A building energy consumption simulation method using BIM technology

Abstract

This invention relates to the field of building energy consumption simulation technology, specifically a method for simulating building energy consumption using BIM technology. The method includes: acquiring a building information model of the target building; extracting building geometric topology, component attributes, and spatial relationship data to construct a structured building dataset; collecting annual meteorological parameters, building operating time, and indoor environmental setting parameters, integrating them to form an environmental and operational dataset; employing an improved finite element mesh generation algorithm to adaptively divide the computational mesh based on component thermal properties and spatial adjacency relationships; constructing dynamic control equations for building thermal processes and assembling an energy consumption simulation model; calling a numerical solver to complete model calculations; outputting hourly energy consumption and indoor thermal environment parameters for the entire year; identifying high-energy-consumption periods and spatial areas through data analysis; and generating an energy-saving optimization report. This solution can improve the accuracy and data refinement of building energy consumption simulation, achieving precise simulation and analysis of the building's annual thermal conditions and energy consumption status.

Description

Technical Field

[0001] This invention relates to the field of building energy consumption simulation technology, and in particular to a method for simulating building energy consumption using BIM technology. Background Technology

[0002] Existing building energy consumption simulations generally rely on building information models to collect basic data and integrate fixed environmental conditions and indoor operating parameters to build a simulation system. Conventional energy consumption simulations generally adopt a unified finite element mesh generation method and use a static calculation mode to complete building thermal-related calculations. Most simulation methods can only carry out energy consumption calculations under short-cycle, fixed operating conditions.

[0003] The standardized grid division method fails to consider the differences in the thermal properties of building components, does not adjust the grid layout based on the spatial adjacency relationships of components, and has a single grid unit division standard, making it unable to match the heat transfer characteristics of different areas of the building. Building thermal process calculations mostly use fixed steady-state equations, and various meteorological data are only selected and substituted into the calculations with fixed values, failing to incorporate the continuously changing meteorological parameter series throughout the year. The building's operating status and indoor environmental parameters cannot be integrated into a unified model.

[0004] To address the current issues of insufficient mesh adaptability and rigid thermal process calculation models, this paper optimizes the mesh generation and processing logic and builds a dynamic thermal process calculation system to meet the actual needs of building full-cycle energy consumption simulation. This system compensates for the shortcomings of conventional simulation methods in dynamic parameter coupling and refined mesh calculation, enabling the complete calculation of building's annual time-series energy consumption and indoor thermal environment. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a building energy consumption simulation method using BIM technology.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a building energy consumption simulation method using BIM technology, comprising:

[0007] Obtain the building information model of the target building, and extract building geometric topology data, component attribute data and spatial relationship data from the building information model to form a structured building dataset;

[0008] The system obtains the annual meteorological parameter sequence of the environment in which the target building is located, as well as the building's operation schedule and indoor environmental setting parameters, to form an environmental and operational dataset.

[0009] An improved finite element mesh generation algorithm is applied to the structured building dataset to generate a physical field calculation mesh suitable for energy consumption simulation. The improved finite element mesh generation algorithm adaptively divides the mesh cells based on the thermal properties and spatial adjacency of building components.

[0010] Based on the physical field calculation grid, a dynamic control equation for the building thermal process is constructed, and combined with the parameters in the environmental and operational datasets, a complete building energy consumption simulation model is assembled.

[0011] The numerical solver is invoked to solve the building energy consumption simulation model, and the hourly energy consumption data and indoor thermal environment parameters of the building under the annual meteorological parameter sequence are output.

[0012] Post-processing and analysis are performed on the hourly energy consumption data and indoor thermal environment parameters to identify high energy consumption periods and spatial areas, and corresponding energy-saving optimization strategy reports are generated.

[0013] As a further aspect of the present invention, the step of obtaining the building information model of the target building includes:

[0014] Extract the original BIM data file containing building geometry information, component attribute information, and spatial relationship information from the original engineering documents provided by the designer or contractor of the target building.

[0015] The original BIM data file is cleaned and standardized to remove redundant information that is irrelevant to energy consumption simulation, and the component names and material properties are unified according to the preset classification and naming standards.

[0016] The cleaned and standardized BIM data is imported into a preset neutral exchange format file to generate a building information model of the target building with a unified structure and clear semantics.

[0017] As a further aspect of the present invention, an improved finite element mesh generation algorithm is applied to the structured building dataset to generate a physical field computation mesh suitable for energy consumption simulation, including:

[0018] The building geometry topology data in the structured building dataset is analyzed to identify the three-dimensional geometric surfaces of all building envelope components;

[0019] Based on the component attribute data, material thermal property parameters are assigned to the surface of each enclosure structure component;

[0020] Using the surfaces of the enclosure structure components as initial boundaries, a spatial adjacency diagram between each surface is established based on the aforementioned spatial relationship data;

[0021] The improved finite element mesh generation algorithm is invoked. Under the constraints of the spatial adjacency graph, the improved finite element mesh generation algorithm initially discretizes the computational domain according to the material thermal property parameters of the component surface. In the region where the material thermal property parameters change drastically, mesh cells of a first preset size are generated, and in the region where the properties are uniform, mesh cells of a second preset size are generated, wherein the second preset size is larger than the first preset size.

[0022] The quality of the initially discretized mesh cells is optimized to ensure that the geometry of the mesh cells meets the requirements of numerical calculation, and finally the physical field calculation mesh composed of nodes and cells is output.

[0023] As a further aspect of the present invention, based on the physical field computational grid, a dynamic control equation for the building thermal process is constructed, and combined with parameters from the environmental and operational datasets, a complete building energy consumption simulation model is assembled, including:

[0024] In each cell of the physical field computation grid, a transient heat conduction differential equation based on energy balance is established, which describes the change of the temperature field inside the cell over time.

[0025] Based on the thermal properties provided by the component attribute data in the structured building dataset, the corresponding thermal conductivity, density, and specific heat capacity parameters are set for the differential equation of each unit.

[0026] The building's interior space is processed into an internal region with time-varying internal heat sources and boundary conditions based on the aforementioned operating schedule and indoor environment setting parameters.

[0027] The annual meteorological parameter sequence is processed into time-varying boundary conditions for the surface of the building's external envelope;

[0028] The differential equations of all units, internal regional conditions and external boundary conditions are coupled and combined to form a set of equations describing the unsteady thermal process of the entire building, which is the complete building energy consumption simulation model.

[0029] As a further aspect of the present invention, a numerical solver is invoked to solve the building energy consumption simulation model, outputting hourly energy consumption data and indoor thermal environment parameters of the building under the annual meteorological parameter sequence, including:

[0030] The differential equations in the building energy consumption simulation model are discretized in time and space, and transformed into a large sparse linear algebraic equation system.

[0031] Set the initial conditions and convergence criteria for the solution. The initial conditions are the initial temperature field assumptions of the building at the start of the simulation.

[0032] The large sparse linear algebraic equation system is solved by calling an iterative numerical solver. In each simulation time step, the temperature value of all grid nodes of the building is calculated.

[0033] Based on the temperature field calculated at each time step, combined with the boundary conditions, the heat transfer through the building envelope and the theoretical load of the air conditioning system are calculated as the hourly energy consumption data.

[0034] Simultaneously, the node temperatures representing the indoor activity areas are extracted from the calculated temperature field and used as the indoor thermal environment parameters.

[0035] As a further aspect of the present invention, post-processing and analysis are performed on the hourly energy consumption data and indoor thermal environment parameters to identify high-energy consumption periods and spatial areas, including:

[0036] The hourly energy consumption data is summarized and statistically analyzed according to the time dimension to calculate the daily cumulative energy consumption, monthly cumulative energy consumption, and annual total energy consumption.

