A method for identifying environmental risks in the site selection of a coating plant and related equipment

Abstract

The embodiment of the present invention provides an environmental risk identification method for the site selection of a coating plant, obtaining an adjacency matrix N*M of a target area, wherein the target area includes a target address and a reference address, and the adjacency matrix includes a central matrix unit corresponding to the target address and an adjacency matrix unit corresponding to the reference address, and each matrix unit in the adjacency matrix corresponds to a historical environmental data B i,j∈N*M , B i,j∈N*M The table is the environmental data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point; the spatiotemporal features of the adjacency matrix N*M are extracted to obtain the spatiotemporal features of the adjacency matrix; based on the spatiotemporal matrix, the environmental risk of the target address is predicted to obtain the environmental risk identification result of the target address. The present invention solves the problem that the existing method does not take into account the evaluation of the health risk of the human body after the pollutants are released into the environment, does not take into account other factors that lead to the occurrence of risk events, and has a greater impact on the environment during the construction and operation of the coating plant.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental pollution risk assessment, and in particular to an environmental risk identification method for coating plant site selection and related equipment. Background Art

[0002] In recent years, with the rapid development of my country's economy, the number of coating factories has continued to increase, and a large number of coating factory areas have been established. Due to the use of a large number of chemicals and complex production processes, in addition to the use of a large number of chemical raw materials, a variety of by-products are also produced in the production process. The "three wastes" generated in the production process are complex in composition and large in quantity. The wastewater, waste residue, and waste gas generated may cause long-term and potential harm to the environment or directly affect people's health. In environmental pollution risk assessment, pollution risk assessment based on existing characteristic indicators often only considers the characteristics of the indicators themselves, while there are many potential correlations between a large number of pollution characteristics, and potential correlations help to achieve more accurate pollution risk assessment.

[0003] At present, the existing methods mainly identify risks by investigating regional environmental basic data and environmental risk sources. They do not take into account the evaluation of human health risks after pollutants are released into the environment, nor do they take into account other factors that lead to risk events. The construction and operation of coating plants have a greater impact on the environment. Summary of the invention

[0004] The embodiment of the present invention provides an environmental risk identification method for the site selection of a coating plant, which aims to solve the problem that the existing existing methods mainly identify risks by investigating regional environmental basic data and environmental risk source items, without considering the evaluation of human health risks after pollutants are released into the environment, without considering other factors that cause risk events, and the construction and operation of coating plants have a greater impact on the environment. By obtaining the adjacency matrix N*M of the target area and extracting the spatiotemporal features of the adjacency matrix N*M, the spatiotemporal features of the adjacency matrix are obtained, and the spatiotemporal matrix is ​​used to predict the environmental risks of the target address, thereby obtaining the environmental risk identification result of the target address, so as to solve the problem that the existing existing methods mainly identify risks by investigating regional environmental basic data and environmental risk source items, without considering the evaluation of human health risks after pollutants are released into the environment, without considering other factors that cause risk events, and the construction and operation of coating plants have a greater impact on the environment.

[0005] In a first aspect, an embodiment of the present invention provides an environmental risk identification method for site selection of a coating plant, the method comprising:

[0006] Obtain an adjacency matrix N*M of the target area, wherein the target area includes a target address and a reference address, the adjacency matrix includes a central matrix unit corresponding to the target address and an adjacency matrix unit corresponding to the reference address, and each matrix unit in the adjacency matrix corresponds to a historical environment data B i,j∈N*M , the B i,j∈N*M The table is the environment data of the address corresponding to the jth matrix unit among the N*M matrix units at the i-th time point;

[0007] Extracting spatiotemporal features from the adjacency matrix N*M to obtain spatiotemporal features of the adjacency matrix;

[0008] Based on the space-time matrix, the environmental risk of the target address is predicted to obtain an environmental risk identification result of the target address.

