Curtain wall lighting performance simulation method and system based on deep learning

Abstract

The invention provides a curtain wall lighting performance simulation method and system based on deep learning, and relates to the technical field of building design and simulation.The curtain wall lighting performance simulation method comprises the steps that firstly, curtain wall design elements (including curtain wall structure arrangement and curtain wall material light transmission performance) and target area illumination action elements (covering illumination trends in different time periods and illumination intensity changes in different time periods) are integrated; inputting a pre-trained daylighting association network construction model, generating illumination undertaking expression distribution of each part of the curtain wall and indoor space illumination diffusion expression distribution, constructing a curtain wall daylighting dynamic association network to present a dynamic action relationship among elements, and performing interactive simulation operation through a deep learning simulation module to obtain an indoor space daylighting dynamic association network; and generating a simulation result containing lighting balance and continuous performance of each indoor area, and generating curtain wall design optimization guidance according to the simulation result so as to adjust curtain wall design elements and optimize indoor lighting performance.

Description

Technical Field

[0001] This invention relates to the field of architectural design and simulation technology, and more specifically, to a method and system for simulating the daylighting performance of curtain walls based on deep learning. Background Technology

[0002] In the field of architectural design and construction, curtain walls, as an important element of modern building facades, not only serve to beautify the building's appearance but also have a crucial impact on interior lighting. A well-designed curtain wall can effectively utilize natural light, providing a comfortable, uniform, and continuous lighting environment indoors, thereby reducing the use of artificial lighting, lowering energy consumption, and aligning with the development concept of green building.

[0003] Currently, traditional methods for analyzing the daylighting performance of curtain walls mainly rely on empirical formulas and simple numerical simulations. Empirical formulas are typically based on simplified assumptions and statistical data, making it difficult to accurately consider the complex dynamic relationships between curtain wall design elements and lighting factors. For example, for different curtain wall structural arrangements and material light transmittance, empirical formulas may not accurately predict the actual daylighting effect under varying light direction and intensity at different times. While simple numerical simulations can simulate light propagation and distribution to some extent, they often neglect the interaction mechanisms between curtain wall design elements and lighting factors, leading to significant deviations between simulation results and actual conditions. Furthermore, traditional methods struggle to comprehensively and dynamically present the correlation between curtain wall design elements, lighting factors, and indoor daylighting performance, failing to provide targeted and forward-looking optimization guidance for curtain wall design. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a method for simulating the daylighting performance of curtain walls based on deep learning, the method comprising: The design elements of the curtain wall are integrated with the lighting effects of the target area. The curtain wall design elements include the curtain wall structure arrangement and the light transmission performance of the curtain wall materials. The lighting effects include the direction of light at different times and the changes in light intensity at different times. The curtain wall design elements and the lighting effect elements are input into a pre-trained daylighting association network to construct a model, generating the distribution of light reception performance of each part of the curtain wall and the distribution of light diffusion performance in the indoor space. A dynamic correlation network for curtain wall lighting is constructed based on the distribution of light reception and the distribution of light diffusion. This dynamic correlation network is used to present the dynamic relationship between curtain wall design elements, lighting effect elements and indoor lighting performance. The deep learning simulation module is activated to perform interactive simulation calculations on the dynamic correlation network of the curtain wall lighting, generating curtain wall lighting simulation results that include the balanced lighting performance of each indoor area and the continuous lighting performance of each indoor area. Based on the curtain wall daylighting simulation results, a curtain wall design optimization guide is generated. The curtain wall design optimization guide is used to adjust the curtain wall design elements to optimize the indoor daylighting performance.

[0005] Furthermore, embodiments of the present invention also provide a deep learning-based curtain wall daylighting performance simulation system, characterized in that it includes: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the aforementioned deep learning-based curtain wall daylighting performance simulation method by executing the machine-executable instructions.

[0006] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, a processor of a computer device reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the computer device to execute the above-described deep learning-based method for simulating the lighting performance of curtain walls.

[0007] Based on the above, by integrating curtain wall design elements with target area lighting elements, encompassing curtain wall structural arrangement, curtain wall material light transmittance, and various factors such as light direction and intensity variations at different times, the integrated elements are input into a pre-trained daylighting network model. This model generates the distribution of light reception at different parts of the curtain wall and the distribution of light diffusion in the interior space, accurately depicting the distribution of light on the curtain wall and in the interior space from both micro and macro perspectives. A dynamic daylighting network for the curtain wall is constructed based on the distribution of light reception at different parts of the curtain wall and the distribution of light diffusion in the interior space. The network presents the dynamic relationship between curtain wall design elements, lighting elements, and indoor lighting performance in an intuitive and scientific way. The deep learning simulation module is activated to perform interactive simulation calculations on the dynamic network. The generated curtain wall lighting simulation results not only include the balanced lighting performance of various indoor areas, but also the continuous lighting performance of various indoor areas, comprehensively and meticulously reflecting the actual situation of indoor lighting. Based on the simulation results, the curtain wall design optimization guidelines can accurately adjust the curtain wall design elements, thereby effectively optimizing the indoor lighting performance and significantly improving the accuracy of curtain wall lighting performance simulation. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the execution flow of the deep learning-based curtain wall daylighting performance simulation method provided in the embodiments of the present invention.

[0009] Figure 2 This is a schematic diagram of exemplary hardware and software components of the deep learning-based curtain wall daylighting performance simulation system provided in an embodiment of the present invention. Detailed Implementation

[0010] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a deep learning-based method for simulating the daylighting performance of curtain walls, as provided in one embodiment of the present invention. The following is a detailed description of this deep learning-based method for simulating the daylighting performance of curtain walls.

[0011] Step S110: Integrate the curtain wall design elements with the target area lighting effect elements. The curtain wall design elements include the curtain wall structure arrangement and the light transmission performance of the curtain wall materials. The lighting effect elements include the light direction at different times and the changes in light intensity at different times.

[0012] In this embodiment, taking the simulation of the daylighting performance of a commercial building's curtain wall as an example, the acquisition of curtain wall design elements requires comprehensive consideration of architectural design drawings, technical parameters of curtain wall materials, etc. Regarding the curtain wall structure arrangement, the overall layout of the curtain wall needs to be clearly defined. For example, the curtain wall of this commercial building is composed of multiple rectangular panels, and the position and dimensions of each panel have detailed design parameters. These parameters can be extracted from the CAD drawings of the architectural design to form curtain wall structure arrangement data containing coordinate attributes and dimensional parameters. Regarding the light transmission performance of the curtain wall materials, parameters such as the light transmittance, reflectance, and absorptivity of the curtain wall materials need to be obtained. These parameters can be obtained through technical documents provided by material suppliers or laboratory test reports. For example, the curtain wall material of this commercial building is a certain type of Low-E glass, and its light transmittance, reflectance, and absorptivity have clearly defined test data.

[0013] Obtaining information on lighting factors requires combining the geographical location and meteorological data of the target area. Regarding the direction of sunlight at different times, based on the geographical location of the commercial building, astronomical algorithms or meteorological software are used to calculate the solar altitude angle and azimuth angle for different seasons, dates, and times, thereby determining the incident direction of sunlight and forming data on the direction of sunlight at different times. Regarding the variation in sunlight intensity at different times, meteorological data for the region is collected, including solar radiation intensity data for different times. This data can be obtained from local weather stations or meteorological databases, thus obtaining data on the variation in sunlight intensity at different times. Integrating the above data on curtain wall structure arrangement, light transmittance of curtain wall materials, direction of sunlight at different times, and variation in sunlight intensity at different times forms a complete dataset.

[0014] Step S120: Input the curtain wall design elements and the lighting effect elements into the pre-trained daylighting association network construction model to generate the light reception performance distribution of each part of the curtain wall and the light diffusion performance distribution of the indoor space.

[0015] Step S121: Perform spatial digital transformation on the curtain wall structure arrangement in the curtain wall design elements to form a digital form of the curtain wall structure with coordinate attributes. Each structural part in the digital form of the curtain wall structure has corresponding position parameters and size parameters.

[0016] In this embodiment, for the curtain wall structure arrangement of the aforementioned commercial building, a spatial coordinate system is first determined. For example, the origin is a point on the ground floor of the building, with the horizontal direction represented by the x-axis and y-axis, and the vertical direction by the z-axis. Then, according to the design drawings of the curtain wall structure, the position and size parameters of each curtain wall panel and supporting frame are converted into numerical representations in this coordinate system. For example, if the coordinates of the lower left corner of a curtain wall panel are (x1, y1, z1) and the coordinates of the upper right corner are (x2, y2, z2), then the position parameter of the panel can be represented by these two coordinate points, and the size parameter can be obtained by calculating the distance between the two coordinate points, such as length x2-x1, width y2-y1, and height z2-z1. By performing the above conversion on all curtain wall structural parts, a digital form of the curtain wall structure with coordinate attributes is finally formed. This digital form is stored in the form of a data structure, for example, a three-dimensional array can be used to represent the position and size parameters of each structural part, with each element of the array corresponding to the coordinate and size information of a structural part.

[0017] Step S122: Perform feature transformation on the light transmission performance of the curtain wall material in the curtain wall design elements to form a material light transmission feature form that can be identified by the light transmission network construction model. The dimension of the material light transmission feature form matches the dimension of the input layer of the light transmission network construction model.

[0018] In this embodiment, the original data for the light transmission performance of the curtain wall material of the aforementioned commercial building includes parameters such as transmittance, reflectance, and absorptivity. To transform these parameters into a feature form recognizable by the daylighting network construction model, standardization is required. For example, assuming the input layer dimension of the daylighting network construction model is 3, corresponding to transmittance, reflectance, and absorptivity respectively. First, the values ​​of transmittance T, reflectance R, and absorptivity A of the curtain wall material are obtained. Then, these values ​​are normalized so that their values ​​are within the range [0, 1]. The normalization method can be min-max normalization, that is, for each parameter, its proportion relative to the possible value range of that parameter is calculated. For example, if the possible value range of transmittance is [0, 1], then the normalized transmittance T' = T / 1; the same applies to reflectance and absorptivity. The normalized T', R', and A' are combined sequentially into a three-dimensional vector. The dimension of this three-dimensional vector matches the input layer dimension of the daylighting network construction model, thus forming the material's light transmission feature form.

[0019] Step S123: The direction of the light direction in the different time periods of the light effect elements is transformed to form a set of light effect directions. Each direction information in the set of light effect directions corresponds to the light incidence direction in a specific time period.

[0020] In this embodiment, the sunlight direction data for different time periods in the area where the aforementioned commercial building is located needs to be transformed into a set of sunlight direction. First, the granularity of time division is determined, for example, dividing the day into 24 time periods using hours as the unit. For each time period, the incident direction of sunlight is transformed into a vector form based on the calculated solar altitude angle and azimuth angle. For example, in a certain time period, if the solar altitude angle is h and the azimuth angle is a, then the incident direction vector of sunlight can be represented as (sin(h)cos(a), sin(h)sin(a), cos(h)), where the x-component represents the horizontal projection, the y-component represents the horizontal projection perpendicular to the x-axis, and the z-component represents the vertical projection. The incident direction vectors of sunlight for each time period are arranged in chronological order to form a set of sunlight direction. Each element in this set of sunlight direction corresponds to the incident direction of sunlight in a specific time period and is stored in vector form for convenient subsequent model processing.

