Indoor light comfort parameter soft measurement method and system for edge calculation

Abstract

The invention relates to an edge calculation-oriented indoor light comfort parameter soft measurement method and system, and the method comprises the steps: collecting key sample variables, carrying out the sampling and processing of the key sample variables, generating a plurality of target samples, and constructing a corresponding typical room scene for each target sample; for each typical room scene, calculating illuminance and daylight glare probability hour by hour, taking the illuminance and daylight glare probability as data labels, and constructing a training data set; training a neural network model by using the training data set to obtain an indoor light comfort soft measurement model, carrying out SHAP sensitivity analysis to obtain the importance of each type of key sample variables, and carrying out lightweight processing based on the importance to obtain an optimized soft measurement model; key variables are collected in real time, and indoor light comfort prediction is carried out by utilizing the optimized soft measurement model based on the key variables; the system is used for implementing the method. Compared with the prior art, the method provided by the invention not only can ensure the accurate measurement of high-precision indoor light comfort parameters, but also can ensure the measurement efficiency.

Description

Technical Field

[0001] This invention relates to the field of architectural design technology, and in particular to a soft measurement method and system for indoor light comfort parameters oriented towards edge computing. Background Technology

[0002] Intelligent control of building lighting environment is a key technology for improving energy efficiency and indoor occupant comfort. Sunlight, as the primary source of indoor lighting, has significant benefits for reducing lighting energy consumption and improving visual comfort and mental health. However, sunlight is dynamic, and improper control can lead to problems such as insufficient illuminance, uneven distribution, or glare. To achieve precise lighting environment control, the core lies in real-time acquisition of key indicators that directly reflect human comfort. Among these, work surface illuminance and daylight glare probability (DGP) are considered two representative parameters. Therefore, existing technologies often use the above two parameters as known quantities when predicting or evaluating indoor lighting comfort. For example, Chinese patent application CN116680779A provides a model feature analysis method based on a prediction model for indoor sunlight perception evaluation in buildings. This method aims to address the problem that existing sunlight perception evaluation studies cannot reflect the subjective feelings of users and the influence weight of environmental parameters. In this method, parameters such as the probability of sunlight glare are obtained based on field experiments in typical building spaces. However, the acquisition methods for the two key parameters, illuminance on the working surface and the probability of sunlight glare, have obvious limitations. Specifically, although illuminance measurement can be achieved through sensors, obtaining detailed distribution data within the space requires the deployment of a large number of sensors, which brings cost and maintenance challenges. The evaluation of DGP is even more complex, requiring panoramic images to be captured through fisheye lenses and analyzed by professional software. This process is cumbersome, requires high-end equipment, and poses a risk of privacy interference. To address the aforementioned technical issues, a physical simulation-based tool for acquiring key parameters has been proposed, such as Chinese patent application CN111259481A. This method provides an evaluation method for building interior lighting environment design that integrates spatial information. To avoid the inconvenience of measured data, it uses simulation software for calculation. While providing accurate results, this method suffers from complex modeling and high computational resource consumption, making it completely unsuitable for real-time control requirements. Furthermore, this type of method also suffers from the following problems: 1) It often employs complex network structures, resulting in high computational overhead and requiring reliance on cloud servers or high-performance workstations for computation, thus limiting its use to offline analysis tools; 2) It lacks a systematic design for lightweight models and fails to consider the resource constraints of edge devices, making it impossible to maintain accuracy while meeting the computational efficiency requirements of real-time inference.

[0003] Therefore, providing a method that can ensure both high-precision measurement of indoor light comfort parameters and measurement efficiency is a technical problem that needs to be solved. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a solution.

[0005] The objective of this invention can be achieved through the following technical solutions: According to a first aspect of the present invention, a soft measurement method for indoor light comfort parameters oriented towards edge computing is provided, comprising: Key sample variables were collected, including geographical and climatic parameters, building spatial parameters, surface material parameters, and spatial characteristics of observation points. The key sample variables are sampled and processed to generate multiple target samples, and a corresponding typical room scene is constructed for each target sample. For each typical room scenario, multiple observation points are deployed, and the illuminance and solar glare probability of each observation point are calculated hourly. The illuminance and solar glare probability are used as data labels for the corresponding key sample variables to construct a training dataset. A neural network model is trained using the training dataset to obtain an indoor light comfort soft measurement model. SHAP sensitivity analysis is performed on the soft measurement model to obtain the importance of each key sample variable. Based on the importance, the soft measurement model is lightweighted to obtain an optimized soft measurement model. Key variables are collected in real time, and indoor light comfort is predicted based on the optimized soft measurement model using these key variables.