[0009] Optionally, the step of obtaining the adjacency matrix N*M of the target area includes:

[0010] Determine a target address in the target area, and determine the construction scale of the target factory building;

[0011] Determine the environmental impact area of ​​the target plant site based on the built-up scale of the target plant;

[0012] In the environmental impact area, K candidate addresses are determined, where K is greater than N*M;

[0013] Among the K candidate addresses, (N*M-1) reference addresses are selected with the target address as the center;

[0014] Acquire the first historical environment data corresponding to the (N*M-1) reference addresses and the second historical environment data corresponding to the target address;

[0015] Based on the (N*M-1) reference addresses, the target address, the first historical environment data and the second historical environment data, an adjacency matrix N*M of the target area is constructed.

[0016] Optionally, within the environmental impact area, the step of determining K candidate addresses includes:

[0017] determining an intersection area between the target area and the environmental impact area;

[0018] If the area ratio of the intersection region to the target region is greater than or equal to a preset ratio threshold, K candidate addresses are determined in the intersection region;

[0019] If the area ratio of the intersection area to the target area is less than a preset ratio threshold, K candidate addresses are determined proportionally in the intersection area, the non-intersection area of ​​the target area, and the non-intersection area of ​​the environmental impact area based on the area ratio.

[0020] Optionally, the step of selecting (N*M-1) reference addresses from the K candidate addresses with the target address as the center includes:

[0021] Calculate the vectors between the target address and each candidate address to obtain K address vectors, each address vector corresponding to one candidate address;

[0022] Based on the address vector, (N*M-1) reference addresses are selected from the K candidate addresses.

[0023] Optionally, the step of selecting (N*M-1) reference addresses from the K candidate addresses based on the address vector includes:

[0024] Sort the K address vectors according to the vector modulus length to obtain an address vector sequence;

[0025] In the address vector sequence, traverse and calculate the vector angle between the first address vector and the second address vector, the first address vector is sorted before the second address vector;

[0026] If the vector angle is less than a preset angle threshold, discarding the second address vector, and determining the next address vector after the second address vector as the second address vector;

[0027] If the vector angle is greater than or equal to a preset angle threshold, determining the second address vector as a reference address;

[0028] If (N*M-1) reference addresses are obtained, stop traversing;

[0029] If the traversal is completed and T reference addresses are obtained, the remaining address vectors are sorted according to the vector modulus length to obtain a remaining address vector sequence, where T is less than (N*M-1);

[0030] Among the address vectors corresponding to the T reference addresses, determining the first second address vector after the first address vector as a new first address vector;

[0031] In the remaining address vector sequence, iterative calculation is performed to determine R reference addresses, R+T=(N*M-1).

[0032] Optionally, the step of selecting (N*M-1) reference addresses from the K candidate addresses with the target address as the center includes:

[0033] Based on the target address and the K candidate addresses, construct a Delaunay triangulation network, wherein the Delaunay triangulation network includes a plurality of Delaunay triangles, and each of the Delaunay triangles does not include any other addresses except the three vertex addresses constituting the Delaunay triangle;

[0034] In the Delaunay triangulation, determining K1 first target Delaunay triangles whose vertex addresses include the target address, where K1 is less than K;

[0035] When the number of vertices of the K1 first target Delaunay triangles other than the target address vertex is less than 8, determine the first target Delaunay triangle with the largest first circumscribed circle area among the K1 first target Delaunay triangles, connect the target address with the center point of the first target Delaunay triangle with the largest first circumscribed circle area, and extend the ray starting from the target address to obtain a discrimination ray;

[0036] Determine the first non-first target Delaunay triangle intersecting the discrimination ray as the first target Delaunay triangle, until the number of vertices of the first target Delaunay triangle, excluding the target address vertex, reaches 8;

[0037] When the number of vertices of the K1 first target Delaunay triangles other than the target address vertex is greater than 8, determining the triangle area of ​​each of the K1 first target Delaunay triangles, and determining the second circumscribed circle area of ​​each of the first target Delaunay triangles;

[0038] Deleting vertices other than the target address vertex in the first target Delaunay triangle whose triangle area and the second circumscribed area are the smallest, until the number of vertices other than the target address vertex of the first target Delaunay triangle reaches 8;

[0039] Using the 8 vertices outside the target address vertex, search outward for new vertices of the adjacent Delaunay triangle until (N*M-1) vertices are obtained;

[0040] In the Delaunay triangulation, corresponding (N*M-1) reference addresses are determined based on (N*M-1) vertices.

