An education building dynamic energy consumption tracing and visualization method and system

Abstract

The application discloses an education building dynamic energy consumption tracing and visualization method and system, belongs to the field of computer graphics image processing and building energy consumption analysis, and comprises the following steps: constructing a semantic voxel grid based on a building information model and a three-dimensional geometric grid, and establishing a spatial coordinate and voxel index mapping; collecting environmental sensing data, establishing an energy conservation discrete model on the semantic voxel grid, and solving according to a time step to obtain a heat flux vector field; obtaining student behavior and indoor number data, aggregating to generate an occupancy behavior feature according to a voxel index and a time index, and forming time series data with the heat flux vector field; obtaining an energy consumption loss weight through association mining and interval mapping to a thermal texture, driving three-dimensional visualization rendering to output an energy consumption tracing scene. The application adopts a semantic voxel grid and multi-source time series data fusion modeling and association mining method, can associate energy consumption loss and occupancy behavior, and improves the education building energy consumption diagnosis and visualization analysis capability.

Description

Technical Field

[0001] This invention relates to the fields of computer graphics and image processing and building energy consumption analysis, and in particular to a method and system for tracing and visualizing dynamic energy consumption of educational buildings. Background Technology

[0002] Building energy consumption monitoring and visualization are crucial components of building energy conservation and smart building management. This is especially true in educational building complexes such as university campuses, where the energy consumption of teaching buildings, laboratories, and other structures is characterized by complex spatial distribution, significant temporal fluctuations, and a high correlation with student and faculty activities. To improve building operation and management, relevant organizations typically install metering devices to collect energy consumption data and combine this data with geographic information systems or building floor plans to visualize the energy consumption and carbon emissions of buildings or facilities. This data is then used for operation management and energy-saving renovation decisions.

[0003] In related technologies, Chinese invention patent CN120182531A discloses a method for drawing a carbon emission monitoring map based on data visualization, relating to the field of carbon emission monitoring. The method includes the following steps: acquiring geographical data of the factory area, carbon emission data from multiple monitoring points, and wind data within the factory area; determining an interpolation point for the factory area; determining multiple neighboring monitoring points from the multiple monitoring points based on the interpolation point; determining the barrier coefficient between the interpolation point and the neighboring monitoring points based on the geographical data of the factory area; determining the carbon emission data of the interpolation point based on the barrier coefficient, the wind data within the factory area, and the carbon emission data of the neighboring monitoring points; and drawing a carbon emission monitoring map based on the interpolation point and the carbon emission data from the multiple monitoring points.

[0004] The inventors believe that most existing visualization methods based on carbon emission monitoring maps or heat maps focus on factory areas or building complexes, paying attention to the distribution of carbon emission intensity within a planar area. They are unable to depict the energy flow path within individual buildings in a fine-grained manner, cannot reflect the impact of door and window construction and building envelope characteristics on energy consumption distribution, and do not correlate information specific to educational scenarios, such as the occupancy of building interiors and the patterns of personnel activities, with changes in energy consumption. Therefore, they still have shortcomings in the dynamic energy consumption tracing and three-dimensional visualization of educational buildings. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for tracing and visualizing dynamic energy consumption in educational buildings. It employs semantic voxel grids and multi-source time-series data fusion modeling and correlation mining methods to correlate energy loss with occupancy behavior, thereby enhancing the energy consumption diagnosis and visualization analysis capabilities of educational buildings.

[0006] The above objectives can be achieved through the following approach:

[0007] A method for dynamic energy consumption tracing and visualization in educational buildings includes: acquiring building information model data and 3D geometric mesh data of the educational building; performing spatial discretization partitioning through an octree to construct a semantic voxel mesh containing door and window opening and closing parameters and material thermal resistance parameters; establishing a bidirectional mapping relationship between physical space coordinates and voxel indices; collecting attitude data from a magnetic angle sensor, fluid data from a wind speed sensor, and thermal image data from an infrared thermal imaging sensor; mapping the attitude data, fluid data, and thermal image data to the semantic voxel mesh; establishing an energy conservation discrete equilibrium equation according to material thermal resistance and convective heat transfer; performing time-step iterative calculations to generate heat flux vector field data; acquiring student entry and exit records, course schedule data, and indoor occupancy monitoring data; mapping the physical space coordinate range to a voxel index set based on the bidirectional mapping relationship; and then... The voxel index and time index are aggregated with the student entry and exit records, the course schedule data, and the indoor occupancy monitoring data to generate an occupancy behavior feature sequence. The heat flux vector field data and the occupancy behavior feature sequence are constructed into a time series tensor according to the voxel index and time index. By performing frequent itemset mining and association weight calculation, an energy loss weight matrix is ​​generated. The energy loss weight matrix and the heat flux vector field data are subjected to interval-based linear mapping calculation at each voxel index to obtain the color component parameters and transparency parameters of each voxel index and time index, generating differential thermal texture data. The differential thermal texture data is mapped to the semantic voxel mesh, and volume rendering calculation is performed on the semantic voxel mesh. Based on the bidirectional mapping relationship, a three-dimensional visualization scene is generated on the three-dimensional geometric mesh data, and the visualization data stream is output.

[0008] Optionally, establishing a bidirectional mapping relationship between physical space coordinates and voxel indexes includes: calling the building information model data to perform attribute parsing on door and window component objects, extracting door and window geometric boundary parameters, opening and closing control parameters, and component material parameters, and generating a door and window component parameter table; performing octree segmentation on the exterior and interior space of the educational building based on the three-dimensional geometric mesh data, determining the voxel unit boundaries of each octree node, and generating a voxel index set; performing spatial overlap judgment on the door and window geometric boundary parameters in the door and window component parameter table and the spatial bounding box coordinates in the voxel index set, writing the corresponding door and window opening and closing parameters and material thermal resistance parameters at the overlapping voxel indexes, and generating a bidirectional mapping relationship between physical space coordinates and voxel indexes based on the spatial bounding box coordinates of the voxel units.

[0009] Optionally, generating heat flux vector field data includes: performing time alignment and interpolation processing on the attitude data, the fluid data, and the thermal imaging data to generate a sensor data sequence; mapping the physical spatial coordinates of the installation positions of each sensor to voxel indices based on the bidirectional mapping relationship; writing wind speed vectors and temperature values ​​extracted from the sensor data sequence at the voxel indices; constructing an energy conservation discrete equilibrium equation by combining the material thermal resistance parameters of the semantic voxel mesh and the voxel adjacency relationship; performing iterative solution on the energy conservation discrete equilibrium equation according to the time step, updating the temperature values ​​of each voxel unit, and calculating the heat flux vector of each voxel unit based on the temperature difference and material thermal resistance parameters of adjacent voxel units to generate heat flux vector field data.

[0010] Optionally, constructing the discrete energy conservation equilibrium equation includes: determining a set of adjacent voxel indices for each voxel index based on the voxel adjacency relationship in the semantic voxel grid, and associating the material thermal resistance parameter with the set of adjacent voxel indices at each voxel index; and establishing a discrete energy conservation equilibrium equation at each voxel index based on the temperature value at the voxel index, the set of adjacent voxel indices, and the material thermal resistance parameter.

[0011] Optionally, generating the occupancy behavior feature sequence includes: performing time alignment and deduplication processing on student entry and exit records, course schedule data, and indoor occupancy monitoring data to generate student entry and exit event sequences, course schedule event sequences, and occupancy statistics sequences; mapping the physical spatial coordinate range of each teaching space to a set of voxel indices based on the bidirectional mapping relationship, and aggregating the student entry and exit event sequences, course schedule event sequences, and occupancy statistics sequences according to the voxel indices and time indices, calculating each voxel index and time index to generate occupancy occupancy parameters and course status markers; sorting the occupancy occupancy parameters and course status markers according to the voxel indices and time indices to generate the occupancy behavior feature sequence.

