A high and large space building non-uniform indoor environment optimization design decision method based on a multi-modal knowledge graph enhanced base model

Abstract

The application provides a high and large space building non-uniform indoor environment optimization design decision method based on a multi-modal knowledge graph enhanced base model. In the design decision process, a green performance optimization design model suitable for the typical engineering scene of the high and large space building non-uniform indoor environment is constructed, the non-uniform indoor environment optimization design problem is understood, and an optimization design decision method is proposed, and numerical values such as building form parameters, space parameters and heating ventilation air conditioning system parameters, green performance image data of a typical section of the high and large space building and the like are output. On this basis, a construction method of an enhanced base model with knowledge representation empowerment is proposed, optimal allocation of computing power resources is realized based on an end-cloud collaborative distributed work architecture, multi-modal knowledge graphs and vector databases are cooperated to improve knowledge retrieval and question and answer efficiency, reduce computing cost and time, and improve computing precision. In combination with a digital sand table, the visualization interaction of the optimization design scheme is realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of green building performance optimization design, and particularly relates to a high and large space building non-uniform indoor environment optimization design decision method based on a multi-modal knowledge graph enhanced base model. BACKGROUND

[0002] In the face of higher requirements for building energy saving, the traditional optimization design decision method relies on performance simulation of green performance data, and has certain limitations in calculation efficiency and calculation cost, cannot simultaneously process multi-modal building green performance data, cannot effectively form multi-platform cooperation of design and simulation, cannot meet the new requirements of energy saving benefits, and cannot effectively respond to the multi-dimensional requirements of buildings.

[0003] The indoor environment of high and large space buildings is non-uniform, is affected by multi-factor coupling, air flow process is complex, environmental parameters are variable, and problems such as heat accumulation and poor air flow are prone to occur. For different functional areas in such buildings, the ventilation, heat load and personnel density are quite different, and local personnel stagnation problems are prone to occur in some functional areas, affecting the comfort of users. How to realize effective temperature and humidity, ventilation control and user comfort adjustment of high and large space buildings is the core problem of green performance optimization decision of such buildings. The traditional design decision method cannot adapt to the situation and the place, and cannot effectively respond to higher comfort and building energy saving requirements.

[0004] For the optimization design decision of non-uniform indoor environment of high and large space buildings, the current methods are mainly based on physical models and machine learning models, which have the following defects:

[0005] 1. It is difficult to process and integrate multi-source, multi-modal heterogeneous data of non-uniform indoor environment. The traditional method relies on static data and models, cannot feedback real-time data, lacks response ability to emergencies, has slow design response and decision lag, and lacks flexibility and adaptability;

[0006] 2. It is difficult to accurately simulate and analyze the green performance data of high and large space building non-uniform indoor environment affected by multi-factor coupling. The traditional method lacks knowledge reasoning ability, cannot effectively take advantage of cloud computing power, has poor learning ability for historical cases and similar schemes, is difficult to effectively propose personalized design decisions for specific scenarios, and has poor flexibility and expandability.

[0007] In recent years, multi-modal large language models (MM-LLMs) have shown great potential in multi-modal data fusion processing. Based on numerical, textual, image, video, and audio inputs, large models can understand user needs and generate decision recommendations and technical reports, but the hallucination problem of large models restricts their accuracy and reliability in engineering scenario design. Using knowledge graph enhanced mode can effectively improve this problem. Relying on the knowledge graph that fits the specific engineering scenario, the design problem is matched with the triples in the knowledge graph through retrieval enhanced generation model, and the wisdom of the large model is empowered. SUMMARY

[0008] The purpose of the present application is to solve the problems in the prior art, and a high and large space building non-uniform indoor environment optimization design decision method based on a multi-modal knowledge graph enhanced base model is proposed.

[0009] The present application is realized by the following technical solutions, and the present application proposes a high and large space building non-uniform indoor environment optimization design decision method based on a multi-modal knowledge graph enhanced base model, which comprises the following steps:

[0010] S1, acquire non-uniform indoor environment multi-modal data knowledge;

[0011] The step S1 comprises the following steps:

[0012] S11, collect non-uniform indoor environment green performance data;

[0013] S12, generate embedded representation of fusion multi-modal non-uniform indoor environment green performance information;

[0014] S13, extract non-uniform indoor environment green performance data information;

[0015] S2, fuse non-uniform indoor environment multi-modal data knowledge, and construct non-uniform indoor environment multi-modal knowledge graph and vector database;

[0016] The step S2 comprises the following steps:

[0017] S21, convert the extracted information into RDF triples;

[0018] S22, construct non-uniform indoor environment multi-modal knowledge graph based on RDF triples;

[0019] S23, determine typical non-uniform indoor environment engineering scenario elements;

[0020] S24, construct and update the corresponding multi-modal knowledge graph and vector database;

[0021] S3, strengthen the optimization of design decision problem knowledge retrieval, and build a green performance optimization design model that fits the typical engineering scenarios of non-uniform indoor environment;

[0022] The step S3 comprises the following steps:

[0023] S31, modular clustering of knowledge in the knowledge graph and obtaining a structured subgraph;

[0024] S32, understanding the design problem and improving the knowledge retrieval efficiency based on the structured subgraph;

[0025] S33, building a green performance optimization design model that fits the typical engineering scenarios of non-uniform indoor environment;

[0026] S4, end-to-cloud collaboration, flexible handling of optimization design problems, and visualization of the optimization design results in the form of a digital sand table;

[0027] The step S4 comprises the following steps:

[0028] S41, building an end-to-cloud collaboration architecture that fits various typical engineering scenarios of non-uniform indoor environment;

[0029] S42, building a digital twin model of high and large space buildings, and displaying the optimization design results through a digital sand table to realize visual interaction.

[0030] Further, in the step S13, the information extraction comprises entity extraction, relationship extraction and event extraction; the entity extraction refers to extracting specific building green performance objects or concepts from the unstructured non-uniform indoor environment green performance data set based on the BiLSTM-CRF deep learning model using the context information through the named entity recognition method; the relationship extraction refers to identifying the relationship between the non-uniform indoor environment green performance entities using the OpenIE and BERT relationship classification model or the dependency syntax analysis method; and the event extraction refers to detecting the spatiotemporal dimension parameters of multiple non-uniform indoor environment green performance entities through the ACE event extraction framework, and identifying the non-uniform indoor environment green performance event trigger words and related information based on the deep learning model.