[0037] On a yearly timescale, a sliding window analysis is performed on the hourly energy consumption data to identify continuous time periods with energy consumption higher than adjacent time periods or higher than a preset threshold, and these continuous time periods are marked as potential high-energy-consumption periods.

[0038] Spatial interpolation is performed on the indoor thermal environment parameters to reconstruct the temperature distribution cloud map of the building's interior space;

[0039] By overlaying and analyzing the temperature distribution cloud map with the spatial relationship data in the structured building dataset, spatial areas where the indoor temperature continuously deviates from the allowable range of the indoor environment setting parameters can be identified.

[0040] Spatiotemporal correlation analysis is performed on the potential high-energy-consumption periods and the temperature anomaly spatial regions to screen out the parts where the time range corresponding to the potential high-energy-consumption periods overlaps with the spatial range corresponding to the temperature anomaly spatial regions. The time range and spatial range corresponding to the overlapping parts are determined as the final high-energy-consumption periods and spatial regions.

[0041] As a further aspect of the present invention, the generation of the corresponding energy-saving optimization strategy report includes:

[0042] For each identified high-energy-consumption period and spatial region, related causal feature data are extracted from the structured building dataset, environmental and operational dataset, and simulation results. The causal feature data includes the building envelope structure of the corresponding region, the outdoor meteorological conditions of the corresponding period, and the corresponding indoor personnel density and equipment power density.

[0043] The causal feature data is input into a pre-trained energy-saving strategy recommendation model, which outputs a series of targeted preliminary optimization suggestions.

[0044] A feasibility assessment is conducted on the preliminary optimization suggestions, and the feasibility assessment is based on preset cost constraints, construction difficulty constraints, and impact on the indoor environment constraints.

[0045] The feasible optimization suggestions after screening are sorted according to their expected energy-saving potential and combined to form a structured text and chart description, ultimately generating the energy-saving optimization strategy report.

[0046] As a further aspect of the present invention, the causal feature data is input into a pre-trained energy-saving strategy recommendation model, which outputs a series of targeted preliminary optimization suggestions, including:

[0047] The energy-saving strategy recommendation model includes a feature encoding network and a strategy decoding network;

[0048] The feature encoding network receives the causal feature data and maps it into a high-dimensional feature vector.

[0049] The policy decoding network receives the high-dimensional feature vector and, based on an attention mechanism, retrieves the candidate policy entries most relevant to the current feature from a predefined energy-saving policy knowledge base.

[0050] The retrieved candidate strategy entries are scored for confidence and conflict detected, and entries with low confidence or that fundamentally conflict with the building's inherent conditions are removed.

[0051] The selected candidate strategy entries are combined with the current causal feature data to instantiate parameters and generate specific and actionable preliminary optimization suggestions.

[0052] As a further aspect of the present invention, the improved finite element mesh generation algorithm is invoked. Under the constraints of the spatial adjacency graph, the improved finite element mesh generation algorithm initially discretizes the computational domain based on the material thermal property parameters of the component surface. It generates mesh elements of a first preset size in regions where the material thermal property parameters change drastically, and generates mesh elements of a second preset size in regions with uniform properties. The second preset size is larger than the first preset size, including:

[0053] The mesh generation process is initiated using the spatial topology of the component surface defined by the spatial adjacency graph as the initial boundary constraint.

[0054] Traverse all surfaces of the building envelope components and read the material thermal properties parameters assigned to each component surface, wherein the material thermal properties parameters include at least the thermal conductivity.

[0055] Within the computational domain, the material thermal property parameters of adjacent or nearby grid potential generation regions are compared. When the gradient of the change of the material thermal property parameters within a unit spatial distance is detected to exceed a preset first threshold, the region where the gradient of the change of the material thermal property parameters exceeds the first threshold is determined to be a region where the material thermal property parameters change drastically.

[0056] In the region where the thermal properties of the material change drastically, initial discretization is performed using a preset first mesh division size to generate mesh cells corresponding to the first mesh division size;

[0057] Within the computational domain, the material thermal property parameters of adjacent or nearby potential generation regions of the grid are compared. When it is detected that the gradient of the change of the material thermal property parameters within a unit spatial distance is lower than a preset second threshold, the region where the gradient of the change of the material thermal property parameters is lower than the second threshold is determined to be a region with uniform properties, wherein the second threshold is less than or equal to the first threshold.

[0058] Within the region with uniform properties, initial discretization is performed using a preset sparse grid partitioning size to generate grid cells corresponding to the second grid partitioning size, wherein the second grid partitioning size is larger than the first grid partitioning size;

[0059] Based on the spatial adjacency diagram, it is ensured that at the boundary between the region where the thermal properties of the material change drastically and the region where the properties are uniform, the generated grid cells of different sizes are connected through a gradual transition, thus completing the initial discretization of the computational domain.

[0060] As a further aspect of the present invention, an iterative numerical solver is invoked to solve the large-scale sparse linear algebraic equation system. Within each simulation time step, the temperature values ​​of all grid nodes of the building are calculated, including:

[0061] Set the initial temperature field for the current simulation time step, and use the grid node temperature values ​​obtained from the previous time step as the initial values ​​for the iterative solution of the current time step;

[0062] The large sparse linear algebraic equation system is expressed as a standard linear system and assembled into a coefficient matrix, an unknown temperature vector, and a vector of constant terms on the right-hand side.

[0063] The preconditional conjugate gradient method is selected as the iterative solution algorithm, and a preconditioner for accelerating convergence is constructed based on the structural characteristics of the coefficient matrix.

[0064] Under the preset maximum number of iterations and residual convergence tolerance constraints, an iterative solution loop is executed. In each iteration, the approximate solution of the unknown temperature vector is updated, and the residual norm corresponding to the current approximate solution is calculated.

[0065] When the residual norm is less than the residual convergence tolerance or the maximum number of iterations is reached, the iteration loop is terminated, and the approximate solution of the unknown temperature vector obtained in the last iteration is output as the final temperature value of all grid nodes in the current time step.

[0066] Using the final temperature field of the current time step as the initial condition, the simulation is progressively carried out to the next simulation time step. The process of setting the initial temperature field of the current simulation time step and outputting the final temperature value of the grid node of the current time step is repeated until the simulation of the entire time span corresponding to the meteorological parameter sequence of the whole year is completed.

[0067] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0068] Based on the thermal properties and spatial adjacency relationships of building components, adaptive grid cell generation is achieved, changing the traditional fixed-specification grid generation method. The arrangement of grid cells conforms to the thermal characteristics of the building components themselves and matches the overall structural layout of the building space. The basic unit layout for physics field calculations is more reasonable, adjusting the adaptability of the overall physics field calculations and mitigating the calculation bias caused by standardized grid generation patterns. The underlying calculation structure for energy consumption simulation conforms to the actual thermal structure of the building, adapting to the basic conditions of heat transfer between different building components, ensuring that the basic framework of thermal calculations matches the actual structural characteristics of the building.