[0021] Step S124: Perform time-series transformation on the changes in light intensity at different times in the light-effect elements to form a time-series feature of light intensity, wherein the time interval of the time-series feature of light intensity is consistent with the time period division of the set of light-effect directions.

[0022] In this embodiment, the data on the variation of light intensity in the area where the aforementioned commercial buildings are located at different times needs to be transformed into a time-series feature of light intensity. First, the solar radiation intensity data for each time period is extracted according to the same time granularity as the set of light-affecting directions, i.e., in hours. Then, the above data is standardized, for example, using Z-standardization, to calculate the difference between the solar radiation intensity of each time period and the mean and standard deviation of the entire time-series, and then divide by the standard deviation to obtain the standardized solar radiation intensity data. The standardized data is then arranged in chronological order to form a time-series sequence. The time interval of this time-series sequence is consistent with the time period division of the set of light-affecting directions, thus forming a time-series feature of light intensity.

[0023] Step S125: Input the digital form of the curtain wall structure, the light transmission characteristics of the material, the set of light direction, and the temporal characteristics of light intensity into the pre-trained daylighting association network to construct the model input layer.

[0024] In this embodiment, the pre-trained lighting association network construction model is a deep learning model, and the structure of its input layer needs to match the dimension of the input data. The obtained digital morphology of the curtain wall structure, the material light transmission characteristics, the set of light direction, and the temporal characteristics of light intensity are combined according to the model's input requirements to form an input data set. For example, the digital morphology of the curtain wall structure can be used as a three-dimensional tensor input, the material light transmission characteristics as a three-dimensional vector input, the set of light direction as a two-dimensional tensor (time × direction vector dimension) input, and the temporal characteristics of light intensity as a one-dimensional vector (time dimension) input.

[0025] Step S126: The structural feature processing layer of the model constructed through the light-relation network performs spatial analysis on the digital morphology of the curtain wall structure to generate the spatial position features of each part of the curtain wall.

[0026] In this embodiment, the structural feature processing layer of the daylighting association network construction model includes multiple convolutional and pooling layers for spatial analysis of the digital morphology of the curtain wall structure. First, the digital morphology of the curtain wall structure is input as a three-dimensional tensor to the first convolutional layer of the structural feature processing layer. This convolutional layer uses multiple convolutional kernels to perform convolution operations on the input tensor, extracting local spatial features of the curtain wall structure. For example, the size of the convolutional kernels can be set according to the level of detail of the curtain wall structure. Through convolution operations, multiple feature maps are obtained, each corresponding to a local spatial feature. Then, pooling operations, such as max pooling or average pooling, are performed on the feature maps to reduce the dimensionality of the feature maps while retaining the main spatial features. After processing by multiple convolutional and pooling layers, the spatial location features of each part of the curtain wall are finally obtained. These spatial location features are represented in the form of high-dimensional vectors, with each vector corresponding to a part of the curtain wall. The dimension of the vector reflects the spatial location information of that part and its spatial relationship with the surrounding structure.

[0027] Step S127: Perform attribute analysis on the light transmission characteristics of the material through the material feature processing layer of the light transmission association network model to generate material light transmission attribute characteristics of each part of the curtain wall.

[0028] In this embodiment, the material feature processing layer of the light-transmitting network construction model contains multiple fully connected layers for property analysis of the material's light-transmitting characteristics. First, the material's light-transmitting characteristics are input as three-dimensional vectors to the first fully connected layer of the material feature processing layer. This fully connected layer performs linear transformations and nonlinear activations on the input vectors to extract the basic light-transmitting property features of the material. For example, a linear transformation maps the three-dimensional vector to a higher-dimensional space, and then a nonlinear activation function (such as the ReLU function) is used to introduce nonlinear features, enhancing the model's expressive power. After processing by multiple fully connected layers, the material light-transmitting property features of each part of the curtain wall are finally obtained. These material light-transmitting property features are represented as high-dimensional vectors, with each vector corresponding to a part of the curtain wall. The dimension of the vector reflects the material's transmittance, reflectance, absorptivity, and other property information of that part, as well as the correlation information between these properties.

[0029] Step S128: The illumination feature processing layer of the model constructed by the light-collecting association network performs spatiotemporal association between the set of illumination directions and the temporal features of illumination intensity to generate illumination features for each time period.

[0030] In this embodiment, the illumination feature processing layer of the illumination association network construction model includes multiple recurrent neural network layers (such as LSTM layers) and convolutional layers, used to spatiotemporally correlate the set of illumination directions with the temporal features of illumination intensity. First, the set of illumination directions is input into the convolutional layer of the illumination feature processing layer as a two-dimensional tensor. This convolutional layer performs a convolution operation on the illumination direction vector for each time period to extract the spatial features of the illumination direction. Simultaneously, the temporal features of illumination intensity are input into the recurrent neural network layer as a one-dimensional vector. This recurrent neural network layer analyzes the temporal changes in illumination intensity to extract the temporal features of illumination intensity. Then, the spatial features obtained from the convolutional layer and the temporal features obtained from the recurrent neural network layer are fused, for example, by concatenating the two feature vectors into a new feature vector. This feature vector simultaneously contains the spatial information of the illumination direction and the temporal information of the illumination intensity. After several of the above processing steps, the illumination characteristics of each time period are finally obtained. These illumination characteristics are represented in the form of high-dimensional vectors, with each vector corresponding to a time period. The dimension of the vector reflects the illumination direction, illumination intensity, and spatiotemporal relationship information between them during that time period.

[0031] Step S129: By constructing the element interaction layer of the lighting association network model, the spatial location features, the material light transmission properties features, and the lighting effect features are integrated across elements to generate comprehensive lighting reception features for each part of the curtain wall.

[0032] In this embodiment, the element interaction layer of the lighting association network construction model includes multiple attention mechanism layers and fully connected layers, used for cross-element fusion of spatial location features, material light transmission properties, and illumination effects. First, the spatial location features, material light transmission properties, and illumination effects are input into the attention mechanism layer, which calculates the attention weight between each feature, thus determining the importance of each feature in the fusion process. For example, the attention weight between spatial location features and material light transmission properties is obtained by calculating their similarity; similarly, the attention weight between spatial location features and illumination effects, and between material light transmission properties and illumination effects, is calculated. Then, based on these attention weights, each feature is weighted, for example, by multiplying the spatial location feature by its corresponding attention weight to obtain the weighted spatial location feature; similarly, the weighted material light transmission properties and illumination effects are obtained. Finally, the three weighted features are spliced ​​together to form the comprehensive light reception features of each part of the curtain wall. The comprehensive light reception features are represented in the form of high-dimensional vectors, with each vector corresponding to a part of the curtain wall. The dimension of the vector reflects the comprehensive influence information of various factors such as the spatial location of the part, the light transmission properties of the material, and the effect of light.

[0033] Step S1210: Based on the spatial distribution of the comprehensive characteristics of light reception in various parts of the curtain wall, generate the distribution of light reception performance in various parts of the curtain wall.

[0034] In this embodiment, the comprehensive characteristics of light reception at various parts of the curtain wall are represented by high-dimensional vectors, with each vector corresponding to a specific part of the curtain wall. First, these high-dimensional vectors are decoded and converted into a form that intuitively represents the light reception performance. For example, certain dimensions of the vector can be mapped to indicators such as light reception intensity, light reception time, and light incidence angle. Then, based on the spatial location information of each part of the curtain wall, the decoded indicators are distributed according to their spatial location, forming a two-dimensional or three-dimensional distribution map. For example, on a two-dimensional plane, based on the planar layout of the curtain wall, each location point corresponds to a part of the curtain wall, and different colors or values ​​are used to represent indicators such as light reception intensity, light reception time, and light incidence angle for that part, thereby generating the distribution of light reception performance at various parts of the curtain wall.

[0035] Step S1211: Based on the spatial parameters of the light reception performance distribution and the digital morphology of the curtain wall structure, the indoor diffusion simulation layer of the model constructed by the light-receiving network simulates the indoor diffusion path and energy change process of light after passing through the curtain wall, and generates the indoor space light diffusion performance distribution.

[0036] In this embodiment, the indoor diffusion simulation layer of the light-receiving network construction model includes multiple physical simulation models and neural network layers to simulate the indoor diffusion path and energy change process of light after passing through the curtain wall. First, based on the light reception distribution of various parts of the curtain wall, the light emission intensity and direction of each curtain wall part are determined. Then, combined with the spatial parameters of the digital morphology of the curtain wall structure, such as the location, size, and shading of the supporting frame, the propagation path of light after exiting the curtain wall part is simulated indoors. In simulating the propagation path, the indoor spatial structure needs to be considered, such as the shape of the room, the reflectivity of the walls, and the shading of furniture. These factors can be obtained through a pre-established indoor spatial model. At the same time, it is also necessary to simulate the energy changes of light during propagation, such as the reflection and absorption of light on the walls and the scattering in the air. These energy changes can be calculated using physical formulas or empirical models, such as calculating the reflected and absorbed energy based on the reflectivity and absorptivity of the wall material, and calculating the scattered energy based on the scattering coefficient of the air. By simulating the light emission from each part of the curtain wall, information such as the light intensity and light coverage of each area in the indoor space is obtained. This information is then distributed according to the spatial location of the indoor areas to form the distribution of light diffusion in the indoor space.

[0037] Step S130: Construct a dynamic correlation network for curtain wall lighting based on the distribution of light reception and the distribution of light diffusion. The dynamic correlation network for curtain wall lighting is used to present the dynamic interaction between curtain wall design elements, lighting effect elements and indoor lighting performance.

[0038] Step S131: Extract the light intensity, light duration, and light incidence angle of each part of the curtain wall from the light reception performance distribution, and use them as the first type of node information of the curtain wall lighting dynamic association network.

[0039] In this embodiment, for the generated distribution of light reception performance of various parts of the curtain wall, key information needs to be extracted as the first type of node information. First, the light reception intensity performance can be obtained by analyzing the light intensity values ​​in the distribution. For example, for each curtain wall part, the maximum, minimum, and average light intensity values ​​at different times are extracted; these values ​​reflect the light reception intensity of that part. The light reception time performance can be obtained by analyzing the light reception time information in the distribution. For example, for each curtain wall part, the effective illumination time and continuous illumination time in a day, month, and year are extracted; this time information reflects the illumination time of that part. The light incidence angle performance can be obtained by analyzing the light incidence angle values ​​in the distribution. For example, for each curtain wall part, the maximum, minimum, and average light incidence angle values ​​at different times are extracted; these angle values ​​reflect the light incidence angle of that part. The extracted light reception intensity performance, light reception time performance, and light incidence angle performance are organized according to the curtain wall part to form the first type of node information, with each curtain wall part corresponding to a set of the above information.

[0040] Step S132: Extract the light intensity, light coverage, and light balance of each indoor area from the light diffusion distribution, and use them as the second type of node information of the curtain wall lighting dynamic association network.