[0031] Further, in the step S21, the non-uniform indoor environment green performance data is converted into RDF triples of "entity-relation-entity" through the RDF toolkit RDFLib of Python, and the triples are stored in a graph database for subsequent processing and query operations;

[0032] In the step S22, the extracted entities are aligned with the existing entities in the non-uniform indoor environment to construct a relationship graph, and a graph matching algorithm is used to link the entities; the records in the non-uniform indoor environment knowledge base that point to the unified entities are merged by using a clustering algorithm or a DeepWalk algorithm to calculate the distance of the entities and relationships embedded in the vector space; an embedding model is used to calculate the relationship similarity, and a graph neural network is used to integrate the relationships between the same entities from different data sources; the integrated data is used to eliminate the conflicting or contradictory knowledge obtained from different data sources and derive new knowledge based on the priority rules set based on the reliability and timestamp of the data source through rule-based reasoning; and the implicit non-uniform indoor environment green performance data relationships are mined based on machine learning reasoning to enhance the knowledge and improve the accuracy of information extraction and the reasoning ability of the knowledge graph.

[0033] Further, in the step S24, the extracted non-uniform indoor environment multi-modal data is preprocessed; the temperature, humidity, air flow rate, light intensity, and energy consumption real-time monitoring numerical structured data are normalized and dimensionally reduced; the project information, user demand, design specification and standard, and policy document text data are embedded into vectors based on a natural language processing model; the building photos, effect drawings, general layout drawings, plan drawings, section drawings, elevation drawings, detail drawings, and optimization simulation result pseudo-color image data are extracted into feature vectors using a ResNet or CNN model, and the user behavior videos and building three-dimensional model videos are extracted into features from key frames using a C3D network and an LSTM model;

[0034] In the step S24, based on the generated embedding vectors, the green performance data feature dimensions are standardized; an HNSW vector index structure is constructed to improve the vector similarity search speed and provide support for the matching of semi-structured data in the multi-modal knowledge graph, and to accelerate the intelligent question and answer process for user design problems;

[0035] In the step S24, for the green performance evaluation indicators of typical engineering scenarios, the influence factors are further determined, including equipment distribution, wall thermal performance, natural lighting, personnel flow, and dynamic load; and the typical model or algorithm elements in calculation or simulation are clarified, including CFD model, daylight analysis model, dynamic energy consumption analysis model, PMV model, PPD model, and energy consumption optimization algorithm, and regional partition control algorithm;

[0036] In the step S24, the green performance portrait of the non-uniform indoor environment of a specific large space engineering scenario is constructed, and the multi-modal green performance database of the specific large space engineering scenario is completed and updated.

[0037] Further, in step S31, the knowledge graph is divided into different communities by identifying the groups of nodes associated with each other through a community detection algorithm; and missing relationships or entities are identified and added through a graph embedding algorithm to enhance the completeness of the knowledge graph and assist knowledge reasoning.

[0038] Further, in step S33, based on the multi-modal knowledge graph of the typical engineering scene of the green performance optimization design of the non-uniform indoor environment, an enhanced base model of the non-uniform indoor environment green performance knowledge representation empowerment is trained based on the base model on the cloud platform, the cloud computing power is utilized to process the coupling relationship between the multi-modal data involved in the non-uniform indoor environment optimization design task, and the decision mode based on the designer's subjective experience is empowered.

[0039] Further, in step S41, based on the designer input green performance optimization design scene text, photo data, the temperature, humidity, air quality, air flow rate parameters are considered, the heat source distribution, personnel distribution, and use scene limiting factors are considered, the non-uniform indoor environment optimization design problem is understood, and the typical problem of high-rise building non-uniform environment optimization design is responded to;

[0040] In step S41, based on the end-cloud collaborative distributed working architecture, for the complex non-uniform indoor environment green performance design decision problem, the data processing capability of the enhanced base model of the cloud knowledge representation empowerment is utilized to obtain a comprehensive understanding of the multi-modal performance data fusion features; a variety of building design, building simulation software or platforms are cooperated to make high-rise space building non-uniform indoor environment green performance optimization design decisions, and the limitations of specific algorithms or software tools in calculation efficiency or calculation cost are overcome;

[0041] In step S41, when the designer inputs the optimization design task instruction to the large model, the cloud side model understands the user demand and issues a simulation instruction to the end side model to guide the end side model to simulate; the end side model combines the cloud side instruction and the engineering scene model or the non-uniform indoor environment parameters to perform green performance simulation optimization and output simulation results to provide decision basis for the cloud side model.

[0042] Further, in step S42, the digital twin model of the high-rise space building is constructed; the parameter and energy consumption changes caused by the optimization scheme modification are analyzed in real time through cloud computing, and the visualization comparison of the optimization design scheme is realized combined with VR and AR technologies, and the real-time collaborative adjustment of the optimization design scheme by multiple persons is supported.

[0043] The application further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method for making non-uniform indoor environment optimization design decisions of high-rise space buildings based on a multi-modal knowledge graph enhanced base model when executing the computer program.

[0044] The application further provides a computer readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the high and large space building non-uniform indoor environment optimization design decision method based on the multi-modal knowledge graph enhanced base model.

[0045] The application has the following beneficial effects:

[0046] In response to the complexity of the high and large space building non-uniform environment, the multi-modal characteristics (such as numerical values, texts, images, videos, etc.) of data are emphasized, and a high and large space building non-uniform indoor environment optimization design decision method based on a multi-modal knowledge graph enhanced base model is proposed. Through knowledge representation of multi-modal data of the high and large space building non-uniform indoor environment, combined with specific application engineering scene types, a non-uniform indoor environment green performance data multi-modal knowledge graph and a vector database are constructed to effectively represent and store the multi-modal data. In the design decision process, a green performance optimization design model that fits the typical engineering scene of the high and large space building non-uniform indoor environment is constructed, the non-uniform indoor environment optimization design problem is understood, and an optimization design decision method is proposed to output numerical values such as building form parameters, space parameters, and heating, ventilation, and air conditioning system parameters, and green performance image data of the typical section of the high and large space building. On this basis, a construction method of an enhanced base model enabled by knowledge representation is proposed, based on an end-cloud collaborative distributed work architecture, optimal allocation of computing resources is realized, multi-modal knowledge graphs and vector databases are cooperated to improve knowledge retrieval and question answering efficiency, reduce computing cost and time, and improve computing accuracy. Combined with a digital sand table, visual interaction of the optimization design scheme is realized. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor based on the provided drawings.

[0048] Figure 1 A flowchart of the high and large space building non-uniform indoor environment optimization design decision method based on the multi-modal knowledge graph enhanced base model.

[0049] Figure 2 A block diagram of the high and large space building non-uniform indoor environment optimization design decision method based on the multi-modal knowledge graph enhanced base model. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of the present application.