[0069] Based on the optimized physical field computational grid, dynamic control equations for building thermal processes are constructed. These equations are integrated with long-term meteorological parameter sequences, building operation schedules, and indoor environmental constraints. The dynamic equations adapt to real-time changes in building thermal state, and integrated calculations are performed in conjunction with long-term environmental parameter fluctuations. Various environmental and operational parameters are processed in a unified manner. Continuous hourly energy consumption data and related indoor thermal environment parameters are output, distinguishing energy consumption performance across different time periods and spaces. This objectively reflects the long-term changes in building thermal state, expands the analytical scope of building energy consumption and indoor environment, and improves the analytical conditions for the building's full-cycle thermal state. Attached Figure Description

[0070] Figure 1 This is a flowchart of a building energy consumption simulation method using BIM technology as described in this invention;

[0071] Figure 2 A flowchart for obtaining the building information model of the target building;

[0072] Figure 3A flowchart for generating a physics computation grid suitable for energy consumption simulation. Detailed Implementation

[0073] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0074] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0075] See Figure 1 This invention provides a method for simulating building energy consumption using BIM technology, the specific method including:

[0076] The process involves: acquiring the building information model (BIM) of the target building; extracting building geometric topology data, component attribute data, and spatial relationship data from the BIM to form a structured building dataset; acquiring the annual meteorological parameter sequence of the target building's environment, as well as the building's operating schedule and indoor environmental setting parameters to form an environment and operation dataset; applying an improved finite element mesh generation algorithm to the structured building dataset to generate a physical field computational mesh suitable for energy consumption simulation, with the improved algorithm adaptively dividing the mesh cells based on the thermal properties and spatial adjacency relationships of building components; constructing dynamic control equations for the building's thermal processes based on the physical field computational mesh, and assembling a complete building energy consumption simulation model by combining the parameters from the environment and operation datasets; solving the building energy consumption simulation model using a numerical solver, outputting hourly energy consumption data and indoor thermal environment parameters of the building under the annual meteorological parameter sequence; performing post-processing and analysis on the hourly energy consumption data and indoor thermal environment parameters to identify high-energy-consumption periods and spatial areas, and generating corresponding energy-saving optimization strategy reports.

[0077] In one embodiment of the present invention, see [reference] Figure 2The process involves extracting original BIM data files containing building geometry, component attribute information, and spatial relationship information from the original engineering documents provided by the design or construction party of the target building; cleaning and standardizing the original BIM data files to remove redundant information unrelated to energy consumption simulation, and unifying component names and material attributes according to preset classification and naming standards; and importing the cleaned and standardized BIM data into a preset neutral exchange format file to generate a building information model of the target building with a unified structure and clear semantics.

[0078] In practice, data extraction was performed from the original engineering files of an office building called "Oasis Building". The original engineering files were provided by the design firm and were in .rvt format. The extraction operation yielded original BIM data files containing building geometry information, component attribute information, and spatial relationship information. The geometry information included the three-dimensional shape and spatial coordinates of all walls, floors, roofs, windows, and doors. The component attribute information included wall construction layer materials, window glass types, and door materials. The spatial relationship information described the adjacency and containment relationships between rooms and the building envelope. In practice, the acquired original BIM data files undergo data cleaning and standardization. The cleaning operation directly removes redundant information unrelated to energy consumption simulation from the original BIM data files. Redundant information includes decorative line textures, furniture models, and construction progress annotations. The standardization process unifies inconsistent component names in the original BIM data files according to the preset "Classification and Naming Standard for Building Energy Consumption Simulation Components". For example, "Exterior Wall_Concrete 200mm" and "Exterior Wall-200 Concrete" are uniformly named "EXTERIORWALL-TYPE01", and the unit of the "thermal conductivity" value in the material properties is unified to W / (m·K). In practice, the cleaned and standardized BIM data is imported into a pre-defined neutral exchange format file. This neutral exchange format file adopts a JSON structure, and its top-level structure contains three main objects: "ProjectInfo", "Spaces", and "BuildingElements". Under the "BuildingElements" object, each building component item is strictly organized according to the field structure of "ID", "Name", "Geometry", "Properties", and "Adjacency" to generate a building information model of the target building with a unified structure and clear semantics.

[0079] In some embodiments, when verifying and completing material property values, if the thermal conductivity attribute value of a component in the original BIM data file is empty or significantly exceeds a reasonable range, the system matches and fills in a standard value from a preset material thermal property library based on the component name. In some embodiments, the standardization process for component names is performed according to a preset mapping rule table. The mapping rule table defines the correspondence between various common original names and standard names, and the processing automatically completes batch conversion through string matching and replacement algorithms.

[0080] Optionally, the preset classification and naming standards are extended based on industry standards, defining hierarchical naming rules for different types of building components. The first level is the component category, the second level is the construction type, and the third level is the specific specification. Optionally, the structure design of the preset neutral exchange format file considers the data interface with subsequent mesh generation and equation assembly modules, ensuring that each field can be unambiguously read and parsed by downstream processing programs. It can be understood that the operation of extracting the original BIM data file from the original project file is implemented by calling the application programming interface of the open-source IFC parsing library or a commercial BIM platform. It can be understood that the cleaning and standardization process can be written as an independent preprocessing script or software module, which accepts various common BIM file formats as input and outputs a neutral exchange format file that meets the requirements. In specific implementation, data standardization processing includes a numerical normalization step. For attributes with numerical ranges, such as the solar heat gain coefficient of a window, it is mapped to the [0,1] interval, and the mapping formula is:

[0081]

[0082] in: Represents the normalized attribute value. Represents the original attribute value. and These represent the minimum and maximum values ​​of the attribute as defined in the preset material library, respectively.

[0083] In one embodiment of the present invention, see [reference] Figure 3The algorithm analyzes the building geometry and topology data in a structured building dataset to identify the 3D geometric surfaces of all building envelope components. Based on component attribute data, it assigns material thermal property parameters to the surface of each building envelope component. Using the surface of the building envelope component as the initial boundary, it establishes a spatial adjacency graph between the surfaces based on spatial relationship data. An improved finite element mesh generation algorithm is then invoked. Under the constraint of the spatial adjacency graph, the algorithm initially discretizes the computational domain based on the material thermal property parameters of the component surfaces. Mesh elements of a first preset size are generated in regions where the material thermal property parameters change drastically, while mesh elements of a second preset size are generated in regions with uniform properties, where the second preset size is larger than the first preset size. The quality of the initially discretized mesh elements is optimized to ensure that the geometry of the mesh elements meets the requirements of numerical computation. Finally, a physical field computation mesh composed of nodes and elements is output. During mesh generation, the spatial topology of the component surfaces defined by the spatial adjacency graph is used as the initial boundary constraint to initiate the mesh generation process. All surfaces of the enclosure structure components are traversed, and the material thermal properties parameters assigned to each component surface are read. These material thermal properties parameters include at least thermal conductivity. Within the computational domain, the material thermal properties parameters of adjacent or nearby potential mesh generation regions are compared. When the gradient of the material thermal properties parameter change within a unit spatial distance exceeds a preset first threshold, the region where the gradient exceeds the first threshold is determined to be a region of drastic change in material thermal properties parameters. Within regions of drastic change in material thermal properties parameters, initial discretization is performed using a preset first mesh partitioning size to generate mesh elements corresponding to the first mesh partitioning size. During computation... Within the domain, the material thermal property parameters of adjacent or nearby potential generation regions are compared. When the gradient of the material thermal property parameters within a unit spatial distance is detected to be lower than a preset second threshold, the region where the gradient of the material thermal property parameters is lower than the second threshold is determined to be a region with uniform properties, where the second threshold is less than or equal to the first threshold. Within the region with uniform properties, initial discretization is performed using a preset sparse grid partitioning size to generate grid cells corresponding to the second grid partitioning size, where the second grid partitioning size is larger than the first grid partitioning size. Based on the spatial adjacency graph, it is ensured that at the boundary between regions with drastic changes in material thermal property parameters and regions with uniform properties, the generated grid cells of different sizes are connected through a gradual transition, thus completing the initial discretization of the computational domain.