[0041] Optionally, the step of extracting spatiotemporal features from the adjacency matrix N*M to obtain the spatiotemporal features of the adjacency matrix includes:

[0042] Extracting spatiotemporal features of the adjacency matrix N*M through the trained capsule network to obtain the spatiotemporal features of the adjacency matrix;

[0043] The step of predicting the environmental risk of the target address based on the spatiotemporal characteristics to obtain the environmental risk identification result of the target address includes:

[0044] The temporal and spatial features are predicted and processed by a trained capsule network to obtain an environmental risk identification result of the target address;

[0045] The capsule network is trained by a data set, and the data set includes a sample adjacency matrix and a label matrix. The sample adjacency matrix is ​​obtained in the same way as the adjacency matrix N*M. Each sample matrix unit in the sample adjacency matrix corresponds to a historical environmental data C from time point T1 to time point T2. i∈[T1，T2],j∈N*M , the C i∈[T1，T2],j∈N*M Indicates the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T1 time point and the T2 time point. Each label matrix unit in the label matrix corresponds to a historical environment data D between the T3 time point and the T4 time point. i∈[T3，T4],j∈N*M , the D i∈[T3，T4],j∈N*M Represents the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T3 time point and the T4 time point, wherein the T1 time point is earlier than the T2 time point, the T3 time point is earlier than the T4 time point, and the T2 time point is earlier than the T3 time point.

[0046] In a second aspect, an embodiment of the present invention further provides an environmental risk identification device for site selection of a coating plant, the environmental risk identification device for site selection of a coating plant comprising:

[0047] The acquisition module is used to acquire the adjacency matrix N*M of the target area, wherein the target area includes the target address and the reference address, and the adjacency matrix includes the central matrix unit corresponding to the target address and the adjacency matrix unit corresponding to the reference address, and each matrix unit in the adjacency matrix corresponds to a historical environment data B i,j∈N*M , the B i,j∈N*M Represents the environment data of the address corresponding to the jth matrix unit among the N*M matrix units at the i-th time point;

[0048] An extraction module, used for performing spatiotemporal feature extraction on the adjacency matrix N*M to obtain the spatiotemporal features of the adjacency matrix;

[0049] The search module is used to predict the environmental risk of the target address based on the spatiotemporal characteristics to obtain an environmental risk identification result of the target address.

[0050] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the environmental risk identification method for paint plant site selection provided in an embodiment of the present invention are implemented.

[0051] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the environmental risk identification method for paint plant site selection provided in the embodiment of the invention are implemented.

[0052] In the embodiment of the present invention, an adjacency matrix N*M of a target area is obtained, the target area includes a target address and a reference address, the adjacency matrix includes a central matrix unit corresponding to the target address and an adjacency matrix unit corresponding to the reference address, and each matrix unit in the adjacency matrix corresponds to a historical environment data B i,j∈N*M , B i,j∈N*M The table is the environmental data of the address corresponding to the jth matrix unit in the N*M matrix units at the ith time point; the spatiotemporal features of the adjacency matrix N*M are extracted to obtain the spatiotemporal features of the adjacency matrix; based on the spatiotemporal matrix, the environmental risk of the target address is predicted to obtain the environmental risk identification result of the target address. The present invention can solve the problem that the existing existing methods mainly identify risks by investigating regional environmental basic data and environmental risk source items, without considering the evaluation of human health risks after the pollutants are released into the environment, without considering other factors that cause risk events, and the construction and operation of the coating plant have a greater impact on the environment. It improves the construction and operation of the coating plant to better consider environmental factors and avoid possible environmental problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 It is a flow chart of an environmental risk identification method for coating plant site selection provided by an embodiment of the present invention;