[0041] In this embodiment, for the generated indoor space light diffusion distribution, key information needs to be extracted as the second type of node information. First, the light arrival intensity can be obtained by analyzing the light intensity values ​​in the distribution. For example, for each indoor area, the maximum, minimum, and average light intensity values ​​at different times are extracted; these values ​​reflect the light arrival intensity of that area. The light coverage area can be obtained by analyzing the light coverage area and shape information in the distribution. For example, for each indoor area, the light coverage area and shape at different times are extracted; this information reflects the light coverage area of ​​that area. The light uniformity can be obtained by analyzing the uniformity of the light intensity distribution. For example, the variance and standard deviation of the light intensity within the area are calculated; a smaller variance indicates a higher degree of light uniformity, and these statistics are used as indicators of light uniformity. The extracted light arrival intensity, light coverage area, and light uniformity are then organized according to indoor areas to form the second type of node information, with each indoor area corresponding to a set of the above information.

[0042] Step S133: Extract the curtain wall panel size representation, panel splicing gap representation, and support frame shading representation from the curtain wall structure arrangement of the curtain wall design elements, as the third type of node information of the curtain wall lighting dynamic association network.

[0043] In this embodiment, for the curtain wall structure arrangement of the aforementioned commercial building, key information needs to be extracted as the third type of node information. First, the curtain wall panel dimensions can be obtained by analyzing the panel size parameters in the curtain wall structure arrangement. For example, for each curtain wall panel, its length, width, and height are extracted, reflecting the panel's dimensions. The panel splicing gaps can be obtained by analyzing the panel splicing method and gap dimensions in the curtain wall structure arrangement. For example, for every two adjacent curtain wall panels, the width, length, and shape of their splicing gaps are extracted, reflecting the gap's condition. The supporting frame obstruction can be obtained by analyzing the position and dimensions of the supporting frames in the curtain wall structure arrangement. For example, for each supporting frame, its projected area on the curtain wall plane and the area of ​​the curtain wall panels it obstructs are extracted, reflecting the obstruction situation of the supporting frame. The extracted curtain wall panel dimensions, panel splicing gaps, and supporting frame obstruction are then organized according to the curtain wall structure sections to form the third type of node information, with each curtain wall structure section corresponding to a set of the above information.

[0044] Step S134: Extract the light transmittance, reflectance, and absorptivity of the curtain wall materials from the light transmittance of the curtain wall design elements, and use them as the fourth type of node information in the dynamic correlation network of curtain wall lighting.

[0045] In this embodiment, regarding the light transmittance performance of the curtain wall materials of the aforementioned commercial building, key information needs to be extracted as the fourth type of node information. First, the light transmittance performance can be obtained by analyzing the light transmittance parameters of the curtain wall material, for example, extracting the light transmittance values ​​of the curtain wall material under different wavelengths of light; these values ​​reflect the light transmittance of the material. The reflectance performance can also be obtained by analyzing the reflectance parameters of the curtain wall material, for example, extracting the reflectance values ​​of the curtain wall material under different wavelengths of light; these values ​​reflect the reflectance of the material. The extracted light transmittance, reflectance, and absorptance performance are then organized according to the curtain wall material section to form the fourth type of node information, with each curtain wall material section corresponding to a set of the above information.

[0046] Step S135: Extract the peak intensity of light and the change of light direction at different times from the light effect elements, as the fifth type of node information of the curtain wall lighting dynamic association network.

[0047] In this embodiment, for the lighting elements of the aforementioned commercial building area, key information needs to be extracted as the fifth type of node information. First, the peak light intensity at different times can be obtained by analyzing light intensity variation data for different time periods. For example, for each time period, the maximum light intensity is extracted, and then these maximum values ​​are arranged in chronological order to form a light intensity peak sequence. These sequences reflect the peak light intensity at different times. The variation in light direction at different times can be obtained by analyzing light direction data for different time periods. For example, the angle between the light direction vectors of adjacent time periods is calculated, and these angles are arranged in chronological order to form a light direction variation sequence. These sequences reflect the variation in light direction at different times. The extracted peak light intensity and light direction variation data for different time periods are then organized by time period to form the fifth type of node information, with each time period corresponding to a set of the above information.

[0048] Step S136: Construct a node group for the dynamic association network of curtain wall lighting. The node group includes curtain wall part nodes, indoor area nodes, curtain wall structure nodes, curtain wall material nodes, and lighting effect nodes. Each type of node corresponds to the above-mentioned different categories of node information.

[0049] In this embodiment, a node group of a dynamic correlation network for curtain wall lighting is constructed based on the five types of node information extracted above. First, the curtain wall component nodes correspond to the first type of node information, with each node containing information such as the light intensity, light duration, and light incidence angle of that component. The indoor area nodes correspond to the second type of node information, with each node containing information such as the light intensity, light coverage, and light uniformity of that area. The curtain wall structure nodes correspond to the third type of node information, with each node containing information such as the curtain wall panel dimensions, panel joint gaps, and support frame obstruction of that structural component. The curtain wall material nodes correspond to the fourth type of node information, with each node containing information such as the material's light transmittance, reflectivity, and absorptivity. The lighting effect nodes correspond to the fifth type of node information, with each node containing information such as the peak light intensity at different times and the change in light direction at different times. The nodes are organized according to the established logical relationships, such as the spatial correspondence between the curtain wall parts and the interior areas, the relationship between the curtain wall parts and the curtain wall structure, the relationship between the curtain wall parts and the curtain wall materials, and the relationship between the curtain wall parts and the effect of lighting, to form a node group.

[0050] Step S137: Analyze the interaction between the nodes of the curtain wall and the nodes of the indoor area. Based on the diffusion path of light from the curtain wall to the indoor area, establish the first type of connection relationship between the nodes of the curtain wall and the nodes of the indoor area. The degree of the effect of the first type of connection relationship is determined based on the energy transfer during the light diffusion process.

[0051] In this embodiment, the indoor area nodes corresponding to each curtain wall node are first determined, for example, by analyzing the diffusion path of light from the curtain wall to the indoor area. For instance, after light is emitted from a certain curtain wall part, it propagates through the indoor space and eventually reaches certain indoor areas; therefore, the curtain wall node interacts with these indoor area nodes. Then, the energy transfer during the light diffusion process is calculated, for example, by analyzing the energy loss of light during propagation. For example, if the energy of light emitted from the curtain wall part is E1, and the energy reaching the indoor area is E2, then the energy transfer efficiency is E2 / E1, which can be used as the degree of interaction of the first type of connection. Based on the above interaction relationships and degrees of interaction, a first type of connection is established between the curtain wall node and the indoor area node, for example, represented by directed edges, with the edge weight representing the degree of interaction.

[0052] Step S138: Analyze the interaction between the curtain wall component nodes and the curtain wall structure nodes. Based on the influence of the curtain wall structure on the light reception of the curtain wall component, establish a second type of connection relationship between the curtain wall component nodes and the curtain wall structure nodes. The degree of influence of the second type of connection relationship is determined based on the influence of structural shading on the light reception.

[0053] In this embodiment, the curtain wall structural nodes corresponding to each curtain wall part node are first determined. This can be achieved, for example, by analyzing the relationship between the curtain wall structure and the curtain wall part. For instance, if a curtain wall part belongs to a curtain wall panel, and that panel is related to certain supporting frames and splicing gaps, then the curtain wall part node has an interaction relationship with these curtain wall structural nodes. Next, the impact of the curtain wall structure on the light reception of the curtain wall part is analyzed. For example, the shading of the supporting frame reduces the light reception area of ​​the curtain wall part, thus affecting the light reception intensity. The degree of influence of structural shading on light reception is calculated, for example, the proportion of the shading area to the area of ​​the curtain wall part. This proportion can be used as the degree of influence of the second type of connection relationship. Based on the above interaction relationships and degrees of influence, a second type of connection relationship is established between the curtain wall part nodes and the curtain wall structural nodes, for example, represented by directed edges, with the edge weight representing the degree of influence.

[0054] Step S139: Analyze the interaction between the curtain wall component nodes and the curtain wall material nodes. Based on the influence of the curtain wall material on the light absorption and reflection of the curtain wall component, establish a third type of connection relationship between the curtain wall component nodes and the curtain wall material nodes. The degree of influence of the third type of connection relationship is determined based on the influence of the material transmittance on the light transmission.

[0055] In this embodiment, the curtain wall material node corresponding to each curtain wall component node is first determined. This can be achieved, for example, by analyzing the relationship between curtain wall materials and curtain wall components. For instance, if a curtain wall component uses a certain type of Low-E glass, then the node of that curtain wall component has an interaction relationship with the curtain wall material node of that glass. Next, the influence of the curtain wall material on the light absorption and reflection of the curtain wall component is analyzed. For example, the material's absorptivity affects the amount of light absorbed, and its reflectivity affects the amount of light reflected, thus affecting the amount of light transmitted. The degree of influence of the material's transmittance on light transmission is calculated. For example, the higher the transmittance, the more light is transmitted; this transmittance can be used as the degree of influence of the third type of connection relationship. Based on the above interaction relationships and degrees of influence, a third type of connection relationship is established between the curtain wall component nodes and the curtain wall material nodes, for example, represented by directed edges, with the edge weight representing the degree of influence.

[0056] Step S1310: Analyze the interaction relationship between the curtain wall nodes and the lighting nodes. Based on the influence of lighting on the lighting reception of the curtain wall, establish a fourth type of connection relationship between the curtain wall nodes and the lighting nodes. The degree of influence of the fourth type of connection relationship is determined based on the influence of lighting intensity and direction on the lighting reception of the curtain wall.

[0057] In this embodiment, the lighting effect node corresponding to each curtain wall node is first determined. This can be achieved, for example, by analyzing the interaction between lighting effect and the curtain wall. For instance, the direction and intensity of light at a certain time period will affect the light reception of a certain curtain wall section, thus establishing an interaction relationship between that curtain wall section node and the lighting effect node for that time period. Then, the impact of lighting effect on the light reception of the curtain wall section is analyzed. For example, stronger light intensity results in higher light reception intensity for the curtain wall section; a more suitable angle between the light direction and the curtain wall section also results in higher light reception intensity. The degree of influence of light intensity and direction on the light reception of the curtain wall section is calculated. This can be achieved, for example, by establishing a function that takes light intensity and direction as input and light reception intensity as output. The output value of this function can be used as the degree of influence of the fourth type of connection relationship. Based on the above interaction relationships and degrees of influence, a fourth type of connection relationship is established between the curtain wall section node and the lighting effect node, represented, for example, by directed edges, with the edge weight representing the degree of influence.

[0058] Step S1311: Analyze the interaction between indoor area nodes and curtain wall structure nodes. Based on the influence of the curtain wall structure on the indoor area light diffusion, establish a fifth type of connection relationship between indoor area nodes and curtain wall structure nodes. The degree of influence of the fifth type of connection relationship is determined based on the influence of the structural arrangement on the indoor light distribution.

[0059] In this embodiment, the curtain wall structure node corresponding to each indoor area node is first determined, for example, by analyzing the spatial relationship between the curtain wall structure and the indoor area. For instance, the supporting frame of a certain curtain wall structure affects the light diffusion path of the indoor area, thus affecting the light distribution of the indoor area. Therefore, the indoor area node has an interaction relationship with the curtain wall structure node. Then, the influence of the curtain wall structure on the light diffusion of the indoor area is analyzed. For example, the position of the supporting frame changes the light diffusion direction, thus affecting the light coverage and light intensity distribution of the indoor area. The degree of influence of the structural arrangement on the indoor light distribution is calculated. For example, the difference in the light distribution of the indoor area with and without the curtain wall structure can be compared. This difference in light distribution can be used as the degree of influence of the fifth type of connection relationship. Based on the above interaction relationships and degrees of influence, a fifth type of connection relationship is established between the indoor area node and the curtain wall structure node, for example, represented by a directed edge, with the edge weight representing the degree of influence.