[0084] In the specific implementation, the structured building dataset from "Oasis Building" is analyzed. This dataset contains building geometric topology data. The identification operation filters out the three-dimensional geometric surfaces of all building envelope components based on geometric entity types and spatial relationships. These surfaces include the inner and outer surfaces of exterior walls, the upper and lower surfaces of floor slabs, and the surfaces of exterior window glass. Each surface is defined by a set of vertex coordinates and facet normal vectors. In the specific implementation, based on the component attribute data in the structured building dataset, material thermal property parameters are assigned to each identified building envelope component surface. For example, the wall surface identified as "EXTERIORWALL-TYPE01" is assigned a thermal conductivity of 0.81 W / (m·K), and the window glass surface identified as "WINDOW-TYPE01" is assigned a thermal conductivity of 1.2 W / (m·K). In the specific implementation, the identified surfaces of the building envelope components are used as initial boundaries. A spatial adjacency graph is established based on the spatial relationship data in the structured building dataset. This graph is an undirected graph data structure where nodes represent component surfaces and edges represent physical adjacencies between two surfaces in three-dimensional space, such as the inner surface of an exterior wall being associated with the boundary of a room. An improved finite element mesh generation algorithm is then invoked. Under the constraints of the spatial adjacency graph, the algorithm initially discretizes the computational domain based on the material thermal properties of the component surfaces. In regions with drastic changes in material thermal properties, a first-preset mesh element with a size of 0.05 meters is generated; in regions with uniform properties, a second-preset mesh element with a size of 0.2 meters is generated. The initially discretized mesh elements are then optimized for quality. This optimization process checks the interior angles, aspect ratio, and Jacobian determinant of each mesh element. Elements that do not meet the preset geometric requirements undergo local node adjustments or face re-division. The output is a physical field computational mesh composed of a list of node coordinates and a list of element connection relationships.

[0085] In practical implementation, the improved finite element mesh generation algorithm follows an adaptive size strategy in its mesh generation process. Within the computational domain, it compares the material thermal properties of adjacent or nearby potential mesh generation regions. These material thermal properties include at least thermal conductivity. When the gradient of change in material thermal properties per unit spatial distance exceeds a preset first threshold of 0.5 (W / (m·K)) / meter, the region is determined to be an area with drastic changes in material thermal properties, such as near the boundary between an exterior wall and a window. Within regions with drastic changes in material thermal properties, an initial discretization using a first mesh size of 0.05 meters is employed. Conversely, within the computational domain, when the gradient of change in material thermal properties per unit spatial distance is lower than a preset second threshold of 0.1 (W / (m·K)) / meter, the region is determined to be a region with uniform properties, such as the central region of a large concrete wall panel. Within regions with uniform properties, an initial discretization using a second mesh size of 0.2 meters is employed. In practice, based on the spatial adjacency diagram, at the boundary between regions where the thermal properties of the material change drastically and regions with uniform properties, the algorithm controls the gradual connection of grid nodes between grid cells of different sizes by inserting transition layer cells. The size of the transition layer cells is linearly interpolated between the sizes of cells in adjacent regions.

[0086] In some embodiments, the spatial adjacency graph is established using an algorithm based on bounding box collision detection and the angle between normal vectors. When the 3D bounding boxes of two component surfaces intersect and the angle between the surface normal vectors is less than a specific angle, an edge is added to the spatial adjacency graph to connect the corresponding two surface nodes. In some embodiments, the gradient calculation of the change of material thermal property parameters within a unit spatial distance uses the central difference method to estimate the spatial derivatives of parameters such as thermal conductivity at regularly sampled spatial points. Optionally, the improved finite element mesh generation algorithm uses the generalization of constrained Delaunay triangulation in 3D space as the basic mesh generation method, and the surfaces defined by the spatial adjacency graph serve as impenetrable constraint boundaries. Optionally, the preset first threshold and second threshold can be set hierarchically according to the typical material property range of building components. For example, one set of thresholds can be set for lightweight composite walls, and another set of thresholds can be set for heavy concrete structures. It is understood that the execution of the improved finite element mesh generation algorithm can be achieved by calling the adaptive mesh generation routines provided in open-source geometry processing libraries such as CGAL and Gmsh, and passing in a custom size field function, which is calculated from the spatial distribution of material thermal property parameters. It is understandable that the mesh quality optimization step can be performed iteratively until all mesh elements meet the shape quality standards required by the numerical solver, such as all elements having interior angles greater than 30 degrees and less than 150 degrees.

[0087] In practical implementation, the improved finite element mesh generation algorithm determines the local mesh size based on the gradient of material thermal property parameters using the following formula:

[0088]

[0089] in: It is the calculated local mesh target size. It is the user-defined base grid size. It is a scaling factor that controls the sensitivity of the mesh size to gradients. Thermal conductivity of the material The gradient vector at the current spatial location, This is the magnitude of the gradient. The formula is used to calculate a desired mesh size for each predetermined location within the computational domain before generating the initial discrete nodes, guiding subsequent cell generation.

[0090] In one embodiment of the present invention, a transient heat conduction differential equation based on energy balance is established on each cell of the physical field computation grid. The differential equation describes the change of the temperature field inside the cell over time. According to the thermal property parameters provided by the component attribute data in the structured building dataset, corresponding thermal conductivity, density, and specific heat capacity parameters are set for the differential equation of each cell. The interior space of the building is processed into an internal region with time-varying internal heat sources and boundary conditions according to the operating schedule and indoor environment parameters. The annual meteorological parameter sequence is processed into time-varying boundary conditions for the surface of the building's external envelope. The differential equations of all cells, the internal region conditions, and the external boundary conditions are coupled and combined to form a set of equations describing the unsteady thermal process of the entire building, i.e., a complete building energy consumption simulation model.

[0091] In practical implementation, based on the physical field computation grid of a building named "Oasis Building," a transient heat conduction differential equation based on energy balance is established on each hexahedral element of the physical field computation grid. The transient heat conduction differential equation describes the change of the temperature field inside the element with time, and its general form is:

[0092]

[0093] in: Representing temperature, it is a function of space and time. Represents time, Represents material density, The specific heat capacity of the representative material Represents the thermal conductivity of the material. Represents the heat source intensity within the unit. It is the gradient operator. It is a divergence operator. In specific implementation, based on the thermal property parameters provided by the component attribute data in the structured building dataset, the corresponding thermal conductivity, density, and specific heat capacity parameters are set for the differential equation of each element. The thermal conductivity, density, and specific heat capacity parameters are mapped from the component attribute data. Each physical field calculation grid element is associated with a specific building component through its spatial location, thereby obtaining the material property value of that component. For example, a grid element located in the outer wall “EXTERIORWALL-TYPE01” has a thermal conductivity of 0.81 W / (m·K), a density of 2200 kg / m³, and a specific heat capacity of 880 J / (kg·K), as shown in Table 1.