[0012] Optionally, generating the energy loss weight matrix includes: aligning the heat flux vector field data and the occupancy behavior feature sequence according to voxel index and time index to generate a voxel behavior energy consumption record set containing heat flux vector, occupancy number parameter, and course status marker; performing discretization encoding on the heat flux vector, occupancy number parameter, and course status marker in the voxel behavior energy consumption record set to generate an itemset marker set, and constructing an itemset record sequence based on the itemset marker set; performing frequent itemset mining and association weight calculation on the itemset record sequence to obtain the metric values ​​of each itemset, and filling the matrix cells with the metric values ​​according to voxel index and time index to generate the energy loss weight matrix.

[0013] Optionally, the generation of the energy loss weight matrix further includes: extracting the matrix unit values ​​corresponding to each voxel index and time index in the energy loss weight matrix and the number of occupants parameter to construct a voxel behavior energy consumption sample set; performing ratio calculation on each voxel index and time index in the voxel behavior energy consumption sample set to generate a behavior energy consumption ratio sequence.

[0014] Optionally, generating differential thermal texture data includes: aligning the energy loss weight matrix, the heat flux vector field data, and the behavior energy ratio sequence according to voxel indices and time indices, forming a triplet sequence containing matrix unit values, heat flux vector magnitude, and ratio values ​​at each voxel index and time index; performing multiplication operations on the matrix unit values, heat flux vector magnitude, and ratio values ​​in the triplet sequence to generate a mapping intensity value sequence, and performing linear mapping calculations on the mapping intensity value sequence through interval partitioning to obtain color component parameters and transparency parameters; arranging the color component parameters and transparency parameters according to voxel indices and time indices to form differential thermal texture data.

[0015] Optionally, the output visualization data stream includes: aligning the differential thermal texture data with the semantic voxel grid according to the voxel index and time index, writing the corresponding color component parameters and transparency parameters at each voxel unit to generate a semantic voxel grid; performing volume rendering calculation on the semantic voxel grid to generate a three-dimensional rendering frame sequence according to the time index; converting the voxel index in the three-dimensional rendering frame sequence into physical space coordinates based on the bidirectional mapping relationship, and writing the corresponding color component parameters and transparency parameters on the three-dimensional geometric grid to generate a three-dimensional visualization scene image sequence and output it as a visualization data stream.

[0016] Based on the same inventive concept, this invention also provides a dynamic energy consumption traceability and visualization system for educational buildings. The system includes: a semantic voxel grid construction module, used to acquire building information model data and three-dimensional geometric grid data of the educational building, perform spatial discretization partitioning through an octree, construct a semantic voxel grid containing door and window opening / closing parameters and material thermal resistance parameters, and establish a bidirectional mapping relationship between physical space coordinates and voxel indices; a heat flux vector field calculation module, used to collect attitude data from a magnetic angle sensor, fluid data from a wind speed sensor, and thermal image data from an infrared thermal imaging sensor, map the attitude data, fluid data, and thermal image data to the semantic voxel grid, establish an energy conservation discrete equilibrium equation according to material thermal resistance and convective heat transfer, perform time-step iterative calculation, and generate heat flux vector field data; and an occupancy behavior feature generation module, used to acquire student entry and exit records, course schedule data, and indoor occupancy monitoring data, and map the physical space coordinate range to the semantic voxel grid based on the bidirectional mapping relationship. The system comprises a voxel index set, which aggregates student entry / exit records, course schedule data, and indoor occupancy monitoring data according to voxel and time indices to generate an occupancy behavior feature sequence; an energy loss weight matrix generation module, which constructs a time-series tensor from the heat flux vector field data and the occupancy behavior feature sequence according to voxel and time indices, and generates an energy loss weight matrix by performing frequent itemset mining and association weight calculation; a differential thermal texture generation module, which performs interval-based linear mapping calculations on the energy loss weight matrix and the heat flux vector field data at each voxel index to obtain the color component parameters and transparency parameters of each voxel index and time index, and generates differential thermal texture data; and a 3D visualization rendering module, which maps the differential thermal texture data to the semantic voxel mesh, performs volume rendering calculations on the semantic voxel mesh, and generates a 3D visualization scene on the 3D geometric mesh data based on the bidirectional mapping relationship, and outputs a visualization data stream.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] 1. By introducing Building Information Modeling (BIM) and 3D geometric mesh, a semantic voxel mesh containing door and window opening and closing parameters and material thermal resistance parameters is constructed based on octree. At the voxel level, discrete energy conservation equilibrium equations and heat flux vector fields are established to achieve a 3D fine-grained characterization of the energy flow path inside the educational building. Compared with the existing technology that only performs carbon emission interpolation on the planar mesh, it can reflect the impact of differences in the building envelope and construction on the local energy consumption distribution.

[0019] 2. By unifying student entry and exit records, course schedule data, indoor occupancy monitoring data, and heat flux vector field into a unified model in the voxel and time dimensions, and generating an energy loss weight matrix and behavioral energy consumption ratio sequence through frequent itemset mining, and driving differential thermal texture and 3D volume rendering, it is possible to associate occupancy behavior with the spatiotemporal distribution of energy consumption anomalies in a unified scene. Compared with the existing technology that only displays the regional carbon emission intensity, it is more conducive to educational building operation and management personnel to identify high energy consumption spaces and time periods and adjust energy use strategies. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a method for tracing and visualizing dynamic energy consumption in educational buildings, according to an embodiment of the present invention.

[0022] Figure 2 This is a diagram showing the weighted relationship between the number of people occupying the space, heat flux, and energy loss in an embodiment of the present invention.

[0023] Figure 3 This is a linear mapping relationship between the mapping intensity value and visual parameters in an embodiment of the present invention.

[0024] Figure 4 This is a schematic diagram of the structure of a dynamic energy consumption traceability and visualization system for educational buildings according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Reference Figure 1 One embodiment of the present invention proposes a method for tracing and visualizing dynamic energy consumption of educational buildings. It adopts semantic voxel grid and multi-source time series data fusion modeling and correlation mining method, which can correlate energy consumption loss and occupancy behavior, thereby improving the energy consumption diagnosis and visualization analysis capabilities of educational buildings.

[0027] The method described in this embodiment specifically includes:

[0028] The building information model data and three-dimensional geometric mesh data of the educational building are obtained. Spatial discretization is performed through octree to construct a semantic voxel mesh containing door and window opening and closing parameters and material thermal resistance parameters. A bidirectional mapping relationship between physical space coordinates and voxel indexes is established.

[0029] Attitude data from a magnetic angle sensor, fluid data from a wind speed sensor, and thermal image data from an infrared thermal imaging sensor are collected. The attitude data, fluid data, and thermal image data are mapped onto the semantic voxel grid. An energy conservation discrete equilibrium equation is established based on material thermal resistance and convective heat transfer. Time-step iterative calculations are performed to generate heat flux vector field data.

[0030] Acquire student entry and exit records, course schedule data, and indoor occupancy monitoring data. Based on the bidirectional mapping relationship, map the physical space coordinate range into a voxel index set. Then, aggregate the student entry and exit records, course schedule data, and indoor occupancy monitoring data according to the voxel index and time index to generate an occupancy behavior feature sequence.

[0031] The heat flux vector field data and the occupancy behavior feature sequence are used to construct a time series tensor according to the voxel index and time index. By performing frequent itemset mining and association weight calculation, an energy loss weight matrix is ​​generated.

[0032] The energy loss weight matrix and the heat flux vector field data are subjected to a linear mapping calculation based on interval partitioning at each voxel index to obtain the color component parameters and transparency parameters of each voxel index and time index, thereby generating differential thermal texture data.

[0033] The differential thermal texture data is mapped to the semantic voxel mesh, volume rendering calculation is performed on the semantic voxel mesh, and a three-dimensional visualization scene is generated on the three-dimensional geometric mesh data based on the bidirectional mapping relationship, and the visualization data stream is output.