[0006] As a preferred technical solution, the geographical and climatic parameters include geographical location, altitude, solar azimuth angle, solar normal direct radiation, and solar horizontal diffuse radiation; the geographical location includes longitude and latitude. The architectural space parameters include the room's length, width, and height, windowsill height, window height, window-to-wall ratio, and window orientation; The surface material parameters include wall reflectivity, ceiling reflectivity, floor reflectivity, and window transmittance; The spatial characteristics of the observation point include the original three-dimensional coordinates of the observation point and the orientation of the observation point; Among them, the window orientation, observation point orientation, and geographical location are discrete variables, while the remaining key sample variables are continuous variables.

[0007] As a preferred technical solution, the method for generating the target sample is as follows: The continuous variables were sampled using the Latin hypercube sampling strategy to obtain a standardized sample. After performing strategic discretization on the discrete variables, they are randomly allocated to each of the standard samples according to a preset ratio to obtain the target samples.

[0008] As a preferred technical solution, the soft measurement model includes an input layer, a feature encoder, a cross-attention mechanism layer, a feature fusion network, and a dual-head output layer. The feature encoder includes a spatial encoder, a radiation environment encoder, and a material and geo-encoder. The soft measurement model performs the following steps to predict illuminance and solar glare probability: A series of data sets are randomly selected from the training dataset and used as input to the soft measurement model; the data sets include data labels and corresponding key sample variables. Feature processing is performed on the key sample variables to obtain spatial geometric features, radiation environment features, and material and geographical features; The aforementioned spatial geometric features, radiation environment features, and material and geographic features are processed using a spatial encoder, a radiation environment encoder, and a material and geographic encoder, respectively. The outputs of the three encoders are fused using the cross-attention mechanism layer to output fused features; the fused features are then processed by a feature fusion network to obtain recognition features. After the aforementioned recognition features are processed by the dual-head output layer, the illuminance and the probability of sunlight glare are output respectively.

[0009] As a preferred technical solution, the spatial geometric features include the room's length, width, and height; window height; windowsill height; window-to-wall ratio; three-dimensional relative coordinates between the observation point and the window; distance between the observation point and the window; and X and Y axis vectors of the observation point's orientation. Based on the key sample features, the following feature processing is used to obtain the three-dimensional relative coordinates between the observation point and the window, the distance between the observation point and the window, and the X and Y axis vectors of the observation point's orientation: A target coordinate system is constructed using the lowest point of the line where the west and south walls of the room intersect as the unified spatial origin. In the target coordinate system, the coordinates of the window center point are calculated according to the window orientation. Based on the coordinates of the window center point and the original three-dimensional coordinates of the observation point, the three-dimensional relative position between the target observation point and the window, and the distance from the target observation point to the window center point are calculated. The X-axis and Y-axis vectors of the observation point orientation are obtained based on the observation point orientation and the original three-dimensional coordinates of the observation point.

[0010] As a preferred technical solution, the radiation environment characteristics include the solar direction vector, solar normal direct radiation, solar horizontal scattered radiation, effective direct radiation component, and effective scattered radiation; based on the key sample characteristics, the solar direction vector, effective direct radiation component, and effective scattered radiation are obtained through the following feature processing: Obtain the solar elevation angle, and convert the solar normal direct radiation into an effective direct radiation component to the horizontal plane based on the sine of the elevation angle; The sky scattering factor is calculated based on the aforementioned elevation angle, and the effective scattered radiation is calculated based on the aforementioned sky scattering factor. The solar direction vector is calculated based on the sine and cosine of the azimuth and altitude angles.

[0011] As a preferred technical solution, the material and geographical features include wall reflectivity, ceiling reflectivity, window transmittance, corrected floor effective reflectivity, longitude spherical coordinates, and latitude spherical coordinates; based on the key sample features, the corrected floor effective reflectivity, longitude spherical coordinates, and latitude spherical coordinates are obtained through the following feature processing: Convert longitude and latitude to spherical coordinates to obtain spherical longitude and spherical latitude; The ground reflectivity is corrected based on the ground reflectivity and window transmittance to obtain the corrected ground reflectivity.