[0094] Table 1: Material Thermophysical Properties

[0095] Material Name Thermal conductivity (W / (m·K)) Density (kg / m³) Specific heat capacity (J / (kg·K)) reinforced concrete 1.74 2500 920 Extruded polystyrene board 0.034 35 1500 Insulating glass 1.2 2500 840 plasterboard 0.25 900 1090

[0096] In practice, the interior space of a building is processed into an internal area with time-varying internal heat sources and boundary conditions based on the operating schedule and indoor environmental setting parameters in the environmental and operational dataset. The operating schedule defines the working and rest patterns of people, lighting, and equipment in the building, and the indoor environmental setting parameters define the temperature and humidity control targets. For example, an office room is marked as "occupied" from 9:00 to 18:00 from Monday to Friday. At this time, the internal heat sources include heat generated by people (5 watts per square meter), heat generated by equipment (15 watts per square meter), and heat generated by lighting (10 watts per square meter). At the same time, the air temperature boundary condition of this area is set to be maintained at 26 degrees Celsius. In practice, the annual meteorological parameter sequence in the environmental and operational dataset is processed into time-varying boundary conditions for the building's external envelope surface. The annual meteorological parameter sequence includes hourly dry-bulb temperature, wet-bulb temperature, direct solar radiation, diffuse solar radiation, wind speed, and wind direction. These parameters are transformed into convective heat transfer boundary conditions and solar radiation heat flux boundary conditions acting on the outer surfaces of the exterior walls, roof, and windows. For example, at a specific moment, the boundary condition for the east-facing exterior wall surface includes a convective heat transfer coefficient calculated from the outdoor dry-bulb temperature and wind speed at that time, and a solar radiation absorption heat flux calculated from the sun's position and direct radiation. In practice, the differential equations of all elements, internal regional conditions and external boundary conditions are coupled and combined. The internal regional conditions are added to the element equations at the corresponding spatial locations in the form of internal heat source terms or third-type boundary conditions. The external boundary conditions are applied to the element nodes on the surface of the building envelope in the form of second-type or third-type boundary conditions. Finally, a large set of partial differential equations describing the unsteady thermal process of the entire building is formed. This set of equations is the complete building energy consumption simulation model.

[0097] In some embodiments, the processing of the internal region can be further refined into multiple hot zones, each associated with a set of operating schedules and indoor environmental setting parameters. In the physical field calculation grid, grid cells belonging to the same hot zone share the same internal heat source and air node coupling conditions. In some embodiments, for scenarios with complex shading components or adjacent buildings, the calculation of solar radiation heat flux boundary conditions needs to consider shading. In this case, an irradiance calculation submodule based on geometric projection is invoked to calculate the accurate instantaneous solar radiation heat gain for each grid patch on the building's exterior surface. Optionally, the time-varying pattern of the internal heat source of the building's interior space can be input in the form of a discrete hourly load density table, which defines the heat generation per unit area per hour of the day for personnel, lighting, and equipment. It can be understood that the process of processing the annual meteorological parameter sequence into time-varying boundary conditions involves parsing and time synchronization of standard meteorological year data files to ensure that the corresponding meteorological parameter vector can be obtained at each simulation time step. It is understandable that a fully assembled building energy consumption simulation model is mathematically represented by generating a large sparse matrix and a right-hand vector, where the non-zero elements of the matrix are determined by the discrete format of the unit equations, and the right-hand vector is determined by the boundary conditions and the internal heat source terms.

[0098] In one embodiment of the present invention, the differential equations in the building energy consumption simulation model are discretized in time and space, transforming them into a large sparse linear algebraic equation system; initial conditions and convergence criteria are set for the solution, with the initial condition being the initial temperature field assumption of the building at the start of the simulation; a numerical solver based on the iterative method is invoked to solve the large sparse linear algebraic equation system, and the temperature values ​​of all grid nodes of the building are calculated in each simulation time step; based on the temperature field calculated in each time step, combined with the boundary conditions, the heat transfer through the building envelope and the theoretical load of the air conditioning system are calculated as hourly energy consumption data; simultaneously, the node temperatures representing the indoor occupancy areas are extracted from the calculated temperature field as indoor thermal environment parameters. In the solution process, the initial temperature field for the current simulation time step is set, and the grid node temperature values ​​obtained from the previous time step are used as the initial values ​​for the iterative solution in the current time step. The large sparse linear algebraic equation system is expressed in standard linear system form and assembled into a coefficient matrix, an unknown temperature vector, and a constant term vector on the right-hand side. The preconditioned conjugate gradient method is selected as the iterative solution algorithm, and a preconditioner for accelerating convergence is constructed based on the structural characteristics of the coefficient matrix. Under the preset maximum number of iterations and residual convergence tolerance constraints, the iterative solution loop is executed, updating the unknown temperature values ​​in each iteration. The approximate solution of the temperature vector is obtained, and the residual norm corresponding to the current approximate solution is calculated. When the residual norm is less than the residual convergence tolerance or the maximum number of iterations is reached, the iteration loop is terminated, and the approximate solution of the unknown temperature vector obtained in the last iteration is output as the final temperature value of all grid nodes in the current time step. The final temperature field of the current time step is used as the initial condition, and the process is repeated from setting the initial temperature field of the current simulation time step to outputting the final temperature value of the grid nodes in the current time step until the simulation of the entire year's meteorological parameter sequence corresponding to the time span is completed.

[0099] In the specific implementation, a solution preparation was carried out for the building energy consumption simulation model of a building named "Oasis Building". The building energy consumption simulation model consists of a system of partial differential equations. The system of differential equations in the building energy consumption simulation model was discretized in time and space. Time discretization was performed using a first-order implicit Euler scheme, and spatial discretization was performed on the physical field computation grid using the finite volume method. The processing transformed the continuous system of partial differential equations into a large sparse linear algebraic equation system concerning the temperature of the grid nodes. In the specific implementation, initial conditions and convergence criteria were set for the solution. The initial conditions were set as the initial temperature field assumption of the building at the start of the simulation (e.g., 0:00 on January 1st), assuming that the temperature field of the entire building is uniform and equal to 20 degrees Celsius. The convergence criterion was set as the L2 norm of the iterative solution residuals being less than 1.0e-6, and the maximum number of iterations was set to 2000.

[0100] In practice, an iterative numerical solver is invoked to solve the large-scale sparse linear algebraic equation system. Within each simulation time step, the temperature values ​​of all grid nodes in the building are calculated. The initial temperature field for the current simulation time step is set, using the grid node temperature values ​​obtained from the previous time step as the initial values ​​for the iterative solution. For the first time step, a user-defined initial temperature field assumption is used. Furthermore, the large-scale sparse linear algebraic equation system is expressed in standard linear system form. And assembled into a coefficient matrix Unknown temperature vector and the vector of the constant term on the right. The coefficient matrix It is a large sparse matrix whose dimension equals the total number of grid nodes. The non-zero elements are determined by the thermal properties of the grid cells and the discretization scheme. The unknown temperature vector is... The vector containing the desired temperature values ​​for all grid nodes at the current time step, and the constant term on the right-hand side. It includes known terms contributed by boundary conditions, internal heat sources, and the temperature field from the previous time step. In specific implementation, the preconditioned conjugate gradient method is selected as the iterative solution algorithm, and based on the coefficient matrix... Based on the structural characteristics, an incomplete Cholesky decomposition preconditioner is constructed to accelerate convergence. In the specific implementation, under the preset maximum number of iterations and residual convergence tolerance constraints, an iterative solution loop is executed, updating the unknown temperature vector in each iteration. Find an approximate solution and calculate the residual norm corresponding to the current approximate solution. ,in This is the approximate solution obtained in the k-th iteration. The iteration loop terminates when the residual norm is less than the residual convergence tolerance or the maximum number of iterations is reached, and the unknown temperature vector obtained in the last iteration is used. The approximate solution is output as the final temperature value of all grid nodes at the current time step. In practice, the final temperature field at the current time step is used as the initial condition, and the simulation is repeated for the next simulation time step, from setting the initial temperature field at the current simulation time step to outputting the final temperature value of the grid nodes at the current time step, until the simulation of the entire annual meteorological parameter sequence corresponding to the time span (8760 hours) is completed.

[0101] In practical implementation, based on the temperature field calculated at each time step and combined with boundary conditions, the heat transfer through the building envelope and the theoretical load of the air conditioning system are calculated as hourly energy consumption data. When calculating heat transfer, the transient heat transfer rate is calculated for each envelope unit, such as exterior walls, roof, and windows, based on the temperature difference and thermal resistance of its inner and outer surface nodes. The theoretical load of the air conditioning system is obtained by performing instantaneous energy balance calculations on all spatial areas that need to maintain a set temperature, summing the heat flow at all boundaries of the spatial area with the internal heat sources. In practical implementation, the node temperatures representing indoor occupancy areas are extracted from the calculated temperature field as indoor thermal environment parameters. For example, the grid node temperature at a height of 1.1 meters above the floor at the center point of each office room is extracted as the hourly indoor dry-bulb temperature of that room (see Table 2).

