow direction; three-dimensional digital twin scenes use three-dimensional models to represent the spatial location of equipment and the real-time operating status of the system.

[0003] Patent document CN104764173A discloses a monitoring method for a heating, ventilation, and air conditioning (HVAC) system, comprising: establishing a 3D HVAC system model based on Building Information Modeling (BIM); finding the actual operating parameters related to each pipe, regulating valve, and equipment in the HVAC system in the 3D HVAC system model and their display methods based on the identifiers of each pipe, regulating valve, and equipment in the HVAC system; displaying the actual operating parameters of the HVAC system at the location of the air conditioning information according to the corresponding display method; the HVAC system model has the actual shape and spatial distribution of the pipes and equipment in the actual HVAC system.

[0004] Patent document CN119902834A discloses a method, device, equipment, and storage medium for visualizing the operation of equipment in a park. The method includes: acquiring the identification information and corresponding location information of each device in the target park, and collecting basic, static, and dynamic information of each device; obtaining the corresponding components from a pre-built component library based on the identification information of each device; deploying the corresponding components of each device on a plan view of the target park by overlaying, integrating, and / or logically connecting them according to the location information of each device, forming a configuration diagram of the target park; binding the basic, static, and dynamic information of each device to the corresponding components in the configuration diagram; and displaying the configuration diagram of the target park to the user in real time.

[0005] However, the existing technology has the following technical defects:

[0006] 1. Directly using large language models to automatically generate is unreliable: Even if a large language model is introduced to automatically generate a visual configuration, if the model is allowed to directly output primitives with geometric coordinates, problems such as primitive overlap, disordered connections, and unusable layouts will occur because the large language model is not good at reasoning about precise geometric coordinates.

[0007] 2. Two-dimensional and three-dimensional interfaces are created independently and are prone to inconsistency: The two interfaces are usually created separately with different tools and different data, resulting in inconsistencies in structure, position, and layout. They also require double maintenance, doubling the maintenance cost.

[0008] 3. Lack of a closed loop from natural language to "verifiable and renderable" configuration: There is a lack of verification between natural language and machine-renderable configuration. The model may output configurations with illegal structures or unresolved point bindings, which cannot directly drive rendering.

[0009] 4. Difficult to replicate quickly, long cycle, and high cost: Two-dimensional configuration diagrams require manual dragging and dropping of points and connecting lines in dedicated configuration software; three-dimensional twins require professional three-dimensional modeling, and must be redone for each new building, making it difficult to replicate quickly, and the cycle is long and the cost is high.

[0010] 5. Risks associated with connecting to existing production control systems: When connecting a running production-level control system to a unified platform, if there are write paths, it may affect the production system. Summary of the Invention

[0011] To address this, this application provides a method for the simultaneous generation of two-dimensional configuration diagrams and three-dimensional digital twins based on natural language, in order to solve the problems of existing technologies that make it difficult to generate two-dimensional configuration diagrams and three-dimensional digital twins simultaneously, that directly generating geometric coordinates from large language models can easily lead to layout errors and rendering failures, and that there is a lack of verification steps.

[0012] To achieve the above objectives, this application provides the following technical solution:

[0013] Firstly, a method for simultaneously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language includes:

[0014] Step 1: Input the natural language instruction, the pre-built unified intermediate representation and the current unified intermediate representation as prompts into the large language model, and constrain the output of the large language model to contain only the device topology relationship and not the geometric coordinates;

[0015] Step 2: Perform structural verification and point resolution verification on the logical topology according to the predefined structural constraints. If the verification fails, the error information is fed back into the large language model for regeneration until the verification passes, and a verified logical topology is obtained.

[0016] Step 3: The deterministic layout engine calculates the geometric coordinates of each node and pipe segment based on the validated logical topology and predefined fixed layout rules, and generates a complete unified intermediate representation containing the geometric coordinates.

[0017] Step 4: Input the complete unified intermediate representation into the two-dimensional vector rendering engine and the three-dimensional rendering engine respectively, and generate two-dimensional configuration diagrams and three-dimensional digital twin scenes simultaneously.

[0018] Preferably, the method further includes: collecting real-time operating data of the equipment, and mapping the real-time data to the state presentation of the corresponding elements in the two-dimensional configuration diagram and the three-dimensional digital twin scene based on the binding information of each node and pipe segment in the complete unified intermediate representation.