[0060] Step S1312: Analyze the interaction between indoor area nodes and curtain wall material nodes. Based on the influence of curtain wall materials on indoor light intensity, establish a sixth type of connection relationship between indoor area nodes and curtain wall material nodes. The degree of influence of the sixth type of connection relationship is determined based on the influence of material transmittance on indoor light intensity.

[0061] In this embodiment, the curtain wall material node corresponding to each indoor area node is first determined. This can be achieved, for example, by analyzing the interaction between the curtain wall material and the indoor area. For instance, the light transmittance of a certain curtain wall material affects the amount of light transmitted, thus affecting the light intensity of the indoor area. Therefore, the indoor area node has an interaction with the curtain wall material node. Then, the influence of the curtain wall material on the light intensity of the indoor area is analyzed. For example, the higher the light transmittance of the material, the higher the light intensity of the indoor area. The degree of influence of the material transmittance on the indoor light intensity is calculated. For example, this can be achieved by establishing a function that takes the material transmittance as input and the indoor area light intensity as output. The output value of this function can be used as the degree of influence of the sixth type of connection. Based on the above interaction and degree of influence, a sixth type of connection is established between the indoor area node and the curtain wall material node, represented by directed edges, with the edge weight representing the degree of influence.

[0062] Step S1313: Integrate the above-mentioned nodes, node information and connection relationships to form a complete dynamic association network for curtain wall lighting. The dynamic association network for curtain wall lighting is used to present the dynamic interaction relationship between curtain wall design elements, lighting effect elements and indoor lighting performance.

[0063] In this embodiment, the obtained curtain wall node, interior area node, curtain wall structure node, curtain wall material node, lighting effect node, and corresponding node information and connection relationships are integrated. First, all nodes are organized according to their category and function to form a set of network nodes. Then, all connection relationships are organized according to their type and degree of influence to form a set of network edges. Finally, the node set and edge set are combined into a complete network structure, which can be represented by a graph data structure, where nodes represent the various types of nodes mentioned above, edges represent the various types of connection relationships mentioned above, and the weight of the edges represents the degree of influence. This forms a complete dynamic correlation network for curtain wall lighting, which can intuitively present the dynamic relationship between curtain wall design elements, lighting effect elements, and interior lighting performance.

[0064] Step S140: Start the deep learning simulation module to perform interactive simulation calculations on the curtain wall lighting dynamic correlation network, and generate curtain wall lighting simulation results that include the balanced lighting performance of each indoor area and the continuous lighting performance of each indoor area.

[0065] Step S141: Perform network structure transformation on the dynamic association network of curtain wall lighting, transforming the node information and connection relationship information in the dynamic association network of curtain wall lighting into a network feature form that can be recognized by the deep learning simulation module. The number of rows in the network feature form corresponds to the number of nodes in the dynamic association network of curtain wall lighting, and the number of columns corresponds to the sum of the node information dimension and the connection relationship feature dimension.

[0066] In this embodiment, the structure of the network feature form that the deep learning simulation module can recognize first needs to be determined. For example, this network feature form can be a two-dimensional matrix, where the number of rows equals the number of nodes in the curtain wall lighting dynamic association network, and the number of columns equals the sum of the information dimension of each node and the connection relationship feature dimension of that node. For each node, its node information is first extracted, such as the light intensity performance, light duration performance, and light incidence angle performance of the nodes in the curtain wall area. This information is converted into a numerical vector, with the dimension of the vector being the node information dimension. Then, the connection relationship information of the node is extracted, such as the degree of influence and type of connection relationship between the node and other nodes. This information is also converted into a numerical vector, with the dimension of the vector being the connection relationship feature dimension. Finally, the node information vector and the connection relationship information vector are concatenated to form the feature row corresponding to the node. The feature rows of all nodes are arranged in the order of the nodes to form a two-dimensional matrix of network feature form.

[0067] Step S142: Input the network feature form into the network feature processing layer of the deep learning simulation module, and perform neighborhood feature integration on the network feature form through network convolution operation to generate neighborhood integrated features for each node.

[0068] Step S1421: Identify the group of directly adjacent nodes of each node in the dynamic association network of the curtain wall lighting, wherein the group of directly adjacent nodes includes all nodes that are directly associated with the node through connection relationships.

[0069] In this embodiment, for each node in the dynamic association network of curtain wall lighting, it is necessary to traverse the network's connection relationships to find all nodes directly associated with that node through these connections. These nodes form the group of directly adjacent nodes of that node. For example, for a curtain wall node, by examining its connection relationships with other nodes, all directly connected indoor area nodes, curtain wall structure nodes, curtain wall material nodes, lighting effect nodes, etc., are identified. These nodes constitute the group of directly adjacent nodes of that curtain wall node.

[0070] Step S1422: Calculate the degree of influence of the connection relationship between each node and each node in the group of its direct neighbors. The degree of influence of the connection relationship is determined based on the degree of influence parameter of the connection relationship.

[0071] In this embodiment, for each node and each node in its group of directly adjacent nodes, it is necessary to obtain the degree of influence parameter of the connection relationship between them. The above parameter has been determined when constructing the dynamic association network of curtain wall lighting, for example, the degree of influence calculated in steps S137-S1312. For each connection relationship, its degree of influence parameter is directly extracted as the degree of influence of the connection relationship between that node and its directly adjacent nodes.

[0072] Step S1423: Standardize the feature row information corresponding to each node in the network feature form and the feature row information corresponding to each node in the group of directly adjacent nodes of that node to eliminate the difference in feature dimensions and generate standardized feature row information.

[0073] In this embodiment, the feature row information of each node and the feature row information of each node in its directly adjacent node group need to be standardized. The standardization method can be Z-standardization, which involves calculating the mean and standard deviation of each feature dimension across all relevant feature rows, then subtracting the mean from each feature value and dividing by the standard deviation to obtain the standardized feature value. For example, for a certain feature dimension, assuming the feature values ​​of the node and each node in its directly adjacent node group are x1, x2, ..., xn, calculate the mean μ = (x1 + x2 + ... + xn) / n and the standard deviation σ = √[((x1 - μ)² + (x2 - μ)² + ... + (xn - μ)²) / (n - 1)]. Then, standardize each feature value xi to xi' = (xi - μ) / σ. This eliminates the dimensional differences between different feature dimensions and generates standardized feature row information.

[0074] Step S1424: Perform element-level operations on the standardized feature row information to generate a set of feature operation results for adjacent nodes.

[0075] In this embodiment, element-wise operations are required for the standardized feature row information of each node and the standardized feature row information of each node in its directly adjacent node group. Element-wise operations can be multiplication, addition, subtraction, etc., and the specific operation method can be determined according to the design requirements of the model. For example, using multiplication, the standardized feature row information of the node is multiplied element-wise with the standardized feature row information of each directly adjacent node, resulting in multiple operation result vectors. These vectors form the set of adjacent node feature operation results.

[0076] Step S1425: Perform a weighted operation on each result in the set of neighboring node feature operations and the corresponding degree of influence of the connection relationship to generate a weighted set of neighboring node features.

[0077] In this embodiment, for each result vector in the set of neighboring node feature operations, it needs to be weighted by the degree of influence of its corresponding connection relationship. The weighting operation involves multiplying each element of the result vector by the degree of influence of the connection relationship to obtain a weighted vector. For example, if a result vector is v = [v1, v2, ..., vm] and its corresponding degree of influence is w, then the weighted vector is v' = [v1×w, v2×w, ..., vm×w]. All these weighted vectors are then combined to form a weighted set of neighboring node features.

[0078] Step S1426: Perform element-wise summation on all features in the weighted neighbor node feature set to generate a neighborhood feature summation result.

[0079] In this embodiment, element-wise summation is required for all vectors in the weighted neighbor node feature set. The element-wise summation is performed by adding the corresponding elements of each vector to obtain a new vector. For example, if the vectors in the weighted neighbor node feature set are v1'=[v11, v12, ..., v1m], v2'=[v21, v22, ..., v2m], ..., vn'=[vn1, vn2, ..., vnm], then the element-wise summed vector is v_sum=[v11+v21+...+vn1, v12+v22+...+vn2, ..., v1m+v2m+...+vnm], and this vector serves as the result of the neighborhood feature summation.

[0080] Step S1427: The summation result of the neighborhood features is normalized to eliminate the difference in feature values ​​caused by the difference in the number of adjacent nodes of different nodes, and normalized neighborhood features are generated.

[0081] In this embodiment, the number of direct neighboring nodes may differ between nodes, leading to variations in the numerical values ​​of the neighborhood feature summation. To eliminate these differences, the neighborhood feature summation result needs to be normalized. Normalization can be achieved by dividing by the number of direct neighboring nodes to obtain an average weighted feature vector. For example, if a node has k direct neighboring nodes and its neighborhood feature summation result is v_sum, then the normalized neighborhood feature vector is v_reg = v_sum / k. This eliminates the numerical differences in features caused by variations in the number of neighboring nodes between different nodes.

[0082] Step S1428: Combine the original feature row information corresponding to each node with the regular neighborhood features of that node to generate the neighborhood integrated features of each node.

[0083] In this embodiment, for each node, its original feature row information (i.e., the feature row information that has not undergone normalization) needs to be combined with the regularized neighborhood features. The combination process can be concatenation, that is, concatenating the vectors of the original feature row information and the regularized neighborhood features to form a new vector, which serves as the neighborhood integration feature of that node. For example, if the vector of the original feature row information is v_original=[o1, o2, ..., op], and the vector of the regularized neighborhood features is v_reg=[r1, r2, ..., rq], then the vector of the neighborhood integration feature is v_integrated=[o1, o2, ..., op, r1, r2, ..., rq].

[0084] Step S143: The neighborhood integration features are analyzed at different levels through the multi-level network analysis layer of the deep learning simulation module. The local correlation features and overall correlation features of the curtain wall lighting dynamic correlation network are extracted respectively, and multi-level network correlation features are generated.

[0085] Step S1431: Set the first-level analysis range, which corresponds to the direct adjacent range of nodes in the curtain wall lighting dynamic association network, and is used to extract local association features.

[0086] In this embodiment, the first-level analysis scope is defined based on the direct adjacency relationships of nodes in the dynamic correlation network of the curtain wall lighting system. For example, for each node, its direct adjacency range includes the group of its direct adjacent nodes and the node itself. Within this range, the association characteristics between nodes are analyzed, and these characteristics reflect the network structure relationships within a local area.

[0087] Step S1432: Within the first level of analysis, the neighborhood integration features of each node are standardized, and the neighboring node feature difference analysis is performed on the standardized neighborhood integration features to extract the feature differences between the node and its direct neighboring nodes and generate local feature difference information.

