Indoor temperature and humidity spatial data prediction method fusing spatial interpolation and graph convolutional network

Abstract

The invention discloses an indoor temperature and humidity spatial data prediction method fusing spatial interpolation and a graph convolutional network, and relates to the field of smart city and environment monitoring, and the method comprises the steps: obtaining indoor multi-point temperature and humidity data actually measured by a sensor, and the spatial position information of a monitoring mark point corresponding to the indoor multi-point temperature and humidity data; the temperature and humidity data, the spatial position information and the time dimension features of the marked points are used as input of Kriging interpolation, and temperature and humidity linear predicted values of part of unmarked points are generated; constructing a multiple linear regression model taking the Kriging prediction result, the measured data of the mark points, the time dimension characteristics, the spatial characteristics and the external meteorological data as independent variables and the target temperature and humidity values as dependent variables, and performing multi-dimensional fitting on the Kriging interpolation result; and constructing a feature matrix by using a linear predicted value output by the regression model and other features, and taking the feature matrix and an adjacent matrix as input of a graph convolutional network to generate indoor temperature and humidity three-dimensional space distribution. According to the invention, high-precision indoor space data prediction can be realized.

Description

Technical Field

[0001] The present invention relates to the field of smart city indoor space data monitoring, and in particular to a method for predicting indoor temperature and humidity spatial data by integrating spatial interpolation and graph convolutional networks. Background Art

[0002] At present, as my country's medium- and long-term development goal, the construction of smart cities requires the digitalization and intelligence of urban buildings to be equipped with advanced indoor environmental monitoring systems to track, analyze and adjust indoor air quality, temperature and humidity and other micro-environmental factors in real time. This can not only improve the comfort and health level inside the building, but also realize early warning and forecasting of indoor disasters such as fire through data analysis.

[0003] Traditional indoor environmental monitoring technology usually uses a single or a small number of sensors, which are placed at the building vents or other easy-to-install locations to achieve environmental monitoring of the entire indoor space. However, this method with low data volume and few monitoring points makes it difficult to accurately monitor and analyze the microenvironment at a specific location, especially in irregular spaces where complex spatial structures exacerbate the limitations of monitoring. Such methods are difficult to cover environmental parameter monitoring in all areas of the building and cannot meet the requirements of smart buildings for refined management.

[0004] With the development of artificial intelligence, more deep learning technologies and spatial interpolation methods are being applied to the field of spatial data monitoring, striving to achieve more comprehensive spatial data acquisition. As a classic geostatistical modeling method, Kriging interpolation is an optimal unbiased interpolation method based on statistical theory. This method is based on the data of known points, combined with spatial position and spatial variation characteristics, to provide estimates of unmeasured points and quantify the uncertainty of the estimates. However, Kriging is highly dependent on spatial autocorrelation. Its basic assumption is regular, flat, and continuous grid estimation. There are boundary effects for irregular spaces. Distance calculations in irregular spaces need to consider spatial geometric structures. Complex geometric shapes make it difficult to accurately model spatial autocorrelation. Graph convolutional neural networks comprehensively consider the global connectivity and local dependencies of measurement points, directly model the non-uniformity and irregular physical structure of space, and can better capture the correlation of complex spaces. However, the effectiveness of graph convolutional neural networks depends on large-scale data for model training and has high requirements for computing resources.

[0005] In summary, the existing indoor temperature and humidity spatial data prediction methods have the following problems:

[0006] (1) Traditional environmental monitoring technology cannot perform detailed monitoring and management of the entire indoor space;

[0007] (2) In irregular spaces, complex geometric shapes make it difficult for traditional methods to accurately model;

[0008] (3) Traditional Kriging methods cannot cope with irregular spaces with boundary effects and complex geometric structures;

[0009] (4) Graph convolutional networks require large amounts of data to train the model, and the effect is poor when the amount of data is insufficient.