[0034] Optionally, establishing a bidirectional mapping relationship between physical space coordinates and voxel indices includes:

[0035] The building information model data is called to perform attribute parsing on the door and window component objects, extract the geometric boundary parameters, opening and closing control parameters and component material parameters of the doors and windows, and generate a door and window component parameter table;

[0036] In establishing the bidirectional mapping relationship between physical space coordinates and voxel indexes, building information model data is invoked to perform attribute parsing on component objects marked as doors and windows. During the parsing process, the geometric vertex coordinates, component outlines, and opening positions of door and window components in the building coordinate system are read to determine the geometric boundary parameters of doors and windows; parameter fields related to door and window opening and closing, such as opening direction, maximum opening angle, and current control status, are read to determine opening and closing control parameters; parameter fields related to door and window materials, such as frame material type, number of glass layers, and infill material type, are read to determine component material parameters, and the above parameters are written into the door and window component parameter table in row record format.

[0037] Based on the three-dimensional geometric mesh data, octree segmentation is performed on the exterior and interior space of the educational building to determine the voxel unit boundary of each octree node and generate a voxel index set.

[0038] After obtaining the parameter table of door and window components, octree segmentation is performed on the exterior and interior space of the educational building based on the 3D geometric mesh data. During the segmentation process, the side lengths of the building's overall bounding box on the three coordinate axes are calculated, the maximum side length is selected, and the upper limit of the voxel unit side length is calculated based on this maximum side length and the target segmentation level. The target segmentation level is calculated by statistically analyzing the meshing experiment results of existing campus building samples, using building area and building height as inputs to establish a lookup table. During actual operation, the corresponding target segmentation level is read from the lookup table based on the building area and height of the educational building. Then, the total number of mesh patches in the 3D geometric mesh data is counted, and the upper limit of the number of mesh patches allowed to be contained in a single voxel unit is calculated based on the target voxel number upper limit. The target voxel number upper limit is also calculated by statistically analyzing the meshing results of existing campus building samples, determining the corresponding voxel number upper limit according to the building area level, and storing it in a lookup table. Using the overall building bounding box as the root node bounding box, the root node is equally divided into eight sub-bounding boxes along the three coordinate axes. For sub-bounding boxes that still satisfy the conditions of having a bounding box side length greater than the upper limit of voxel element side length and an internal mesh patch count greater than the upper limit of mesh patch count, the same division process is continued until the division conditions are no longer met. For each node where division stops, the minimum bounding box coordinate range of the space covered by that node is calculated and used as the voxel element boundary. A unique voxel index is assigned to each voxel element, thus generating a voxel index set.

[0039] For example, when processing the three-dimensional geometric mesh data of a teaching building, the side lengths of the circumscribed cuboid of the building along the three coordinate axes are first calculated. Assume the maximum side length is... The target segmentation level number is retrieved from the lookup table. ,Will Divide by The upper limit of the voxel element side length is obtained. Then, the total number of mesh patches in the 3D geometric mesh of the teaching building is counted. The maximum number of target voxels that can be retrieved from the lookup table is [number]. ,Will Divide by The upper limit of the number of mesh faces that a single voxel element can contain is obtained. Taking the circumscribed cuboid of the entire teaching building as the root node bounding box, octree partitioning is recursively performed on nodes that satisfy the condition that the side length of the bounding box is greater than the upper limit of the voxel element side length and the number of internal mesh faces is greater than the upper limit of the number of mesh faces. The resulting series of terminal bounding boxes are used as voxel elements, and voxel index sets are generated by numbering them sequentially.

[0040] The door and window geometric boundary parameters in the door and window component parameter table are used to perform spatial overlap judgment with the spatial bounding box coordinates in the voxel index set. The corresponding door and window opening and closing parameters and material thermal resistance parameters are written at the overlapping voxel indexes. A bidirectional mapping relationship between physical space coordinates and voxel indexes is generated based on the spatial bounding box coordinates of the voxel unit.

[0041] After obtaining the parameter table and voxel index set for door and window components, a spatial overlap judgment is performed between the geometric boundary parameters of the door and window components in the parameter table and the spatial bounding box coordinates in the voxel index set. During the spatial overlap judgment process, for each door and window component record, its bounding box in the architectural coordinate system is calculated based on the geometric boundary parameters. Intersection tests are then performed sequentially with the spatial bounding box coordinates of each voxel unit to determine the voxel indexes where intersection occurs. The corresponding door / window opening / closing parameters and material thermal resistance parameters are written at the intersecting voxel indexes, enabling the voxel unit to possess the opening / closing state and thermal characteristics related to doors and windows. Simultaneously, the coordinates of the center point of the spatial bounding box of the voxel unit are used as the physical spatial coordinates of that voxel unit, establishing a mapping relationship from physical spatial coordinates to voxel indexes. This mapping from voxel indexes to physical spatial coordinates is recorded in the reverse index table, thus forming a bidirectional mapping relationship.

[0042] Optionally, the generation of heat flux vector field data includes:

[0043] Time alignment and interpolation are performed on the attitude data, fluid data, and thermal imaging data to generate a sensor data sequence;

[0044] In generating heat flux vector field data, the attitude data, fluid data, and thermal imaging data are first time-aligned and interpolated. For time alignment, the sampling period of the sensor with the smallest sampling period among the three types of sensors is used as the unified time step, and the raw timestamps of all sensors are quantized onto a unified time index. For sensor data lacking observations at a certain time index, linear interpolation based on two adjacent valid sampling points is used to fill in the value at that time index. The interpolated attitude data, wind speed vector, and temperature values ​​are then arranged according to the time index and sensor number, forming an ordered sequence of sensor data by time index.

[0045] Based on the bidirectional mapping relationship, the physical space coordinates of the installation positions of each sensor are mapped to voxel indices. Wind speed vectors and temperature values ​​extracted from the sensor data sequence are written at the voxel indices. Energy conservation discrete equilibrium equations are constructed by combining the material thermal resistance parameters of the semantic voxel mesh and the voxel adjacency relationship.

[0046] After obtaining the sensor data sequence, the physical spatial coordinates of each sensor's installation location are mapped to voxel indices based on a bidirectional mapping relationship. For each time index, the wind speed vector and temperature value corresponding to that time index are read from the sensor data sequence, and these values ​​are written into the voxel index corresponding to the sensor's installation location. For voxel elements in the semantic voxel mesh where sensors are not directly deployed, their adjacent voxel index set is determined based on voxel adjacency relationships. The temperature values ​​written on adjacent voxel elements are combined with material thermal resistance parameters, and an energy conservation discrete equilibrium equation is established for each voxel element using a controlled volume method. This ensures that the temperature change of the voxel element within a time step matches the conductive and convective heat transfer through its boundary. Voxel adjacency relationships are established by selecting six voxel elements in the 3D mesh that share a facet with the current voxel element as the set of adjacent voxel elements, and associating a heat conduction path and the corresponding material thermal resistance parameter between each pair of adjacent voxel elements.

[0047] The energy conservation discrete equilibrium equation is iteratively solved according to the time step, the temperature value of each voxel unit is updated, and the heat flux vector of each voxel unit is calculated based on the temperature difference between adjacent voxel units and the material thermal resistance parameters to generate heat flux vector field data.

[0048] After establishing the discrete equilibrium equation for energy conservation, iterative solutions are performed on this equation using a uniform time step. During each time iteration, the conductive heat flux between each pair of adjacent voxel units is calculated based on the temperature value and material thermal resistance parameters of each voxel unit at the current time index. Convective heat flux is calculated based on the surface temperature value of the voxel unit in contact with the external environment and the wind speed vector. Then, the net heat change of each voxel unit is accumulated based on the conductive and convective heat fluxes, and the temperature value of the voxel unit is updated according to its volume and equivalent specific heat parameters. After the temperature update, the heat flux vector components of each voxel unit are calculated along the three coordinate directions based on the temperature difference between adjacent voxel units, the spatial vector between voxel centers, and the material thermal resistance parameters. The temperature update results and heat flux vector components are stored in the corresponding time index to form heat flux vector field data.

