Intelligent indoor disinfection effect prediction method and system

Abstract

The invention discloses an intelligent indoor disinfection effect prediction method and system. The method comprises the steps of disinfection data acquisition, data preliminary optimization, disinfection data embedding, effect prediction model construction and disinfection effect prediction. The invention relates to the technical field of disinfection effect data processing, in particular to an intelligent indoor disinfection effect prediction method and system, and the method comprises the steps: obtaining original disinfection effect data through disinfection data collection; a data optimization method of missing value processing, spatio-temporal data alignment, data standardization and data set segmentation is adopted; a material physical property characteristic embedding method is adopted, the influence of local indoor material difference is reduced, and the prediction performance of the model is improved through interaction of disinfection parameters and material physical properties; a physical enhancement graph convolutional network model is adopted as an effect prediction model, and deviation caused by disinfection effect lag is solved through combination of disinfectant forward diffusion process characteristics and disinfectant reverse residual process characteristics and through a time compensation mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of disinfection effect data processing, and specifically refers to an intelligent indoor disinfection effect prediction method and system. Background Art

[0002] Indoor disinfection effect prediction refers to estimating the effect after disinfection treatment in an indoor environment through scientific methods and models, aiming to evaluate the removal effect of disinfection methods on pathogenic microorganisms and the usage efficiency of disinfectants; it can help formulate scientific and reasonable disinfection strategies, improve the hygiene and safety of public places or home environments, reduce the risk of infectious disease transmission, and has important application value especially in disease prevention and control and hospital infection control.

[0003] However, traditional indoor disinfection effect prediction methods have technical problems such as using a single temporal convolutional kernel or fixed-period assumption, being unable to distinguish the long-term trend and short-term fluctuations of disinfectant diffusion, and relying on uniform grid division for spatial modeling, ignoring the influence of the non-uniform distribution of material physical properties within the region on the disinfection effect; traditional indoor disinfection effect prediction methods have technical problems such as being unable to model the reverse process of disinfectant degradation, being unable to capture the non-linear influence of environmental parameters such as ultraviolet intensity, temperature, and humidity on the disinfection effect lag, and having errors in time delay compensation. Summary of the Invention

[0004] In view of the above situation, to overcome the defects of the prior art, the present invention provides an intelligent indoor disinfection effect prediction method and system. Aiming at the technical problems of traditional indoor disinfection effect prediction methods, such as using a single temporal convolutional kernel or fixed-period assumption, being unable to distinguish the long-term trend and short-term fluctuations of disinfectant diffusion, and relying on uniform grid division for spatial modeling, ignoring the influence of the non-uniform distribution of material physical properties within the region on the disinfection effect, this solution creatively adopts the method of embedding material physical property characteristics. Through temporal multi-dimensional decomposition, spatial multi-granularity representation, and embedding of material physical properties, the influence of local indoor material differences is reduced, and through the interaction between disinfection parameters and material physical properties, the prediction performance of the model is improved; aiming at the technical problems of traditional indoor disinfection effect prediction methods, such as being unable to model the reverse process of disinfectant degradation, being unable to capture the non-linear influence of environmental parameters such as ultraviolet intensity, temperature, and humidity on the disinfection effect lag, and having errors in time delay compensation, this solution creatively adopts a physically enhanced graph convolutional network model as the effect prediction model, which can fully combine the characteristics of the forward diffusion process of disinfectants and the characteristics of the reverse residual process of disinfectants under the condition of conforming to physical laws, and adjust the historical cumulative effect through a time compensation mechanism, solving the deviation caused by the disinfection effect lag.

[0005] The technical solution adopted by the present invention is as follows: An intelligent indoor disinfection effect prediction method provided by the present invention, the method includes the following steps:

[0006] Step S1: Disinfection data collection;

[0007] Step S2: Preliminary data optimization;

[0008] Step S3: Disinfection data embedding;

[0009] Step S4: Construction of the effect prediction model;

[0010] Step S5: Prediction of disinfection effect.

[0011] Further, in step S1, the disinfection data collection is used to collect the original data required for predicting the indoor disinfection effect. Specifically, the disinfection data is collected from the disinfection log system, the building information system, and the environmental sensor system to obtain the original disinfection effect data set. The original disinfection effect data set specifically includes the historical disinfection effect original set and the real-time disinfection effect original set. Both the historical disinfection effect original set and the real-time disinfection effect original set include spatio-temporal reference data, disinfection-related data, and environmental data. The historical disinfection effect original set further includes historical disinfection effect label data.