[0102] Table 2: Parameter Table for Iterative Solution Process

[0103] Parameter name symbol Example values Convergence tolerance 1.0e-6 Maximum number of iterations 2000 Initial temperature assumption 20.0°C Time step 3600s

[0104] In some embodiments, time discretization can employ a higher-order scheme, such as the Crank-Nicolson scheme, to improve the accuracy of time integration. In this case, the constant term vector on the right-hand side... The assembly method needs to be adjusted accordingly to include a weighted average of information from the current and previous time steps. In some embodiments, the construction of the preconditioner can employ an algebraic multigrid method instead of incomplete Cholesky decomposition to handle more complex coefficient matrices or coefficients with larger condition numbers. Optionally, in assembling the coefficient matrix... When storing non-zero elements of a matrix, a compressed sparse row format can be used to save memory and improve the efficiency of matrix-vector multiplication. Optionally, for buildings with periodic operating characteristics, a simulation result from a typical period can be used as the initial temperature field to reduce the simulation warm-up time required to reach dynamic equilibrium from a uniform initial field. It is understood that calling iterative method-based numerical solvers can be achieved by linking numerical computing libraries such as PETSc and Eigen, which provide optimized implementations of iterative solvers such as the preconditioned conjugate gradient method. It is understood that calculating the theoretical load of the air conditioning system is a post-processing step based on the temperature field results; the load calculation model can be a simple heat balance model or coupled with a simplified air handling unit model.

[0105] In one embodiment of the present invention, hourly energy consumption data is summarized and statistically analyzed according to the time dimension to calculate daily cumulative energy consumption, monthly cumulative energy consumption, and annual total energy consumption. On the annual time scale, a sliding window analysis is performed on the hourly energy consumption data to identify continuous time periods with energy consumption higher than adjacent time periods or higher than a preset threshold, and these continuous time periods are marked as potential high-energy-consumption periods. Spatial interpolation is performed on indoor thermal environment parameters to reconstruct the temperature distribution cloud map of the building's interior space. The temperature distribution cloud map is overlaid with the spatial relationship data in the structured building dataset to identify spatial areas where the indoor temperature continuously deviates from the allowable range of the indoor environment setting parameters. Spatiotemporal correlation analysis is performed on the potential high-energy-consumption periods and the spatial areas with abnormal temperatures to screen out the parts where the time range corresponding to the potential high-energy-consumption periods overlaps with the spatial range corresponding to the spatial areas with abnormal temperatures, and the time range and spatial range corresponding to the overlapping parts are determined as the final high-energy-consumption periods and spatial areas. For each identified high-energy-consumption period and spatial area, relevant causal feature data is extracted from structured building datasets, environmental and operational datasets, and simulation results. These causal feature data include the building envelope structure of the corresponding area, outdoor meteorological conditions during the corresponding period, and the corresponding indoor occupancy density and equipment power density. The causal feature data is then input into a pre-trained energy-saving strategy recommendation model, which outputs a series of targeted preliminary optimization suggestions. The feasibility of these preliminary optimization suggestions is assessed based on preset cost constraints, construction difficulty constraints, and constraints on the impact on the indoor environment. The selected feasible optimization suggestions are then ranked according to their expected energy-saving potential and combined to form structured text and chart descriptions, ultimately generating an energy-saving optimization strategy report. The energy-saving strategy recommendation model comprises a feature encoding network and a strategy decoding network. The feature encoding network receives causal feature data and maps it into a high-dimensional feature vector. The strategy decoding network receives the high-dimensional feature vector and, based on an attention mechanism, retrieves candidate strategy entries most relevant to the current features from a predefined energy-saving strategy knowledge base. The retrieved candidate strategy entries are then scored for confidence and conflict detected, eliminating entries with low confidence or those fundamentally conflicting with the building's inherent conditions. Finally, the filtered candidate strategy entries are combined with the current causal feature data to instantiate parameters and generate specific and actionable preliminary optimization suggestions.

[0106] In the specific implementation, the simulated output of an office building named "Oasis Building" is processed. Hourly energy consumption data is summarized and statistically analyzed according to the time dimension. The calculation operation accumulates the hourly air conditioning load data of 8760 hours to obtain the daily cumulative energy consumption value. All daily cumulative energy consumption values ​​are added together by month to obtain the monthly cumulative energy consumption. The monthly cumulative energy consumption of the twelve months is added together to obtain the total annual energy consumption. In the specific implementation, on the annual time scale, a sliding window analysis is performed on the hourly energy consumption data. The width of the sliding window is set to 24 hours, and the sliding step is 1 hour. The average value and standard deviation of the energy consumption data in each window are calculated. Continuous time periods that identify the window's average energy consumption as being more than 20% higher than the average energy consumption of the three adjacent windows before and after it, or whose absolute value of the window's average energy consumption is higher than a preset threshold, are marked as potentially high-energy-consumption periods. In practical implementation, spatial interpolation is performed on indoor thermal environment parameters, which are the hourly temperatures of monitoring points at the center of each room. Using the Kriging spatial interpolation algorithm, the temperature values ​​at these discrete points are used as observations to reconstruct a continuous temperature distribution cloud map of each floor of the building at a specific time. In practical implementation, the temperature distribution cloud map is overlaid and analyzed with spatial relationship data from a structured building dataset. This spatial relationship data defines the spatial boundaries and functions of each room. The program determines which two-dimensional pixel areas in the temperature distribution cloud map correspond to which room in three-dimensional space, and counts the number of hours in a typical week when the temperature in each room exceeds the allowable range of the indoor environment parameters, identifying spatial areas where the indoor temperature consistently deviates from the allowable range. For example, a conference room might have a temperature above 27°C for more than 30% of its working hours. In practice, a spatiotemporal correlation analysis is performed on potential high-energy-consumption periods and temperature anomaly spatial areas to screen out the overlapping parts between the time range corresponding to potential high-energy-consumption periods and the spatial range corresponding to temperature anomaly spatial areas. Specifically, it is checked whether there are any identified temperature anomaly spatial areas with both abnormal energy consumption and temperature records within each marked potential high-energy-consumption period. The time range and spatial range corresponding to the overlapping parts are determined as the final high-energy-consumption period and spatial area. For example, "July 15th 13:00-16:00" is determined as a high-energy-consumption period, and the corresponding "open office area on the third floor of the East Zone" is determined as a high-energy-consumption spatial area.

[0107] In practical implementation, for each identified high-energy-consumption period and spatial area, relevant causal feature data is extracted from structured building datasets, environmental and operational datasets, and simulation results. The extracted causal feature data includes the building envelope structure of the corresponding area, outdoor meteorological conditions during the corresponding period, and the corresponding indoor occupancy density and equipment power density. In practical implementation, the causal feature data is input into a pre-trained energy-saving strategy recommendation model. The model outputs a series of targeted preliminary optimization suggestions. For example, regarding the causes of high energy consumption in the "open-plan office area on the third floor of the East Zone" in the afternoon, the model might output preliminary optimization suggestions such as "increasing the shading coefficient of exterior windows," "optimizing the air conditioning start-stop schedule," and "suggesting the installation of indoor zone temperature sensors." In practical implementation, a feasibility assessment is conducted on the preliminary optimization suggestions. The feasibility assessment is based on preset cost constraints, construction difficulty constraints, and impact constraints on the indoor environment. The cost constraint requires an investment payback period of less than 5 years; the construction difficulty constraint requires that the renovation should not affect the normal use of the main functional areas of the building; and the impact constraint on the indoor environment requires that the indoor daylight coefficient after the renovation should not be less than 90% of the original design value. Preliminary optimization suggestions that do not meet any of these constraints are eliminated. In practice, the feasible optimization suggestions after screening are sorted according to their expected energy-saving potential. The expected energy-saving potential is estimated by substituting the parameters of each suggestion into a simplified energy consumption model and then combining them to form a structured text and chart description. The text includes problem diagnosis, optimization suggestions, expected benefits and key points of implementation. The charts include a high energy consumption spatiotemporal distribution map, an optimization measure diagram and a comparison chart of expected energy-saving curves, and finally generate a complete energy-saving optimization strategy report.