[0010] In view of the above problems, the present invention discloses a method for predicting indoor temperature and humidity spatial data by integrating spatial interpolation and graph convolutional network. Summary of the invention

[0011] An object of the present invention is to solve at least the above problems and to provide at least the advantages which will be described hereinafter.

[0012] The present invention aims to solve the problem that traditional monitoring technology is difficult to achieve accurate and comprehensive monitoring and modeling in indoor spaces with complex geometric structures.

[0013] In order to achieve the above technical objectives, the technical solution of the present invention is: a method for predicting indoor temperature and humidity spatial data by integrating spatial interpolation and graph convolutional network, and the specific steps are as follows:

[0014] Step 1) Set up sensors at multiple points in the target research area, each sensor corresponds to a measurement mark point, and record the temperature and humidity data of each sensor and its measurement time dimension characteristics. At the same time, obtain the spatial location information of all measurement mark points and pre-process the data;

[0015] Step 2) using the data obtained in step 1) as the input of Kriging interpolation, constructing a spatial covariance model by capturing the linear correlation of space, and performing linear prediction of temperature and humidity at unlabeled points;

[0016] Step 3) construct a multivariate linear regression model with Kriging linear prediction results, measured data of marked points, time dimension characteristics, spatial characteristics and external meteorological data as independent variables and target temperature and humidity values ​​as dependent variables, perform multidimensional fitting on the Kriging interpolation results, and generate processed temperature and humidity spatial prediction values ​​according to the fitted multivariate linear regression model;

[0017] Step 4) The predicted values ​​processed by the multivariate linear regression model are subjected to residual fusion, and feature splicing is performed with the temperature and humidity data measured at the marking points to construct a feature matrix. At the same time, based on the spatial position information of the marking points, the geometric relationship and physical distance between the sensor nodes are obtained to construct an adjacency matrix;

[0018] Step 5) Build a graph convolutional network model, input the feature matrix and adjacency matrix of step 4) into the graph convolutional network model for training, and use the fitted prediction model to obtain the three-dimensional spatial point-surface distribution of indoor temperature and humidity in the study area.

[0019] The present invention has the following beneficial effects:

[0020] 1. Compared with other indoor environmental monitoring methods, the multi-point distributed data acquisition method can obtain more comprehensive and precise spatial point measurement data, rely on the Kriging method to capture the best linear unbiased estimate of environmental spatial data, and obtain the overall data distribution of the target area.

[0021] 2. There is autocorrelation in the space of ambient temperature and humidity. Especially for irregular indoor scenes, Kriging interpolation is given priority. Considering the correlation between temperature and humidity and the influence of time dimension characteristics, spatial characteristics and outdoor meteorological data variables, a multivariate linear regression model is established to improve the prediction accuracy of the graph convolutional network for the target.

[0022] 3. By processing the initial estimated linear temperature and humidity values ​​and the measured temperature and humidity values, the fusion data features are obtained to improve the forward quality of the data, improve the estimation accuracy of the graph convolution model, and solve the problem of nonlinear prediction of spatial distribution.

[0023] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the overall technical route of the present invention;

[0025] Figure 2 This is a diagram showing the deployment of multi-point distributed sensors of the present invention;

[0026] Figure 3 The variability of the point pair of the present invention when the direction is xy=0°, z=0°;

[0027] Figure 4 The variability of the point pair of the present invention is xy = 30°, z = 0°;

[0028] Figure 5 The variability of the point pair direction of the present invention is xy=60°, z=0°;

[0029] Figure 6 The variability of the point pair of the present invention is xy = 90°, z = 0°;

[0030] Figure 7 The variability of the point pair of the present invention is xy = 120°, z = 0°;

[0031] Figure 8 The variability of the point pair of the present invention is xy = 150°, z = 0°;

[0032] Fig. 9 The variability of the point pair of the present invention is xy = 180°, z = 0°;

[0033] Fig.10 It is a three-dimensional point data distribution diagram of the spatial prediction result of the present invention;