[0012] As a preferred technical solution, the number of neurons in the input layer is equal to the number of key sample variables in the input soft measurement model; the number of neurons in the spatial encoder is equal to the number of spatial geometric features; the number of neurons in the radiation environment encoder is equal to the number of radiation environment features; and the number of neurons in the material and geospatial encoder is equal to the number of material and geospatial features.

[0013] As a preferred technical solution, the importance includes the importance of predicting the contrast ratio and the probability of solar glare for each type of key sample variable; the lightweight processing based on the importance is as follows: For illuminance prediction and solar glare probability prediction, the importance of each type of key sample variable is ranked. Key sample indicators with less than preset values ​​in illuminance prediction and solar glare probability prediction were selected respectively to construct a first set of redundant variables and a second set of redundant variables. Select the same key sample indicators from the first and second sets of redundant variables as redundant variables; Remove the input layer neurons, spatial encoder neurons, radiation environment encoder neurons, and material and geo-encoder neurons corresponding to redundant variables in the aforementioned soft measurement model.

[0014] According to a second aspect of the present invention, an edge computing-oriented soft measurement system for indoor light comfort parameters is provided for implementing the above-described method.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1) To address the problems existing in the prior art, this invention provides a soft measurement method for indoor light comfort. The method replaces measured values ​​with illuminance and solar glare probability obtained through simulation calculations. Specifically, it collects easily accessible key sample features, including geographical climate, architectural space, surface materials, and spatial characteristics of the observation point. Based on these key sample features, it simulates typical room scenarios and calculates solar glare probability and illuminance based on the simulated typical room scenarios. This avoids the need for deploying a large number of sensors, overcoming the limitations in obtaining illuminance and solar glare probability. The simulated illuminance and solar glare probability are used as data labels, and training data is constructed with the corresponding key sample variables to train the model. This allows the model to effectively learn the mapping relationship between key variables and illuminance and solar glare probability, thus enabling prediction of both based solely on easily obtainable parameters.

[0016] 2) This invention addresses edge computing applications by introducing a lightweight model processing mechanism based on feature importance analysis. This significantly reduces the model input dimension and computational complexity while ensuring prediction accuracy, enabling soft measurement models to achieve real-time inference and stable operation on edge devices with limited computing resources. Compared to existing technologies that rely on cloud servers or high-performance computing platforms, this invention reduces latency and network dependence caused by data transmission, improves system response speed and operational reliability, and is suitable for engineering applications such as real-time monitoring and intelligent control of indoor lighting environments. Attached Figure Description

[0017] Figure 1 This is a framework diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention; Figure 3 This is a model structure diagram of the indoor light comfort soft measurement model of the present invention; Figure 4 A flowchart illustrating the soft measurement of indoor light comfort parameters using the soft measurement model for indoor light comfort of this invention; Figure 5 This is a schematic diagram of the deployment of the indoor light comfort soft measurement system of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0019] To address the problems existing in the prior art, this invention provides a soft measurement method for indoor lighting comfort parameters oriented towards edge computing. The entire process includes, as follows: Figure 1 The four stages shown are data preparation and feature construction, model development and training, model optimization, and edge deployment. The detailed process is as follows: Figure 2 As shown, it includes: S1. Collect key sample variables.

[0020] This step corresponds to the data preparation stage. To ensure the soft sensor model has good generalization ability and spatial adaptability, it is necessary to select key variables with spatial and semantic expressive capabilities. In this embodiment, 21 key variables affecting the indoor lighting environment were selected, mainly covering four categories: geographical and climatic parameters, architectural spatial parameters, surface material parameters, and spatial characteristics of observation points. These features can comprehensively cover most physical factors, ensuring that the model has strong transferability and physical rationality. By selecting key sample parameters, it is ensured that all factors affecting the indoor lighting environment can be comprehensively captured, and that the model can be effectively predicted against brightness and DGP. The detailed data types of various key sample variables are shown in Table 1.

[0021] Table 1. Detailed Types of Key Sample Variables For ease of calculation, the window orientation, observation point orientation, and geographical location were simplified and set as discrete variables, while the remaining key sample variables were continuous variables.