[0088] In this embodiment, the neighborhood integration features of each node are first standardized, similar to step S1423. This involves calculating the mean and standard deviation of the neighborhood integration features of the node and each of its directly adjacent nodes in each dimension, followed by Z-standardization. Then, a neighboring node feature difference analysis is performed on the standardized neighborhood integration features. For example, the Euclidean distance or cosine similarity between the standardized neighborhood integration features of the node and the standardized neighborhood integration features of each directly adjacent node is calculated. These distances or similarities reflect the feature differences between the node and its directly adjacent nodes. The above differences are then organized according to a set format to generate local feature difference data.

[0089] Step S1433: Perform feature filtering on the local feature differences, retain key feature content that can reflect local correlation, and generate local correlation features.

[0090] In this embodiment, the method for feature selection based on local feature differences can be based on feature importance or relevance. For example, the relevance of each feature to the local association can be calculated, and features with high relevance can be retained. Alternatively, principal component analysis can be used to extract the main feature components, which can reflect the key feature content of the local association. The selected feature content is then organized to generate local association features.

[0091] Step S1434: Set the second-level analysis range, which corresponds to the indirect adjacent range of nodes in the curtain wall lighting dynamic association network, that is, the direct adjacent range of directly adjacent nodes, and is used to extract intermediate-level association features.

[0092] In this embodiment, the second-level analysis scope is defined based on the indirect adjacency relationships of nodes in the dynamic correlation network of the curtain wall lighting. For example, for each node, its indirect adjacency scope is the group of directly adjacent nodes of its directly adjacent nodes (excluding the node itself). Within this scope, the association characteristics between nodes are analyzed, and these characteristics reflect the network structure relationships within the intermediate-level region.

[0093] Step S1435: Within the second-level analysis scope, the neighborhood integration features of each node are standardized, and the indirect neighbor node feature transmission analysis is performed on the standardized neighborhood integration features. The influence of the indirect neighbor node features transmitted by the node through the direct neighbor nodes is extracted, and the intermediate-level feature transmission situation is generated.

[0094] In this embodiment, the neighborhood integration features of each node are first standardized, using a method similar to step S1423. Then, the standardized neighborhood integration features are analyzed for indirect neighbor node feature transfer. For example, by analyzing the connection relationships and feature transfer paths between directly adjacent nodes and indirect adjacent nodes, the influence of indirect neighbor node features transferred by the node through its directly adjacent nodes is calculated. For instance, for each directly adjacent node, the feature transfer coefficient between it and its indirect adjacent nodes is calculated, and then this coefficient is multiplied by the standardized neighborhood integration features of the indirect adjacent nodes to obtain the influence of indirect neighbor node features transferred by the node through its directly adjacent nodes. The above influence information is then organized according to a set format to generate intermediate-level feature transfer information.

[0095] Step S1436: Extract features from the intermediate level feature transmission situation, retain key feature content that can reflect the intermediate level association relationship, and generate intermediate level association features.

[0096] In this embodiment, the method for feature extraction of intermediate-level feature transmission is similar to step S1433, that is, based on the importance or relevance of features, key feature content that can reflect the intermediate-level association relationship is selected. For example, the relevance of each feature to the intermediate-level association relationship is calculated, and features with high relevance are retained; or principal component analysis is used to extract the main feature components. The extracted feature content is then organized to generate intermediate-level association features.

[0097] Step S1437: Set the third-level analysis range, which corresponds to the range of all nodes in the curtain wall lighting dynamic association network, and is used to extract the overall association features.

[0098] In this embodiment, the third-level analysis scope is defined as the entire node range of the curtain wall lighting dynamic association network. Within this scope, the association characteristics between all nodes are analyzed, and these characteristics reflect the structural relationships of the entire network.

[0099] Step S1438: Within the scope of the third-level analysis, the neighborhood integration features of all nodes are standardized, and global feature statistics are performed on the standardized neighborhood integration features to extract the overall distribution and mutual influence of global node features, and generate global feature sorting results.

[0100] In this embodiment, the neighborhood integration features of all nodes are first standardized, using a method similar to step S1423. Then, global feature statistics are performed on the standardized neighborhood integration features. For example, the mean, standard deviation, maximum, and minimum values ​​of the neighborhood integration features of all nodes are calculated in each dimension. These statistics reflect the overall distribution of global node features. Simultaneously, the covariance or correlation coefficient between nodes is calculated; these coefficients reflect the mutual influence of global node features. The above statistics and coefficients are then organized to generate a global feature analysis result.

[0101] Step S1439: Perform feature integration on the global feature sorting results, transform the global feature distribution and mutual influence into feature forms that can reflect the overall correlation, and generate overall correlation features.

[0102] In this embodiment, the method for integrating the global feature analysis results involves comprehensively processing the global feature distribution and mutual influence to form a feature form that reflects the overall correlation. For example, the statistics of the global feature distribution and the coefficients of mutual influence are concatenated to form a high-dimensional vector, which serves as the overall correlation feature.

[0103] Step S14310: Integrate the local association features, the intermediate-level association features, and the overall association features to generate multi-level network association features.

[0104] In this embodiment, local association features, intermediate-level association features, and global association features are integrated. This integration can be achieved by concatenation, where the vectors of these three features are concatenated to form a new vector, which serves as the multi-level network association feature. For example, the vector of local association features is v_local=[l1, l2, ..., lp], the vector of intermediate-level association features is v_middle=[m1, m2, ..., mq], the vector of global association features is v_global=[g1, g2, ..., gr], and the vector of multi-level network association features is v_multi=[l1, l2, ..., lp, m1, m2, ..., mq, g1, g2, ..., gr].

[0105] Step S144: Input the multi-level network association features into the feature fusion layer of the deep learning simulation module, dynamically calculate the fusion ratio of different levels of network association features through the attention mechanism, and perform weighted fusion of the multi-level network association features to generate global fusion network features.

[0106] In this embodiment, the feature fusion layer of the deep learning simulation module includes an attention mechanism module for dynamically calculating the fusion ratio of network-related features at different levels. First, multi-level network-related features are input into the attention mechanism module, which assigns an attention weight to each level of network-related feature. The magnitude of the weight reflects the importance of that level of feature in the fusion process. The attention weight can be calculated based on the importance or relevance of the features, for example, by calculating the relevance between each level of feature and the final output. Then, based on the aforementioned attention weights, the multi-level network-related features are weighted and fused. The weighted fusion method involves multiplying each level of network-related feature by its corresponding attention weight, and then concatenating all the weighted features to form a globally fused network feature. For example, if the attention weight for local association features is w_local, the attention weight for intermediate-level association features is w_middle, and the attention weight for global association features is w_global, then the vector of global fused network features is v_fused=[w_local×v_local,w_middle×v_middle,w_global×v_global] (here, multiplication is each element of the vector multiplied by its weight, and then concatenated).

[0107] Step S145: Input the global fusion network features into the first simulation branch of the deep learning simulation module to perform a uniformity analysis of the light distribution in various indoor areas, extract the changes in light intensity and light coverage in various indoor areas at different times, and generate a uniform lighting performance in various indoor areas.

[0108] Step S1451: Extract the feature parts related to the light distribution in each indoor area from the global fusion network features. The feature parts include the light intensity feature content and the light coverage feature content of each indoor area at different times.

[0109] In this embodiment, the global fusion network features include network association features at multiple levels. The features related to the light distribution in different indoor areas need to be extracted. These features can be determined by analyzing the dimensions and meanings of the global fusion network features. For example, features of certain dimensions correspond to the light intensity features of different indoor areas at different times, while features of other dimensions correspond to the light coverage features of different indoor areas at different times. By performing feature selection or dimensional division on the global fusion network features, these features related to the light distribution in different indoor areas can be extracted.

[0110] Step S1452: Classify the features according to indoor area division and time period division to generate regional time period feature groups corresponding to different time periods for each indoor area.

[0111] In this embodiment, indoor areas can be divided according to the functional areas of the building. For example, the indoor areas of a commercial building can be divided into shop areas, corridor areas, public rest areas, etc.; time periods can be divided according to hours, days, months, years, etc. The extracted features related to the light distribution of each indoor area are categorized according to indoor area and time period. For example, for each indoor area, the light intensity features and light coverage features for different time periods are categorized separately to form regional time period feature groups corresponding to that indoor area for different time periods.

[0112] Step S1453: Perform time-series change analysis on the light intensity characteristics of each regional time period characteristic group, extract the light intensity change content of the same indoor area in adjacent time periods, and generate a regional time period light intensity change sequence.

[0113] In this embodiment, time-series variation analysis is required for the light intensity characteristics within each regional time-segment characteristic group. The method for time-series variation analysis can be to calculate the difference or rate of change of light intensity between adjacent time periods. For example, if the light intensity characteristic of a certain indoor area at a certain time period t is I_t, and the light intensity characteristic of time period t+1 is I_t+1, then the change in light intensity between adjacent time periods is ΔI_t = I_t+1 - I_t (or the rate of change is ΔI_t / I_t). Arranging these changes in light intensity between adjacent time periods in chronological order generates a regional time-segment light intensity variation sequence.

[0114] Step S1454: Perform trend analysis on the light intensity change sequence of the area over a time period, and extract the significant changes, average changes, and frequency of changes in the change sequence as the light intensity changes in different areas of the indoor space at different times.

[0115] In this embodiment, the method for analyzing the trend of light intensity variation sequences in a region over a given time period can be based on statistical analysis or machine learning algorithms. For example, significant changes can be extracted by calculating the extreme values ​​in the change sequence, i.e., finding the maximum and minimum values ​​in the change sequence; the changes corresponding to these extreme values ​​are the significant changes. Average changes can be obtained by calculating the mean of the change sequence, which reflects the average level of light intensity changes in adjacent time periods. Change frequency can be obtained by calculating the number of transitions in the direction of change (increase or decrease) in the change sequence; the more transitions, the higher the change frequency. The extracted significant changes, average changes, and change frequencies are then compiled to represent the light intensity variations in different areas of the indoor space at different times.

[0116] Step S1455: Perform spatial variation analysis on the light coverage range characteristics of each regional time period characteristic group, extract the light coverage area variation and light coverage pattern variation of the same indoor area at different time periods, and generate a regional time period light coverage variation sequence.

[0117] In this embodiment, spatial variation analysis is required for the light coverage range characteristics of each regional time period characteristic group. The spatial variation analysis method can be to calculate the difference in light coverage area and the difference in light coverage pattern between different time periods. For example, if the light coverage area of ​​a certain indoor area at a certain time period t is A_t, and the light coverage area at time period t+1 is A_t+1, then the change in light coverage area is ΔA_t = A_t+1 - A_t. The change in light coverage pattern can be measured by calculating the shape difference of the light coverage area between the two time periods, for example, using a shape similarity index (such as Hausdorff distance). Arranging the above-mentioned changes in light coverage area and light coverage pattern at different time periods in chronological order generates a regional time period light coverage variation sequence.

[0118] Step S1456: Perform change feature analysis on the light coverage change sequence of the area during the time period, and extract the maximum change content, average change content, and change duration content in the change sequence as the change of light coverage range of each area in the room at different time periods.