[0034] Fig.11 It is a three-dimensional surface data distribution diagram of the spatial prediction result of the present invention. DETAILED DESCRIPTION

[0035] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0036] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials are commercially available unless otherwise specified; in the description of the present invention, the terms "lateral", "longitudinal", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0037] Combination Figure 1 The technical route adopted by the present invention is described. First, the temperature and humidity data of the marked points in the target study area are obtained, and the Kriging interpolation method is used to estimate the linear prediction value; then, based on these prediction results, a multivariate linear regression model is constructed to perform multi-dimensional fitting on the linear changes of temperature and humidity; finally, the spatial data is processed using a graph convolutional neural network to further obtain the comprehensive point-surface distribution of indoor space temperature and humidity, and achieve accurate prediction. The specific implementation includes the following steps:

[0038] Step 1) Set up sensors at multiple points in the target research area, each sensor corresponds to a measurement mark point, obtain indoor temperature and humidity data, record the measurement time dimension characteristics, and perform data preprocessing; Combined with Figure 2Describe the situation of the multi-point distributed sensors: For example, taking the satellite image cloud service research room of the Guangxi Key Laboratory of Spatial Information and Surveying and Mapping as the research area, its spatial information is x = 7.4m, y = 8.75m, z = 3m. Sensors are arranged in a 4*5 matrix on the xy ground plane; sensors are arranged in a 4*4 matrix on each of the two zy and zx planes in space; sensors with random spatial parameters xyz are arranged in the central area of space. The total number of measurement marking points is 60. Then, data is uploaded to the server side through LoRa wireless communication technology for data storage and subsequent processing. The data preprocessing includes: dealing with missing values, data smoothing, noise reduction, and normalization of the temperature and humidity data of each sensor; specifically as follows: For the missing values of the temperature and humidity data, an interpolation method combined with mean filling is used to ensure data integrity; to eliminate the noise in the data, an exponential smoothing method is used for data smoothing; since the fitting effect of the graph convolutional network is closely related to the data quality, it is necessary to normalize the spatial position information and the temperature and humidity data; the spatial coordinates of the marking points are scaled to the interval [0,1]; at the same time, the numerical ranges of temperature and humidity are quite different, and the two need to be normalized to [0,1] respectively to eliminate the dimension difference and ensure data consistency.

[0039] In step 2), when performing Kriging interpolation, it is necessary to capture the spatial correlation between the marking point pairs. First, calculate the distance between the point pairs and perform directional grouping, then calculate and use the semi-variogram to analyze the spatial variability in different directions, identify the strength of the correlation in each direction, the direction with strong correlation is determined as the main axis, and the direction with weak correlation is the secondary axis. Finally, based on the constructed spatial covariance model, perform linear interpolation prediction on the temperature and humidity of the unmarked points. The specific steps are as follows:

[0040] Calculate the Euclidean distance d in the horizontal direction between each pair of marking point pairs xy and the height difference d in the vertical direction z , to obtain the point pair distance. The formula is as follows:

[0041]

[0042] d z =丨z i -z j ∣

[0043] where x i ,y i and x j ,y j are the coordinates of the marking points P i and P j on the XY plane respectively, and z i and z j are the installation heights of the sensors. Group according to the point pair distance intervals.

[0044] After completing the distance grouping, for irregular indoor spaces, the marker points need to be further grouped by area based on the distance grouping, and the directions need to be divided based on geometric features, the distance between adjacent points, and the angle. Specifically, the direction is divided every 15° on the XY plane, ranging from 0° to 180°, for a total of 13 directions, and then the Z-axis vertical angle is divided into 7 different angles of 0°, 15°, 30°, 45°, 60°, 75°, and 90°. Combined with the XY plane direction and the Z-axis angle, anisotropy analysis of different directions in three-dimensional space is performed.