[0051] In combination with Figures 1-2 The present application provides a high and large space building non-uniform indoor environment optimization design decision method based on a multi-modal knowledge graph enhanced base model, which comprises the following steps:

[0052] S1, acquire multi-modal data knowledge of non-uniform indoor environment; comprehensively apply a LoRa-based MEMS thermal environment monitoring sensor array fixed on a building facade or installed on an intelligent cleaning or service robot, a thermal imaging camera carried on a UAV, and indoor environment physical parameter acquisition equipment such as a multi-wavelength high-density point cloud LiDAR, and user data acquisition equipment such as a high-resolution camera, a quantum dot infrared imager for multi-spectral imaging, an ultra-wideband (UWB) multiple-input multiple-output (MIMO) radar, and a flexible high-speed organic light detection device, to collect and model multi-modal data of non-uniform indoor environment of high and large space buildings.

[0053] The step S1 comprises the following steps:

[0054] S11, collect non-uniform indoor environment green performance data; in the step S11, based on dynamic acquisition of four types of data information of non-uniform indoor environment values, texts, pictures, and videos on a BIM, IoT, and other integrated platforms, a non-uniform indoor environment green performance data set is constructed, and the data modal type and corresponding data format requirements in the data analysis process are determined. BIM platform data is acquired through IFC format, and IoT platform data is acquired through JSON format. Numerical data is input through CSV, JSON, etc., textual data is input through TXT, CSV, JSON, etc., image data is input through JPEG, PNG, etc., and video data is input through MP4, AVI, etc.

[0055] S12, generating an embedded representation of the fusion multi-modal non-uniform indoor environment green performance information; using CLIP, MMV, etc. cross-modal alignment algorithm and multi-modal Transformer to process building green performance data of different modalities, and converting these data into high-dimensional vectors that retain their multi-modal data characteristics, embedding a shared vector space to realize information fusion and alignment, retaining and strengthening the correlation between different modalities of building green performance data, such as the correlation between images and text descriptions, etc., to assist in understanding the semantic relationship of multi-modal data, so that the model can consider the information of multiple modalities when processing multi-modal tasks.

[0056] S13, extracting non-uniform indoor environment green performance data information; the information extraction in step S13 includes entity extraction, relationship extraction and event extraction; the entity extraction refers to extracting specific building green performance objects or concepts from unstructured non-uniform indoor environment green performance data set based on BiLSTM-CRF deep learning model using named entity recognition (NER) method and context information; the relationship extraction refers to identifying the relationship between the above-mentioned non-uniform indoor environment green performance entities using OpenIE and BERT relationship classification model or dependency syntax analysis method; the event extraction refers to the spatio-temporal dimension parameters of multiple non-uniform indoor environment green performance entities, which are detected by ACE event extraction framework and identified by BiLSTM-CRF, BERT, etc. deep learning model based on non-uniform indoor environment green performance event trigger words and related information.

[0057] S2, fusing multi-modal data knowledge of non-uniform indoor environment, and constructing multi-modal knowledge graph and vector database of non-uniform indoor environment; applying multi-dimensional knowledge graph to expand the knowledge representation of temperature, humidity, noise level, air quality, air flow rate, passenger flow distribution, passenger behavior, etc. data of non-uniform indoor environment of high and large space building, combining with engineering scene types such as industrial plant, stadium, airport terminal, large exhibition center, etc. to construct multi-modal knowledge graph and vector database of non-uniform indoor environment green performance of high and large space building.

[0058] The step S2 includes the following steps:

[0059] S21, converting the extracted information into RDF triples; in step S21, the non-uniform indoor environment green performance data is converted into RDF triples of "entity-relation-entity" by using RDF toolkit RDFLib of Python, and the triples can be stored in graph database for subsequent processing and query operations, etc.

[0060] S22, constructing a multi-modal knowledge graph of the non-uniform indoor environment based on RDF triples; in the step S22, the extracted entities are aligned with the existing entities of the non-uniform indoor environment to construct a relationship graph between the entities, and a PageRank, Random Walk or other graph matching algorithm is used to link the entities. The records in the non-uniform indoor environment knowledge base that point to the unified entities are merged by using a K-means, DBSCAN or DeepWalk algorithm to calculate the distance of the entities and relationships embedded in the vector space. A Word2Vec, GloVe or other embedding model is used to calculate the relationship similarity, and a GCN, RGCNs or other graph neural network is used to integrate the relationships between the same entities from different data sources. The integrated data is used to eliminate the conflicting or contradictory knowledge obtained from different data sources and derive new knowledge based on rule-based reasoning according to the reliability, time stamp and other priority rules set by the data source; the implicit non-uniform indoor environment green performance data relationship is mined based on GNN or other machine learning reasoning, the knowledge is enhanced, and the accuracy of information extraction and the reasoning ability of the knowledge graph are improved.

[0061] S23, determining typical non-uniform indoor environment engineering scene elements; for industrial plants, sports venues, airport terminals and large exhibition centers with non-uniform indoor environment characteristics, the engineering scene type is identified, and the corresponding green performance evaluation index is extracted, including power intensity, equipment energy consumption ratio, building energy consumption, indoor air quality, thermal comfort, lighting energy efficiency, etc.

[0062] S24, construct and update the corresponding multi-modal knowledge graph and vector database; in the step S24, the extracted non-uniform indoor environment multi-modal data is preprocessed. The real-time monitoring numerical structured data such as temperature, humidity, air flow rate, illumination intensity and energy consumption is normalized and dimensionally reduced; the text data such as project information, user demand, design specification and standard, policy document is based on BERT, TF-IDF and other natural language processing models to generate embedded vectors; the image data such as building photos, effect drawing, general plan, plan, section drawing, elevation drawing, detail drawing and optimization simulation result pseudo-color drawing are extracted by using ResNet, CNN model to extract feature vectors, and the user behavior video, building three-dimensional model and other video data are extracted from key frames by using C3D network and LSTM model. Based on the generated embedded vectors, the standardization green performance data feature dimension is constructed. The HNSW vector index structure is constructed to improve the vector similarity search speed, and to provide support for the matching of semi-structured data in the multi-modal knowledge graph, and to speed up the intelligent question and answer process for user design problems. For the green performance evaluation index of the typical engineering scene, further determine its influencing factors, including equipment distribution, wall heat insulation performance, natural lighting, personnel flow, dynamic load, etc.; and clarify the typical model or algorithm elements in calculation or simulation, including CFD model, sunshine analysis model, dynamic energy consumption analysis model, PMV model, PPD model and energy consumption optimization algorithm, regional partition control algorithm, etc. The green performance portrait of non-uniform indoor environment of specific large space engineering scene is constructed, and the multi-modal green performance database of specific large space engineering scene is completed and updated.

[0063] S3, strengthen the optimization design decision problem knowledge retrieval, and construct a green performance optimization design model suitable for the typical engineering scene of non-uniform indoor environment; based on the non-uniform indoor environment multi-modal knowledge graph and vector database, a green performance optimization design model of non-uniform indoor environment typical engineering scene is constructed, which cooperates with the indoor environment physical parameter acquisition device, the user data acquisition device, the HVAC equipment, the intelligent operation and maintenance robot, the edge computing device and the cloud computing platform.