[0108] In its implementation, the energy-saving strategy recommendation model comprises a feature encoding network and a strategy decoding network. The feature encoding network receives structured causal feature data vectors and maps them into 128-dimensional high-dimensional feature vectors through a three-layer fully connected neural network. The strategy decoding network receives these high-dimensional feature vectors and, based on an attention mechanism, calculates the similarity between these vectors and the feature vectors of each strategy entry in a predefined energy-saving strategy knowledge base. It then retrieves the five candidate strategy entries most relevant to the current feature from the knowledge base. Finally, the retrieved candidate strategy entries undergo confidence scoring and conflict detection. The confidence score is obtained after attention weight normalization. Conflict detection checks whether the building conditions required by the candidate strategy entry match the inherent conditions (such as structure type and equipment system type) extracted from the structured building dataset, eliminating entries with a confidence score below 0.6 or those that fundamentally conflict with the building's inherent conditions. In practice, the selected candidate strategy items are combined with the current causal characteristic data to instantiate parameters and generate specific and actionable preliminary optimization suggestions. For example, the general item "add external shading" is instantiated as "install a 0.8-meter-wide horizontal fixed external shading plate above the east-facing window on the third floor of the East Zone. The shading plate is made of aluminum alloy and the solar radiation reflectivity must be greater than 0.7".

[0109] In some embodiments, when using sliding window analysis to identify potentially high-energy-consumption periods, the window width and judgment threshold can be parameterized according to building type and climate zone. For example, for hotel buildings, the window width can be set to 12 hours to capture the differences in energy consumption patterns between day and night. In some embodiments, the pre-trained energy-saving strategy recommendation model adopts a sequence-to-sequence model trained based on a historical renovation case library, the feature encoding network is a bidirectional long short-term memory network, and the strategy decoding network is a recurrent neural network based on an attention mechanism. Optionally, spatiotemporal correlation analysis can adopt a grid-based spatiotemporal cube indexing method to discretize the time and spatial dimensions, efficiently querying events that simultaneously exhibit energy consumption anomalies and temperature anomalies within a specific time slice and spatial voxel. Optionally, the energy-saving strategy knowledge base is organized in the form of a graph database, with strategy entries as nodes. Node attributes include applicable conditions, energy-saving mechanisms, and reference parameters, and the relationships between nodes represent the combination, mutual exclusion, or sequential order of strategies. It can be understood that the operation of spatial interpolation to reconstruct the temperature distribution cloud map can be performed at a selected series of representative moments, without having to perform it for all 8760 hours, in order to balance accuracy and computational cost. It is understandable that constraints in feasibility assessments can be configured in a separate assessment rule configuration file, allowing users to adjust constraint thresholds or enable / disable specific constraints based on the specific project situation.

[0110] In practical implementation, the calculation of confidence scores for the retrieved candidate strategy items can be expressed as follows:

[0111]

[0112] in: This represents the confidence score of the c-th candidate strategy item. The original association score between the c-th candidate policy entry and the input high-dimensional feature vector is calculated by the policy decoding network through the attention mechanism. Represents the original association score of the i-th candidate strategy entry. This represents the total number of candidate strategy entries retrieved from the knowledge base. The formula is used to transform the attention weights into a normalized probability distribution, which serves as the basis for the confidence score of the strategy entries.

[0113] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A building energy consumption simulation method using a BIM technique, characterized by, The method includes: Obtain the building information model of the target building, and extract building geometric topology data, component attribute data and spatial relationship data from the building information model to form a structured building dataset; The system obtains the annual meteorological parameter sequence of the environment in which the target building is located, as well as the building's operation schedule and indoor environmental setting parameters, to form an environmental and operational dataset. An improved finite element mesh generation algorithm is applied to the structured building dataset to generate a physical field calculation mesh suitable for energy consumption simulation. The improved finite element mesh generation algorithm adaptively divides the mesh cells based on the thermal properties and spatial adjacency of building components. Based on the physical field calculation grid, a dynamic control equation for the building thermal process is constructed, and combined with the parameters in the environmental and operational datasets, a complete building energy consumption simulation model is assembled. The numerical solver is invoked to solve the building energy consumption simulation model, and the hourly energy consumption data and indoor thermal environment parameters of the building under the annual meteorological parameter sequence are output. Post-processing and analysis are performed on the hourly energy consumption data and indoor thermal environment parameters to identify high energy consumption periods and spatial areas, and corresponding energy-saving optimization strategy reports are generated.

2. The building energy consumption simulation method using BIM technology according to claim 1, wherein, The acquisition of the building information model of the target building includes: Extract the original BIM data file containing building geometry information, component attribute information, and spatial relationship information from the original engineering documents provided by the designer or contractor of the target building. The original BIM data file is cleaned and standardized to remove redundant information that is irrelevant to energy consumption simulation, and the component names and material properties are unified according to the preset classification and naming standards. The cleaned and standardized BIM data is imported into a preset neutral exchange format file to generate a building information model of the target building with a unified structure and clear semantics. 3.The building energy consumption simulation method using BIM technology according to claim 1, wherein, An improved finite element mesh generation algorithm is applied to the structured building dataset to generate a physical field computation mesh suitable for energy consumption simulation, including: The building geometry topology data in the structured building dataset is analyzed to identify the three-dimensional geometric surfaces of all building envelope components; Based on the component attribute data, material thermal property parameters are assigned to the surface of each enclosure structure component; Using the surfaces of the enclosure structure components as initial boundaries, a spatial adjacency diagram between each surface is established based on the aforementioned spatial relationship data; The improved finite element mesh generation algorithm is invoked. Under the constraints of the spatial adjacency graph, the improved finite element mesh generation algorithm initially discretizes the computational domain according to the material thermal property parameters of the component surface. In the region where the material thermal property parameters change drastically, mesh cells of a first preset size are generated, and in the region where the properties are uniform, mesh cells of a second preset size are generated, wherein the second preset size is larger than the first preset size. The quality of the initially discretized mesh cells is optimized to ensure that the geometry of the mesh cells meets the requirements of numerical calculation, and finally the physical field calculation mesh composed of nodes and cells is output. 4.The building energy consumption simulation method using BIM technology according to claim 1, wherein, Based on the physical field computational grid, dynamic control equations for building thermal processes are constructed, and combined with parameters from the environmental and operational datasets, a complete building energy consumption simulation model is assembled, including: In each cell of the physical field computation grid, a transient heat conduction differential equation based on energy balance is established, which describes the change of the temperature field inside the cell over time. Based on the thermal properties provided by the component attribute data in the structured building dataset, the corresponding thermal conductivity, density, and specific heat capacity parameters are set for the differential equation of each unit. The building's interior space is processed into an internal region with time-varying internal heat sources and boundary conditions based on the aforementioned operating schedule and indoor environment setting parameters. The annual meteorological parameter sequence is processed into time-varying boundary conditions for the surface of the building's external envelope; The differential equations of all units, internal regional conditions and external boundary conditions are coupled and combined to form a set of equations describing the unsteady thermal process of the entire building, which is the complete building energy consumption simulation model.