[0049] For example, when performing iterative solutions on the semantic voxel grid of a teaching area, a uniform time step of 120 seconds is selected, dividing the two-hour runtime into 60 time steps. For the calculation within each time step, the wind speed vector and temperature value at the start of that time period are first read from the sensor data sequence and written into the voxel element corresponding to the sensor installation location. Then, based on the current temperature of each voxel element and the temperatures of adjacent voxel elements, the net conductive heat and convective heat transfer of each voxel element within that time step are calculated using the material thermal resistance parameters, and the voxel element temperature is updated. After the temperature update is completed, the heat flux component along the direction of the line connecting the center points of adjacent voxel elements is calculated using the temperature difference between them and the corresponding material thermal resistance parameters. The heat flux components in the three coordinate directions are then combined into a heat flux vector. By repeating the above process over all time steps, heat flux vector field data covering the entire teaching area and the entire calculation period can be obtained.

[0050] Optionally, the construction of the discrete energy conservation equilibrium equation includes:

[0051] Based on the voxel adjacency relationship in the semantic voxel grid, a set of adjacent voxel indices is determined for each voxel index, and the material thermal resistance parameter is associated with the set of adjacent voxel indices at each voxel index.

[0052] In constructing the discrete equilibrium equations for energy conservation, the existing spatial partitioning results in the semantic voxel mesh are first used to determine the set of neighboring voxel indices for each voxel index. To this end, based on the spatial bounding box of each voxel element, other voxel elements sharing a common surface with that bounding box in the 3D coordinate system are retrieved. These voxel elements sharing a common surface are considered neighboring voxel elements, and their voxel indices are recorded, forming the set of neighboring voxel indices corresponding to the current voxel index. Simultaneously, the material thermal resistance parameter corresponding to the current voxel element is associated with this set of neighboring voxel indices, so that when discretizing the heat transfer relationship later, the material thermal resistance parameter corresponding to each heat transfer path can be directly called.

[0053] Based on the temperature value at the voxel index, the set of adjacent voxel indices, and the material thermal resistance parameter, an energy conservation discrete equilibrium equation is established at each voxel index.

[0054] After determining the set of adjacent voxel indices corresponding to each voxel index and completing the material thermal resistance parameter association, an energy conservation discrete equilibrium equation is established at each voxel index based on the temperature value at the voxel index, the set of adjacent voxel indices, and the material thermal resistance parameter. When establishing the energy conservation discrete equilibrium equation, a control volume method is used, treating each voxel unit as a control volume. For each voxel unit, within a given time step, the heat transferred through conduction from the common surface connecting adjacent voxel units and the convective heat transferred with the ambient fluid are accumulated to form the total heat entering and leaving the voxel unit; simultaneously, the energy change inside the voxel unit is calculated based on the temperature change, volume, and equivalent specific heat parameter. By setting the total heat entering the voxel unit minus the total heat leaving the voxel unit equal to the energy change inside the voxel unit, an energy conservation discrete equilibrium equation is obtained at each voxel index. For voxel units that are in direct contact with the outside air, the equivalent convective heat transfer parameters are determined based on empirical heat transfer correlations using the wind speed vector provided by the wind speed sensor and the temperature value provided by the infrared thermal imaging sensor. These convective heat transfer parameters, along with the voxel unit surface temperature, are then substituted into the convective heat transfer term in the energy conservation discrete equilibrium equation.

[0055] For example, when solving the discrete energy conservation equation for the exterior wall region of a single classroom, the thickness direction of the exterior wall is divided into several voxel elements. For voxel elements adjacent to the interior side, their adjacent voxel index set includes voxel element indices pointing to the interior air and indices pointing to voxel elements inside the wall. In the discrete energy conservation equation for this voxel element, a portion of the conductive heat is calculated using the material thermal resistance parameter and the temperature difference between the voxel elements inside the wall, while the other portion of the convective heat transfer is calculated using the classroom air temperature, wall surface temperature, and equivalent convective heat transfer parameters. For voxel elements adjacent to the exterior side, their discrete energy conservation equation includes a conduction term pointing to the voxel elements inside the wall and a convective heat transfer term pointing to the exterior air. The exterior air temperature and wind speed are provided by sensors installed near the exterior wall. By establishing the discrete energy conservation equation of the above form at all voxel indices, a complete set of discrete equations is provided for iterative solution at the time step.

[0056] Optionally, the generation of the occupancy behavior feature sequence includes:

[0057] Perform time alignment and deduplication on student entry and exit records, course schedule data and indoor number monitoring data to generate student entry and exit event sequences, course schedule event sequences and number of people statistics sequences;

[0058] In generating the occupancy behavior feature sequence, the student entry / exit records, course schedule data, and indoor occupancy monitoring data are first time-aligned and deduplicated. During time alignment, the data with the highest time resolution among the three types is selected as the unified time step, and all original timestamps are mapped to discrete time indices by rounding down. For records of the same student ID repeatedly swiping their card on the same access control device within the same time index, only the first valid record is retained according to the timestamp order to obtain the student entry / exit event sequence. For course schedule data, the course start and end times are mapped to start and end time indices according to the teaching space identifier and start / end time, forming the course schedule event sequence. For indoor occupancy monitoring data, the sampling timestamps are mapped to time indices, and multiple sensor readings under the same time index are merged, for example, by taking the arithmetic mean or the maximum value, to obtain the occupancy statistics sequence.

[0059] For example, in a certain teaching building, student entry and exit records are stored in the form of access control card swipe records, including student ID, card swipe time, and access control number; course schedule data includes course number, classroom, class date, and session information; indoor attendance monitoring data is periodically output by infrared counting devices installed at classroom entrances, including timestamps and current count values. For time alignment, one minute is selected as the uniform time step, mapping the original timestamps of the three types of data to time indices at one-minute intervals; for multiple consecutive card swipes by the same student at the same access control device, only the first successful swipe is retained; course session information is converted into start and end time indices through a pre-configured session schedule; the count values ​​from multiple infrared counting devices at the same time index are taken as the arithmetic mean to form the attendance statistics for the corresponding time index.

[0060] Based on the bidirectional mapping relationship, the physical space coordinate range of each teaching space is mapped to a set of voxel indices. The student entry and exit event sequence, the course arrangement event sequence and the number of people statistics sequence are aggregated according to the voxel index and time index. Each voxel index and time index is calculated to generate the number of people occupied and the course status marker.

[0061] After obtaining the student entry / exit event sequence, course schedule event sequence, and headcount sequence, the physical spatial coordinate range of each teaching space is mapped to a voxel index set based on a bidirectional mapping relationship. The physical spatial coordinate range of the teaching space is obtained through the classroom geometric boundary parameters in the Building Information Model (BIM). For each teaching space, the 3D bounding box of the teaching space in the building coordinate system is determined using the geometric outer contour coordinates recorded in the BIM. Then, voxel units that spatially overlap with the 3D bounding box are queried through the bidirectional mapping relationship, and the corresponding voxel indexes are added to the voxel index set of the teaching space. Subsequently, under each time index, the voxel index set belonging to a certain teaching space is traversed. The student entry / exit event sequence and headcount sequence corresponding to that teaching space are aggregated according to the time index. By counting and accumulating the entry and exit events or using the count value of the headcount sequence, the occupied headcount parameter corresponding to each voxel index and time index is calculated. At the same time, based on whether there is a valid course schedule and course type within the time index according to the course schedule event sequence, a corresponding course status flag is generated. The course status flags are encoded according to states such as in class, vacant, and self-study to reflect the usage status of the teaching space under different time indices.

[0062] For example, a teaching space identified as "Classroom A" has a clearly defined room number and geometric boundary coordinates in the Building Information Model (BIM). Through a bidirectional mapping relationship, a set of voxel indices intersecting with the bounding box of Classroom A's geometric boundary are obtained, for example, voxel indices from 100 to 150. For the tenth time index of a certain day, if the course schedule event sequence shows that a "Mathematical Analysis" course exists in Classroom A at that time index, the course status is set to "In Class"; simultaneously, if the student entry / exit event sequence shows that three students entered the classroom at the previous time index and no students left, and the headcount sequence shows that the current count is thirty, then thirty is used as the occupied headcount parameter for Classroom A at that time index, and this occupied headcount parameter is copied and written to the records corresponding to voxel indices 100 to 150.