[0064] The step S3 includes the following steps:

[0065] S31, modular clustering of knowledge in the knowledge graph and obtaining a structured subgraph; in the step S31, the knowledge graph is divided into different communities by identifying the node groups associated with each other through a community detection algorithm. The missing relationships or entities are identified and added through a graph embedding algorithm to enhance the completeness of the knowledge graph and assist knowledge reasoning.

[0066] S32, understand the design problem, improve the efficiency of knowledge retrieval based on structured subgraph; in the step S32, based on the design problem text, real scene photos, flat section engineering drawings and other green performance optimization design engineering scene data input by the designer, the potential green performance hidden problem of the specific optimization design scene is understood, and the indoor physical environment parameters such as temperature, humidity, illumination, air quality, air pollutant concentration and ventilation volume are considered in combination with the restriction factors such as equipment distribution, wall heat insulation performance, natural lighting, personnel flow and dynamic load. The hierarchical clustering method such as hierarchical clustering and k-means clustering is used to extract the abstract of the multi-dimensional scale non-uniform indoor environment typical engineering scene, and the community is analyzed hierarchically. The structured subgraph and its abstract are converted into high-dimensional vector representation by Node2Vec, DeepWalk, GraphSAGE and other algorithms. The efficiency of machine learning tasks such as clustering and retrieval is improved by using FAISS, HNSW and other algorithms to assist high-dimensional vector retrieval.

[0067] S33, construct a green performance optimization design model for non-uniform indoor environment typical engineering scene; in the step S33, based on the multi-modal knowledge graph of the green performance optimization design typical engineering scene of the non-uniform indoor environment, the coupling relationship between the multi-modal data involved in the non-uniform indoor environment optimization design task is processed by combining multi-modal Transformer, GNN, RNN and other neural network models, and the traditional decision mode based on the designer's subjective experience is enabled. In the step S33, the main output data includes numerical and image data, so as to provide design alternatives at the levels of building form, space layout and energy system design, respond to the climate, environment and cultural requirements of the design task, and assist in putting forward optimization design decision. In the step S33, the model cooperates with various building design and building simulation software or platforms to process multi-modal building green performance data, and gets rid of the limitations of specific algorithms or software tools in computing efficiency or computing cost. In the step S33, the output numerical data includes building form parameters such as building shape coefficient and window-wall ratio, building space parameters such as open space, depth, layer height, equipment running time, power consumption and other HVAC system parameters; the output image data includes green performance prediction for high space building typical section such as high space building temperature gradient prediction graph, flow field prediction graph and glare simulation graph.

[0068] The green performance optimization design model of the non-uniform indoor environment typical engineering scene includes:

[0069] The monitoring module comprises a physical parameter acquisition submodule and a user data acquisition submodule. The physical parameter acquisition submodule comprises a LoRa-based MEMS thermal environment monitoring sensor array fixed on a building facade or mounted on an intelligent cleaning or service robot, which acquires data such as non-uniform indoor environment temperature, humidity, air quality, and air flow rate of a high and large space building; a pipeline detection robot, which detects cracks, blockages, corrosion, and other problems of HVAC equipment pipelines; a high-motion performance wheeled robot, which detects abnormal conditions such as fire and smoke and monitors indoor safety; and a drone, which monitors and analyzes the thermal distribution of the facade of a high and large building. The user data acquisition submodule comprises a high-resolution camera, which identifies the motion of a person based on a computer vision algorithm; a quantum dot infrared imager for multi-spectral imaging, which monitors the flow density of people in a space in real time; an ultra-wideband (UWB) multiple-input multiple-output (MIMO) radar, which obtains high-quality images of respiratory signals in multi-person scenes such as entrance halls, security check areas, waiting areas, and commercial areas; and a flexible high-speed organic light detection device, which acquires physiological signals such as the respiration, heart rate, and skin electricity of users in an office area.

[0070] The retrieval module uses the data processing capabilities of the cloud platform based on cross-modal alignment technology to obtain a comprehensive understanding of the multi-modal performance data fusion features and design problems. For a design problem input by a user, the design problem is vectorized based on the "knowledge retrieval" and "retrieval enhancement generation" technologies, and similarity retrieval and comparison are performed with the multi-modal knowledge graph and vector database. The knowledge graph structured semantic problem is used to determine the relationship between feature vectors and entities, relationships, and attributes in the graph database; the vector database is used to store data feature vectors to find the most relevant match of green performance multi-modal data. Based on similar cases and other knowledge at the semantic level, the ability to capture key design problems is enhanced, calculation simulation instructions are generated, and the knowledge graph structured subgraph involved in the optimization design problem is determined.

[0071] The calculation module cooperates with the retrieval module, integrates the element features of typical engineering scenarios of high and large space buildings based on the calculation simulation instructions and the knowledge graph structured subgraph, and uses lightweight and personalized models deployed on the edge side to realize hierarchical lightweight decision-making. The green performance evaluation indicators of the non-uniform indoor environment are dynamically analyzed, multi-modal data are fused based on data acquisition, thermal simulation software is used for simulation of heat transfer, hot air flow, and light intensity in buildings, and CFD simulation air flow patterns, thermal distribution, and temperature gradients are analyzed to respond to indoor environments in different seasons and different use scenarios. Optimization design decisions such as air conditioning design, ventilation system layout, and heat source distribution are analyzed and made, and the corresponding building form, space, and heating, ventilation, and air conditioning system design are output.

[0072] The interaction module: combined with the output result of the calculation module, the digital twin model of the high and large space building is constructed. Combined with VR and AR technology, the typical flat and vertical section temperature, humidity, air quality, air flow rate, passenger flow distribution, passenger behavior and other decision support data images are projected on the corresponding model surface. Designers manipulate the digital sand table by wearing VR equipment or moving the equipment components in the physical model. The cloud platform calculates and analyzes the changes in the parameters and energy consumption caused by the operation in real time, and recommends the arrangement position of the equipment in the alternative optimization design scheme for the designer to mark in the digital twin model, realizes the visualization comparison of the optimization design scheme, and supports real-time collaborative adjustment of the optimization design scheme by multiple personnel.

[0073] The digital twin model of the high and large space building includes: outputting the numerical and image data of the temperature, humidity, air quality, air flow rate and other values of the high and large space building non-uniform indoor environment by the MEMS thermal environment monitoring sensor array, outputting the passenger flow density image data by the infrared imager, outputting the building form parameters such as building shape coefficient, window-wall ratio, and other numerical data by LiDAR, outputting the green performance prediction image data such as high and large space building temperature gradient prediction map and flow field prediction map for the typical section of high and large space building by thermal simulation and CFD simulation, and outputting the data such as HVAC equipment operation time and power consumption.

