CN-121980646-A - Edge calculation-oriented indoor light comfort parameter soft measurement method and system

CN121980646ACN 121980646 ACN121980646 ACN 121980646ACN-121980646-A

Abstract

The invention relates to an edge-calculation-oriented indoor light comfort parameter soft measurement method and system, wherein the method comprises the steps of collecting key sample variables, sampling and processing the key sample variables to generate a plurality of target samples, constructing corresponding typical room scenes for each target sample, calculating illuminance and sunlight glare probability for each typical room scene as data labels one by one, constructing a training data set, training a neural network model by utilizing 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 variable, carrying out light weight treatment based on the importance to obtain an optimized soft measurement model, collecting the key variables in real time, and carrying out indoor light comfort prediction by utilizing the optimized soft measurement model based on the key variables. Compared with the prior art, the invention provides a method capable of ensuring accurate measurement of high-precision indoor light comfort parameters and ensuring measurement efficiency.

Inventors

ZHU HAN
Cao Wenzan
LI ZHENGRONG

Assignees

同济大学

Dates

Publication Date: 20260505
Application Date: 20251230

Claims (10)

1. An edge-calculation-oriented indoor light comfort parameter soft measurement method is characterized by comprising the following steps of: Collecting key sample variables, wherein the key sample variables comprise geographic climate parameters, building space parameters, surface material parameters and observation point space characteristics; sampling and processing the key sample variables to generate a plurality of target samples, and constructing a corresponding typical room scene for each target sample; the method comprises the steps of deploying multi-point observation points for each typical room scene, calculating the illuminance and sunlight glare probability of each observation point one by one, taking the illuminance and sunlight glare probability as data labels of corresponding key sample variables, 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, performing SHAP sensitivity analysis on the soft measurement model to obtain the importance of each type of key sample variable, and performing light weight treatment on the soft measurement model based on the importance to obtain an optimized soft measurement model; And acquiring key variables in real time, and predicting indoor light comfort by using an optimized soft measurement model based on the key variables.
2. The edge-oriented computing indoor light comfort parameter soft measurement method of claim 1, wherein the geographic climate parameters comprise geographic location, altitude, solar azimuth, solar normal direct radiation and solar horizontal scattered radiation, and the geographic location comprises longitude and latitude; the building space parameters comprise the length, width and height of a room, the height of a windowsill, the height of a window, the ratio of window walls and the orientation of the window; the surface material parameters comprise wall reflectivity, ceiling reflectivity, ground reflectivity and window transmissivity; the observation point space features comprise an original three-dimensional coordinate of the observation point and an orientation of the observation point; The window orientation, the observation point orientation and the geographic position are discrete variables, and the rest key sample variables are continuous variables.
3. The method for soft measuring indoor light comfort parameters for edge-oriented computing according to claim 2, wherein the method for generating the target sample is as follows: sampling the continuous variable by using Latin hypercube sampling strategy to obtain a standard sample; and after strategic discretization processing is carried out on the discrete variables, randomly distributing the discrete variables into each standard sample according to a preset proportion to obtain a target sample.
4. The method for soft measurement of indoor light comfort parameters for edge-oriented computing as set forth in claim 1, wherein the soft measurement model comprises an input layer, a feature encoder, a cross-attention mechanism layer, a feature fusion network and a double-headed output layer, the feature encoder comprises a space encoder, a radiation environment encoder and a material and geography encoder, and the soft measurement model performs the following steps of: randomly selecting a series of data sets from the training data set as input of a soft measurement model, wherein the data sets comprise data labels and corresponding key sample variables; performing feature processing on the key sample variables to obtain space geometric features, radiation environment features, materials and geographic features; processing the space geometric feature, the radiation environment feature, the material and the geographic feature by using a space encoder, a radiation environment encoder and a material and geographic encoder respectively; The output of the three encoders is subjected to feature fusion by utilizing the cross attention mechanism layer to output fusion features; and after the identification features are processed by the double-end output layer, respectively outputting illuminance and sunlight glare probability.
5. The method for soft measurement of indoor light comfort parameters for edge-oriented computing as set forth in claim 4, wherein the spatial geometrical features include length, width and height of a room, window height, sill height, window wall ratio, three-dimensional relative coordinates of an observation point-window, a X, Y-axis vector of an observation point-window distance and an observation point orientation, and the three-dimensional relative coordinates of the observation point-window, the X, Y-axis vector of the observation point-window distance and the observation point orientation are obtained based on the key sample features through the following feature processing: constructing a target coordinate system by taking the bottommost point of the boundary line between the west wall and the south wall of the room as a unified space origin; Under the target coordinate system, calculating window center point coordinates according to the window orientation, and calculating the three-dimensional relative position of the target observation point and the window based on the window center point coordinates and the original three-dimensional coordinates of the observation point, wherein the distance from the target observation point to the window center point is calculated; And obtaining X-axis and Y-axis vectors of the observation point orientation based on the observation point orientation and the original three-dimensional coordinates of the observation point.
6. The edge-calculation-oriented indoor light comfort parameter soft measurement method of claim 4, wherein the radiation environmental features comprise a solar direction vector, a solar normal direct radiation, a solar horizontal scattered radiation, an effective direct component and an effective scattered radiation, and the solar direction vector, the effective direct component and the effective scattered radiation are obtained based on the key sample features through the following feature processing: Acquiring a solar altitude, converting solar normal direct radiation into an effective direct component to the horizontal based on the sine of the altitude; calculating a sky-scattering factor based on the altitude angle, and calculating effective scattered radiation based on the sky-scattering factor; and calculating a solar direction vector based on the azimuth angle and the sine and cosine of the altitude angle.
7. The method for soft measurement of edge-oriented computing indoor light comfort parameters of claim 4, wherein the material and geographic features include wall reflectivity, ceiling reflectivity, window transmissivity, corrected floor effective reflectivity, longitude spherical coordinates and latitude spherical coordinates, and the corrected floor effective reflectivity, longitude spherical coordinates and latitude spherical coordinates are obtained based on the key sample features through the following feature processing: Converting the longitude and the latitude into spherical coordinates to obtain spherical longitude and spherical latitude; and correcting the ground reflectivity based on the ground reflectivity and the window transmissivity to obtain corrected ground reflectivity.
8. The edge-calculation-oriented indoor light comfort parameter soft measurement method of claim 4, wherein the number of neurons in the input layer is equal to the number of types of key sample variables in the input soft measurement model, the number of neurons in the space encoder is equal to the number of types of space geometric features, the number of neurons in the radiation environment encoder is equal to the number of types of radiation environment features, and the number of neurons in the material and the geo-encoder is equal to the number of types of material and the geo-features.
9. The edge-calculation-oriented indoor light comfort parameter soft measurement method of claim 1, wherein the importance comprises importance of each type of key sample variable contrast prediction and sunlight glare probability prediction, and the light weight processing based on the importance is as follows: the importance of various key sample variables is respectively sequenced for illuminance prediction and sunlight glare probability prediction; key sample indexes with importance smaller than a preset value in illuminance prediction and sunlight glare probability prediction are respectively screened, and a first redundant variable set and a second redundant variable set are constructed; Selecting the same key sample index in the first redundant variable set and the second redundant variable set as a redundant variable; And removing input layer neurons, space encoder neurons, radiation environment encoder neurons, materials and geo-encoder neurons corresponding to the redundant variables in the soft measurement model.
10. An edge-calculation-oriented indoor light comfort parameter soft measurement system, characterized in that the system is used for implementing the method according to any one of claims 1-9.