[0019] Preferably, the status presentation includes: valve open or closed status, unit operation, stop or fault status, sensor readings, and pipeline flow direction animation derived from topological conditions.

[0020] Preferably, the derivation method of the pipe segment flow direction animation is as follows: determine whether the pipe segment is in a flowing state based on the topological conditions. The topological conditions are that when at least one water supply valve is in the fully open state, the corresponding water supply pipe segment is in a flowing state.

[0021] As a preferred option, it also includes: taking the current complete unified intermediate representation and the new natural language instructions as input for the next round, and repeating steps 1 to 4 to incrementally edit the generated two-dimensional configuration diagram and three-dimensional digital twin scene.

[0022] Preferably, in step 1, the pre-constructed unified intermediate representation includes: data structure constraints, a primitive library directory, and a point binding naming convention; the primitive library directory is indexed by primitive category kind, and each kind corresponds to a two-dimensional vector primitive mapping and a three-dimensional parametric component mapping; the nodes in the complete unified intermediate representation include primitive category kind, display label label, and status binding bind, and the pipe segment includes medium type medium and flow direction binding bind.flow.

[0023] Preferably, in step 1, the device topology relationship includes: a device list, the connection relationship between devices, and the role of each device.

[0024] Preferably, in step 3, the predefined fixed layout rules include: arranging from one side to the other according to the topological order from cold source to end, arranging by source and layer, arranging the supply and return water main pipes in the center, and arranging similar elements in rows.

[0025] Preferably, the large language model is accessed through a unified interface and can be switched to different large language models according to the deployment mode. The deployment modes include: private deployment mode and cloud service deployment mode; the large language model under the private deployment mode includes an open source model that can be self-hosted on the customer's intranet.

[0026] Secondly, a device for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language includes:

[0027] The dialogue and topology generation module is used to input natural language instructions, pre-built unified intermediate representations and the current unified intermediate representations as prompts into the large language model, and to constrain the output of the large language model to contain only device topology relationships and not geometric coordinates.

[0028] The structure and point verification module is used to perform structure verification and point parsing verification on the logical topology according to the predefined structural constraints. If the verification fails, the error information is fed back to the large language model for regeneration until the verification passes, and the verified logical topology is obtained.

[0029] The deterministic layout module is used by the deterministic layout engine to calculate the geometric coordinates of each node and pipe segment based on the validated logical topology and predefined fixed layout rules, and generate a complete unified intermediate representation containing the geometric coordinates.

[0030] Two-dimensional and three-dimensional rendering modules are used to input the complete unified intermediate representation into the two-dimensional vector rendering engine and the three-dimensional rendering engine respectively, and simultaneously generate two-dimensional configuration diagrams and three-dimensional digital twin scenes.

[0031] Compared with the prior art, this application has at least the following beneficial effects:

[0032] 1. This application provides a method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language, including: inputting natural language instructions, a pre-constructed unified intermediate representation, and the current unified intermediate representation as prompts into a large language model, and outputting a logical topology; performing structural verification and point resolution verification on the logical topology according to predefined structural constraints; if the verification fails, the error information is fed back into the large language model for regeneration until the verification passes, resulting in a verified logical topology; a deterministic layout engine calculates the geometric coordinates for each node and pipe segment based on the verified logical topology and predefined fixed layout rules, generating a complete unified intermediate representation containing the geometric coordinates; and inputting the complete unified intermediate representation into a two-dimensional vector rendering engine and a three-dimensional rendering engine respectively, synchronously generating two-dimensional configuration diagrams and three-dimensional digital twin scenes. This application achieves synchronous generation of two-dimensional configuration diagrams and three-dimensional digital twins, and uses deterministic layout to replace the large language model to directly generate coordinates to avoid errors, while introducing a verification closed loop to ensure that the configuration is renderable.

[0033] 2. By using the current unified intermediate representation and new instructions as input for the next round and repeatedly executing the generation process, dialogic incremental editing of the generated 2D and 3D scenes is realized, thereby avoiding starting from scratch, reducing modification costs, and maintaining synchronous updates and context continuity at both ends. Attached Figure Description

[0034] To more intuitively illustrate the prior art and this application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary drawings, those skilled in the art are able to easily make conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0035] Figure 1 This is a flowchart of a method for synchronously generating a two-dimensional configuration diagram and a three-dimensional digital twin based on natural language, provided in Embodiment 1 of this application. Detailed Implementation

[0036] The present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0037] In the description of this application: unless otherwise stated, "a plurality of" means two or more. The terms "first," "second," "third," etc., in this application are intended to distinguish the objects referred to and do not have any special meaning in terms of technical connotation (e.g., they should not be construed as an emphasis on importance or order). Expressions such as "including," "comprising," and "having" also mean "not limited to" (certain units, components, materials, steps, etc.).