[0074] S4, construct a non-uniform indoor environment design decision cloud platform model, coordinate cloud side and end side model, and flexibly handle optimization design problems; based on the non-uniform indoor environment optimization design model, integrate the high and large space building flat and vertical section wind field map collected by the MEMS thermal environment monitoring sensor array, the thermal image generated by the infrared imager capturing thermal radiation, the high and large space building three-dimensional model generated based on the high-precision spatial data of LiDAR, the thermal simulation software and CFD simulation results, and construct a visual high and large space building non-uniform indoor environment performance digital twin model based on VR and AR technology; for the HVAC equipment arrangement position scheme of typical functional spaces such as entrance hall, security check area, waiting area, commercial area and office area, realize the visual interaction of designers in the form of digital sand table.

[0075] The step S4 includes the following steps:

[0076] S41, build an end-cloud collaborative architecture that fits various non-uniform indoor environment typical engineering scenarios.

[0077] In the step S41, based on the text, photos and other green performance optimization design scene basic data input by the designer, for the parameters such as temperature, humidity, air quality and air flow rate, taking into account the distribution of heat sources, personnel distribution, use scenarios and other limiting factors, understanding the non-uniform indoor environment optimization design problem, and responding to the typical problems of high and large building non-uniform environment optimization design.

[0078] Based on the end-cloud collaborative distributed working architecture, for the complex non-uniform indoor environment green performance design decision problem, the data processing capability of the enhanced base model is utilized to obtain a comprehensive understanding of the multi-modal performance data fusion features based on the cloud knowledge representation empowerment. Collaborate with various building design, building simulation software or platforms to develop high space building non-uniform indoor environment green performance optimization design decision, and get rid of the limitations of specific algorithms or software tools in computing efficiency or computing cost.

[0079] When the designer inputs the optimization design task instruction to the large model, the cloud side model understands the user demand and issues simulation instructions to the end side model to guide the end side model to simulate; the end side model combines the cloud side instruction and the engineering scene model or non-uniform indoor environment parameters to perform green performance simulation optimization and provide simulation results and other decision basis for the cloud side model.

[0080] S42, a digital twin model of the high space building is constructed, and the optimization design result is displayed through a digital sand table to realize visual interaction.

[0081] In step S42, the digital twin model of the high space building is constructed; the parameter and energy consumption changes caused by the optimization scheme modification are analyzed in real time through cloud computing, and the visualization comparison of the optimization design scheme is realized combined with VR and AR technologies, and the real-time collaborative adjustment of the optimization design scheme by multiple personnel is supported.

[0082] For industrial plants, sports venues, airport terminals, large exhibition centers and other high and large space buildings with non-uniform indoor environment characteristics, the MEMS thermal environment monitoring sensor array based on LoRa, camera, infrared imager hardware device data are integrated, the time variation trend of sensor data is analyzed, the peak period of personnel flow is analyzed to analyze the time sequence and spatial characteristics of the engineering scene, the engineering scene type is identified and classified based on the deep learning model analysis of historical data, the green performance evaluation index is extracted, the influencing factors and typical models or algorithms in calculation or simulation are determined, the multi-modal green performance database is completed and updated, and is stored in a graph database. Based on the deep learning algorithm for processing graph structure data, the green performance data connected to each other is traversed, a unified representation method of cross-scale multi-modal information such as text, raster image, three-dimensional model, plane vector and building performance in the field of building is established, the input discrete data is converted into a unified multi-modal shared vector space high-dimensional vector, and according to the engineering scene, a multi-modal knowledge graph of typical engineering scenes of non-uniform indoor environment green performance optimization design is constructed. And by generating embedded vectors, standardizing the green performance data feature dimension, constructing an HNSW vector index structure, improving the vector similarity search speed, and constructing a non-uniform indoor environment green performance data vector database, support is provided for matching semi-structured data in the multi-modal knowledge graph.

[0083] The indoor environment monitoring sensor array includes: a flexible resistance temperature sensor, a flexible capacitance humidity sensor, a tunable diode laser absorption spectrum sensor for monitoring CO2, CO, methane and other gas concentrations, an optical particle counter for monitoring PM2.5, PM10 and other suspended particulate matters in the air, and a particle image velocimeter for monitoring air flow rate.

[0084] The knowledge graph and vector database support multi-modal data input of numerical values (including indoor environment data, building form data, green performance data, etc.), texts (project information, user requirements, design specifications and standards, policy documents, etc.), images (building photos, effect drawings, general layout, plan, section, elevation, detail drawing, optimization simulation result pseudo-color map, etc.), and videos (user behavior videos, building three-dimensional models, etc.).

[0085] The green performance evaluation index includes pollutant concentration, time and area ratio of indoor environment parameters in the adaptive thermal comfort area under natural ventilation or composite ventilation working conditions of main functional rooms, temperature range of inner surface of outer wall and roof, average natural ventilation frequency of main functional rooms under typical working conditions in transition season, energy consumption of cold and heat sources, energy consumption of transmission and distribution system, renewable energy utilization rate, power intensity, etc.

[0086] Typical problems in the optimization design of the indoor environment of tall buildings include: uneven distribution of longitudinal heat sources in tall spaces, requiring a balance between energy consumption and comfort, analysis of heat convection, and determination of local heating and fresh air system solutions; large temperature and humidity gradients caused by height differences, requiring coordination of temperature and humidity and development of zoning design schemes for air conditioning systems; and uneven airflow, requiring the identification of air stagnation zones, installation of air circulation fans, and a combination of natural and mechanical ventilation to overcome spatial height differences and airflow stratification.

[0087] Example

[0088] like Figure 1 The illustration shows a non-uniform indoor environment optimization design decision-making method based on a multimodal knowledge graph-enhanced base model, applicable to tall, open-air buildings such as airport terminals. These terminals have complex indoor environments with uneven passenger flow, densely populated areas with alternating functions, and significant variations in usage frequency and demand. The temperature and humidity distribution within the indoor environment is also uneven, and issues such as heat accumulation and poor air circulation are prone to occur. The design must balance energy conservation and comfort while ensuring smooth passenger flow, achieving intelligent control.

[0089] S1. Acquire knowledge of multimodal data related to non-uniform indoor environments;

[0090] In this embodiment, data collection is based on BIM and IoT platforms. Data collected using the BIM platform includes: geometric spatial data such as building dimensions, location, spatial layout, and window-to-wall ratio; material physical properties such as thermal conductivity, thermal resistance, and acoustic properties of building materials; basic parameters and pipeline layout of the building's HVAC system and other electrical equipment; energy monitoring data such as daylighting and energy consumption; and building operation and maintenance data such as equipment maintenance records and operation schedules. Based on the IoT platform, real-time dynamic data of the terminal building can be obtained, specifically including: indoor physical environment parameters such as temperature and humidity, air quality, noise level, and illuminance on work surfaces; energy consumption monitoring data such as real-time total energy consumption of HVAC system equipment or real-time energy consumption of specific areas and specific equipment; and user-related data such as pedestrian flow distribution and passenger behavior. Simultaneously, it can also record dynamic changes in outdoor physical environment parameters such as meteorological data, wind speed, and wind direction, and supplement project information, user requirements, design specifications and standards, and policy documents.