[0055] Figure 2 It is a structural schematic diagram of an environmental risk identification device for coating plant site selection provided by an embodiment of the present invention;

[0056] Figure 3It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] like Figure 1 As shown, Figure 1 This is a method flow chart of an environmental risk identification method for coating plant site selection provided by an embodiment of the present invention. The environmental risk identification method for coating plant site selection includes the following steps:

[0059] 101. Obtain the adjacency matrix N*M of the target area.

[0060] In an embodiment of the present invention, the above-mentioned environmental risk identification method for site selection of a coating plant can be applied to smart devices. The above-mentioned smart devices can be understood as smart devices that need to perform environmental risk identification for site selection of a coating plant, such as various types of mobile terminal devices.

[0061] The target area includes a target address and a reference address. The target address can be understood as the site selection location of the coating plant, and the reference address can be understood as the adjacent geographical location of the target address. The reference address includes multiple addresses.

[0062] The adjacency matrix includes a central matrix unit corresponding to the target address and an adjacency matrix unit corresponding to the reference address. Each matrix unit in the adjacency matrix corresponds to a historical environment data B i,j∈N*M , B i,j∈N*M Represents the environment data of the address corresponding to the j-th matrix unit in the N*M matrix units at the i-th time point.

[0063] The adjacency matrix is ​​a data structure representing a graph structure, which is used to store the connection relationship between each vertex in the graph. The adjacency matrix is ​​used to represent the relationship structure between the target address and the reference address. If the straight-line distance between the target address and the reference address is shorter, the corresponding element value in the adjacency matrix is ​​higher, and vice versa, the corresponding element value in the adjacency matrix is ​​lower.

[0064] The above historical environmental data can be understood as the environmental conditions of the geographical location over a period of history.

[0065] The above environmental data includes industrial environmental information, natural environmental information and human settlement environment information. The above industrial environmental information can be understood as various types of information related to industrial production and manufacturing, including information on industrial development status and market demand. The above natural environmental information can be understood as various types of information related to the natural environment, including climate, topography, water sources, soil, vegetation, animals, etc. The above human settlement environment information can be understood as various types of information related to the human settlement environment, including population distribution, residents' living standards, traffic conditions, etc.

[0066] The above N can be understood as the data of the reference address, N is a quantifier, and can be 3, 4, 5, 6, etc. The above M can be understood as the number of adjacent areas around each reference address, M is a quantifier, and can be 1, 2, 3, 4, 5, etc.

[0067] It should be noted that by obtaining the adjacency matrix N*M of the target area, it is convenient to query and analyze environmental data at different time points and addresses.

[0068] 102. Perform spatiotemporal feature extraction on the adjacency matrix N*M to obtain the spatiotemporal features of the adjacency matrix.

[0069] In the embodiment of the present invention, the above-mentioned spatiotemporal feature extraction can be understood as the process of extracting feature information with temporal and spatial correlation from the adjacency matrix N*M. It can be obtained by training the data set, and then the environmental risk of the target address is predicted to obtain the environmental risk identification result of the target address.

[0070] The trained capsule network is obtained by training the untrained capsule network with the data set, and the data set includes the sample adjacency matrix and the label matrix. The above data set includes the sample adjacency matrix and the label matrix. The sample adjacency matrix is ​​obtained in the same way as the adjacency matrix N*M of the target address above. Each sample matrix unit in the sample adjacency matrix corresponds to a historical environmental data C from time point T1 to time point T2. i∈[T1，T2],j∈N*M , C i∈[T1，T2],j∈N*M Indicates the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T1 time point and the T2 time point. Each label matrix unit in the label matrix corresponds to a historical environment data D between the T3 time point and the T4 time point. i∈[T3，T4],j∈N*M , D i∈[T3，T4],j∈N*M Represents the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T3 time point and the T4 time point, where the T1 time point is earlier than the T2 time point, the T3 time point is earlier than the T4 time point, and the T2 time point is earlier than the T3 time point.