[0063] The number of people occupying the course and the course status marker are sorted by voxel index and time index to generate an occupation behavior feature sequence.

[0064] After obtaining the number of occupants and course status markers corresponding to each voxel index and time index, the occupants and course status markers are sorted by voxel index and time index to generate an occupancy behavior feature sequence. During the sorting process, the voxel index is used as the first sorting key and the time index as the second sorting key, arranging all records from smallest to largest voxel index, and within the same voxel index, from smallest to largest time index. For each combination of voxel index and time index, an occupancy behavior feature vector containing the number of occupants and course status markers is constructed, and these occupancy behavior feature vectors are written into the storage structure in sorted order to form the occupancy behavior feature sequence. The occupancy behavior feature sequence provides the input data foundation for subsequent construction of time-series tensors and the execution of frequent itemset mining.

[0065] Optionally, the generation of the energy loss weight matrix includes:

[0066] Align the heat flux vector field data with the occupancy behavior feature sequence according to the voxel index and time index to generate a voxel behavior energy consumption record set containing heat flux vector, occupancy number parameter and course status marker;

[0067] In generating the energy loss weight matrix, the heat flux vector field data and the occupancy behavior feature sequence are first aligned using voxel indices and time indices. During alignment, the voxel indices in the semantic voxel grid are used as spatial indices, and the time indices obtained by dividing the data with a uniform time step are used as time axes. For each combination of voxel and time indices, the corresponding heat flux vector is read from the heat flux vector field data, and the corresponding number of occupants and course status markers are read from the occupancy behavior feature sequence to form a record. By performing the above operations on all combinations of voxel and time indices, a set of voxel behavior energy consumption records consisting of multiple records is obtained. Each record contains at least a voxel index, a time index, a heat flux vector, a number of occupants parameter, and a course status marker.

[0068] The heat flux vector, the number of people occupying the voxel behavior energy consumption record set, and the course status marker are discretized and encoded to generate an itemset marker set, and an itemset record sequence is constructed based on the itemset marker set;

[0069] After obtaining the voxel behavior energy consumption record set, the heat flux vector, occupancy parameter, and course status marker in the voxel behavior energy consumption record set are discretized and encoded to generate an itemset marker set. For the heat flux vector, the magnitude of the three-dimensional heat flux vector is calculated, and the heat flux vector magnitude is divided into several continuous intervals according to the numerical range. For example, the entire sample is divided into several intervals based on the quantile of the heat flux vector magnitude sample, and a unique heat flux level marker is assigned to each interval. For the occupancy parameter, the occupancy value is divided into several intervals according to the classroom's designed capacity. For example, it is divided into low occupancy interval, medium occupancy interval, and high occupancy interval by percentage, and a unique occupancy level marker is assigned to each interval. For the course status marker, status texts such as "in class," "empty," and "studying" are used directly, or these statuses are mapped to integer codes. For each record in the voxel behavior energy consumption record set, the corresponding heat flux level label, number of occupants level label and course status label are combined together to form an itemset label. All itemset labels are collected into an itemset label set, and then arranged according to time index order or record order to form an itemset record sequence.

[0070] Frequent itemset mining and association weight calculation are performed on the itemset record sequence to obtain the metric values ​​of each itemset. The matrix cells are then filled with the metric values ​​according to the voxel index and time index to generate the energy loss weight matrix.

[0071] After constructing the itemset record sequence, frequent itemset mining and association weight calculation are performed on the itemset record sequence to generate an energy loss weight matrix. Frequent itemset mining uses a frequent pattern growth algorithm to scan the itemset record sequence and determine the total number of records in the itemset record sequence, denoted as [the number of records is missing here]. Then, a minimum support count threshold is set based on the length of the itemset record sequence. For example, the minimum support count threshold is... Set it to one percent of the total number of records, that is:

[0072] ,

[0073] For each candidate itemset, count the number of records appearing in the itemset record sequence, and denote this number as . ,like Not less than If a candidate itemset is selected, it is added to the frequent itemset set, and the support of that frequent itemset is calculated. The support of a frequent itemset is calculated using the following formula:

[0074] ,

[0075] in, Indicates the first The support of each frequent itemset is a dimensionless value. This indicates that the itemset record sequence contains the first... The number of records in each itemset is a non-negative integer; This represents the total number of records in the itemset record sequence, and is a positive integer. After obtaining the support for each frequent itemset, the support is combined with the heat flux vector and occupancy parameters in the corresponding record to calculate the energy loss association weight. For voxel indexes... Time index is Let the record be the heat flux vector corresponding to the record. The magnitude of the heat flux vector is The corresponding number of occupants is: The frequent itemset to which this record belongs is the [number]th [item set]. A frequent itemset with a support of A reference heat flux modulus is obtained by statistically analyzing the heat flux vector modulus samples in all records. The reference heat flux modulus is taken as the 95th percentile value of the heat flux vector modulus sample, meaning that 95% of the sample values ​​in all heat flux vector modulus samples are not greater than the reference heat flux modulus. A reference occupancy parameter is obtained by statistically analyzing the occupancy parameter samples in all records. The maximum number of occupants is taken from the entire sample as the reference occupancy parameter. Therefore, the voxel index is defined as follows: Time index is The corresponding energy loss association weight of the record for:

[0076] ,

[0077] in, voxel index Time Index The corresponding energy loss correlation weight is a dimensionless value. The support of the frequent itemset to which the record belongs reflects the frequency of the combination of heat flux level, number of occupants level and course status in all records. is the heat flux vector magnitude, which physically represents the heat flux density per unit area; The reference heat flux modulus is used to normalize the heat flux vector modulus; The "Number of Occupants" parameter represents the number of students in the teaching space corresponding to the current voxel index and time index. To normalize the occupancy parameter, the upper limit of the designed capacity of the teaching space or floor, or the maximum number of occupants in the collected samples, is taken. The physical meaning of this formula is that when a record corresponding to a voxel index and a time index simultaneously satisfies the characteristics of frequent occurrence, high heat flux modulus, and high occupancy, its energy loss correlation weight increases, indicating that the space unit is more likely to experience energy loss related to occupancy behavior during that time period.

[0078] For example, suppose that in the itemset record sequence of a certain teaching building, the total number of records is... For 10,000 entries, the minimum support count threshold can be obtained according to the aforementioned rules. For a given frequent itemset, if the frequent itemset appears 500 times in the sequence of itemset records, then the support of the frequent itemset is... For a record with a voxel index of 100 and a time index of 300, assume the corresponding heat flux vector magnitude is... Reference heat flux modulus Number of people occupying Reference number of occupants The energy loss associated with this record is weighted. The calculation process is as follows: First, calculate the heat flux normalization factor. Calculate the normalization factor for the number of people occupying the space. Substitute Therefore, the energy loss correlation weights corresponding to the voxel index and the time index can be obtained. The value is 0.03. Finally, the energy loss corresponding to each voxel index and time index is associated with a weight. By filling in matrix cells with voxel indices as row indices and time indices as column indices, an energy loss weight matrix can be generated, providing the input data foundation for subsequent differential thermal texture generation and energy consumption source tracing analysis of educational buildings, such as... Figure 2 As shown, the energy loss correlation weight is determined by the normalized number of occupants, the normalized heat flux modulus, and the support of frequent itemsets. The position of the bubbles in the figure represents the level of occupancy and heat flux, the size of the bubbles represents the support, and the shade of gray of the bubbles reflects the final calculated weight; the darker the color, the higher the correlation of energy loss in that area.

[0079] Optionally, the generation of the energy loss weight matrix further includes:

[0080] Extract the matrix unit values ​​corresponding to each voxel index and time index in the energy loss weight matrix, as well as the number of people occupying the energy, and construct a voxel behavior energy consumption sample set.