[0045] Based on the grouping situation, the semivariance value of each marker point pair is calculated to analyze the spatial correlation. First, the semivariance value of each marker point pair (x i ,x j ) of the interpolated square of the temperature and humidity values ​​[z(x i )-z(x j )] 2 , where z(x) is the temperature or humidity value of the marker point x. For a given distance h, the average squared difference of all pairs of points with a distance h is calculated to calculate the semivariate value, and the formula is as follows:

[0046]

[0047] Among them, γ(h k ) is the distance h k The semivariogram value between the point pairs, N(h k ) is the distance h k The number of point pairs.

[0048] Based on the calculated semi-variance values, an empirical semi-variogram is drawn in each direction and fitted using a Gaussian model, which is in the following form:

[0049]

[0050] Then, by comparing the variability in each direction, the main axis and the secondary axis are determined. Figure 3 to Figure 9 This is an empirical semi-variogram of the present invention grouped in different directions, which illustrates the change of the variability of spatial data with distance and direction. Figure 3 It is the variability of the point with respect to the direction xy = 0°, z = 0°; Figure 4 It is the variability of the point pair with the direction xy = 30°, z = 0°; Figure 5 It is the variability of the point pair with the direction xy = 60°, z = 0°; Figure 6 It is the variability of the point pair with the direction xy = 90°, z = 0°; Figure 7It is the variability of the point pair with the direction xy = 120°, z = 0°; Figure 8 It is the variability of the point pair in the direction of xy = 150°, z = 0°; Fig. 9 is the variability change of the point to the direction of xy = 180°, z = 0°; by comparing and analyzing the variability changes in different directions, the direction with stronger variability is selected as the main axis, and the direction with weaker variability is selected as the secondary axis. Figure 4 The xy=30°, z=0° point pair direction is used as the main axis, and the Figure 5 The point direction of xy=60°, z=0° is used as the secondary axis.

[0051] By fitting the semivariogram model, the spatial covariance model is defined, which is reflected as

[0052]

[0053] For example, the target area z=0 plane obtains 20 measurement mark point positions x1, x2, …, x n , obtain the covariance matrix through the covariance function Cov(h)

[0054] For the target unmarked point, the covariance between it and all known marked points is calculated to construct the covariance vector.

[0055] For example, the target area z = 0 marker point P is calculated by fitting the semivariogram function γ(h) and the covariance function Cov(h). a With the unmarked point P u The covariance values ​​of

[0056] Solve the linear system Cw=C u , calculate the Kriging weight w, and get the weight w of each known point to the unlabeled point i .

[0057] Creating the kriging equations Calculate the linear prediction results of temperature and humidity at unmarked points in the indoor space, where Z(x i ) is a known point x i The temperature and humidity values ​​of the unmarked point x0 are given in Table 1. Z(x0) is the linear prediction result of the temperature and humidity of the unmarked point x0. i To assign x to each group i x u The weight coefficient of .

[0058] Step 3) A multivariate linear regression model is constructed with the Kriging linear prediction results, measured data of the marked points, time dimension characteristics, spatial characteristics and outdoor meteorological data as independent variables, and the target predicted temperature and humidity as dependent variables. By fitting the multivariate linear regression model, the predicted values ​​of the Kriging interpolation and other independent variables are multi-dimensionally fitted to adjust the preliminary Kriging linear prediction results.

[0059] Step 4) Based on the multivariate linear regression prediction results, calculate the residual between the model prediction value and the measured value of the marked point i , and correct the predicted value, the residual ò i It is expressed as:

[0060] ò i =Y i r -Y i p

[0061] Corrected forecast value for:

[0062]

[0063] The prediction results after residual correction are concatenated with other features to form a complete feature matrix. Specifically, the prediction results after residual correction are generated by using residuals to adjust the multivariate linear regression, and then concatenated with other features such as temperature and humidity data measured based on the marker points. For example, the spatial location information α of the marker point, the time dimension feature t, the outdoor temperature β1, the outdoor humidity β2, the rainfall ε, and the processed temperature and humidity data prediction value Y s Fusion of the complete feature matrix Ensure that all important spatial, temporal, meteorological characteristics and temperature and humidity data are effectively integrated to provide a complete feature set for subsequent modeling.