[0071] Furthermore, the trained capsule network can be used to extract spatiotemporal features from the adjacency matrix N*M. The trained capsule network can extract spatiotemporal features from the adjacency matrix N*M. The capsule network will learn the connection relationship between nodes and the attribute information of edges from the adjacency matrix N*M, thereby extracting a vector representation with spatiotemporal features.

[0072] The spatiotemporal characteristics of the adjacency matrix mentioned above can reflect the changing trends and laws of the adjacency matrix at different time points, and can reflect the mutual influence relationship and spatial distribution law between different addresses in the adjacency matrix.

[0073] It should be noted that the time series analysis and spatial correlation analysis of the elements in the adjacency matrix N*M are performed through the trained capsule network to obtain the spatiotemporal characteristics of the adjacency matrix.

[0074] 103. Based on the temporal and spatial characteristics, the environmental risks of the target address are predicted to obtain the environmental risk identification results of the target address.

[0075] In an embodiment of the present invention, the environmental risk of the target address can be predicted by the spatiotemporal characteristics of the adjacency matrix through a trained capsule network.

[0076] It should be noted that the present invention predicts the environmental risk of the target address through the capsule network, thereby obtaining an environmental risk identification result. The environmental risk identification result can better understand the environmental conditions and potential risks of the target address.

[0077] In the embodiment of the present invention, an adjacency matrix N*M of a target area is obtained, the target area includes a target address and a reference address, the adjacency matrix includes a central matrix unit corresponding to the target address and an adjacency matrix unit corresponding to the reference address, and each matrix unit in the adjacency matrix corresponds to a historical environment data B i,j∈N*M , B i,j∈N*MThe table is the environmental data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point; the spatiotemporal feature extraction of the adjacency matrix N*M is performed to obtain the spatiotemporal feature of the adjacency matrix; based on the spatiotemporal matrix, the environmental risk of the target address is predicted to obtain the environmental risk identification result of the target address. The present invention can solve the problem that the existing site selection method relies on the experience of the investigators to select the site, resulting in the subjectivity and uncertainty of the site selection result, and the construction and operation of the coating plant have a greater impact on the environment. By extracting the spatiotemporal feature of the adjacency matrix N*M, the spatiotemporal feature of the adjacency matrix is ​​obtained, and the spatiotemporal matrix is ​​used to predict the environmental risk of the target address, so as to obtain the environmental risk identification result of the target address, so as to solve the problem that the existing method mainly identifies risks by investigating the basic data of the regional environment and the source of environmental risks, without considering the evaluation of the health risks of human beings after the pollutants are released into the environment, and without considering other factors that cause the occurrence of risk events, and the construction and operation of the coating plant have a greater impact on the environment. It improves the construction and operation of the coating plant, better considers environmental factors, and avoids possible environmental problems.

[0078] Optionally, the step of obtaining the adjacency matrix N*M of the target area includes:

[0079] Determine a target address in a target area, and determine the built-up scale of a target plant; determine an environmental impact area of ​​the target plant based on the built-up scale of the target plant; determine K candidate addresses in the environmental impact area, where K is greater than N*M; among the K candidate addresses, select (N*M-1) reference addresses with the target address as the center; obtain first historical environmental data corresponding to the (N*M-1) reference addresses, and second historical environmental data corresponding to the target address; construct an adjacency matrix N*M of the target area based on the (N*M-1) reference addresses, the target address, the first historical environmental data, and the second historical environmental data.

[0080] In the embodiment of the present invention, the above-mentioned target address can be understood as the site selection location of the coating plant.