[0081] In constructing the voxel behavior energy consumption sample set, the row index of the energy loss weight matrix is ​​first used as the voxel index, and the column index is used as the time index. For each voxel index and time index combination, the energy loss associated weight value in the corresponding matrix cell is read. Simultaneously, the number of occupants corresponding to the voxel index and time index is retrieved from the occupancy behavior feature sequence. The voxel index, time index, energy loss associated weight value, and number of occupants parameter are combined into a voxel behavior energy consumption sample record, and each voxel behavior energy consumption sample record is added to the voxel behavior energy consumption sample set.

[0082] For example, for a voxel index of 100 and a time index of 300, the matrix cell value 0.03 is read from the energy loss weight matrix as the energy loss association weight value corresponding to the voxel index and the time index, and the number of occupants parameter 45 is read from the occupancy behavior feature sequence. A voxel behavior energy consumption sample record is formed by voxel index 100, time index 300, energy loss association weight value 0.03, and number of occupants parameter 45, and this sample record is added to the voxel behavior energy consumption sample set. By repeating the above steps for all voxel indexes and time index combinations, a voxel behavior energy consumption sample set covering the entire semantic voxel grid and all time indices is obtained.

[0083] The ratio of each voxel index to the time index in the voxel behavior energy consumption sample set is calculated to generate a behavior energy consumption ratio sequence.

[0084] After obtaining the voxel-based behavioral energy consumption sample set, the ratio of each voxel index to the time index in the voxel-based behavioral energy consumption sample set is calculated to generate a behavioral energy consumption ratio sequence. For each voxel-based behavioral energy consumption sample record, the energy loss associated weight value and the number of occupants parameter are read. When the number of occupants parameter is greater than zero, the energy loss associated weight value is divided by the number of occupants parameter to obtain the behavioral energy consumption ratio corresponding to that voxel index and time index; when the number of occupants parameter is equal to zero, the behavioral energy consumption ratio corresponding to that voxel index and time index is set to zero. Then, the behavioral energy consumption ratios corresponding to each voxel index and time index are arranged in the order of voxel index and time index and written into the behavioral energy consumption ratio sequence, so that the behavioral energy consumption ratio sequence and the energy loss weight matrix are consistent in both voxel index and time index dimensions.

[0085] Optionally, the generation of differential thermal texture data includes:

[0086] Align the energy loss weight matrix, the heat flux vector field data, and the behavior energy ratio sequence according to the voxel index and time index, and form a triplet sequence containing the matrix unit value, the heat flux vector magnitude, and the ratio value at each voxel index and time index.

[0087] In generating differential thermal texture data, the voxel index in the semantic voxel mesh is first used as the row index, and the time index obtained by dividing the data with a unified time step is used as the column index. The energy loss weight matrix, heat flux vector field data, and behavioral energy ratio sequence are then aligned. For each combination of voxel index and time index, the corresponding matrix element value is read from the energy loss weight matrix and used as the energy loss associated weight value. The three-dimensional heat flux vector of the corresponding voxel element under that time index is read from the heat flux vector field data, and the heat flux vector magnitude is calculated based on the components in the three coordinate directions. The corresponding behavioral energy ratio value is read from the behavioral energy ratio sequence. The energy loss associated weight value, heat flux vector magnitude, and behavioral energy ratio value are combined into a triplet record and written into the triplet sequence in the order of voxel index and time index.

[0088] For example, for a voxel cell with a voxel index of 100 and a time index of 300, the matrix cell value, for example, 0.03, is read from the energy loss weight matrix as the energy loss associated weight value for that voxel cell; the heat flux components of that voxel cell in the three coordinate directions are read from the heat flux vector field data, and the heat flux vector magnitude is calculated accordingly, for example, a value of 240; the behavioral energy ratio value of that voxel cell at that time index is read from the behavioral energy ratio sequence, for example, an order of magnitude of [missing value]. The ratio values. These three values ​​are combined into a triplet record and stored in the triplet sequence in association with the voxel index 100 and the time index 300.

[0089] Multiplication is performed on the matrix unit values, heat flux vector magnitude and ratio values ​​in the triplet sequence to generate a mapping intensity value sequence. The mapping intensity value sequence is then subjected to linear mapping calculation through interval division to obtain color component parameters and transparency parameters.

[0090] After obtaining the triplet sequence, multiplication is performed on the matrix unit values, heat flux vector magnitude, and behavior energy consumption ratio values ​​in the triplet sequence to generate a mapping intensity value sequence. The mapping intensity value is understood as a comprehensive scalar index representing the degree of energy loss, heat flux magnitude, and the degree of influence per unit of occupied behavior under the same voxel index and time index. The mapping intensity values ​​corresponding to all voxel indices and time indices are statistically analyzed to obtain the minimum and maximum values. This value range can be divided into multiple continuous intervals, such as equal-width intervals, based on the distribution of the mapping intensity values. For each mapping intensity value, the interval it falls into is determined, and linear interpolation is performed proportionally between the interval's start and end values, mapping this proportion to the values ​​of the color component parameters and transparency parameters.

[0091] For example, when statistically analyzing the mapping intensity values ​​of a certain teaching area, the range of all mapping intensity values ​​is divided into five continuous intervals. For mapping intensity values ​​falling into the lowest interval, they are linearly mapped to color component parameters and transparency parameters with higher blue components, lower red components, and transparency values ​​close to 0.2; for mapping intensity values ​​falling into the highest interval, they are linearly mapped to color component parameters and transparency parameters with higher red components, lower blue components, and transparency values ​​close to 0.8; for mapping intensity values ​​in the middle interval, color component parameters and medium transparency are obtained according to their interval position ratio, falling between cool and warm tones. In this way, the color component parameters and transparency parameters continuously change numerically and reflect the magnitude of the mapping intensity values, such as... Figure 3 As shown, this diagram illustrates the mapping relationship between the calculated mapping intensity values ​​and the visual grayscale and transparency parameters required for the final visualization. The trend of the lines in the diagram reflects the linear mapping logic: low intensity values ​​are mapped to dark grayscale and low transparency, while high intensity values ​​are mapped to light grayscale and high transparency.

[0092] The color component parameters and the transparency parameters are arranged according to the voxel index and the time index to form differential thermal texture data.

[0093] After calculating the color component parameters and transparency parameters, these parameters are arranged according to voxel index and time index to form differential thermal texture data. During arrangement, the voxel index is used as the first index dimension and the time index as the second index dimension. The color component parameters and transparency parameters corresponding to each voxel index and time index are encoded into four-channel or three-channel texture unit values ​​in a fixed order. For example, the red component, green component, blue component, and transparency parameter are stored sequentially in the same texture unit. All texture units are written to the storage structure sequentially according to the traversal order of the voxel index and time index, forming differential thermal texture data that can be directly loaded by the graphics rendering module. This differential thermal texture data is aligned and mapped with the semantic voxel mesh during subsequent rendering, thus displaying the differences in energy consumption and behavioral impact at different spatial locations and time periods in the form of color and transparency variations in the 3D visualization scene.

[0094] Optionally, the output visualization data stream includes:

[0095] Align the differential thermal texture data with the semantic voxel grid according to the voxel index and time index, and write the corresponding color component parameters and transparency parameters at each voxel unit to generate the semantic voxel grid.

[0096] In the process of outputting the visualization data stream, the differential thermal texture data is first aligned with the semantic voxel mesh using voxel indices and temporal indices. For each combination of voxel index and temporal index, the corresponding color component parameters and transparency parameters are read from the differential thermal texture data. The color component parameters are written into the color attribute of that voxel cell in the semantic voxel mesh, and the transparency parameters are written into the opacity attribute of that voxel cell, so that each voxel cell has explicit color component parameters and transparency parameters under different temporal indices. This yields a semantic voxel mesh with color component parameters and transparency parameters, providing input for volumetric rendering calculations.