[0119] In this embodiment, the method for analyzing the change characteristics of the regional time-segment light coverage change sequence is similar to step S1454. The maximum change content can be extracted by calculating the extreme values ​​in the change sequence, i.e., finding the maximum and minimum values ​​in the change sequence; the change content corresponding to these extreme values ​​is the maximum change content. The average change content can be obtained by calculating the mean of the change sequence, which reflects the average level of light coverage change in different time periods. The duration of change can be obtained by calculating the number of consecutive time periods with the same direction of change (increase or decrease) in the change sequence; the more consecutive time periods, the longer the duration of change. The extracted maximum change content, average change content, and duration of change content are then organized to represent the changes in light coverage in different areas of the indoor space at different time periods.

[0120] Step S1457: Integrate the changes in light intensity and light coverage of different areas of the indoor space at different times to generate a balanced lighting performance for each area of ​​the indoor space.

[0121] In this embodiment, the changes in light intensity and light coverage in different areas of the room at different times are integrated. The integration method can be splicing, that is, splicing the feature vectors of the two cases to form a new vector, which serves as the representation of the balanced lighting in different areas of the room. For example, if the vector of light intensity change is v_intensity=[i1, i2, ..., ip], and the vector of light coverage change is v_coverage=[c1, c2, ..., cq], then the vector of the balanced lighting in different areas of the room is v_balance=[i1, i2, ..., ip, c1, c2, ..., cq].

[0122] Step S146: Input the global fusion network features into the second simulation branch of the deep learning simulation module to perform time-series analysis on the lighting duration of each indoor area, extract the effective lighting time and continuous lighting time of each indoor area in different seasons and on different dates, and generate the lighting duration performance of each indoor area.

[0123] Step S1461: Extract the feature parts that are continuously related to the illumination of each indoor area from the global fusion network features. The feature parts include the illumination start time feature content and illumination end time feature content of each indoor area in different seasons and on different dates.

[0124] In this embodiment, the features related to the duration of illumination in different indoor areas need to be extracted from the global fusion network features. These features can be determined by analyzing the dimensions and meanings of the global fusion network features. For example, some dimensions of features correspond to the illumination start time features of different indoor areas in different seasons and on different dates, while other dimensions of features correspond to the illumination end time features of different indoor areas in different seasons and on different dates. By performing feature selection or dimensional division on the global fusion network features, these features related to the duration of illumination in different indoor areas can be extracted.

[0125] Step S1462: Classify the features according to indoor area division, season division, and date division to generate regional seasonal date feature groups for each indoor area corresponding to different seasons and dates.

[0126] In this embodiment, the methods for dividing indoor areas, seasons, and dates are similar to those in step S1452. The extracted features related to the continuous illumination of each indoor area are categorized according to indoor area, season, and date. For example, for each indoor area, the features of the start time and end time of illumination for different seasons and dates are categorized separately to form regional seasonal date feature groups corresponding to that indoor area for different seasons and dates.

[0127] Step S1463: Calculate the time span of the light start time feature and light end time feature in the seasonal date feature group of each region, and extract the total light time content of each indoor region in the corresponding season and date.

[0128] In this embodiment, for the illumination start time and illumination end time features within each regional seasonal date feature group, a time span calculation is required. The method for calculating the time span is to subtract the illumination start time from the illumination end time to obtain the total illumination time. For example, if the illumination start time for a certain season and a certain date in a certain indoor area is T_start and the illumination end time is T_end, then the total illumination time is T_total = T_end - T_start.

[0129] Step S1464: Set an effective light intensity standard. Based on the light intensity characteristics in the seasonal date characteristic group of the region, extract the time content in which the light intensity of each indoor area meets the effective light intensity standard within the total light time content of the corresponding season and date.

[0130] In this embodiment, the effective illuminance standard needs to be determined based on the usage requirements of the indoor space and the comfort requirements of the human body. For example, for the shop area of ​​a commercial building, the effective illuminance standard can be set to 300 lux (lx) or higher. Then, based on the illuminance characteristics in the seasonal and date characteristic groups of the area, the time periods within the total illuminance time content that meet the effective illuminance standard are identified. For example, for each time t within the total illuminance time content, it is determined whether its illuminance I_t is greater than or equal to the effective illuminance standard I_standard. If so, that time belongs to the time content that meets the standard. The above time content that meets the standard is organized to obtain the time content that meets the effective illuminance standard for each indoor area in the corresponding season and date.

[0131] Step S1465: Calculate the total duration of the time content that meets the effective light intensity standard, as the effective light duration of each indoor area in different seasons and on different dates.

[0132] In this embodiment, for each indoor area, the total duration of time that meets the effective light intensity standard for the corresponding season and date needs to be calculated. The total duration is calculated by adding the durations of all time periods that meet the standard. For example, if the time periods that meet the standard are [t1_start, t1_end], [t2_start, t2_end], ..., [tn_start, tn_end], then the total duration is Σ(ti_end - ti_start), where i ranges from 1 to n. The above total duration is then organized to represent the effective light intensity time of each indoor area in different seasons and on different dates.

[0133] Step S1466: Perform a continuity analysis on the changes in light intensity for each indoor area within the total illumination time for the corresponding season and date, and extract the time period when the light intensity continuously meets the effective light intensity standard without interruption.

[0134] In this embodiment, a continuity analysis is required for the changes in light intensity in each indoor area within the total illumination time for the corresponding season and date. The method for continuity analysis is to identify time periods where the light intensity consistently meets the effective light intensity standard without interruption. For example, starting from the beginning of the total illumination time, the light intensity at each moment is sequentially checked to ensure it meets the standard. When multiple consecutive moments meet the standard, the start and end times of that time period are recorded, forming an uninterrupted time segment. These uninterrupted time segments are then organized to obtain the time segments where the light intensity in each indoor area consistently meets the effective light intensity standard without interruption for the corresponding season and date.

[0135] Step S1467: Calculate the longest and average duration of the uninterrupted time content as the continuous illumination time of various indoor areas in different seasons and on different dates.

[0136] In this embodiment, for each indoor area where the light intensity continuously meets the effective light intensity standard without interruption for the corresponding season and date, it is necessary to calculate its longest and average duration. The longest duration is calculated by finding the longest duration among all uninterrupted time periods; the average duration is calculated by adding the durations of all uninterrupted time periods and then dividing by the number of uninterrupted time periods. The above longest and average durations are then compiled to represent the continuous light intensity duration for each indoor area in different seasons and dates.

[0137] Step S1468: Integrate the effective lighting time of each indoor area in different seasons and on different dates with the continuous lighting time of each indoor area in different seasons and on different dates to generate the continuous lighting performance of each indoor area.

[0138] In this embodiment, the effective illumination time and continuous illumination time of various indoor areas under different seasons and dates are integrated. The integration method can be splicing, that is, splicing the feature vectors of the two cases to form a new vector, which is used as the continuous illumination performance of each indoor area. For example, if the vector of effective illumination time is v_effective=[e1, e2, ..., ep], and the vector of continuous illumination time is v_continuous=[c1, c2, ..., cq], then the vector of continuous illumination performance of each indoor area is v_duration=[e1, e2, ..., ep, c1, c2, ..., cq].

[0139] Step S147: Integrate the balanced lighting performance of each indoor area with the continuous lighting performance of each indoor area to generate a curtain wall lighting simulation result that includes the balanced lighting performance and the continuous lighting performance of each indoor area.

[0140] In this embodiment, the balanced lighting performance and the continuous lighting performance of each indoor area are integrated. The integration method can be splicing, that is, splicing the feature vectors of these two performances to form a new vector, which is used as the lighting simulation result of the curtain wall. For example, if the vector of the balanced lighting performance of each indoor area is v_balance=[b1, b2, ..., bp], and the vector of the continuous lighting performance of each indoor area is v_duration=[d1, d2, ..., dq], then the vector of the lighting simulation result of the curtain wall is v_simulation=[b1, b2, ..., bp, d1, d2, ..., dq].

[0141] Step S150: Generate curtain wall design optimization guidelines based on the curtain wall daylighting simulation results. The curtain wall design optimization guidelines are used to adjust curtain wall design elements to optimize indoor daylighting performance.

[0142] Step S151: Extract the lighting balance performance of each indoor area from the curtain wall lighting simulation results, extract the changes in light intensity and light coverage of each indoor area at different times, classify them according to the functional type of the indoor area, and generate the regional functional lighting balance sorting results.

[0143] In this embodiment, the functional types of indoor areas can be divided according to the building's usage. For example, the functional types of indoor areas in commercial buildings can be divided into shop areas, office areas, and public rest areas. The lighting balance performance of each indoor area is extracted from the curtain wall lighting simulation results. Then, based on the functional type of each indoor area, the changes in light intensity and light coverage at different times are classified. For example, the changes in light intensity and light coverage of all shop areas are grouped into one category, all office areas into another, and so on. The classified results are then organized to generate a comprehensive analysis of the lighting balance of different functional areas.

[0144] Step S152: Extract the continuous lighting performance of each indoor area from the curtain wall lighting simulation results, extract the effective illumination time and continuous illumination time of each indoor area in different seasons and on different dates, classify them according to the usage frequency of the indoor areas, and generate the continuous lighting performance analysis results of the area usage.

[0145] In this embodiment, the frequency of indoor area usage can be divided according to the building's usage. For example, the frequency of indoor area usage in commercial buildings can be divided into high-frequency usage areas (such as shop areas), medium-frequency usage areas (such as corridor areas), and low-frequency usage areas (such as equipment rooms). The continuous lighting performance of each indoor area is extracted from the curtain wall lighting simulation results. Then, based on the usage frequency of each indoor area, the effective illumination time and continuous illumination time for different seasons and dates are classified. For example, the effective illumination time and continuous illumination time of all high-frequency usage areas are grouped into one category, all medium-frequency usage areas into another, and so on. The classified results are then organized to generate a comprehensive analysis of the continuous lighting performance of each area.

[0146] Step S153: Compare the results of the regional functional lighting balance analysis with the preset regional functional lighting balance standard, extract the indoor areas that do not meet the standard and the corresponding feature content that does not meet the standard, and generate the balance optimization direction.

[0147] In this embodiment, the preset regional functional lighting balance standard needs to be determined based on the functional requirements and human comfort requirements of the indoor area. For example, for the shop area, the preset standard deviation of light intensity variation should not exceed a certain value, and the maximum value of light coverage variation should not exceed a certain value. The results of the regional functional lighting balance analysis are compared with the preset standard to identify indoor areas that do not meet the standard. Then, the characteristics of these non-compliant indoor areas are analyzed, such as excessive light intensity variation or excessive light coverage variation. Based on these characteristics, the direction for balance optimization is determined, such as adjusting the curtain wall structure to reduce light intensity variation or adjusting the curtain wall material to optimize light coverage variation.

[0148] Step S154: Compare the results of continuous lighting management of the area with the preset continuous lighting standards of the area, extract the indoor areas that do not meet the continuous lighting standards and the corresponding feature content that does not meet the standards, and generate continuous optimization directions.

[0149] In this embodiment, the preset standard for continuous daylighting in a given area needs to be determined based on the frequency and needs of use of the indoor area. For example, for high-frequency use areas, the preset effective illumination time should not be less than a certain duration, and the average duration of continuous illumination time should not be less than a certain duration. The results of the analysis of continuous daylighting in different areas are compared with the preset standard to identify indoor areas where the daylighting continuity does not meet the standard. Then, the characteristics of these indoor areas that do not meet the standard are analyzed, such as insufficient effective illumination time or insufficient continuous illumination time. Based on these characteristics, the direction for continuous optimization is determined, such as adjusting the curtain wall structure to increase the effective illumination time or adjusting the curtain wall material to extend the continuous illumination time.