[0012] Further, in step S2, the preliminary data optimization is used to preprocess the collected original disinfection effect data to optimize the data. Specifically, it includes the following steps:

[0013] Step S21: Missing value processing, which is used to remove missing values. Specifically, the missing values in the historical disinfection effect original set and the real-time disinfection effect original set are removed to obtain the roughly processed historical data set and the roughly processed real-time data set;

[0014] Step S22: Spatio-temporal data alignment, which is used to align spatio-temporal data. Specifically, the spatial data in the roughly processed historical data set and the roughly processed real-time data set is converted into a unified coordinate system, and the time data is synchronized to obtain the aligned historical data set and the aligned real-time data set;

[0015] Step S23: Data standardization, which is used to standardize the aligned data. Specifically, the minimum-maximum standardization method is used to process the aligned historical data set and the aligned real-time data set to obtain the standardized historical data set and the standardized real-time data set;

[0016] Step S24: Data set segmentation, which is used to segment the data set. Specifically, the standardized historical data set is segmented into a prediction training set and a prediction test set.

[0017] Further, in step S3, the disinfection data embedding is used to model the physical interaction between disinfection parameters and indoor building materials. Specifically, it includes the following steps;

[0018] Step S31: Temporal dimension decomposition, which is used to separate the multi-scale features of the time series. Specifically, the temporal dimension features are decomposed into a trend term, a periodic term, and a transient term;

[0019] Step S32: Spatial hierarchical encoding, which is used to construct a multi-granularity spatial representation. Specifically, spatial hierarchical encoding is performed at the grid level, regional level, and global level;

[0020] Step S33: Parameter cross-embedding, which is used to interact the disinfection parameters and the physical parameters of indoor building materials. The formula used is as follows:

[0021] ;

[0022] In the formula, represents the disinfection material interaction feature, represents the softmax function, represents the disinfection parameter feature, represents the indoor building material attribute feature, which has the same dimension as the disinfection parameter feature, represents the dimension of the disinfection parameter feature;

[0023] Step S34: Construct the feature set. Specifically, through the temporal dimension decomposition, the spatial hierarchical encoding, and the parameter cross-embedding, the standardized real-time data set, the prediction training set, and the prediction test set are processed to obtain a real-time feature set, a training feature set, and a test feature set.

[0024] Furthermore, in step S4, the construction of the effect prediction model is used to build a model required for predicting the indoor disinfection effect. Specifically, a physically enhanced graph convolutional network model is constructed as the effect prediction model;

[0025] The construction of the effect prediction model specifically includes the following steps:

[0026] Step S41: Design the activation function, which is used to introduce physical constraints. Specifically, by designing a physically enhanced activation function, the specific formula of the physically enhanced activation function is as follows:

[0027] ;

[0028] In the formula, represents the physically enhanced activation function, x represents the independent variable of the physically enhanced activation function, represents the ventilation rate data of region j, represents the absorption coefficient of the material in region j, represents the sigmoid function, represents the Euclidean distance from region m to region j, represents the environmental humidity data of region m, Denote the dynamic temperature coefficient of area j, Denote the ReLU function, Denote the disinfectant concentration in area j, Denote the saturated concentration of the disinfectant;

[0029] Step S42: Design the forward diffusion process to simulate the active propagation of the disinfectant under ventilation and material constraints. The steps include:

[0030] Step S421: Calculate the forward dynamic adjacency matrix using the following formula:

[0031] ;

[0032] In the formula, Denote the forward dynamic adjacency matrix, Denote the weight matrix for calculating the forward dynamic adjacency matrix, Denote the connectivity feature between area m and area j, Denote the air velocity data from area m to area j, Denote the distance decay exponent, Denote the connectivity existence feature from area m to area j, Denote an indicator function, which has a value of 1 when there is connectivity between area m and area j, and a value of 0 when there is no connectivity between area m and area j;

[0033] Step S422: Implement physical constraints using the following formula:

[0034] ;

[0035] In the formula, Denote the disinfectant release amount in area m, Denote the volume of area m, Denote the density of the disinfectant, Denote a small value used to prevent division by zero, Denote the forward dynamic adjacency matrix under constraints;

[0036] Step S423: Forward graph convolution update;