[0038] The terms used in this application, such as "upper," "lower," "left," "right," and "middle," are generally used to indicate the general relative positional relationship for the purpose of intuitive understanding by referring to the accompanying drawings, and are not absolute limitations on the positional relationship in the actual product.

[0039] Example 1

[0040] This embodiment provides a method for the simultaneous generation of two-dimensional configuration diagrams and three-dimensional digital twins based on natural language. The method pre-constructs a unified intermediate representation (IR) that is independent of the rendering method, and achieves reliable generation from natural language to visualization through a method chain of "large language model generating only logical topology, deterministic engine calculating geometric layout, single IR driving rendering from both ends, real-time point state mapping, and multi-turn dialogue incremental editing".

[0041] Please see Figure 1 This embodiment provides a method for simultaneously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language, including:

[0042] S1, Dialogue Understanding and Logical Topology Generation: Natural language instructions, pre-built unified intermediate representations, and the current unified intermediate representations are used as prompts input to the large language model, and the output of the large language model is constrained to contain only device topology relationships and not geometric coordinates.

[0043] Specifically, the pre-built unified intermediate representation includes: data structure constraints, a primitive library directory, and a point binding naming convention; among which, the primitive library directory is indexed by primitive category kind, and each kind corresponds to a two-dimensional vector primitive mapping and a three-dimensional parametric component mapping.

[0044] The nodes and segments of the Unified Intermediate Representation (IR) include at least:

[0045] Nodes: Each node describes a graphic element and includes at least: the element category kind (e.g., chiller unit, air-cooled module, cooling tower, air conditioning unit, fan coil unit, circulating pump, water supply pump, valve, supply / return water header and sensor, etc.), the display label, and the status binding bind (binding the graphic element's running, on / off, fault or reading status to the device point key).

[0046] Pipes: Each pipe segment must include at least: medium type (e.g., chilled water supply, chilled water return), connection relationship (which nodes / connection points are connected), and flow direction binding bind.flow (binding whether the pipe segment is flowing to a condition derived from the equipment status).

[0047] Example of IR data structure:

[0048] Node examples (field illustrations):

[0049] { id: 'PARK', kind: 'Cooling source - water-cooled unit', label: 'Park cooling source', bind: { ...}};

[0050] { id: 'CH', kind: 'Cold Source - Air-cooled Module', label: 'Air-cooled Unit', bind: { run: 'pt:CH.run', fault: 'pt:CH.fault'}};

[0051] { id: 'VP1', kind: 'valve', label: 'park valve', medium: 'chilled water supply', bind: { open: 'pt:V.p1'}};

[0052] { id: 'PUMP', kind: 'pump', label: 'circulation pump', bind: { run: 'pt:PUMP.run'}};

[0053] { id: 'SH', kind: 'mother pipe', label: 'water supply', medium: 'chilled water supply'};

[0054] { id: 'HALL', kind: 'End Load', label: 'Venue', ...};

[0055] {id: 'T1', kind: 'sensor', label: 'water supply temperature', bind: { value: 'pt:T.sup'}}.

[0056] Pipes example (field illustration):

[0057] { id: 'p.sup', medium: 'chilled water supply', connection: [water source → junction point → main water supply pipe → end], bind: { flow: 'pt:f.sup'} ;

[0058] { id: 'p.ret', medium: 'frozen water return', connection: [end → return water header → branch point → each water source], bind: { flow: 'pt:f.ret'}}.

[0059] This step provides the natural language instructions, along with the IR data structure constraints (schema), the directory of available primitive categories (primitive library directory), the point binding naming conventions, and the current IR (initially empty), to the large language model. This constrains the large language model to output the logical topology, namely the equipment list, the connection relationships between equipment, and the roles of each equipment (cold source / terminal / pipe segment, etc.), without outputting the geometric coordinates of the primitives.