[0150] Step S155: Extract the curtain wall design elements that need to be adjusted based on the balance optimization direction, determine the specific direction for adjusting the curtain wall structure arrangement or the light transmission performance of the curtain wall materials, and generate structural material adjustment suggestions.

[0151] In this embodiment, the design elements of the curtain wall that need adjustment are analyzed based on the direction of balance optimization. For example, if the direction of balance optimization is to reduce changes in light intensity, it may be necessary to adjust the curtain wall structure arrangement, such as changing the angle or size of the curtain wall panels to optimize the incident and reflected light; or to adjust the light transmittance of the curtain wall materials, such as selecting materials with more stable light transmittance to reduce changes in light intensity. The specific direction of adjustment is determined; for example, adjusting the curtain wall structure arrangement could involve increasing or decreasing the tilt angle of the curtain wall panels, and adjusting the light transmittance of the curtain wall materials could involve increasing or decreasing the light transmittance of the materials. These adjustment suggestions are then compiled to generate structural material adjustment suggestions.

[0152] Step S156: Extract the curtain wall design elements that need to be adjusted based on the continuous optimization direction, determine the specific direction for adjusting the curtain wall structure arrangement or the light transmission performance of the curtain wall materials, and generate supplementary structural material adjustment suggestions.

[0153] In this embodiment, the design elements of the curtain wall that need adjustment are analyzed based on the direction of continuous optimization. For example, if the direction of continuous optimization is to increase the effective illumination time, it may be necessary to adjust the curtain wall structure arrangement, such as changing the orientation or area of ​​the curtain wall to increase the incident time of light; or to adjust the light transmittance of the curtain wall material, such as selecting a material with higher light transmittance to increase the amount of light transmitted. The specific direction of adjustment is determined. For example, adjusting the curtain wall structure arrangement could mean changing the orientation of the curtain wall to south, and adjusting the light transmittance of the curtain wall material could mean replacing it with glass with higher light transmittance. The above adjustment suggestions are compiled to generate supplementary structural material adjustment suggestions.

[0154] Step S157: Integrate the structural material adjustment suggestions and the supplementary structural material adjustment suggestions, determine the priority order and specific scope of the adjustments, and generate curtain wall design optimization guidelines.

[0155] In this embodiment, structural material adjustment suggestions and supplementary structural material adjustment suggestions are integrated. First, the importance and urgency of each adjustment suggestion are analyzed to determine its priority. For example, adjustments that significantly impact human comfort have a higher priority; adjustments that are easy to implement can also have a higher priority. Then, the specific scope of each adjustment suggestion is determined. For example, adjustments to the curtain wall structure arrangement could involve adjusting the angles of certain curtain wall panels, while adjustments to the light transmittance of curtain wall materials could involve replacing materials in certain areas. The priority and specific scope of the adjustments are then organized to generate a curtain wall design optimization guide. This guide provides curtain wall designers with specific adjustment directions and implementation scopes to optimize indoor lighting performance.

[0156] Based on the same inventive concept, please refer to Figure 2 The diagram shows a schematic block diagram of a deep learning-based curtain wall lighting performance simulation system 100 provided in this application embodiment for performing the above-described deep learning-based curtain wall lighting performance simulation method. The deep learning-based curtain wall lighting performance simulation system 100 may include a communication unit 110, a machine-readable storage medium 120, and a processor 130.

[0157] In this embodiment, both the machine-readable storage medium 120 and the processor 130 are located within the deep learning-based curtain wall daylighting performance simulation system 100 and are separately configured. However, it should be understood that the machine-readable storage medium 120 may also be independent of the deep learning-based curtain wall daylighting performance simulation system 100 and may be accessed by the processor 130 via a bus interface. Alternatively, the machine-readable storage medium 120 may be integrated into the processor 130 and may communicate with external systems via the communication unit 110.

[0158] The processor 130 is the control center of the deep learning-based curtain wall lighting performance simulation system 100. It connects various parts of the system via various interfaces and lines, and performs overall monitoring of the deep learning-based curtain wall lighting performance simulation system 100 by running or executing software programs and / or modules stored in the machine-readable storage medium 120, and by calling data stored in the machine-readable storage medium 120. Optionally, the processor 130 may include one or more processing cores; for example, the processor 130 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into the processor. The machine-readable storage medium 120 is used to store machine-executable instructions for executing the scheme of this application, and the processor 130 is used to execute the machine-executable instructions stored in the machine-readable storage medium 120 to implement the deep learning-based curtain wall daylighting performance simulation method provided in the aforementioned method embodiments.

[0159] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for simulating the daylighting performance of curtain walls based on deep learning, characterized in that, The method includes: The design elements of the curtain wall are integrated with the lighting effects of the target area. The curtain wall design elements include the curtain wall structure arrangement and the light transmission performance of the curtain wall materials. The lighting effects include the direction of light at different times and the changes in light intensity at different times. The curtain wall design elements and the lighting effect elements are input into a pre-trained daylighting association network to construct a model, generating the distribution of light reception performance of each part of the curtain wall and the distribution of light diffusion performance in the indoor space. A dynamic correlation network for curtain wall lighting is constructed based on the distribution of light reception and the distribution of light diffusion. This dynamic correlation network is used to present the dynamic relationship between curtain wall design elements, lighting effect elements and indoor lighting performance. The deep learning simulation module is activated to perform interactive simulation calculations on the dynamic correlation network of the curtain wall lighting, generating curtain wall lighting simulation results that include the balanced lighting performance of each indoor area and the continuous lighting performance of each indoor area. Based on the curtain wall daylighting simulation results, a curtain wall design optimization guide is generated. The curtain wall design optimization guide is used to adjust the curtain wall design elements to optimize the indoor daylighting performance.

2. The method for simulating the daylighting performance of curtain walls based on deep learning according to claim 1, characterized in that, The process of inputting the curtain wall design elements and the lighting effect elements into a pre-trained daylighting association network to construct a model generates the distribution of light reception characteristics at various parts of the curtain wall and the distribution of light diffusion characteristics in the interior space, including: The curtain wall structure arrangement in the curtain wall design elements is spatially digitally transformed to form a digital form of the curtain wall structure with coordinate attributes. Each structural part in the digital form of the curtain wall structure has corresponding position parameters and size parameters. The light transmission performance of the curtain wall material in the curtain wall design elements is transformed to form a material light transmission feature form that can be identified by the light-transmitting network construction model. The dimension of the material light transmission feature form matches the dimension of the input layer of the light-transmitting network construction model. The direction of the light in different time periods in the light effect elements is transformed to form a set of light effect directions. Each direction information in the set of light effect directions corresponds to the light incident direction in a specific time period. The changes in light intensity at different times in the light-affecting elements are transformed into a time sequence to form a time sequence feature of light intensity. The time interval of the time sequence feature of light intensity is consistent with the time segmentation of the set of light-affecting directions. The digital form of the curtain wall structure, the light transmission characteristics of the material, the set of light direction, and the temporal characteristics of light intensity are all input into a pre-trained lighting association network to construct the model input layer. The structural feature processing layer of the model constructed by the light-tracing association network performs spatial analysis on the digital morphology of the curtain wall structure, generating the spatial position features of each part of the curtain wall. The material feature processing layer of the model constructed by the light-transmitting network performs attribute analysis on the light-transmitting feature form of the material to generate the material light-transmitting attribute features of each part of the curtain wall. The illumination feature processing layer of the model constructed by the light-collecting association network performs spatiotemporal correlation between the set of illumination directions and the temporal characteristics of illumination intensity to generate illumination features for each time period. The element interaction layer of the model constructed by the lighting association network integrates the spatial location features, the material light transmission properties, and the lighting effect features across elements to generate comprehensive lighting reception features for each part of the curtain wall. Based on the spatial distribution of the comprehensive characteristics of light reception in various parts of the curtain wall, the distribution of light reception performance in various parts of the curtain wall is generated; The indoor diffusion simulation layer, constructed through a light-receiving network model, simulates the indoor diffusion path and energy change process of light after it passes through the curtain wall, based on the light reception performance distribution and the spatial parameters of the digital morphology of the curtain wall structure, thereby generating the indoor space light diffusion performance distribution.

3. The method for simulating the daylighting performance of curtain walls based on deep learning according to claim 1, characterized in that, The construction of a dynamic correlation network for curtain wall lighting based on the distribution of light reception and the distribution of light diffusion includes: The light intensity, light duration, and light incidence angle of each part of the curtain wall are extracted from the light reception performance distribution and used as the first type of node information of the curtain wall lighting dynamic association network. Extract the light arrival intensity, light coverage, and light balance of each indoor area from the light diffusion distribution, and use them as the second type of node information in the curtain wall lighting dynamic association network. The curtain wall panel size, panel splicing gap, and supporting frame shading are extracted from the curtain wall structural arrangement of the curtain wall design elements and used as the third type of node information in the dynamic association network of curtain wall lighting. Extract the light transmittance, reflectance, and absorptivity of the curtain wall materials from the light transmittance of the curtain wall design elements, and use them as the fourth type of node information in the dynamic correlation network of curtain wall lighting. The peak intensity of light and the change of light direction at different times are extracted from the light effect elements and used as the fifth type of node information in the dynamic correlation network of curtain wall lighting. A node group is formed to construct a dynamic correlation network for curtain wall lighting. The node group includes curtain wall part nodes, indoor area nodes, curtain wall structure nodes, curtain wall material nodes, and lighting effect nodes. Each type of node corresponds to the above-mentioned different categories of node information. Analyze the interaction between the nodes of the curtain wall and the nodes of the interior area. Based on the diffusion path of light from the curtain wall to the interior area, establish the first type of connection relationship between the nodes of the curtain wall and the nodes of the interior area. The degree of the first type of connection relationship is determined based on the energy transfer during the light diffusion process. The interaction between the curtain wall component nodes and the curtain wall structure nodes is analyzed. Based on the influence of the curtain wall structure on the light reception of the curtain wall component, a second type of connection relationship between the curtain wall component nodes and the curtain wall structure nodes is established. The degree of the second type of connection relationship is determined based on the influence of structural shading on the light reception. The interaction between the curtain wall component nodes and the curtain wall material nodes is analyzed. Based on the influence of the curtain wall material on the light absorption and reflection of the curtain wall component, a third type of connection relationship between the curtain wall component nodes and the curtain wall material nodes is established. The degree of the effect of the third type of connection relationship is determined based on the influence of the material transmittance on the light transmission. The interaction between the nodes of the curtain wall and the nodes affected by lighting is analyzed. Based on the influence of lighting on the lighting reception of the curtain wall, a fourth type of connection relationship between the nodes of the curtain wall and the nodes affected by lighting is established. The degree of influence of the fourth type of connection relationship is determined based on the influence of lighting intensity and lighting direction on the lighting reception of the curtain wall. The interaction between indoor area nodes and curtain wall structure nodes is analyzed. Based on the influence of the curtain wall structure on the light diffusion in the indoor area, a fifth type of connection relationship between indoor area nodes and curtain wall structure nodes is established. The degree of influence of the fifth type of connection relationship is determined based on the influence of the structural arrangement on the indoor light distribution. The interaction between indoor area nodes and curtain wall material nodes is analyzed. Based on the influence of curtain wall materials on indoor light intensity, a sixth type of connection relationship between indoor area nodes and curtain wall material nodes is established. The degree of influence of the sixth type of connection relationship is determined based on the influence of material transmittance on indoor light intensity. By integrating the above-mentioned nodes, node information, and connection relationships, a complete dynamic correlation network for curtain wall lighting is formed. This dynamic correlation network is used to present the dynamic interaction between curtain wall design elements, lighting effect elements, and indoor lighting performance.