[0037] Step S43: Design the reverse residual propagation to simulate the passive diffusion of the disinfectant under the influence of concentration gradient and human contact. The steps include:

[0038] Step S431: Calculate the half-life of the disinfectant components;

[0039] Step S432: Calculate the reverse dynamic adjacency matrix using the following formula:

[0040] ;

[0041] In the formula, represents the reverse dynamic adjacency matrix, represents the disinfectant concentration in area m, represents the half-life of the disinfectant component in area n, represents the personnel contact frequency from area m to area j, represents the surface area of area j;

[0042] Step S433: Reverse graph convolution update;

[0043] Step S44: Bidirectional gated fusion, which is used to balance forward propagation and reverse residuals. Specifically, a gating mechanism is designed to fuse forward propagation features and reverse residual features;

[0044] Step S45: Obtain the model output. The formula used is as follows:

[0045] ;

[0046] In the formula, represents the preliminary predicted output of the model, represents the model output weight, represents the time delay coefficient, represents the weight for calculating the delay coefficient, represents the ultraviolet intensity at time represents the model predicted output at time t, Tm represents the total number of time steps, represents the preliminary predicted output of the model at time

[0047] Step S46: Construct and train the model. Specifically, through the designed activation function, the designed forward diffusion process, the designed reverse residual propagation, the bidirectional gated fusion, and the obtaining of the model output, a physically enhanced graph convolutional network model is constructed. The model is trained based on the training feature set, and the model performance is verified based on the test feature set to obtain a physically enhanced graph convolutional network model, which is used as an effect prediction model.

[0048] Furthermore, in step S5, the disinfection effect prediction is specifically to use the effect prediction model to predict the indoor disinfection effect based on the real-time feature set, obtain the disinfection effect prediction reference data, and comprehensively evaluate the indoor disinfection effect based on the disinfection effect prediction reference data.

[0049] An intelligent indoor disinfection effect prediction system provided by the present invention includes a disinfection data acquisition module, a data preliminary optimization module, a disinfection data embedding module, an effect prediction model construction module, and a disinfection effect prediction module;

[0050] The disinfection data acquisition module is used for disinfection data acquisition. Through disinfection data acquisition, an original disinfection effect data set is obtained, and the original disinfection effect data set is sent to the data preliminary optimization module;

[0051] The data preliminary optimization module is used for data preliminary optimization. Through data preliminary optimization, a standardized real-time data set, a prediction training set, and a prediction test set are obtained, and the standardized real-time data set, the prediction training set, and the prediction test set are sent to the disinfection data embedding module;

[0052] The disinfection data embedding module is used for disinfection data embedding. Through disinfection data embedding, the physical interaction between disinfection parameters and indoor building materials is modeled, a real-time feature set, a training feature set, and a test feature set are obtained, the real-time feature set is sent to the disinfection effect prediction module, and the training feature set and the test feature set are sent to the effect prediction model construction module;

[0053] The effect prediction model construction module is used for constructing an effect prediction model. By constructing a physical enhanced graph convolutional network model, an effect prediction model is obtained, and the effect prediction model is sent to the disinfection effect prediction module;

[0054] The disinfection effect prediction module is used for predicting the disinfection effect. By using the effect prediction model to predict the indoor disinfection effect, disinfection effect prediction reference data is obtained.

[0055] The beneficial effects obtained by the present invention by adopting the above scheme are as follows:

[0056] (1) Aiming at the technical problems that the traditional indoor disinfection effect prediction method uses a single time convolution kernel or a fixed-period hypothesis, cannot distinguish the long-term trend and short-term fluctuation of disinfectant diffusion, and the spatial modeling depends on uniform grid division, ignoring the influence of the non-uniform distribution of material physical properties in the region on the disinfection effect, this scheme creatively adopts the method of embedding material physical property characteristics. Through time multi-dimensional decomposition, spatial multi-granularity representation, and embedding of material physical properties, the influence of local indoor material differences is reduced, and through the interaction between disinfection parameters and material physical properties, the prediction performance of the model is improved.