[0064] Step 5) Build a graph convolutional network model, input the feature matrix and adjacency matrix of step 4) into the graph convolutional network model for training, and use the fitted prediction model to obtain the three-dimensional spatial point-surface distribution of indoor temperature and humidity in the study area. The specific steps are as follows:

[0065] According to the spatial position information of the marker points processed in step 2), the geometric relationship and physical distance between each sensor node are obtained, and the adjacency matrix is ​​constructed based on these distances.

[0066] The Gaussian decay function is used to adjust the weight according to the distance between nodes to provide a smooth transition. Define the weight. Where d ijis the distance between nodes i and j, σ is used to control the distance decay rate. The larger σ is, the slower the weight decays; the smaller σ is, the faster the weight decays. According to the distance d between all measured marked nodes ij , apply the Gaussian decay function to construct the adjacency matrix Set the minimum self-loop weight ε to deal with the node overlap problem, and normalize the adjacency matrix Balance the degree differences between different nodes.

[0067] Build a graph convolutional network model and transform the feature matrix and the adjacency matrix As the input of the graph convolutional network GCN, in order to ensure that the node's own features participate in the convolution, the adjacency matrix Add to the self-loop identity matrix I to obtain the adjacency matrix with self-loops According to the degree of each node Constructed degree matrix And perform symmetric normalization to obtain For each layer l (l = 1, 2, 3, ..., L), by Implement feature aggregation, where W (l) It is a weight matrix, and the nonlinear ability of the model is introduced through the nonlinear activation function σ=max(0,x).

[0068] The optimization is performed by defining the mean square error MSE as the loss function. The function formula is as follows:

[0069]

[0070] Use the gradient descent algorithm to minimize the loss, update the network weights, and finally output the feature matrix G of the layer after multiple layers of graph convolution. (L) Represented as the predicted value of temperature and humidity for each unlabeled node

[0071] Use the fitted prediction model to obtain the temperature and humidity prediction values ​​of unlabeled nodes The predicted data and spatial location information are integrated with the three-dimensional real scene model of the study area to obtain the three-dimensional spatial point-surface distribution of indoor temperature and humidity in the study area, as follows:

[0072] The distribution of three-dimensional target point data is presented in the form of three-dimensional scattered points. The three-dimensional mapping position of each target point in the real-life three-dimensional model is determined according to the three-dimensional coordinates (x, y, z). The temperature or humidity of the target point is mapped to a color through the color mapping function to show the data level. Fig.10 The mapping distribution result of the point data of the target point shows the spatial distribution of the point data.

[0073] The three-dimensional target surface data distribution is divided into temperature and humidity prediction values ​​through spatial slicing Rendering is performed to show the spatial distribution of indoor temperature and humidity. The study area is divided into multiple xy planes to construct multiple levels of distribution. Each layer generates the x and y range grid of the real-life 3D model of the corresponding study area. Applying a gradient color map will Mapped to the corresponding color parameters, such as Fig.11 The rendering mapping result of the spatial slice plane shows the spatial distribution of the surface data.

[0074] The indoor temperature and humidity spatial data prediction method proposed in this invention, which integrates spatial interpolation and graph convolutional network, can effectively consider the correlation between temperature and humidity and the influence of time dimension characteristics, spatial characteristics and outdoor meteorological data variables, and use the Kriging spatial interpolation method to extract deep spatial information characteristics through graph convolutional network to solve the problem of nonlinear prediction of spatial distribution. It has been verified that the fusion method of spatial interpolation and graph convolutional network can well capture the spatial characteristics of indoor temperature and humidity, and can better reflect the spatial and temporal three-dimensional point-surface distribution of indoor temperature and humidity.