Description

Edge calculation-oriented indoor light comfort parameter soft measurement method and system Technical Field The invention relates to the technical field of building design, in particular to an edge calculation-oriented indoor light comfort parameter soft measurement method and system. Background The intelligent regulation and control of the building light environment is a key technical field for improving energy efficiency and indoor personnel comfort level. The sunlight is used as a main indoor lighting source, and the reasonable utilization of the sunlight has remarkable benefits for reducing illumination energy consumption and improving visual comfort and mental health of personnel. However, sunlight has a dynamic variation characteristic, and improper control may cause problems such as insufficient illuminance, uneven distribution, or glare. To achieve accurate light environment regulation, the core is to acquire key indexes capable of directly reflecting human comfort in real time, wherein the working surface illuminance and the sunlight glare probability (DGP) are considered as representative two parameters. Therefore, in the prior art, when indoor light environment comfort level prediction or evaluation is performed, the two parameters are often taken as known quantities, for example, a Chinese patent application CN116680779A, which provides a model feature analysis method based on a building indoor sunlight perception evaluation prediction model, so as to solve the problem that sunlight perception evaluation research in the prior art cannot reflect subjective feeling of a user and influence weight of environmental parameters, in the method, parameters such as sunlight glare probability and the like are acquired based on field experiments in a typical building space, and the acquisition modes of two key parameters such as working surface illuminance and sunlight glare probability have obvious limitations, in particular, illuminance measurement can be realized through sensors, but a large number of sensors need to be deployed to acquire refined distribution data in the space, so that cost and maintenance problems are brought; the evaluation of DGP is more complicated, panoramic images are shot through a fisheye lens and analyzed by professional software, and the process is complex in operation, high in equipment requirement and has privacy interference risk. In order to solve the above technical problems, a tool based on physical simulation is provided to achieve key parameter acquisition, such as chinese patent application CN111259481a, which provides a method for evaluating the design of the indoor light environment of a building to integrate spatial information, in which in order to avoid the inconvenience caused by actually measured data, the method is calculated by simulation software, and although the method can provide accurate results, the modeling is complex, the consumption of calculation resources is large, and the real-time control requirement cannot be satisfied completely. The method also has the problems that 1) a complex network structure is adopted, the calculation cost is large, the cloud server or the high-performance workstation is required to be relied on to finish calculation, the method can be used as an offline analysis tool only, 2) the model light-weight systematic design is lacking, the resource constraint characteristic of the edge equipment is not considered, and the calculation efficiency requirement of real-time reasoning can not be met while the precision is kept. Therefore, the method for ensuring the accurate measurement of the high-precision indoor light comfort parameter and the measurement efficiency is a technical problem to be solved. Disclosure of Invention The present invention has for its object to overcome the above-mentioned drawbacks of the prior art by providing a method. The aim of the invention can be achieved by the following technical scheme: according to a first aspect of the present invention, there is provided an edge-oriented computing indoor light comfort parameter soft measurement method, comprising: Collecting key sample variables, wherein the key sample variables comprise geographic climate parameters, building space parameters, surface material parameters and observation point space characteristics; sampling and processing the key sample variables to generate a plurality of target samples, and constructing a corresponding typical room scene for each target sample; the method comprises the steps of deploying multi-point observation points for each typical room scene, calculating the illuminance and sunlight glare probability of each observation point one by one, taking the illuminance and sunlight glare probability as data labels of corresponding key sample variables, 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, performing SHAP sens