[0060] S2, Structural verification and closed-loop retry: The logical topology is verified for structural integrity and parsability based on predefined structural constraints. If the verification fails, the error information is fed back into the large language model for regeneration until the verification passes and the logical topology that has passed the verification is obtained.

[0061] Specifically, this step validates the output of the large language model according to the predefined structural constraints (schema) and verifies whether the point binding keys can be resolved to real device points. If the validation fails, the error information is fed back to the large language model for regeneration until it passes, thereby ensuring that the output is valid, renderable, and the points can be resolved.

[0062] S3, Deterministic Automatic Layout: The deterministic layout engine calculates the geometric coordinates of each node and pipe segment based on the validated logical topology and predefined fixed layout rules, and generates a complete and unified intermediate representation containing the geometric coordinates;

[0063] Specifically, this step involves a deterministic layout engine calculating the geometric coordinates of each node and pipe segment based on predefined fixed layout rules (e.g., arranging elements from one side to the other in a topological order of "cold source—pipeline—terminal"; source / layer separation; centered supply and return water headers; and evenly distributed similar elements in rows). This process yields a complete IR with coordinates from the logical topology. Because the coordinates are calculated by the deterministic engine rather than provided by a large language model, element overlap and connection misalignment are avoided.

[0064] Example of deterministic layout mapping:

[0065] The layout engine maps logical topology to coordinates. For example, cold source nodes are arranged on one side of the scene, terminal loads are arranged on the other side, supply / return water headers are placed horizontally in the center, and similar terminal elements (such as fan coil units) are evenly distributed in rows along the wall; two-dimensional plane coordinates ( , The coordinates (X, Z) of the IR are converted to the reference according to a fixed ratio to ensure that the relative position of the same IR is consistent in two dimensions and three dimensions.

[0066] S4, Single IR Dual-End Rendering: Inputs the complete and unified intermediate representation into the 2D vector rendering engine and the 3D rendering engine respectively, and generates 2D configuration diagrams and 3D digital twin scenes simultaneously.

[0067] Specifically, this step uses the same IR as the sole data source, and generates two-dimensional configuration diagrams (e.g., SVG process flow diagrams) through a two-dimensional vector rendering engine, and generates three-dimensional digital twin scenes through a three-dimensional rendering engine combined with a parametric primitive library (selecting the corresponding three-dimensional components by looking up the table by kind). The two share the same IR and the same set of layout calculation results, thereby maintaining strong consistency between structure and layout, and only a single data source needs to be maintained.

[0068] This embodiment provides a method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language, which also includes real-time state mapping: collecting real-time operating data of the equipment, and mapping the real-time data to the state presentation of corresponding elements in the two-dimensional configuration diagram and three-dimensional digital twin scene according to the binding information of each node and pipe segment in the complete unified intermediate representation.

[0069] The status presentation includes: valve opening or closing status, unit operation, stop or fault status, sensor readings, and pipe flow direction animation derived from topological conditions; the derivation method of the pipe flow direction animation is: to determine whether the pipe section is in a flow state based on the topological conditions. The topological condition is that when at least one water supply valve is in the fully open state, the corresponding water supply pipe section presents a flow state.

[0070] It should be noted that during real-time status mapping, after the large language model provides a mapping suggestion of "original location → standard semantic binding key" based on the imported list of original device locations, it can be manually confirmed or corrected on the interface and written into the IR, reducing the cost of manual configuration item by item.

[0071] Real-time state mapping example:

[0072] State mapping functions map device points to bound values, for example:

[0073] When the valve status code is "fully open", the corresponding valve element is displayed in the open color, and the flow direction condition of its water supply pipe section is set.

[0074] The flow direction conditions of the pipe section are derived using topology, such as "Is the water supply pipe section flowing = (Park supply valve open or unit supply valve open) ? 1 : 0", so that the flow direction animation reflects the real cooling path, rather than simply binding a single water pump;

[0075] Unit operation / shutdown / fault are mapped to different colors or materials (it can be agreed that faults will not be represented by red, but by amber, etc.);

[0076] Sensor readings are written to the corresponding node in real time.

[0077] The two-dimensional configuration diagram and the three-dimensional digital twin are driven by the same state mapping result, ensuring consistency between the two ends.

[0078] This embodiment provides a method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language. It also includes multi-round dialogue incremental editing: the current complete unified intermediate representation and the new natural language instructions are used as the input for the next round, and steps S1 to S4 are repeated to incrementally edit the generated two-dimensional configuration diagram and three-dimensional digital twin scene.