[0075] As described above, for professionals in the field, the essence of the present invention goes far beyond the specific details of the exemplary implementation. The implementation forms of its core spirit and features can be diversified and are not limited to the examples currently shown. Therefore, the embodiments should be regarded as exemplary rather than restrictive. The scope of the present invention is defined by the attached claims, and it is intended that all variants and implementations that are consistent with the principle of equivalence of the claims be included in the scope of protection. Any graphical mark in the claims should not be regarded as a limitation of the relevant claims.

[0076] It is worth noting that although this specification is presented through specific implementation methods, experts in the relevant technical field should recognize that these descriptions are only for clear expression, and the technical solutions in each implementation method can be appropriately combined to form other implementation forms that adapt to specific situations and materials without departing from the purpose and scope of the present invention. The essence of the present invention is not limited to the specific embodiments disclosed, and all embodiments formed after reasonable changes or equivalent substitutions within the spirit and scope of the present invention belong to the protection scope of the present invention.

[0077] In summary, the present invention aims to convey its core concepts and technical framework through exemplary examples, while emphasizing the flexibility of innovation and the breadth of application, ensuring that all variants and implementation methods that are consistent with its spirit and principles are covered.

Claims

1. A method for predicting indoor temperature and humidity spatial data by integrating spatial interpolation and graph convolutional network, characterized in that: include: Step 1) Distribute sensors at multiple points in the target research area, each sensor corresponds to a measurement mark point, and record the temperature and humidity data of each sensor and its measurement time dimension characteristics; at the same time, obtain the spatial location information of all measurement mark points and pre-process the data; Step 2), using the data obtained in step 1) as the input of Kriging interpolation, by capturing the linear correlation of space, constructing a spatial covariance model, and performing linear prediction of temperature and humidity at unlabeled points; Step 3) Construct a multivariate linear regression model with Kriging linear prediction, measured data of marked points, time dimension characteristics, spatial characteristics and external meteorological data as independent variables and target temperature and humidity values ​​as dependent variables, perform multidimensional fitting on the Kriging prediction results, and generate processed temperature and humidity spatial prediction values ​​based on the fitted multivariate linear regression model; Step 4) Perform residual fusion on the predicted values ​​processed by the multivariate linear regression model, and perform feature splicing with the temperature and humidity data measured at the marking points to construct a feature matrix. At the same time, based on the spatial position information of the marking points, obtain the geometric relationship and physical distance between each sensor node and construct an adjacency matrix; Step 5) Build a graph convolutional network model, input the feature matrix and adjacency matrix of step 4) into the graph convolutional network model for training, and use the fitted prediction model to obtain the three-dimensional spatial point distribution of indoor temperature and humidity in the study area.

2. The indoor temperature and humidity spatial data prediction method integrating spatial interpolation and graph convolutional network according to claim 1 is characterized in that: In step 1), the temperature and humidity data of each sensor are processed for missing values, data smoothing, noise reduction and normalization; specifically, for the missing values ​​of the temperature and humidity data, the interpolation method is combined with the mean filling method to ensure data integrity; to eliminate the noise in the data, the exponential smoothing method is used for data smoothing; since the fitting effect of the graph convolutional network is closely related to the data quality, the spatial position information and the temperature and humidity data need to be normalized; the spatial coordinates of the marked points are scaled to the interval [0,1]; at the same time, the numerical ranges of temperature and humidity are quite different, and they need to be normalized to [0,1] respectively to eliminate the dimension difference and ensure the consistency of the data.

3. The indoor temperature and humidity spatial data prediction method integrating spatial interpolation and graph convolutional network according to claim 1 is characterized in that: In step 2), the semivariograms at different horizontal and vertical angles are calculated, and empirical semivariograms in corresponding directions are drawn based on these semivariograms to capture how spatial correlation changes with distance in different directions. According to the variability of each direction in the semivariogram, the direction with stronger variability is determined as the main axis, and the direction with weaker variability is determined as the secondary axis. Then, a spatial covariance model is constructed based on the semivariogram fitted by the Gaussian model, and the temperature and humidity of unmarked points are spatially predicted using the Kriging interpolation method.