[0022] S2. Sample and process key sample variables to generate multiple target samples, and construct a corresponding typical room scene for each target sample.

[0023] S21. The Latin hypercube sampling strategy is used to sample continuous variables to obtain a standardized sample.

[0024] S22. After performing strategic discretization on the discrete variables, they are randomly assigned to each standardized sample according to a preset ratio to obtain the target sample; this ensures that all categories are representative and avoids category bias.

[0025] S3. For each typical room scenario, deploy multiple observation points and calculate the illuminance and solar glare probability of each observation point hourly; use the illuminance and solar glare probability as data labels for the corresponding key sample variables to construct a training dataset.

[0026] For each sample, a typical room scenario is defined, and a model is built using simulation software. Multiple observation points are deployed in the scenario, and a simulation engine is used to calculate the illuminance and glare indices of each point every hour throughout the year. After screening, a large-scale training dataset including training and validation data is finally constructed.

[0027] The data in the training dataset is preprocessed, including cleaning outliers and removing invalid samples caused by simulation errors or data anomalies, to ensure the trainability of the target variable.

[0028] S4. Train a neural network model using the training dataset to obtain a soft measurement model for indoor light comfort. Perform SHAP sensitivity analysis on the soft measurement model to obtain the importance of each key sample variable. Based on the importance, perform lightweight processing on the soft measurement model to obtain an optimized soft measurement model.

[0029] In this embodiment, the trained soft measurement model structure includes an input layer, a feature encoder, a cross-attention mechanism layer, a feature fusion network, and a dual-head output layer. The number of neurons in the input layer is equal to the number of key sample variables in the input soft measurement model; the number of neurons in the spatial encoder is equal to the number of spatial geometric features; the number of neurons in the radiation environment encoder is equal to the number of radiation environment features; and the number of neurons in the material and geography encoder is equal to the number of material and geography features. Specifically, from... Figure 3 As can be seen, the feature encoder includes a spatial encoder, a radiation environment encoder, and a material and geo-encoder; the feature fusion network part includes three fully connected layers; the dual-head output layer includes an illuminance output head and a solar glare probability output head.

[0030] During model training, this embodiment employs a specific optimization strategy. Specifically, since glare from the sun has a significant negative impact on work when determining lighting comfort, and DGP is a non-linear risk indicator based on visual perception, the Huber loss from illuminance prediction and the binary cross-entropy loss from DGP prediction are used together. The composite loss function, composed of weights, is used in conjunction with the AdamW optimizer for parameter optimization; and the initial learning rate is set to... The learning rate is dynamically adjusted using the ReduceLROnPlateau strategy, and the numerical stability of the training process is ensured by gradient pruning.

[0031] Furthermore, the model continuously monitors its performance on the validation set during training and saves the model parameters at the optimal moment for validation loss. In this embodiment, the model's optimal weights on the validation set occur at epoch 417, at which point the illumination prediction on the validation set... The R² of the model is 0.9794, the RMSE is 341.49 lx, and the MAE is 63.28 lx; the R² of the DGP prediction is 0.9953, the RMSE is 0.0109, and the MAE is 0.0050. This indicates that the constructed soft sensor model achieves a goodness of fit close to 1 and a low error level on both target tasks. The overall accuracy can meet the requirements of indoor light environment prediction and control. The neural network model at the corresponding time point is selected as the soft sensor model.

[0032] After the above training is completed, the SHAP sensitivity analysis method is used to perform feature importance analysis on the trained model. This importance includes the importance of each key sample variable in predicting saturation and solar glare probability. Based on the importance, lightweight processing is performed as follows: For illuminance prediction and solar glare probability prediction, the importance of each type of key sample variable is ranked.

[0033] 1. Select the components with less importance than illuminance prediction and solar glare probability prediction respectively. Based on the key sample indicators, construct the first set of redundant variables and the second set of redundant variables.

[0034] Select the same key sample indicators from the first and second sets of redundant variables as redundant variables.

[0035] Remove redundant variables from the input layer neurons, spatial encoder neurons, radiation environment encoder neurons, and material and geo-encoder neurons in the soft measurement model.

[0036] After the above lightweighting process, in this embodiment, the redundant variables selected based on the results include wall reflectivity, ceiling reflectivity, window transmittance, and the corrected effective floor reflectivity. Therefore, in the actual inference process, it is not necessary to collect the above four types of data, and the corresponding neurons in the input layer and feature encoder are removed, thus achieving model lightweighting.