[0091] The geometric space data includes parameter value data, descriptive text data, and architectural photos, renderings, general plans, floor plans, cross-sectional views, elevations, detail drawings, and other image data, material physical properties include value data and descriptive text data, energy monitoring data includes value data with time series information, descriptive text data, pseudo-color images, and video data, indoor physical environment data includes value data with time series information, pseudo-color images, and video data, user-related data includes questionnaire text data, photo image data, and behavior video data.

[0092] The raw data of different data sources is preprocessed to extract structured information. For tabular numerical data, no preprocessing is required. For text data, semantic segmentation and part-of-speech tagging are achieved based on NLP or LLM models.

[0093] Using a multi-modal Transformer deep learning architecture, models such as ViLBERT and UNITER are used to embed numerical, textual, pictorial, and video data in vector space through different encoders. Self-attention mechanisms are used to assign weights between different modalities, enhancing the understanding of data between different modalities, such as the relationship between the terminal cross-sectional view and the corresponding annotation text; generating representations related to both modalities while preserving the time series information of dynamic data.

[0094] For the preprocessed data, extract the entity and relationship data. Apply deep learning-based models such as BERT, RoBERTa, and spaCy to classify and identify entities, which include terminal green performance object or concept-related category information, serving as the subject or object of RDF triples. Use OpenIE, BERT, and other relationship classification models or dependency syntax analysis methods to identify relationships between the above-mentioned terminal indoor environment green performance entities, outputting the predicate of RDF triples.

[0095] For event data involving multiple entities, attributes, and time series information, use the ACE event extraction framework to mark entities and relationships, and use deep learning models such as BiLSTM-CRF and BERT to identify event trigger words and roles, and convert them into triples.

[0096] S2, fuse non-uniform indoor environment multi-modal data knowledge, and construct a non-uniform indoor environment multi-modal knowledge graph and vector database;

[0097] Convert the terminal indoor environment green performance data into RDF triples of "entity-relation-entity" using the RDF toolkit RDFLib of Python, and store them in the Neo4j graph database.

[0098] When new indoor environment green performance knowledge of the terminal is extracted, the above steps are repeated to extract the entity, relationship, and attribute information of the knowledge, and the newly extracted indoor environment green performance entity of the terminal is identified and labeled using context information, embedded into the vector space, and linked based on semantic similarity calculation. For text information, entities can be matched based on string similarity. For records in the graph database pointing to the same entity, based on clustering algorithms such as K-means, DBSCAN, or machine learning methods such as logistic regression and random forest, items with similar names or identical IDs are filtered, and it is determined whether the relationships of different inputs are the same through pattern matching or rule reasoning; the newly extracted indoor environment green performance entity of the terminal and its relationships and attributes are embedded into the vector space using graph embedding methods such as DeepWalk and Node2Vec, and repeated entities are identified and merged through distance calculation. Based on this, it is determined whether the relationships between entities are consistent through semantic similarity calculation, and similar relationships from different sources are fused through graph convolutional neural networks or relationship attention networks. In the case of conflicts between entities or attributes, the data with the highest priority is selected as the trusted data based on the set terminal data source reliability, timestamp, value range, and other factors, and the logical constraints between the data are verified through ontology reasoning algorithms to satisfy objective laws and architectural common sense. Based on this, new knowledge is further inferred through GNN algorithms.

[0099] For the terminal engineering scene, considering its complex indoor environment, uneven distribution of human flow, and time-varying characteristics, the evaluation indicators include: indoor temperature, relative humidity, air flow rate, air quality (carbon dioxide concentration, PM2.5 concentration), working surface illumination, different regional heating, ventilation, and air conditioning system, lighting system power intensity, equipment energy consumption, user thermal comfort, and human flow density changes.

[0100] The cold and heat load distribution of the terminal space is complex and is affected by its large-scale depth, bay, and height. Different regions have different influencing factors, including: 1. Whether close to the curtain wall, the area close to the curtain wall is greatly affected by outdoor environmental solar radiation, penetration wind, etc., and the area far from the curtain wall is mainly affected by personnel mobility and density, and equipment heat dissipation; 2. Net height, the space has a large temperature difference between the upper and lower through regions due to the upward floating of hot air; 3. Different use mode regions, the main regions of the terminal include the entrance hall, check-in area, security check area, waiting area, baggage claim area, and commercial area. Among them, the entrance hall has strong personnel mobility, the check-in area has the characteristics of high-density personnel gathering for a short time, the security check area and the baggage claim area have personnel retention and flow patterns, the waiting area has a long personnel retention time, and the commercial area has large personnel density fluctuations.

[0101] The data is preprocessed, and different regions or heights, different use scenarios, temperature, humidity, air flow rate, light intensity, energy consumption, and other real-time monitoring numerical structured data are normalized and dimensionally reduced. The features of image data such as typical regional profile optimization simulation result pseudo-color maps and user behavior videos are extracted. The text data such as design specifications and standards, policy documents, and other common design specifications and standards of the terminal building are subjected to natural language processing to generate embedding vectors. The image data such as building photos, effect maps, general layout maps, sectional maps, elevation maps, detail sample maps, and other image data are subjected to ResNet and CNN model feature extraction, and the building three-dimensional model is subjected to C3D network and LSTM model feature extraction from key frames. On this basis, an HNSW vector index structure is constructed, and a vector database of the terminal building indoor environment multi-modal knowledge graph is constructed to assist in the green performance optimization design decision of the large space building non-uniform indoor environment, such as non-uniform longitudinal heat source distribution, large temperature and humidity gradient, uneven air flow, local glare or insufficient lighting, and other green performance optimization design decisions.

[0102] S3, strengthening optimization design decision problem knowledge retrieval, constructing green performance optimization design model suitable for typical engineering scene of non-uniform indoor environment;

[0103] For the green performance data knowledge graph of the terminal building, community detection algorithms such as Louvain and Label Propagation are used for rough division to efficiently segment the community structure in the large-scale knowledge graph. Node2Vec and DeepWalk random walk are used to embed text nodes and graph nodes respectively to enhance the completeness of the knowledge graph and assist in knowledge reasoning.

[0104] Hierarchical clustering, k-means clustering, and other hierarchical clustering methods are used to extract terminal building engineering scene abstracts. Node2Vec, DeepWalk, and GraphSAGE algorithms are used to convert entities and relationships of the terminal building green performance knowledge into high-dimensional vectors. FAISS and HNSW algorithms are used to store the embedded communities and entities in the vector database to assist in retrieval.