4. The indoor temperature and humidity spatial data prediction method integrating spatial interpolation and graph convolutional network according to claim 3 is characterized in that: The specific steps are as follows: Calculate the Euclidean distance between each pair of marked points, and obtain the point pair (x i ,x j ) space distance h ij ; Then, group the point pairs in horizontal and vertical directions, and calculate the point pairs in different directions (x i ,x j ) of the interpolated square of the temperature and humidity values ​​[z(x i )-z(x j )] 2 , where z(x) is the temperature or humidity value of the marked point x; For a given distance h, the average squared difference of all pairs of points with a distance h is calculated to calculate the semivariate value. The formula is as follows: Among them, γ(h k ) is the distance h k The semivariogram value between the point pairs, N(h k ) is the distance h k The number of point pairs; Based on the calculated semivariogram, an empirical semivariogram in each direction is drawn and fitted using a Gaussian model. The major and minor axes are determined by comparing the variability in each direction. Finally, the fitted semivariogram is used to define the spatial covariance model, which is reflected by For the target unmarked point, calculate the covariance between it and all known marked points and construct the covariance vector; Calculate the target area marker point P using the fitted semivariogram γ(h) and covariance function Cov(h) a With the unmarked point P u The covariance values ​​of By solving the linear system Cw = C u , calculate the Kriging weight w, and get the weight w of each known point to the unlabeled point i ; Finally, create the Kriging equation Get the linear prediction results of temperature and humidity at unmarked points in the indoor space, where Z(x i ) is a known point x i The temperature and humidity values ​​of the unmarked point x0 are given in Table 1. Z(x0) is the linear prediction result of the temperature and humidity of the unmarked point x0. i To assign x to each group i x u The weight coefficient of .

5. The indoor temperature and humidity spatial data prediction method integrating spatial interpolation and graph convolutional network according to claim 1 is characterized in that: In step 5), the feature matrix and adjacency matrix obtained in step 4) are input into the graph convolutional network model for training, and the feature matrix and the adjacency matrix As the input of the graph convolutional network GCN, in order to ensure that the sensor node’s own features participate in the convolution, the adjacency matrix Add to the self-loop identity matrix I to obtain the adjacency matrix with self-loops According to the degree of each sensor node Constructed degree matrix And perform symmetric normalization to obtain For each layer l (l = 1, 2, 3, ..., L), by Implement feature aggregation, where W (l) is a weight matrix, which introduces the nonlinear capability of the model through the nonlinear activation function σ=max(0,x); The optimization is performed by defining the mean square error as the loss function. The function formula is as follows: Use the gradient descent algorithm to minimize the loss, update the network weights, and finally output the feature matrix G of the layer after multiple layers of graph convolution. (L) Represented as the predicted value of temperature and humidity for each unlabeled node 6. The indoor temperature and humidity spatial data prediction method integrating spatial interpolation and graph convolutional network according to claim 1 or 5 is characterized in that: In step 5), the fitted prediction model is used to obtain the three-dimensional spatial point-surface distribution of indoor temperature and humidity in the study area, as follows: The distribution of three-dimensional target point data is presented in the form of three-dimensional scattered points. The three-dimensional mapping position of each target point in the real-scene three-dimensional model is determined according to the three-dimensional coordinates (x, y, z). The temperature or humidity of the target point is mapped to a color through the color mapping function to show the data level. The three-dimensional target surface data distribution is divided into temperature and humidity prediction values ​​through spatial slicing Rendering is performed to display the spatial distribution of indoor temperature and humidity; the study area is divided into multiple xy planes to construct multiple levels of distribution, and each layer generates the x and y range grid of the real-life 3D model of the corresponding study area. Applying a gradient color map will Map to the corresponding color parameters to obtain the three-dimensional surface data distribution of the indoor space.

Images