[0037] Soft measurement model execution according to as follows Figure 4 The following steps are shown to predict illuminance and solar glare probability: S41. Randomly select a series of data groups from the training dataset as the input of the soft measurement model input layer. Each data group includes data labels and corresponding key sample variables, and the input layer includes 25 neurons, each neuron corresponding to one type of input data.

[0038] S42. Perform feature processing on key sample variables to obtain spatial geometric features, radiation environment features, and material and geographical features.

[0039] First, the input data is separated into three categories: spatial geometry data, radiation environment data, and other data types.

[0040] i) The spatial geometric features are 12-dimensional features, including the room's length, width, and height, window height, windowsill height, window-to-wall ratio, three-dimensional relative coordinates between the observation point and the window, distance between the observation point and the window, and the X and Y axis vectors of the observation point's orientation.

[0041] Based on key sample features, the following feature processing is used to obtain the three-dimensional relative coordinates of the observation point and the window, the distance between the observation point and the window, and the X and Y axis vectors of the observation point's orientation: 1. Use the lowest point of the line where the west and south walls of the room meet as the origin of the unified space to construct the target coordinate system.

[0042] In the target coordinate system, calculate the coordinates of the window center point based on the window orientation. Calculate the three-dimensional relative position between the target observation point and the window, and the distance from the target observation point to the window center point, based on the window center point coordinates and the original three-dimensional coordinates of the observation point.

[0043] Based on the orientation of the observation point and its original three-dimensional coordinates, obtain the X-axis and Y-axis vectors of the observation point orientation.

[0044] By following the steps above, the geometric relationship between the point location and the center of the window can be explicitly described.

[0045] ii) The radiation environment features are 7-dimensional features, including solar direction vector, solar normal direct radiation, solar horizontal scattered radiation, effective direct component and effective scattered radiation.

[0046] Based on key sample features, the solar direction vector, effective direct radiation component, and effective scattered radiation are obtained through the following feature processing: Obtain the solar altitude angle, and convert the solar normal direct radiation into an effective direct radiation component to the horizontal plane based on the sine of the altitude angle.

[0047] The sky scattering factor is calculated based on the elevation angle, and the effective scattered radiation is calculated based on the sky scattering factor.

[0048] In detail, the sky scattering factor includes: , in, This indicates the solar altitude angle.

[0049] The solar direction vector is calculated based on the sine and cosine of the azimuth and altitude angles.

[0050] The above processing provides information about the radiation incident direction for the model.

[0051] iii) The material and geographical features are 6-dimensional features, including wall reflectivity, ceiling reflectivity, window transmittance, corrected floor effective reflectivity, longitude spherical coordinates, and latitude spherical coordinates.

[0052] Based on the key sample features, the following feature processing is performed to obtain the corrected effective reflectance, longitude spherical coordinates, and latitude spherical coordinates of the floor: Converting longitude and latitude into spherical coordinates to obtain spherical longitude and spherical latitude avoids the periodic boundary problems caused by directly using angles and more accurately represents the influence of the Earth's surface position on the sun's path.

[0053] The ground reflectivity is corrected based on the ground reflectivity and window transmittance to obtain the corrected ground reflectivity.

[0054] Based on the characteristic that floor reflective energy easily leaks out under strong direct sunlight, a combined term of ground reflectivity and glass transmittance is introduced to correct the floor reflectivity, making it more consistent with the physical laws of indoor light distribution. Specifically, the correction is as follows: Because the effect of floor reflection on the indoor lighting environment is mainly reflected in the window system reducing the background brightness formed after direct sunlight, rather than the direct sunlight itself; The term "item" indicates the relative effectiveness of floor reflection after the window suppresses direct sunlight.

[0055] S43. The spatial geometric features, radiation environment features, and material and geographic features are processed using the spatial encoder, radiation environment encoder, and material and geographic encoder, respectively, and the features at 128 latitudes are output.

[0056] The spatial geometry encoder is responsible for extracting 12-dimensional spatial geometric features. It uses a three-layer fully connected network with the GELU activation function, which helps to improve gradient flow and enhance training stability.

[0057] The radiation environment encoder processes 7-dimensional radiation environment features through a two-layer fully connected network.