[0079] In other words, this embodiment uses the current IR and the new natural language instructions as input for the next round. The large language model produces the incrementally modified logical topology, which is then updated and re-rendered through subsequent steps to achieve conversational iterative editing (such as "change the fan coil units to 25 per side" or "add a cooling tower").

[0080] It should be noted that during multi-round incremental editing, relatively fixed parts (IR structural constraints, primitive library catalog, point specifications) are used as cacheable prefixes, and only "current IR + new instruction" changes in each round to reduce the cost of calling.

[0081] Example of incremental editing in multi-turn dialogue:

[0082] User: "Replace the fan coil units in the venue with 25 units on each of the two side walls."

[0083] Method: The large language model generates an incremental topology (changes in the number of end primitives and grouping) under the current IR constraints → verification → layout recalculation → synchronous update of 2D and 3D to 25 units per side.

[0084] This embodiment provides a method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language. The large language model is accessed through a unified interface (e.g., a protocol compatible with standard dialogue interfaces) and can be switched to different large language models according to the deployment mode. The deployment modes include: private deployment mode and cloud service deployment mode. The large language model in the private deployment mode includes an open source model that can be self-hosted on the customer's intranet.

[0085] It should be noted that when connecting data from an existing production control system, only a one-way read-only data channel is established (e.g., read-only database account, read-only cache access), with no write path, thereby avoiding impact on the production system in a "construction-based read-only" manner.

[0086] The method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language provided in this embodiment has the following advantages:

[0087] 1. The creation of 2D configuration diagrams and 3D digital twins has been reduced from relying on drag-and-drop / modeling by professionals to "natural language dialogue + a small amount of manual confirmation", which greatly shortens the production cycle, lowers the threshold, and facilitates the rapid replication of new buildings.

[0088] 2. The large language model only generates logical topology, while the geometric coordinates are calculated by the deterministic engine. This avoids primitive overlap and connection disorder caused by the model directly generating coordinates, making the generated results reliable and repeatable.

[0089] 3. A single IR can drive both 2D and 3D simultaneously, and the layout can be calculated once and shared at both ends, so that the 2D configuration and 3D twin maintain strong consistency and the maintenance cost is halved.

[0090] 4. A closed loop of structural verification and point parsing verification ensures that the configuration generated by natural language is legal, renderable, and the points are parsable.

[0091] 5. One-way read-only access avoids impacting existing production control systems.

[0092] In summary, this embodiment achieves the synchronous generation of two-dimensional configuration diagrams and three-dimensional digital twins, and uses deterministic layout to directly generate coordinates instead of large language models to avoid errors. At the same time, a verification closed loop is introduced to ensure that the configuration can be rendered.

[0093] Example 2

[0094] This embodiment provides a device for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language, including:

[0095] The dialogue and topology generation module is used to input natural language instructions, pre-built unified intermediate representations and the current unified intermediate representations as prompts into the large language model, and to constrain the output of the large language model to contain only device topology relationships and not geometric coordinates.

[0096] The structure and point verification module is used to perform structure verification and point parsing verification on the logical topology according to the predefined structural constraints. If the verification fails, the error information is fed back to the large language model for regeneration until the verification passes, and the verified logical topology is obtained.

[0097] The deterministic layout module is used by the deterministic layout engine to calculate the geometric coordinates of each node and pipe segment based on the validated logical topology and predefined fixed layout rules, and generate a complete unified intermediate representation containing the geometric coordinates.

[0098] Two-dimensional and three-dimensional rendering modules are used to input the complete unified intermediate representation into the two-dimensional vector rendering engine and the three-dimensional rendering engine respectively, and simultaneously generate two-dimensional configuration diagrams and three-dimensional digital twin scenes.

[0099] This embodiment provides a device for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language. It further includes: a real-time data acquisition and status mapping module, an IR storage module, a point mapping human-machine verification module, a read-only access adaptation module, and a primitive library management module. Each module can be deployed as a private service (customer intranet, including a self-hosted open-source large language model) or a cloud service.

[0100] For details on the implementation of each module in a natural language-based two-dimensional configuration diagram and three-dimensional digital twin synchronous generation device, please refer to the above description of the limitations of a natural language-based two-dimensional configuration diagram and three-dimensional digital twin synchronous generation method, which will not be repeated here.