[0057] (2) Aiming at the technical problems that the traditional indoor disinfection effect prediction method cannot model the reverse process of disinfectant degradation, cannot capture the non-linear influence of environmental parameters such as ultraviolet intensity, temperature and humidity on the disinfection effect lag, and there are errors in time delay compensation, this scheme creatively adopts a physical enhanced graph convolutional network model as the effect prediction model, which can fully combine the characteristics of the forward diffusion process of the disinfectant and the characteristics of the reverse residue process of the disinfectant under the condition of conforming to physical laws, and adjusts the historical cumulative effect through a time compensation mechanism, solving the deviation caused by the disinfection effect lag. Brief Description of the Drawings

[0058] Figure 1 It is a schematic flowchart of an intelligent indoor disinfection effect prediction method provided by the present invention;

[0059] Figure 2 It is a schematic module diagram of an intelligent indoor disinfection effect prediction system provided by the present invention;

[0060] Figure 3 It is a schematic flowchart of the preliminary optimization of the data in step S2;

[0061] Figure 4 It is a schematic flowchart of the embedding of disinfection data in step S3;

[0062] Figure 5 It is a schematic flowchart of the construction of the effect prediction model in step S4.

[0063] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. Detailed Embodiments

[0064] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present invention.

[0065] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention.

[0066] Embodiment 1, refer to Figure 1 , the technical solutions adopted by the present invention are as follows: An intelligent indoor disinfection effect prediction method provided by the present invention, the method includes the following steps:

[0067] Step S1: Collection of disinfection data;

[0068] Step S2: Preliminary optimization of the data;

[0069] Step S3: Embedding of disinfection data;

[0070] Step S4: Construction of the effect prediction model;

[0071] Step S5: Prediction of disinfection effect.

[0072] Example 2. Refer to Figure 1 and Figure 2 In step S1, the disinfection data collection is used to collect the original data required for predicting the indoor disinfection effect. Specifically, the original disinfection effect dataset is obtained by collecting disinfection data from the disinfection log system, the building information system, and the environmental sensor system. The original disinfection effect dataset specifically includes the historical disinfection effect original set and the real-time disinfection effect original set. Both the historical disinfection effect original set and the real-time disinfection effect original set include spatio-temporal reference data, disinfection-related data, and environmental data. The historical disinfection effect original set also includes historical disinfection effect label data. The spatio-temporal reference data specifically includes disinfection operation timestamp data, indoor area space division data, door and window connection data, and ventilation duct layout data. The disinfection-related data specifically includes disinfectant type data and disinfectant spraying data. The environmental data specifically includes indoor building material property data, environmental temperature data, environmental humidity data, environmental air flow velocity data, and indoor ventilation rate data.

[0073] Example 3. Refer to Figure 1 、 Figure 2 and Figure 3 Based on the above example, in step S2, the data preliminary optimization is used to preprocess the collected original disinfection effect data to optimize the data, and specifically includes the following steps:

[0074] Step S21: Missing value processing, which is used to remove missing values. Specifically, the missing values in the historical disinfection effect original set and the real-time disinfection effect original set are removed to obtain the roughly processed historical dataset and the roughly processed real-time dataset;

[0075] Step S22: Spatio-temporal data alignment, which is used to align spatio-temporal data. Specifically, the spatial data in the roughly processed historical dataset and the roughly processed real-time dataset is converted into a unified coordinate system, and the time data is clock synchronized to obtain the aligned historical dataset and the aligned real-time dataset;

[0076] Step S23: Data standardization, which is used to standardize the aligned data. Specifically, the minimum-maximum standardization method is used to process the aligned historical dataset and the aligned real-time dataset to obtain the standardized historical dataset and the standardized real-time dataset;

[0077] Step S24: Dataset segmentation, which is used to segment the dataset. Specifically, the standardized historical dataset is segmented into a prediction training set and a prediction test set.

[0078] Example 4, refer to Figure 1 , Figure 2 and Figure 4 , based on the above example, in step S3, the disinfection data embedding is used to model the physical interaction between disinfection parameters and indoor building materials, and specifically includes the following steps;

[0079] Step S31: Time dimension decomposition, which is used to separate the multi-scale features of the time series. Specifically, the time dimension features are decomposed into a trend term, a periodic term, and a transient term. The formula used is as follows:

[0080] ;

[0081] In the formula, represents the time dimension feature, represents the trend term, represents the periodic term, represents the transient term, represents the average pooling function, represents extracting the first k main frequency components using the fast Fourier transform, represents the multi-scale time feature;

[0082] Step S32: Spatial hierarchical encoding, which is used to construct a multi-granularity spatial representation. Specifically, spatial hierarchical encoding is performed at the grid level, regional level, and global level. The formula used is as follows:

[0083] ;

[0084] In the formula, represents the spatial feature at position i, represents the grid-level spatial feature, represents the regional-level spatial feature, represents the global-level spatial feature, represents the grid-level linear transformation weight matrix, represents the set of positions in region j, and Global represents the global position set;

[0085] Step S33: Parameter cross-embedding, which is used to interact the disinfection parameters and the physical parameters of indoor building materials. The formula used is as follows:

[0086] ;

[0087] In the formula, represents the disinfection material interaction feature, represents the softmax function, represents the disinfection parameter feature, represents the indoor building material attribute feature, which has the same dimension as the disinfection parameter feature Represents the characteristic dimension of disinfection parameters;

[0088] Step S34: Construct a feature set. Specifically, through the above-mentioned time dimension decomposition, spatial hierarchical encoding, and parameter cross-embedding, process the standardized real-time data set, the prediction training set, and the prediction test set to obtain a real-time feature set, a training feature set, and a test feature set.

[0089] By performing the above operations, for the technical problem that the traditional indoor disinfection effect prediction method uses a single time convolution kernel or a fixed-period hypothesis, unable to distinguish the long-term trend and short-term fluctuations of disinfectant diffusion, and the spatial modeling depends on uniform grid division, ignoring the influence of the non-uniform distribution of material physical properties within the region on the disinfection effect, this solution creatively adopts the method of embedding material physical property characteristics. Through time multi-dimensional decomposition, spatial multi-granularity representation, and embedding of material physical properties, the influence of local indoor material differences is reduced, and through the interaction between disinfection parameters and material physical properties, the prediction performance of the model is improved.

[0090] Example 5, refer to Figure 1 、 Figure 2 and Figure 5 , based on the above example, in step S4, the construction of the effect prediction model is used to construct the model required for predicting the indoor disinfection effect. Specifically, construct a physical enhanced graph convolutional network model as the effect prediction model;

[0091] The construction of the effect prediction model specifically includes the following steps:

[0092] Step S41: Design an activation function to introduce physical constraints. Specifically, by designing a physical enhanced activation function, the specific formula of the physical enhanced activation function is as follows:

[0093] ;

[0094] In the formula, represents the physical enhanced activation function, x represents the independent variable of the physical enhanced activation function, represents the ventilation rate data of region j, represents the absorption coefficient of the material in region j, represents the sigmoid function, represents the Euclidean distance from region m to region j, represents the environmental humidity data of region m, represents the dynamic temperature coefficient of region j, represents the ReLU function, represents the disinfectant concentration in region j, represents the disinfectant saturation concentration;

[0095] Step S42: Design a forward diffusion process to simulate the active propagation of disinfectant under ventilation and material constraints. The steps include:

[0096] Step S421: Calculate the forward dynamic adjacency matrix. The formula used is as follows:

[0097] ;

[0098] In the formula, represents the forward dynamic adjacency matrix, represents the weight matrix for calculating the forward dynamic adjacency matrix, represents the connectivity feature between region m and region j, represents the air velocity data from region m to region j, represents the distance decay exponent, represents the connectivity existence feature from region m to region j, represents an indicator function. When there is connectivity between region m and region j, its value is 1; when there is no connectivity between region m and region j, its value is 0;

[0099] Step S422: Implement physical constraints. The formula used is as follows:

[0100] ;

[0101] In the formula, represents the disinfectant release amount in region m, represents the volume of region m, represents the density of the disinfectant, represents a small value used to prevent division by zero, represents the forward dynamic adjacency matrix under constraints;

[0102] Step S423: Forward graph convolution update. The formula used is as follows:

[0103] ;

[0104] In the formula, represents the forward graph convolution feature of the (l + 1)-th layer, represents the graph convolution operation function, represents the forward graph convolution feature of the l-th layer;

[0105] Step S43: Design reverse residual propagation to simulate the passive diffusion of disinfectant under the influence of concentration gradient and human contact. The steps include:

[0106] Step S431: Calculate the half-life of the disinfectant components. The formula used is as follows:

[0107] ;

[0108] In the formula, represents the half-life of the disinfectant component in area j, represents the standard half-life of the disinfectant component, represents the influence coefficient of ultraviolet intensity, represents the ultraviolet intensity data in area j, represents the temperature in area j;

[0109] Step S432: Calculate the reverse dynamic adjacency matrix, and the formula used is as follows:

[0110] ;

[0111] In the formula, represents the reverse dynamic adjacency matrix, represents the disinfectant concentration in area m, represents the half-life of the disinfectant component in area n, represents the personnel contact frequency from area m to area j, represents the surface area of area j;

[0112] Step S433: Reverse graph convolution update, and the formula used is as follows:

[0113] ;

[0114] In the formula, represents the reverse graph convolution feature of the (l + 1)-th layer, represents the reverse graph convolution feature of the l-th layer;

[0115] Step S44: Bidirectional gated fusion, which is used to balance forward propagation and reverse residuals. Specifically, a gated mechanism is designed to fuse forward propagation features and reverse residual features, and the formula used is as follows:

[0116] ;

[0117] In the formula, Gw represents the gated weight, represents the calculation weight of the gated weight, represents the forward propagation feature, represents the reverse residual feature, represents the fused feature;

[0118] Step S45: Obtain the model output, and the formula used is as follows:

[0119] ;

[0120] In the formula, represents the preliminary predicted output of the model, represents the model output weight, denote time delay coefficient denote the weight for calculating the delay coefficient denote ultraviolet intensity at a certain time denote the predicted output of the model at time t, and Tm denotes the total number of time steps denote the preliminary predicted output of the model at a certain time

[0121] Step S46: Construct and train the model. Specifically, construct a physically enhanced graph convolutional network model through the designed activation function, the designed forward diffusion process, the designed reverse residual propagation, the bidirectional gated fusion, and the obtaining of the model output. Train the model based on the training feature set, and verify the model performance based on the test feature set to obtain a physically enhanced graph convolutional network model, which is used as the effect prediction model

[0122] By performing the above operations, for the technical problems existing in the traditional indoor disinfection effect prediction method, that is, it cannot model the reverse process of disinfectant degradation, and cannot capture the non-linear effects of environmental parameters such as ultraviolet intensity, temperature, and humidity on the time lag of disinfection effect, and there are errors in time delay compensation. This solution creatively uses a physically enhanced graph convolutional network model as the effect prediction model, which can fully combine the characteristics of the forward diffusion process of the disinfectant and the characteristics of the reverse residual process of the disinfectant under the condition of conforming to physical laws, and adjust the historical cumulative effect through the time compensation mechanism, thus solving the deviation caused by the time lag of disinfection effect

[0123] Example Six, refer to Figure 1 and Figure 2 , based on the above example, in step S5, the disinfection effect prediction is specifically to use the effect prediction model to predict the indoor disinfection effect based on the real-time feature set, obtain the reference data for disinfection effect prediction, and comprehensively evaluate the indoor disinfection effect based on the reference data for disinfection effect prediction

[0124] Example Seven, refer to Figure 1 and Figure 2 , based on the above example, an intelligent indoor disinfection effect prediction system provided by the present invention includes a disinfection data acquisition module, a data preliminary optimization module, a disinfection data embedding module, an effect prediction model construction module, and a disinfection effect prediction module

[0125] The disinfection data acquisition module is used for disinfecting data acquisition. Through disinfecting data acquisition, an original disinfection effect data set is obtained, and the original disinfection effect data set is sent to the data preliminary optimization module

[0126] The data preliminary optimization module is used for preliminary data optimization. Through preliminary data optimization, a standardized real-time data set, a prediction training set, and a prediction test set are obtained, and the standardized real-time data set, the prediction training set, and the prediction test set are sent to the disinfection data embedding module;

[0127] The disinfection data embedding module is used for disinfection data embedding. Through disinfection data embedding, the physical interaction between disinfection parameters and indoor building materials is modeled to obtain a real-time feature set, a training feature set, and a test feature set, and the real-time feature set is sent to the disinfection effect prediction module, and the training feature set and the test feature set are sent to the effect prediction model construction module;

[0128] The effect prediction model construction module is used for constructing an effect prediction model. By constructing a physical enhanced graph convolutional network model, an effect prediction model is obtained, and the effect prediction model is sent to the disinfection effect prediction module;

[0129] The disinfection effect prediction module is used for predicting the disinfection effect. By using the effect prediction model to predict the indoor disinfection effect, disinfection effect prediction reference data is obtained.

[0130] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "including", "comprising" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0131] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made in these embodiments without departing from the principles and spirit of the present invention.