[0058] The material and geoencoder processes 6-dimensional feature materials and geographic features through a two-layer fully connected network.

[0059] S44. The outputs of the three encoders are fused using a cross-attention mechanism layer to output fused features; the fused features are then processed by a feature fusion network to obtain the recognition features.

[0060] S45. After the recognition features are processed by the dual-head output layer, the illuminance and the probability of sunlight glare are output respectively.

[0061] Finally, illuminance and DGP are predicted separately using independent dual output heads. Illuminance prediction uses the Softplus function to ensure that the output is non-negative, while DGP prediction uses the Sigmoid function to constrain the probability to the range of 0-1.

[0062] S5. Real-time acquisition of key variables, and prediction of indoor light comfort based on optimized soft measurement models using key variables.

[0063] Finally, in the edge deployment stage, the lightweight model, after feature optimization and structural pruning, is deployed to the measurement platform, such as... Figure 5 As shown, the front-end and back-end systems were developed within the measurement platform environment: the front-end uses a web interface for user interaction and result visualization, while the back-end uses Python to write a model calling interface, realizing the complete process of environmental data reception, feature processing, model inference, and prediction result return. This system can achieve real-time and accurate prediction of indoor light comfort parameters in a resource-constrained edge computing environment, providing a practical technical solution for building light environment monitoring.

[0064] This invention also provides a soft measurement system for indoor light comfort parameters oriented towards edge computing, used to implement the above-mentioned method. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process described herein can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0065] The present invention also provides an electronic device including a central processing unit (CPU) that can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) or loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0066] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0067] The processing unit executes the various methods and processes described above, such as methods S1 to S5. For example, in some embodiments, methods S1 to S5 may be implemented as computer software programs tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of methods S1 to S5 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S5 by any other suitable means (e.g., by means of firmware).

[0068] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0069] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0070] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0071] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A soft measurement method for indoor lighting comfort parameters oriented towards edge computing, characterized in that, include: Key sample variables were collected, including geographical and climatic parameters, building spatial parameters, surface material parameters, and spatial characteristics of observation points. The key sample variables are sampled and processed to generate multiple target samples, and a corresponding typical room scene is constructed for each target sample. For each typical room scenario, multiple observation points are deployed, and the illuminance and solar glare probability of each observation point are calculated hourly. The illuminance and solar glare probability are used as data labels for the corresponding key sample variables to construct a training dataset. A neural network model is trained using the training dataset to obtain an indoor light comfort soft measurement model. SHAP sensitivity analysis is performed on the soft measurement model to obtain the importance of each key sample variable. Based on the importance, the soft measurement model is lightweighted to obtain an optimized soft measurement model. Key variables are collected in real time, and indoor light comfort is predicted based on the optimized soft measurement model using these key variables.

2. The soft measurement method for indoor light comfort parameters oriented towards edge computing according to claim 1, characterized in that, The geographical and climatic parameters include geographical location, altitude, solar azimuth, solar normal radiation, and solar horizontal diffuse radiation; the geographical location includes longitude and latitude. The architectural space parameters include the room's length, width, and height, windowsill height, window height, window-to-wall ratio, and window orientation; The surface material parameters include wall reflectivity, ceiling reflectivity, floor reflectivity, and window transmittance; The spatial characteristics of the observation point include the original three-dimensional coordinates of the observation point and the orientation of the observation point; Among them, the window orientation, observation point orientation, and geographical location are discrete variables, while the remaining key sample variables are continuous variables.

3. The soft measurement method for indoor light comfort parameters oriented towards edge computing according to claim 2, characterized in that, The method for generating the target sample is as follows: The continuous variables were sampled using the Latin hypercube sampling strategy to obtain a standardized sample. After performing strategic discretization on the discrete variables, they are randomly allocated to each of the standard samples according to a preset ratio to obtain the target samples.