4. The method for simulating the daylighting performance of curtain walls based on deep learning according to claim 1, characterized in that, The deep learning simulation module is activated to perform interactive simulation calculations on the curtain wall daylighting dynamic correlation network, generating curtain wall daylighting simulation results that include the balanced and continuous daylighting performance of various indoor areas, including: The dynamic association network for curtain wall lighting is transformed into a network structure, converting the node information and connection relationship information in the dynamic association network for curtain wall lighting into a network feature form that can be recognized by the deep learning simulation module. The number of rows in the network feature form corresponds to the number of nodes in the dynamic association network for curtain wall lighting, and the number of columns corresponds to the sum of the node information dimension and the connection relationship feature dimension. The network feature form is input into the network feature processing layer of the deep learning simulation module, and the network feature form is integrated with the neighborhood features through network convolution operation to generate the neighborhood integrated features of each node. The neighborhood integration features are analyzed at different levels by using the multi-level network analysis layer of the deep learning simulation module. The local correlation features and overall correlation features of the curtain wall lighting dynamic correlation network are extracted respectively, and multi-level network correlation features are generated. The multi-level network association features are input into the feature fusion layer of the deep learning simulation module. The fusion ratio of the network association features at different levels is dynamically calculated through the attention mechanism. The multi-level network association features are then weighted and fused to generate global fused network features. The global fusion network features are input into the first simulation branch of the deep learning simulation module to perform a uniformity analysis of the light distribution in various indoor areas, extract the changes in light intensity and light coverage in various indoor areas at different times, and generate a uniform lighting performance in various indoor areas. The global fusion network features are input into the second simulation branch of the deep learning simulation module to perform time-series analysis on the lighting duration of each indoor area, extract the effective lighting time and continuous lighting time of each indoor area in different seasons and on different dates, and generate the lighting duration performance of each indoor area. By integrating the balanced lighting performance and the continuous lighting performance of each indoor area, a curtain wall lighting simulation result is generated that includes both the balanced lighting performance and the continuous lighting performance of each indoor area.

5. The method for simulating the daylighting performance of curtain walls based on deep learning according to claim 4, characterized in that, The process of integrating neighborhood features in the network feature processing layer of the deep learning simulation module to generate neighborhood integrated features for each node includes: Identify the group of directly adjacent nodes for each node in the dynamic association network of curtain wall lighting, wherein the group of directly adjacent nodes includes all nodes that are directly associated with that node through connection relationships; Calculate the degree of influence of the connection relationship between each node and each node in its group of directly adjacent nodes. The degree of influence of the connection relationship is determined based on the degree parameter of the connection relationship. The feature row information corresponding to each node in the network feature form is standardized with the feature row information corresponding to each node in the group of nodes directly adjacent to that node to eliminate the difference in feature dimensions and generate standardized feature row information. Perform element-level operations on the standardized feature row information to generate a set of feature operation results for adjacent nodes; Each result in the set of neighboring node feature operations is weighted and its corresponding connection relationship influence is weighted to generate a weighted set of neighboring node features. Element-wise summation is performed on all features in the weighted neighbor node feature set to generate a neighborhood feature summation result. The summation result of the neighborhood features is normalized to eliminate the difference in feature values ​​caused by the difference in the number of adjacent nodes of different nodes, and normalized neighborhood features are generated. The original feature row information corresponding to each node is combined with the regular neighborhood features of that node to generate the neighborhood integrated feature of each node.

6. The deep learning-based method for simulating the daylighting performance of curtain walls according to claim 4, characterized in that, The method involves performing network structure analysis at different levels on the neighborhood integration features through a multi-level network analysis layer of the deep learning simulation module, extracting local and overall correlation features of the curtain wall lighting dynamic correlation network, and generating multi-level network correlation features, including: A first-level analysis range is defined, which corresponds to the direct adjacent range of nodes in the dynamic association network of curtain wall lighting, and is used to extract local association features. Within the scope of the first level of analysis, the neighborhood integration features of each node are standardized, and the neighboring node feature difference analysis is performed on the standardized neighborhood integration features to extract the feature differences between the node and its direct neighboring nodes and generate local feature difference information. Feature filtering is performed on the local feature differences to retain key feature content that can reflect local correlations and generate local correlation features; A second-level analysis range is defined, which corresponds to the indirect adjacent range of nodes in the dynamic association network of curtain wall lighting, i.e. the direct adjacent range of directly adjacent nodes, and is used to extract intermediate-level association features. Within the second-level analysis scope, the neighborhood integration features of each node are standardized, and the indirect neighbor node feature transmission analysis is performed on the standardized neighborhood integration features. The influence of the indirect neighbor node features transmitted by the node through the direct neighbor node is extracted to generate the intermediate-level feature transmission situation. The intermediate-level feature transmission is refined, and key feature content that can reflect the intermediate-level correlation is retained to generate intermediate-level correlation features. A third-level analysis scope is defined, which corresponds to the entire node range of the curtain wall lighting dynamic association network, and is used to extract overall association features; Within the scope of the third-level analysis, the neighborhood integration features of all nodes are standardized, and global feature statistics are performed on the standardized neighborhood integration features to extract the overall distribution and mutual influence of global node features, generating global feature analysis results. The global feature analysis results are integrated to transform the global feature distribution and mutual influence into a feature form that can reflect the overall correlation, thereby generating overall correlation features. By integrating the local association features, the intermediate-level association features, and the overall association features, a multi-level network association feature is generated.

7. The deep learning-based method for simulating the daylighting performance of curtain walls according to claim 4, characterized in that, The process involves inputting the global fusion network features into the first simulation branch of the deep learning simulation module to perform a uniformity analysis of the light distribution in different indoor areas. This includes extracting the changes in light intensity and light coverage in different indoor areas at different times, and generating a balanced lighting performance for each indoor area. From the global fusion network features, extract the feature parts related to the light distribution in various indoor areas. The feature parts include the light intensity feature content and the light coverage feature content of various indoor areas at different times. The feature components are categorized according to indoor area division and time period division to generate regional time period feature groups corresponding to different time periods for each indoor area; A time-series change analysis of the light intensity characteristics in each regional time period characteristic group was performed to extract the light intensity change content of the same indoor area in adjacent time periods and generate a regional time period light intensity change sequence. The change trend analysis of the light intensity change sequence of the area during the time period is carried out, and the significant change content, average change content and change frequency content in the change sequence are extracted as the light intensity change of each area of ​​the indoor space at different time periods. Spatial variation analysis was performed on the light coverage range characteristics of each regional time period characteristic group to extract the changes in light coverage area and light coverage pattern of the same indoor area at different time periods, and a regional time period light coverage change sequence was generated. The change characteristics of the light coverage change sequence in the area during different time periods are analyzed, and the maximum change content, average change content, and change duration content in the change sequence are extracted as the changes in the light coverage range of different indoor areas at different time periods. By integrating the changes in light intensity and light coverage of different areas of the indoor space at different times, a balanced lighting performance is generated for each area of ​​the indoor space.

8. The method for simulating the daylighting performance of curtain walls based on deep learning according to claim 4, characterized in that, The process involves inputting the global fusion network features into the second simulation branch of the deep learning simulation module to perform a time-series analysis of the lighting duration in various indoor areas. This process extracts the effective lighting time and continuous lighting time for different seasons and dates in each indoor area, generating a representation of the lighting duration in each area. This includes: Extract the feature portion that is continuously related to the illumination of each indoor area from the global fusion network features. The feature portion includes the illumination start time feature content and illumination end time feature content of each indoor area in different seasons and on different dates. The feature components are categorized according to indoor area division, season division, and date division to generate regional seasonal date feature groups corresponding to different seasons and dates for each indoor area. The time span of the light start time feature and light end time feature in the seasonal date feature group of each region is calculated, and the total light time of each indoor area in the corresponding season and date is extracted. An effective light intensity standard is set, and based on the light intensity characteristics in the seasonal and date characteristic groups of the region, the time content in which the light intensity of each indoor area meets the effective light intensity standard is extracted within the total light time content of the corresponding season and date. Calculate the total duration of the time content that meets the effective light intensity standard, and use it as the effective light duration of various indoor areas in different seasons and on different dates; A continuous analysis of the changes in light intensity in each indoor area within the total illumination time of the corresponding season and date was conducted to extract the time period in which the light intensity continuously met the effective light intensity standard without interruption. Calculate the longest and average duration of the uninterrupted time content to represent the continuous illumination time of different areas of the room in different seasons and on different dates; By integrating the effective lighting time of each indoor area in different seasons and on different dates with the continuous lighting time of each indoor area in different seasons and on different dates, a continuous lighting performance of each indoor area is generated.

9. The method for simulating the daylighting performance of curtain walls based on deep learning according to claim 1, characterized in that, The process of generating curtain wall design optimization guidelines based on the curtain wall daylighting simulation results includes: Extract the lighting balance performance of each indoor area from the curtain wall lighting simulation results, extract the changes in light intensity and light coverage of each indoor area at different times, classify them according to the functional type of the indoor area, and generate the regional functional lighting balance sorting results. Extract the continuous lighting performance of each indoor area from the curtain wall lighting simulation results, extract the effective lighting time and continuous lighting time of each indoor area in different seasons and on different dates, classify them according to the usage frequency of indoor areas, and generate the continuous lighting analysis results of the area usage. The results of the regional functional lighting balance analysis are compared with the preset regional functional lighting balance standards. Indoor areas that do not meet the standards and their corresponding non-compliant features are extracted to generate a balance optimization direction. The results of continuous lighting analysis of the area are compared with the preset continuous lighting standards for the area. Indoor areas that do not meet the continuous lighting standards and their corresponding non-compliant features are extracted to generate continuous optimization directions. Based on the aforementioned balance optimization direction, extract the curtain wall design elements that need to be adjusted, determine the specific direction for adjusting the curtain wall structure arrangement or the light transmission performance of the curtain wall materials, and generate structural material adjustment suggestions; Based on the continuous optimization direction, extract the curtain wall design elements that need to be adjusted, determine the specific direction for adjusting the curtain wall structure arrangement or the light transmission performance of the curtain wall materials, and generate supplementary structural material adjustment suggestions; By integrating the proposed structural material adjustment suggestions with the supplementary structural material adjustment suggestions, the priority order and specific scope of the adjustments are determined, and a curtain wall design optimization guideline is generated.

10. A deep learning-based system for simulating the daylighting performance of curtain walls, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the deep learning-based curtain wall daylighting performance simulation method according to any one of claims 1 to 9 by executing the machine-executable instructions.

Images