5. The building energy consumption simulation method using BIM technology according to claim 1, wherein, The numerical solver is invoked to solve the building energy consumption simulation model, outputting hourly energy consumption data and indoor thermal environment parameters of the building under the annual meteorological parameter series, including: The differential equations in the building energy consumption simulation model are discretized in time and space, and transformed into a large sparse linear algebraic equation system. Set the initial conditions and convergence criteria for the solution. The initial conditions are the initial temperature field assumptions of the building at the start of the simulation. The large sparse linear algebraic equation system is solved by calling an iterative numerical solver. In each simulation time step, the temperature value of all grid nodes of the building is calculated. Based on the temperature field calculated at each time step, combined with the boundary conditions, the heat transfer through the building envelope and the theoretical load of the air conditioning system are calculated as the hourly energy consumption data. Simultaneously, the node temperatures representing the indoor activity areas are extracted from the calculated temperature field and used as the indoor thermal environment parameters.

6. The building energy consumption simulation method using BIM technology according to claim 1, wherein, Post-processing and analysis are performed on the hourly energy consumption data and indoor thermal environment parameters to identify high-energy-consumption periods and spatial areas, including: The hourly energy consumption data is summarized and statistically analyzed according to the time dimension to calculate the daily cumulative energy consumption, monthly cumulative energy consumption, and annual total energy consumption. On a yearly timescale, a sliding window analysis is performed on the hourly energy consumption data to identify continuous time periods with energy consumption higher than adjacent time periods or higher than a preset threshold, and these continuous time periods are marked as potential high-energy-consumption periods. Spatial interpolation is performed on the indoor thermal environment parameters to reconstruct the temperature distribution cloud map of the building's interior space; By overlaying and analyzing the temperature distribution cloud map with the spatial relationship data in the structured building dataset, spatial areas where the indoor temperature continuously deviates from the allowable range of the indoor environment setting parameters can be identified. Spatiotemporal correlation analysis is performed on the potential high-energy-consumption periods and the temperature anomaly spatial regions to screen out the parts where the time range corresponding to the potential high-energy-consumption periods overlaps with the spatial range corresponding to the temperature anomaly spatial regions. The time range and spatial range corresponding to the overlapping parts are determined as the final high-energy-consumption periods and spatial regions.

7. The building energy consumption simulation method using BIM technology according to claim 6, wherein, The generation of the corresponding energy-saving optimization strategy report includes: For each identified high-energy-consumption period and spatial region, related causal feature data are extracted from the structured building dataset, environmental and operational dataset, and simulation results. The causal feature data includes the building envelope structure of the corresponding region, the outdoor meteorological conditions of the corresponding period, and the corresponding indoor personnel density and equipment power density. The causal feature data is input into a pre-trained energy-saving strategy recommendation model, which outputs a series of targeted preliminary optimization suggestions. A feasibility assessment is conducted on the preliminary optimization suggestions, and the feasibility assessment is based on preset cost constraints, construction difficulty constraints, and impact on the indoor environment constraints. The feasible optimization suggestions after screening are sorted according to their expected energy-saving potential and combined to form a structured text and chart description, ultimately generating the energy-saving optimization strategy report.

8. The building energy consumption simulation method using BIM technology according to claim 7, wherein, The causal feature data is input into a pre-trained energy-saving strategy recommendation model, which outputs a series of targeted preliminary optimization suggestions, including: The energy-saving strategy recommendation model includes a feature encoding network and a strategy decoding network; The feature encoding network receives the causal feature data and maps it into a high-dimensional feature vector. The policy decoding network receives the high-dimensional feature vector and, based on an attention mechanism, retrieves the candidate policy entries most relevant to the current feature from a predefined energy-saving policy knowledge base. The retrieved candidate strategy entries are scored for confidence and conflict detected, and entries with low confidence or that fundamentally conflict with the building's inherent conditions are removed. The selected candidate strategy entries are combined with the current causal feature data to instantiate parameters and generate specific and actionable preliminary optimization suggestions. 9.The building energy consumption simulation method using BIM technology according to claim 3, wherein, The improved finite element mesh generation algorithm is invoked. Under the constraints of the spatial adjacency graph, the improved finite element mesh generation algorithm initially discretizes the computational domain based on the material thermal property parameters of the component surface. It generates mesh elements of a first preset size in regions where the material thermal property parameters change drastically, and generates mesh elements of a second preset size in regions with uniform properties. The second preset size is larger than the first preset size, including: The mesh generation process is initiated using the spatial topology of the component surface defined by the spatial adjacency graph as the initial boundary constraint. Traverse all surfaces of the building envelope components and read the material thermal properties parameters assigned to each component surface, wherein the material thermal properties parameters include at least the thermal conductivity. Within the computational domain, the material thermal property parameters of adjacent or nearby grid potential generation regions are compared. When the gradient of the change of the material thermal property parameters within a unit spatial distance is detected to exceed a preset first threshold, the region where the gradient of the change of the material thermal property parameters exceeds the first threshold is determined to be a region where the material thermal property parameters change drastically. In the region where the thermal properties of the material change drastically, initial discretization is performed using a preset first mesh division size to generate mesh cells corresponding to the first mesh division size; Within the computational domain, the material thermal property parameters of adjacent or nearby potential generation regions of the grid are compared. When it is detected that the gradient of the change of the material thermal property parameters within a unit spatial distance is lower than a preset second threshold, the region where the gradient of the change of the material thermal property parameters is lower than the second threshold is determined to be a region with uniform properties, wherein the second threshold is less than or equal to the first threshold. Within the region with uniform properties, initial discretization is performed using a preset sparse grid partitioning size to generate grid cells corresponding to the second grid partitioning size, wherein the second grid partitioning size is larger than the first grid partitioning size; Based on the spatial adjacency diagram, it is ensured that at the boundary between the region where the thermal properties of the material change drastically and the region where the properties are uniform, the generated grid cells of different sizes are connected through a gradual transition, thus completing the initial discretization of the computational domain.

10. The building energy consumption simulation method using BIM technology according to claim 5, wherein, The large sparse linear algebraic equation system is solved using an iterative numerical solver. Within each simulation time step, the temperature values ​​of all grid nodes of the building are calculated, including: Set the initial temperature field for the current simulation time step, and use the grid node temperature values ​​obtained from the previous time step as the initial values ​​for the iterative solution of the current time step; The large sparse linear algebraic equation system is expressed as a standard linear system and assembled into a coefficient matrix, an unknown temperature vector, and a vector of constant terms on the right-hand side. The preconditional conjugate gradient method is selected as the iterative solution algorithm, and a preconditioner for accelerating convergence is constructed based on the structural characteristics of the coefficient matrix. Under the preset maximum number of iterations and residual convergence tolerance constraints, an iterative solution loop is executed. In each iteration, the approximate solution of the unknown temperature vector is updated, and the residual norm corresponding to the current approximate solution is calculated. When the residual norm is less than the residual convergence tolerance or the maximum number of iterations is reached, the iteration loop is terminated, and the approximate solution of the unknown temperature vector obtained in the last iteration is output as the final temperature value of all grid nodes in the current time step. Using the final temperature field of the current time step as the initial condition, the simulation is progressively carried out to the next simulation time step. The process of setting the initial temperature field of the current simulation time step and outputting the final temperature value of the grid node of the current time step is repeated until the simulation of the entire time span corresponding to the meteorological parameter sequence of the whole year is completed.

Images