4. The soft measurement method for indoor light comfort parameters oriented towards edge computing according to claim 1, characterized in that, The soft measurement model includes an input layer, a feature encoder, a cross-attention mechanism layer, a feature fusion network, and a dual-head output layer. The feature encoder includes a spatial encoder, a radiation environment encoder, and a material and geo-encoder. The soft measurement model performs the following steps to predict illuminance and solar glare probability: A series of data sets are randomly selected from the training dataset and used as input to the soft measurement model; the data sets include data labels and corresponding key sample variables. Feature processing is performed on the key sample variables to obtain spatial geometric features, radiation environment features, and material and geographical features; The aforementioned spatial geometric features, radiation environment features, and material and geographic features are processed using a spatial encoder, a radiation environment encoder, and a material and geographic encoder, respectively. The outputs of the three encoders are fused using the cross-attention mechanism layer to output fused features. The fused features are processed by a feature fusion network to obtain the recognition features; After the aforementioned recognition features are processed by the dual-head output layer, the illuminance and the probability of sunlight glare are output respectively.

5. The soft measurement method for indoor light comfort parameters oriented towards edge computing according to claim 4, characterized in that, The spatial geometric features include the room's length, width, and height; window height; windowsill height; window-to-wall ratio; three-dimensional relative coordinates between the observation point and the window; distance between the observation point and the window; and X and Y axis vectors of the observation point's orientation. Based on these key sample features, the following feature processing is used to obtain the three-dimensional relative coordinates between the observation point and the window, the distance between the observation point and the window, and the X and Y axis vectors of the observation point's orientation: A target coordinate system is constructed using the lowest point of the line where the west and south walls of the room intersect as the unified spatial origin. In the target coordinate system, the coordinates of the window center point are calculated according to the window orientation. Based on the coordinates of the window center point and the original three-dimensional coordinates of the observation point, the three-dimensional relative position between the target observation point and the window, and the distance from the target observation point to the window center point are calculated. The X-axis and Y-axis vectors of the observation point orientation are obtained based on the observation point orientation and the original three-dimensional coordinates of the observation point.

6. The soft measurement method for indoor light comfort parameters oriented towards edge computing according to claim 4, characterized in that, The radiation environment characteristics include the solar direction vector, solar normal direct radiation, solar horizontal scattered radiation, effective direct radiation component, and effective scattered radiation; based on the key sample characteristics, the solar direction vector, effective direct radiation component, and effective scattered radiation are obtained through the following feature processing: Obtain the solar elevation angle, and convert the solar normal direct radiation into an effective direct radiation component to the horizontal plane based on the sine of the elevation angle; The sky scattering factor is calculated based on the aforementioned elevation angle, and the effective scattered radiation is calculated based on the aforementioned sky scattering factor. The solar direction vector is calculated based on the sine and cosine of the azimuth and altitude angles.

7. The soft measurement method for indoor light comfort parameters oriented towards edge computing according to claim 4, characterized in that, The aforementioned material and geographical features include wall reflectivity, ceiling reflectivity, window transmittance, corrected floor effective reflectivity, longitude spherical coordinates, and latitude spherical coordinates; based on the aforementioned key sample features, the corrected floor effective reflectivity, longitude spherical coordinates, and latitude spherical coordinates are obtained through the following feature processing: Convert longitude and latitude to spherical coordinates to obtain spherical longitude and spherical latitude; The ground reflectivity is corrected based on the ground reflectivity and window transmittance to obtain the corrected ground reflectivity.

8. A soft measurement method for indoor light comfort parameters oriented towards edge computing according to claim 4, characterized in that, The number of neurons in the input layer is equal to the number of key sample variables in the input soft measurement model; the number of neurons in the spatial encoder is equal to the number of spatial geometric features; the number of neurons in the radiation environment encoder is equal to the number of radiation environment features; and the number of neurons in the material and geography encoder is equal to the number of material and geography features.

9. A soft measurement method for indoor light comfort parameters oriented towards edge computing according to claim 1, characterized in that, The importance mentioned includes the importance of predicting the illuminance of each key sample variable and the probability of solar glare; the lightweighting process based on this importance is as follows: For illuminance prediction and solar glare probability prediction, the importance of each type of key sample variable is ranked. Key sample indicators with less than preset values ​​in illuminance prediction and solar glare probability prediction were selected respectively to construct a first set of redundant variables and a second set of redundant variables. Select the same key sample indicators from the first and second sets of redundant variables as redundant variables; Remove the input layer neurons, spatial encoder neurons, radiation environment encoder neurons, and material and geo-encoder neurons corresponding to redundant variables in the aforementioned soft measurement model.

10. A soft measurement system for indoor lighting comfort parameters oriented towards edge computing, characterized in that, The system is used to implement the method as described in any one of claims 1 to 9.

Images