[0097] Perform volume rendering calculations on the semantic voxel mesh and generate a 3D rendering frame sequence according to the time index;

[0098] After obtaining the semantic voxel mesh with color component parameters and transparency parameters, volumetric rendering calculations are performed on the semantic voxel mesh. During volumetric rendering calculations, a sampling ray is emitted for each screen pixel according to the current viewing angle. Sampling is performed at a fixed step size within the spatial intersection region of the sampling ray and the semantic voxel mesh. At each sampling point, the color component parameters and transparency parameters of the corresponding voxel unit are read, and forward cumulative blending is performed sequentially from the front end to the back end of the ray, superimposing the color component parameters and transparency parameters into the color buffer of the current pixel. By performing volumetric rendering calculations on all screen pixels, a 3D rendered image is generated at each time index. Arranging the 3D rendered images corresponding to all time indices in time index order yields the 3D rendered frame sequence.

[0099] For example, when performing volumetric rendering on a teaching area, a screen resolution of 1024 x 1024 is selected. For each screen pixel, a sampling ray is emitted into the semantic voxel grid along the viewing direction, with the sampling step size set to half the minimum side length of the voxel unit. When the sampling ray passes through a high-energy-dissipation leakage area, because the color component parameters of the voxel units in that area are mapped to bright warm tones, the forward cumulative blending process will cause that area to appear as a distinct bright color block in the rendered image. As the time index changes, the above volumetric rendering process is repeated for all time indices to obtain a 3D rendering frame sequence reflecting the energy consumption evolution process.

[0100] Based on the bidirectional mapping relationship, the voxel indexes in the 3D rendering frame sequence are converted into physical space coordinates, and the corresponding color component parameters and transparency parameters are written on the 3D geometric mesh to generate a 3D visualization scene image sequence and output it as a visualization data stream.

[0101] After obtaining the 3D rendering frame sequence, the voxel indices in the sequence are converted into physical space coordinates based on a bidirectional mapping relationship, and the corresponding color component parameters and transparency parameters are written into the 3D geometric mesh data. During the conversion process, for each visible voxel index in each frame of the rendered image, the bidirectional mapping relationship is used to query the physical space coordinates corresponding to that voxel index. Furthermore, the mesh facets or vertices in the 3D geometric mesh data that intersect with those physical space coordinates are queried, and the color component parameters and transparency parameters of that voxel index under that time index are written into the corresponding mesh vertex color attribute or fragment color attribute. To make the superposition of the building material color and the energy consumption tracing color more intuitive, a linear blending of the building base color and the differential thermal color can be performed during rendering. Therefore, the following color blending formula is introduced:

[0102] ,

[0103] in, This represents the final output color vector written onto the three-dimensional geometric mesh, which can be understood as a three-dimensional color vector containing red, green, and blue components. Voxel index In time index The corresponding transparency parameter is a real number between zero and one, where zero represents complete transparency and one represents complete opacity. This transparency parameter comes from differential thermal texture data. This represents the base color vector of the building body at the corresponding physical location. The base color vector is determined based on the material information of components such as walls, doors and windows recorded in the building information model. For example, concrete walls are set to grayish and glass curtain walls are set to bluish. Voxel index In time index The color vector formed by the color component parameters given in the lower differential thermal texture data is used to reflect the energy consumption intensity of that location during that time period. The physical meaning of this formula is that it linearly interpolates the base color of the building and the thermal color obtained from energy consumption tracing according to the transparency parameter. The smaller the transparency parameter, the closer the output color is to the base color of the building; the larger the transparency parameter, the closer the output color is to the thermal color, thus visually achieving a highlight overlay display of high-energy-consumption areas. After the color writing is completed, all 3D visualization scene images under all time indices are arranged in ascending order of time index and encapsulated at a fixed frame rate to form a continuously playable visualization data stream. For example, setting the number of time indices per second to thirty, encapsulating thirty consecutive frames of 3D visualization scene images into a one-second video clip, and sequentially connecting all 3D visualization scene images within the entire operating cycle, a visualization data stream continuously displaying the energy consumption tracing results of educational buildings in the time dimension can be obtained, allowing maintenance personnel to browse, play back, and analyze it on terminal devices.

[0104] For example, for a certain combination of voxel index and time index, assume the transparency parameter The base color vector of the building body is 0.75. The red, green, and blue components are 0.5, 0.5, and 0.5 respectively, representing the differential thermal color vector. The red, green, and blue components are 1, 0.2, and 0.1, respectively. According to the above formula, the red component of the output color vector is... The green component is The blue component is As can be seen, the output color vector is clearly biased towards red and has high brightness, which can clearly show the high energy consumption state at this location in the 3D visualization scene image.

[0105] Based on the same inventive concept, this invention also provides a dynamic energy consumption traceability and visualization system for educational buildings, such as... Figure 4 As shown, the system includes:

[0106] The semantic voxel mesh construction module is used to acquire the building information model data and three-dimensional geometric mesh data of educational buildings. It performs spatial discretization partitioning through an octree, constructs a semantic voxel mesh containing door and window opening and closing parameters and material thermal resistance parameters, and establishes a bidirectional mapping relationship between physical space coordinates and voxel indices.

[0107] The heat flux vector field calculation module is used to collect attitude data from the magnetic angle sensor, fluid data from the wind speed sensor, and thermal image data from the infrared thermal imaging sensor. It maps the attitude data, fluid data, and thermal image data to the semantic voxel grid, establishes an energy conservation discrete equilibrium equation according to the material thermal resistance and convective heat transfer, performs time-step iterative calculation, and generates heat flux vector field data.

[0108] The occupancy behavior feature generation module is used to acquire student entry and exit records, course schedule data and indoor number of people monitoring data. Based on the bidirectional mapping relationship, the physical space coordinate range is mapped to a set of voxel indices. The student entry and exit records, course schedule data and indoor number of people monitoring data are aggregated according to the voxel index and time index to generate an occupancy behavior feature sequence.

[0109] The energy loss weight matrix generation module is used to construct a time series tensor by combining the heat flux vector field data and the occupancy behavior feature sequence according to the voxel index and time index, and generate the energy loss weight matrix by performing frequent itemset mining and association weight calculation.

[0110] The differential thermal texture generation module is used to perform a linear mapping calculation based on interval partitioning on the energy loss weight matrix and the heat flux vector field data at each voxel index to obtain the color component parameters and transparency parameters of each voxel index and time index, and generate differential thermal texture data.

[0111] The 3D visualization rendering module is used to map the differential thermal texture data to the semantic voxel mesh, perform volume rendering calculations on the semantic voxel mesh, generate a 3D visualization scene on the 3D geometric mesh data based on the bidirectional mapping relationship, and output a visualization data stream.

[0112] It should be noted that the functional division and information interaction between the various modules described above are logical, but in terms of physical implementation, they can be integrated on the same software platform or deployed in a distributed manner. The connections between them represent data flow and control flow, aiming to collaboratively achieve the objectives of this invention. The above descriptions are merely exemplary embodiments of this invention and should not be construed as limiting the scope of protection of this invention.

Claims

1. A method for tracing and visualizing dynamic energy consumption in educational buildings, characterized in that, The method includes: The building information model data and three-dimensional geometric mesh data of the educational building are obtained. Spatial discretization is performed through octree to construct a semantic voxel mesh containing door and window opening and closing parameters and material thermal resistance parameters. A bidirectional mapping relationship between physical space coordinates and voxel indexes is established. Attitude data from a magnetic angle sensor, fluid data from a wind speed sensor, and thermal image data from an infrared thermal imaging sensor are collected. The attitude data, fluid data, and thermal image data are mapped onto the semantic voxel grid. An energy conservation discrete equilibrium equation is established based on material thermal resistance and convective heat transfer. Time-step iterative calculations are performed to generate heat flux vector field data. Acquire student entry and exit records, course schedule data, and indoor occupancy monitoring data. Based on the bidirectional mapping relationship, map the physical space coordinate range into a voxel index set. Then, aggregate the student entry and exit records, course schedule data, and indoor occupancy monitoring data according to the voxel index and time index to generate an occupancy behavior feature sequence. The heat flux vector field data and the occupancy behavior feature sequence are used to construct a time series tensor according to the voxel index and time index. By performing frequent itemset mining and association weight calculation, an energy loss weight matrix is ​​generated. The energy loss weight matrix and the heat flux vector field data are subjected to a linear mapping calculation based on interval partitioning at each voxel index to obtain the color component parameters and transparency parameters of each voxel index and time index, thereby generating differential thermal texture data. The differential thermal texture data is mapped to the semantic voxel mesh, volume rendering calculation is performed on the semantic voxel mesh, and a three-dimensional visualization scene is generated on the three-dimensional geometric mesh data based on the bidirectional mapping relationship, and the visualization data stream is output.