[0101] The technical features of the above embodiments can be combined in any way (as long as there is no contradiction in the combination of these technical features). For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; these embodiments not explicitly written should also be considered to be within the scope of this specification.

Claims

1. A method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language, characterized in that, include: Step 1: Input the natural language instruction, the pre-built unified intermediate representation and the current unified intermediate representation as prompts into the large language model, and constrain the output of the large language model to contain only the device topology relationship and not the geometric coordinates; Step 2: Perform structural verification and point resolution verification on the logical topology according to the predefined structural constraints. If the verification fails, the error information is fed back into the large language model for regeneration until the verification passes, and a verified logical topology is obtained. Step 3: The deterministic layout engine calculates the geometric coordinates of each node and pipe segment based on the validated logical topology and predefined fixed layout rules, and generates a complete unified intermediate representation containing the geometric coordinates. Step 4: Input the complete unified intermediate representation into the two-dimensional vector rendering engine and the three-dimensional rendering engine respectively, and generate two-dimensional configuration diagrams and three-dimensional digital twin scenes simultaneously.

2. The method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language according to claim 1, characterized in that, Also includes: The system collects real-time operating data from the equipment and maps the real-time data to the state presentation of corresponding elements in the two-dimensional configuration diagram and the three-dimensional digital twin scene based on the binding information of each node and pipe segment in the complete unified intermediate representation.

3. The method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language according to claim 2, characterized in that, The status presentation includes: valve opening or closing status, unit operation, shutdown or fault status, sensor readings, and pipeline flow direction animation derived from topological conditions.

4. The method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language according to claim 3, characterized in that, The derivation method of the pipe segment flow direction animation is as follows: determine whether the pipe segment is in a flowing state based on the topological conditions. The topological conditions are that when at least one water supply valve is in the fully open state, the corresponding water supply pipe segment is in a flowing state.

5. The method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language according to claim 1, characterized in that, Also includes: The current complete unified intermediate representation and the new natural language instructions are used as input for the next round. Steps 1 to 4 are repeated to incrementally edit the generated two-dimensional configuration diagram and three-dimensional digital twin scene.

6. The method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language according to claim 1, characterized in that, In step 1, the pre-built unified intermediate representation includes: data structure constraints, a primitive library directory, and a point binding naming convention; the primitive library directory is indexed by primitive category kind, and each kind corresponds to a two-dimensional vector primitive mapping and a three-dimensional parametric component mapping; the nodes in the complete unified intermediate representation include primitive category kind, display label label, and state binding bind, and the pipe segment includes medium type medium and flow direction binding bind.flow.

7. The method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language according to claim 1, characterized in that, In step 1, the device topology relationship includes: device list, device connection relationship and each device role.

8. The method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language according to claim 1, characterized in that, In step 3, the predefined fixed layout rules include: arranging from one side to the other according to the topological order from cold source to terminal, arranging by source and layer, arranging the supply and return water main pipes in the center, and arranging similar elements in rows.

9. The method for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language according to claim 1, characterized in that, The large language model is accessed through a unified interface and can be switched to different large language models according to the deployment mode. The deployment modes include: private deployment mode and cloud service deployment mode; the large language model under the private deployment mode includes an open source model that can be self-hosted on the customer's intranet.

10. A device for synchronously generating two-dimensional configuration diagrams and three-dimensional digital twins based on natural language, characterized in that, include: The dialogue and topology generation module is used to input natural language instructions, pre-built unified intermediate representations and the current unified intermediate representations as prompts into the large language model, and to constrain the output of the large language model to contain only device topology relationships and not geometric coordinates. The structure and point verification module is used to perform structure verification and point parsing verification on the logical topology according to the predefined structural constraints. If the verification fails, the error information is fed back to the large language model for regeneration until the verification passes, and the verified logical topology is obtained. The deterministic layout module is used by the deterministic layout engine to calculate the geometric coordinates of each node and pipe segment based on the validated logical topology and predefined fixed layout rules, and generate a complete unified intermediate representation containing the geometric coordinates. Two-dimensional and three-dimensional rendering modules are used to input the complete unified intermediate representation into the two-dimensional vector rendering engine and the three-dimensional rendering engine respectively, and simultaneously generate two-dimensional configuration diagrams and three-dimensional digital twin scenes.

Images