[0132] The above description of the present invention and its embodiments is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and design similar structural methods and embodiments without creative efforts without departing from the purpose of the present invention, they should all fall within the protection scope of the present invention.

Claims

1. An intelligent indoor disinfection effect prediction method, characterized in that: The method includes the following steps: S1: Disinfection data collection. Through disinfection data collection, an original disinfection effect dataset is obtained. The original disinfection effect dataset specifically includes a historical disinfection effect original set and a real-time disinfection effect original set; S2: Preliminary data optimization. The collected original data is preprocessed to optimize the data, obtaining a standardized real-time dataset, a prediction training set, and a prediction test set; S3: Disinfection data embedding. It is used to model the physical interaction between disinfection parameters and indoor building materials. Through time dimension decomposition, spatial hierarchical coding, and parameter cross-embedding, a real-time feature set, a training feature set, and a test feature set are obtained; S4: Construction of an effect prediction model. It is used to construct a model required for predicting the indoor disinfection effect. Specifically, a physically enhanced graph convolutional network model is constructed as the effect prediction model; In step S4, step S41: Design an activation function. It is used to introduce physical constraints. Specifically, a physically enhanced activation function is designed. The specific formula of the physically enhanced activation function is as follows: ； In the formula, represents a physical enhancement activation function, x represents the independent variable of the physical enhancement activation function, represents the ventilation rate data of region j, represents the absorption coefficient of the material in region j, represents the sigmoid function, represents the Euclidean distance from region m to region j, represents the environmental humidity data of region m, represents the dynamic temperature coefficient of region j, represents the ReLU function, represents the disinfectant concentration in region j, represents the saturated disinfectant concentration; S5: Disinfection effect prediction. Specifically, the indoor disinfection effect is predicted through the effect prediction model, obtaining disinfection effect prediction reference data.

2. The intelligent indoor disinfection effect prediction method according to claim 1, wherein: The construction of the effect prediction model specifically includes the following steps: Step S41: Design an activation function; Step S42: Design a forward diffusion process. It is used to simulate the active propagation of disinfectants under ventilation and material constraints. The steps include: Step S421: Calculate the forward dynamic adjacency matrix. The formula used is as follows: ； In the formula, represents the forward dynamic adjacency matrix, represents the weight matrix for calculating the forward dynamic adjacency matrix, represents the connectivity feature between region m and region j, represents the air velocity data from region m to region j, represents the distance decay exponent, represents the connectivity existence feature from region m to region j, represents an indicator function, which has a value of 1 when there is connectivity between region m and region j, and a value of 0 when there is no connectivity between region m and region j; Step S422: Implement physical constraints. The formula used is as follows: ； In the formula, represents the disinfectant release amount in area m, represents the volume of area m, represents the density of the disinfectant, represents a small value used to prevent division by zero, represents the positive dynamic adjacency matrix under constraints; Step S423: Forward graph convolutional update; Step S43: Design a reverse residual propagation. It is used to simulate the passive diffusion of disinfectants under the influence of concentration gradients and human contact. The steps include: Step S431: Calculate the half-life of disinfectant components; Step S432: Calculate the reverse dynamic adjacency matrix. The formula used is as follows: ； In the formula, represents the reverse dynamic adjacency matrix, represents the disinfectant concentration in area m, represents the half-life of the disinfectant component in area n, represents the personnel contact frequency from area m to area j, represents the surface area of area j; Step S433: Reverse graph convolutional update; Step S44: Bidirectional gated fusion. It is used to balance forward propagation and reverse residue. Specifically, a gated mechanism is designed to fuse forward propagation features and reverse residue features; Step S45: Obtain the model output. The formula used is as follows: ； In the formula, represents the preliminary prediction output of the model, represents the output weight of the model, represents the time delay coefficient, represents the calculation weight of the delay coefficient, represents the ultraviolet intensity at time represents the predicted output of the model at time t, Tm represents the total number of time steps, represents the preliminary predicted output of the model at time Step S46: Model construction and training. Specifically, through the designed activation function, the designed forward diffusion process, the designed reverse residual propagation, the bidirectional gated fusion, and the obtaining of the model output, the construction of the physically enhanced graph convolutional network model is carried out. The model is trained based on the training feature set, and the model performance is verified based on the test feature set, obtaining the physically enhanced graph convolutional network model and using it as the effect prediction model.