[0105] Based on the design problem text input in the question and answer process, the specific scene photo of the terminal, the terminal plane and vertical section engineering drawing, the typical section simulation diagram of the terminal, and other green performance optimization design engineering scene materials, the specific optimization design scene type (entrance hall, check-in area, security area, waiting area, luggage pickup area, commercial area) is judged, the potential green performance hidden problem of the specific optimization design scene is understood, the relevant area plane and typical section temperature, relative humidity, air flow rate, air quality (carbon dioxide concentration, PM2.5 concentration) distribution, regional heating, ventilation and air conditioning system, lighting system power intensity, equipment energy consumption, and user thermal comfort feeling, people flow density change are searched, and for the check-in area and the security area, the work surface illuminance information needs to be considered.

[0106] The CFD model, the sunshine analysis model, the dynamic energy consumption analysis model, the PMV model, the PPD model, the energy consumption optimization algorithm, the regional partition control algorithm and the like are called in cooperation to process the specific green performance simulation analysis problem, the indoor physical environment data of the multiple prediction models are coupled with the outdoor physical environment elements of the site, the terminal green performance problem decision is supported, and the designer is assisted in scheme comparison and selection. The output terminal local form or space layout optimization opinion, energy system or technical platform optimization design opinion and the like alternative scheme, and the building form parameters such as terminal shape coefficient or window-wall ratio, building space parameters such as specific area opening or depth after optimization of the space layout, regional equipment running time, power consumption and the like heating, ventilation and air conditioning system parameters of the multiple optimization schemes, the typical section building temperature gradient prediction diagram, the flow field prediction diagram, the glare simulation diagram and the like supporting these data are used to get rid of the limitations of specific algorithms or software tools in calculation efficiency or calculation cost.

[0107] S4, a non-uniform indoor environment design decision cloud platform model is constructed, and the cloud side and the end side model are cooperated to flexibly process the optimization design problem;

[0108] On the basis of the data set obtained by using the green performance optimization design model matched with the typical engineering scene of the terminal to perform design simulation, decision and the like, the computing power advantage and the data processing capacity of the cloud platform are exerted, the Mixtral8x78 large model is used to construct an enhanced base model enabled by green performance knowledge representation of the terminal, the multi-modal building information involved in the terminal scene is comprehensively understood, and the reasoning analysis ability of the green performance optimization design model to cope with complex problems is better supported.

[0109] On this basis, the collaborative analysis capability of the cloud-side enhanced base model and the terminal-side green performance optimization design model is strengthened. The calculation cost and time are reduced, and the calculation efficiency and accuracy are improved. When the designer inputs the optimization design task instruction to the large model, the cloud-side model understands the user demand and sends simulation instructions to the terminal-side model to guide the terminal-side model to simulate. The terminal-side model combines the cloud-side instructions and the engineering scene model or the non-uniform indoor environment parameters to perform green performance simulation optimization and output simulation results to provide decision basis for the cloud-side model.

[0110] For real-time updated terminal indoor environment green performance data of IoT, etc., the Internet of Things sensors are used for real-time collection, the analysis interval time is determined according to the measurement accuracy requirement, data analysis is performed, and corresponding decision reference scheme is provided. The designer or terminal operation and maintenance personnel compares and selects or modifies the alternative scheme to formulate an indoor physical environment regulation device operation scheme. On the basis of excluding abnormal data, the average value, standard deviation and other statistical parameters of the data in different time sequences are analyzed, and the phased law characteristics are summarized. When the deviation exceeds the preset critical value, the knowledge graph is updated.

[0111] In the process of solving the green performance optimization design, the green performance portrait of the terminal indoor environment is constructed, and the multi-modal green performance database of the terminal is completed and updated.

[0112] The application further provides an electronic device including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method when executing the computer program.

[0113] The application further provides a computer readable storage medium for storing computer instructions, and the computer instructions implement the steps of the method when executed by a processor.

[0114] The memory in the embodiments of the present application can be a volatile memory or a nonvolatile memory, or can include both volatile and nonvolatile memory. Among them, the nonvolatile memory can be a read only memory (ROM), a programmable ROM (PROM), an erasable programmable ROM (EPROM), an electrically EPROM (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM) used as an external cache. By way of example, and not limitation, many forms of RAM can be used, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM). It is to be noted that the memory described with the methods of the present application is intended to include, but not be limited to, these and any other suitable types of memory.

[0115] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media. The available media can be magnetic media (such as floppy disk, hard disk, magnetic tape), optical media (such as high-density digital video disc (digital video disc, DVD)), or semiconductor media (such as solid state disc (solid state disc, SSD)) and the like.

[0116] In the implementation process, each step of the above method can be completed by integrated logic circuit of hardware in the processor or instruction in the form of software. The steps of the method disclosed in the embodiments of the present application can be directly embodied as hardware processor execution, or executed by combination of hardware and software modules in the processor. The software module can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, register or other mature storage medium in the art. The storage medium is located in the memory, and the processor reads the information in the memory and combines the hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0117] It should be noted that the processor in the embodiments of the present application can be an integrated circuit chip with processing capability of signals. In the implementation process, each step of the method embodiments can be completed by integrated logic circuits or instructions in the form of software of the hardware in the processor. The processor mentioned above can be a general processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can be any conventional processor. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as hardware code processor execution completion, or executed by hardware and software module combination in the code processor. The software module can be located in random access memory, flash memory, read only memory, programmable read only memory or electrically erasable programmable memory, register and other mature storage media in the art. The storage medium is located in the memory, and the processor reads the information in the memory, and combines the hardware to complete the steps of the above method.

[0118] The above describes in detail the high and large space building non-uniform indoor environment optimization design decision method based on the multi-modal knowledge graph enhanced base model. The principle and implementation mode of the present application are described by using specific examples. The above embodiment is only used to help understand the method and core idea of the present application; at the same time, for the general technical personnel in the art, according to the idea of the present application, the specific implementation mode and application range will be changed, and the above description should not be understood as the limitation of the present application.