2. The method for dynamic energy consumption tracing and visualization of educational buildings according to claim 1, characterized in that, The establishment of a bidirectional mapping relationship between physical space coordinates and voxel indices includes: The building information model data is called to perform attribute parsing on the door and window component objects, extract the geometric boundary parameters, opening and closing control parameters and component material parameters of the doors and windows, and generate a door and window component parameter table; Based on the three-dimensional geometric mesh data, octree segmentation is performed on the exterior and interior space of the educational building to determine the voxel unit boundary of each octree node and generate a voxel index set. The door and window geometric boundary parameters in the door and window component parameter table are used to perform spatial overlap judgment with the spatial bounding box coordinates in the voxel index set. The corresponding door and window opening and closing parameters and material thermal resistance parameters are written at the overlapping voxel indexes. A bidirectional mapping relationship between physical space coordinates and voxel indexes is generated based on the spatial bounding box coordinates of the voxel unit.

3. The method for dynamic energy consumption tracing and visualization of educational buildings according to claim 1, characterized in that, The generated heat flux vector field data includes: Time alignment and interpolation are performed on the attitude data, fluid data, and thermal imaging data to generate a sensor data sequence; Based on the bidirectional mapping relationship, the physical space coordinates of the installation positions of each sensor are mapped to voxel indices. Wind speed vectors and temperature values ​​extracted from the sensor data sequence are written at the voxel indices. Energy conservation discrete equilibrium equations are constructed by combining the material thermal resistance parameters of the semantic voxel mesh and the voxel adjacency relationship. The energy conservation discrete equilibrium equation is iteratively solved according to the time step, the temperature value of each voxel unit is updated, and the heat flux vector of each voxel unit is calculated based on the temperature difference between adjacent voxel units and the material thermal resistance parameters to generate heat flux vector field data.

4. The method for dynamic energy consumption tracing and visualization of educational buildings according to claim 3, characterized in that, The construction of the energy conservation discrete equilibrium equation includes: Based on the voxel adjacency relationship in the semantic voxel grid, a set of adjacent voxel indices is determined for each voxel index, and the material thermal resistance parameter is associated with the set of adjacent voxel indices at each voxel index. Based on the temperature value at the voxel index, the set of adjacent voxel indices, and the material thermal resistance parameter, an energy conservation discrete equilibrium equation is established at each voxel index.

5. The method for dynamic energy consumption tracing and visualization of educational buildings according to claim 1, characterized in that, The generated occupancy behavior feature sequence includes: Perform time alignment and deduplication on student entry and exit records, course schedule data and indoor number monitoring data to generate student entry and exit event sequences, course schedule event sequences and number of people statistics sequences; Based on the bidirectional mapping relationship, the physical space coordinate range of each teaching space is mapped to a set of voxel indices. The student entry and exit event sequence, the course arrangement event sequence and the number of people statistics sequence are aggregated according to the voxel index and time index. Each voxel index and time index is calculated to generate the number of people occupied and the course status marker. The number of people occupying the course and the course status marker are sorted by voxel index and time index to generate an occupation behavior feature sequence.

6. The method for dynamic energy consumption tracing and visualization of educational buildings according to claim 1, characterized in that, The generated energy loss weight matrix includes: Align the heat flux vector field data with the occupancy behavior feature sequence according to the voxel index and time index to generate a voxel behavior energy consumption record set containing heat flux vector, occupancy number parameter and course status marker; The heat flux vector, the number of people occupying the voxel behavior energy consumption record set, and the course status marker are discretized and encoded to generate an itemset marker set, and an itemset record sequence is constructed based on the itemset marker set; Frequent itemset mining and association weight calculation are performed on the itemset record sequence to obtain the metric values ​​of each itemset. The matrix cells are then filled with the metric values ​​according to the voxel index and time index to generate the energy loss weight matrix.

7. The method for dynamic energy consumption tracing and visualization of educational buildings according to claim 5, characterized in that, The generated energy loss weight matrix also includes: Extract the matrix unit values ​​corresponding to each voxel index and time index in the energy loss weight matrix, as well as the number of people occupying the energy, and construct a voxel behavior energy consumption sample set. The ratio of each voxel index to the time index in the voxel behavior energy consumption sample set is calculated to generate a behavior energy consumption ratio sequence.

8. The method for dynamic energy consumption tracing and visualization of educational buildings according to claim 7, characterized in that, The generated differential thermal texture data includes: Align the energy loss weight matrix, the heat flux vector field data, and the behavior energy ratio sequence according to the voxel index and time index, and form a triplet sequence containing the matrix unit value, the heat flux vector magnitude, and the ratio value at each voxel index and time index. Multiplication is performed on the matrix unit values, heat flux vector magnitude and ratio values ​​in the triplet sequence to generate a mapping intensity value sequence. The mapping intensity value sequence is then subjected to linear mapping calculation through interval division to obtain color component parameters and transparency parameters. The color component parameters and the transparency parameters are arranged according to voxel index and time index to form differential thermal texture data.

9. The method for dynamic energy consumption tracing and visualization of educational buildings according to claim 1, characterized in that, The output visualization data stream includes: Align the differential thermal texture data with the semantic voxel grid according to the voxel index and time index, and write the corresponding color component parameters and transparency parameters at each voxel unit to generate the semantic voxel grid. Perform volume rendering calculations on the semantic voxel mesh and generate a 3D rendering frame sequence according to the time index; Based on the bidirectional mapping relationship, the voxel indexes in the 3D rendering frame sequence are converted into physical space coordinates, and the corresponding color component parameters and transparency parameters are written on the 3D geometric mesh to generate a 3D visualization scene image sequence and output it as a visualization data stream.

10. A dynamic energy consumption tracing and visualization system for educational buildings, applied to the dynamic energy consumption tracing and visualization method for educational buildings as described in any one of claims 1-9, characterized in that, The system includes: The semantic voxel mesh construction module is used to acquire the building information model data and three-dimensional geometric mesh data of educational buildings. It performs spatial discretization partitioning through an octree, constructs a semantic voxel mesh containing door and window opening and closing parameters and material thermal resistance parameters, and establishes a bidirectional mapping relationship between physical space coordinates and voxel indices. The heat flux vector field calculation module is used to collect attitude data from the magnetic angle sensor, fluid data from the wind speed sensor, and thermal image data from the infrared thermal imaging sensor. It maps the attitude data, fluid data, and thermal image data to the semantic voxel grid, establishes an energy conservation discrete equilibrium equation according to the material thermal resistance and convective heat transfer, performs time-step iterative calculation, and generates heat flux vector field data. The occupancy behavior feature generation module is used to acquire student entry and exit records, course schedule data and indoor number of people monitoring data. Based on the bidirectional mapping relationship, the physical space coordinate range is mapped to a set of voxel indices. The student entry and exit records, course schedule data and indoor number of people monitoring data are aggregated according to the voxel index and time index to generate an occupancy behavior feature sequence. The energy loss weight matrix generation module is used to construct a time series tensor by combining the heat flux vector field data and the occupancy behavior feature sequence according to the voxel index and time index, and generate the energy loss weight matrix by performing frequent itemset mining and association weight calculation. The differential thermal texture generation module is used to perform a linear mapping calculation based on interval partitioning on the energy loss weight matrix and the heat flux vector field data at each voxel index to obtain the color component parameters and transparency parameters of each voxel index and time index, and generate differential thermal texture data. The 3D visualization rendering module is used to map the differential thermal texture data to the semantic voxel mesh, perform volume rendering calculations on the semantic voxel mesh, generate a 3D visualization scene on the 3D geometric mesh data based on the bidirectional mapping relationship, and output a visualization data stream.

Images