3. The intelligent indoor disinfection effect prediction method according to claim 1, wherein: The disinfection data embedding specifically includes the following steps: Step S31: Time dimension decomposition. It is used to separate the multi-scale features of the time series. Specifically, the time dimension features are decomposed into a trend term, a periodic term, and a transient term; Step S32: Spatial hierarchical coding. It is used to construct a multi-granularity spatial representation. Specifically, spatial hierarchical coding is carried out according to the grid level, the regional level, and the global level; Step S33: Parameter cross-embedding. It is used to interact the disinfection parameters and the physical parameters of indoor building materials. The formula used is as follows: ； In the formula, represents the disinfection material interaction feature, represents the softmax function, represents the disinfection parameter feature, represents the indoor building material attribute feature, which has the same dimension as the disinfection parameter feature, represents the dimension of the disinfection parameter feature; Step S34: Construct a feature set. Specifically, through the time - dimension decomposition, the spatial - level encoding, and the parameter cross - embedding, process the standardized real - time data set, the prediction training set, and the prediction test set to obtain a real - time feature set, a training feature set, and a test feature set.

4. An intelligent indoor disinfection effect prediction method according to claim 1, characterized in that: The disinfection data collection is used to collect the original data required for predicting the indoor disinfection effect. Specifically, collect disinfection data from the disinfection log system, the building information system, and the environmental sensor system to obtain the original disinfection effect data set. The original disinfection effect data set specifically includes a historical disinfection effect original set and a real - time disinfection effect original set. Both the historical disinfection effect original set and the real - time disinfection effect original set include spatio - temporal reference data, disinfection - related data, and environmental data. The historical disinfection effect original set also includes historical disinfection effect label data.

5. The intelligent indoor disinfection effect prediction method according to claim 1, characterized in that: The initial data optimization specifically includes the following steps: Step S21: Missing - value processing, which is used to remove missing values. Specifically, remove the missing values in the historical disinfection effect original set and the real - time disinfection effect original set to obtain a roughly processed historical data set and a roughly processed real - time data set. Step S22: Spatio - temporal data alignment, which is used to align spatio - temporal data. Specifically, convert the spatial data in the roughly processed historical data set and the roughly processed real - time data set into a unified coordinate system and synchronize the time data to obtain an aligned historical data set and an aligned real - time data set. Step S23: Data standardization, which is used to standardize the aligned data. Specifically, use the min - max standardization method to process the aligned historical data set and the aligned real - time data set to obtain a standardized historical data set and a standardized real - time data set. Step S24: Data - set segmentation, which is used to segment the data set. Specifically, segment the standardized historical data set into a prediction training set and a prediction test set.

6. The intelligent indoor disinfection effect prediction method according to claim 1, characterized in that: The disinfection - effect prediction is specifically to use the effect - prediction model to predict the indoor disinfection effect based on the real - time feature set to obtain disinfection - effect prediction reference data, and comprehensively evaluate the indoor disinfection effect based on the disinfection - effect prediction reference data.

7. An intelligent indoor disinfection effect prediction system for implementing an intelligent indoor disinfection effect prediction method according to any one of claims 1-6, characterized in that: It includes a disinfection data collection module, an initial data optimization module, a disinfection data embedding module, an effect - prediction model construction module, and a disinfection - effect prediction module.

8. An intelligent indoor disinfection effect prediction system according to claim 7, characterized in that: The disinfection data collection module is used for disinfection data collection. Through disinfection data collection, obtain the original disinfection effect data set and send the original disinfection effect data set to the initial data optimization module. The initial data optimization module is used for initial data optimization. Through initial data optimization, obtain a standardized real - time data set, a prediction training set, and a prediction test set, and send the standardized real - time data set, the prediction training set, and the prediction test set to the disinfection data embedding module. The disinfection data embedding module is used for disinfection data embedding. Through disinfection data embedding, the physical interaction between disinfection parameters and indoor building materials is modeled to obtain a real-time feature set, a training feature set, and a test feature set, and the real-time feature set is sent to the disinfection effect prediction module, and the training feature set and the test feature set are sent to the effect prediction model construction module; The effect prediction model construction module is used for constructing an effect prediction model. By constructing a physical enhanced graph convolutional network model, an effect prediction model is obtained, and the effect prediction model is sent to the disinfection effect prediction module; The disinfection effect prediction module is used for predicting the disinfection effect. By using the effect prediction model to predict the indoor disinfection effect, disinfection effect prediction reference data is obtained.

Images