Claims

1. A high and large space building non-uniform indoor environment optimization design decision method based on a multi-modal knowledge graph enhanced base model, characterized in that, The method comprises the following steps: S1, acquiring non-uniform indoor environment multi-modal data knowledge; The step S1 comprises the following steps: S11, collecting non-uniform indoor environment green performance data; S12, generating embedded representation of fusion multi-modal non-uniform indoor environment green performance information; S13, extracting non-uniform indoor environment green performance data information; S2, fusing non-uniform indoor environment multi-modal data knowledge, constructing non-uniform indoor environment multi-modal knowledge graph and vector database; The step S2 comprises the following steps: S21, converting the extracted information into RDF triples; S22, constructing non-uniform indoor environment multi-modal knowledge graph based on RDF triples; S23, determining typical non-uniform indoor environment engineering scene elements; S24, constructing and updating corresponding multi-modal knowledge graph and vector database; S3, strengthening and optimizing design decision problem knowledge retrieval, and constructing green performance optimization design model suitable for typical engineering scene of non-uniform indoor environment; The step S3 comprises the following steps: S31, modular clustering the knowledge in the knowledge graph and obtaining a structured subgraph; S32, understanding the design problem and improving the knowledge retrieval efficiency based on the structured subgraph; S33, constructing green performance optimization design model suitable for typical engineering scene of non-uniform indoor environment; S4, end-cloud collaboration, flexible processing of optimization design problems, and realizing the visualization interaction of optimization design results in the form of digital sand table; The step S4 comprises the following steps: S41, constructing end-cloud collaboration architecture suitable for various typical engineering scenes of non-uniform indoor environment; S42, constructing digital twin model of high and large space building, and realizing the visualization interaction by displaying the optimization design results through digital sand table; In the step S24, the extracted non-uniform indoor environment multi-modal data is preprocessed; the temperature, humidity, air flow rate, illumination intensity and energy consumption real-time monitoring numerical structured data are normalized and dimensionally reduced; the project information, user demand, design specification and standard and policy file text data are embedded into vectors based on a natural language processing model; the building photo, effect drawing, general plan, plan, section drawing, elevation drawing, detail drawing and optimization simulation result pseudo-color image data are extracted into feature vectors by using ResNet or CNN model, and the user behavior video and building three-dimensional model video data are extracted into features from key frames by using C3D network and LSTM model; In the step S24, based on the generated embedded vectors, the green performance data feature dimension is standardized; the HNSW vector index structure is constructed to improve the vector similarity search speed and provide support for matching of semi-structured data in the multi-modal knowledge graph, and accelerate the intelligent question and answer process for user design problems; In step S24, the influencing factors of the typical engineering scenario green performance evaluation index are further determined, including equipment distribution, wall thermal performance, natural lighting, personnel flow and dynamic load; and the typical model or algorithm elements in calculation or simulation are clarified, including CFD model, sunshine analysis model, dynamic energy consumption analysis model, PMV model, PPD model and energy consumption optimization algorithm, regional partition control algorithm; In step S24, the green performance image of the non-uniform indoor environment of the specific large space engineering scenario is constructed, and the multi-modal green performance database of the specific large space engineering scenario is completed and updated.

2. The method of claim 1, wherein, In step S13, the information extraction includes entity extraction, relation extraction and event extraction; the entity extraction refers to extracting specific building green performance objects or concepts from the unstructured non-uniform indoor environment green performance dataset based on the context information by using the named entity recognition method based on the BiLSTM-CRF deep learning model; the relation extraction refers to identifying the relationship between the above-mentioned non-uniform indoor environment green performance entities by using the OpenIE and BERT relation classification model or the dependency syntax analysis method; and the event extraction refers to the spatio-temporal dimension parameters of multiple non-uniform indoor environment green performance entities, which are detected by the ACE event extraction framework and identified by the deep learning model based on the non-uniform indoor environment green performance event trigger words and related information.

3. The method of claim 1, wherein, In step S21, the non-uniform indoor environment green performance data is converted into RDF triples of "entity-relation-entity" by using the RDF toolkit RDFLib of Python, and the triples are stored in a graph database for subsequent processing and query operations; In step S22, the extracted entities are aligned with the existing entities of the non-uniform indoor environment to construct a relationship graph, and the entities are linked by using a graph matching algorithm; the records in the non-uniform indoor environment knowledge base that point to the unified entities are merged by using a clustering algorithm or a DeepWalk algorithm to calculate the distance of the entities and relationships embedded in the vector space; the relationship similarity is calculated by using an embedding model, and the relationships between the same entities of different data sources are integrated by using a graph neural network; The integrated data is used for reasoning based on rules, and according to the priority rules based on the reliability and timestamp of the data source, the conflicting or contradictory knowledge obtained from different data sources is eliminated, and new knowledge is derived; Based on machine learning reasoning, the implicit non-uniform indoor environment green performance data relationship is mined, the knowledge is enhanced, and the accuracy of information extraction and the reasoning ability of the knowledge graph are improved.

4. The method of claim 1, wherein, In step S31, the knowledge graph is divided into different communities by using a community detection algorithm to identify the node groups associated with each other; and the missing relationships or entities are identified and added by using a graph embedding algorithm to enhance the completeness of the knowledge graph and assist knowledge reasoning.

5. The method of claim 1, wherein, In step S33, based on the multi-modal knowledge graph of the typical engineering scene of the green performance optimization design of the non-uniform indoor environment, an enhanced base model of the non-uniform indoor environment green performance knowledge representation empowerment is trained based on the base model on the cloud platform, taking advantage of the cloud computing power to process the coupling relationship between the multi-modal data involved in the non-uniform indoor environment optimization design task, and empowering the decision-making mode based on the designer's subjective experience.

6. The method of claim 1, wherein, In step S41, based on the green performance optimization design scene text and photo data input by the designer, the temperature, humidity, air quality, and air flow rate parameters are considered, and the heat source distribution, personnel distribution, and use scene limiting factors are taken into account to understand the non-uniform indoor environment optimization design problem and respond to the typical problems of high-rise building non-uniform environment optimization design. In step S41, based on the end-cloud collaborative distributed work architecture, the data processing capability of the enhanced base model of the cloud knowledge representation empowerment is used to obtain a comprehensive understanding of the multi-modal performance data fusion features for complex non-uniform indoor environment green performance design decision problems. Collaborate with various building design and building simulation software or platforms to make green performance optimization design decisions for high-rise space buildings, and get rid of the limitations of specific algorithms or software tools in computing efficiency or computing cost. In step S41, when the designer inputs the optimization design task instruction to the large model, the cloud side model understands the user's demand and issues simulation instructions to the end side model to guide the end side model to simulate; the end side model combines the cloud side instructions and the engineering scene model or non-uniform indoor environment parameters to perform green performance simulation optimization and output simulation results to provide decision basis for the cloud side model.

7. The method of claim 1, wherein, In step S42, the digital twin model of the high-rise space building is constructed; the cloud computing analyzes the parameter and energy consumption changes caused by the optimization scheme modification in real time, and combines with VR and AR technologies to realize the visualization comparison of the optimization design scheme and support real-time collaborative adjustment of the optimization design scheme by multiple personnel. 8.An electronic device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to realize the steps of the method of any one of claims 1-7.

9. A computer readable storage medium for storing computer instructions, characterized in that, The computer instructions are executed by the processor to realize the steps of the method of any one of claims 1-7.

Images