[0081] The construction scale of the above-mentioned factory building includes overall planning, facilities and equipment, etc. The above-mentioned overall planning includes the building area of ​​the factory building, such as building area, building planning, etc. The above-mentioned facilities and equipment include various equipment required for the factory building and the required raw material information, such as spraying equipment, drying equipment, conveying equipment, etc., and the required raw materials such as coatings, solvents, additives, etc.

[0082] Furthermore, the environmental areas that may be affected are determined based on the scale and nature of the target plant. For example, if the plant is large or has special properties (such as a coating plant, chemical plant, etc.), then the area of ​​its impact may be larger, and may include surrounding air, water, land and other environmental elements. For example, a large coating plant may pollute the surrounding air and water, so an environmental impact area needs to be determined to facilitate environmental monitoring and protection. Similarly, large chemical plants may also have similar environmental impacts. In addition, if the plant is located in a densely populated area or an ecologically fragile area, then the area of ​​its impact may be larger.

[0083] The above-mentioned environmental impact area can be understood as the area affected by the environment generated by the target factory building. The environmental impacts include factors such as pollutant emissions, noise, and ecological damage. Among them, pollutant emissions include waste gas, waste water, waste residue, etc.

[0084] The above K candidate addresses can be understood as K locations that can be selected as the target plant location within the environmental impact area. These candidate addresses can be pre-set or determined according to actual conditions.

[0085] The above historical environmental data refers to the environmental conditions of the geographical location over a period of time in the past.

[0086] The above K is a quantifier, which can be 1, 2, 3, 4, etc.

[0087] It should be noted that the target address and the built-up scale of the target plant can be determined in the target area, and the environmental impact area of ​​the target plant can be determined based on the built-up scale of the target plant, and K candidate addresses can be determined within the environmental impact area. Among the K candidate addresses, (N*M-1) reference addresses are selected with the target address as the center, and the first historical environmental data corresponding to the (N*M-1) reference addresses and the second historical environmental data corresponding to the target address are obtained. The adjacency matrix N*M of the target area is constructed based on the (N*M-1) reference addresses, the target address, the first historical environmental data and the second historical environmental data.

[0088] Optionally, within the environmental impact area, the step of determining K candidate addresses includes:

[0089] Determine the intersection area of ​​the target area and the environmental impact area; if the area ratio of the intersection area to the target area is greater than or equal to a preset ratio threshold, then determine K candidate addresses in the intersection area; if the area ratio of the intersection area to the target area is less than the preset ratio threshold, then based on the area ratio, determine K candidate addresses in proportion in the intersection area, the non-intersection area of ​​the target area, and the non-intersection area of ​​the environmental impact area.

[0090] In the embodiment of the present invention, the above-mentioned intersection area can be understood as the intersection of the target area and the environmental impact area, that is, the common area of ​​the target area and the environmental impact area.

[0091] The above-mentioned preset ratio threshold is a ratio threshold pre-set by the system and is used as a basis for judgment.

[0092] It should be noted that the present invention determines the intersection of the target area and the environmental impact area. If the ratio of the area of ​​the intersection to the area of ​​the target area is greater than or equal to a preset ratio threshold, then K candidate addresses are determined in the intersection. If the area ratio of the intersection area to the target area is less than the preset ratio threshold, then K candidate addresses are determined in proportion to the intersection area, the non-intersection area of ​​the target area, and the non-intersection area of ​​the environmental impact area according to the area ratio.

[0093] Optionally, the step of selecting (N*M-1) reference addresses from the K candidate addresses with the target address as the center includes:

[0094] Calculate the vectors between the target address and each candidate address to obtain K address vectors, each address vector corresponds to a candidate address; based on the address vectors, select (N*M-1) reference addresses from the K candidate addresses.

[0095] In the embodiment of the present invention, the vectors between the target address and each candidate address may reflect the distance or direction relationship between the target address and the candidate address. Each address vector corresponds to a candidate address.

[0096] It should be noted that by calculating the distance or direction relationship from the target address to each candidate address, K address vectors are obtained, each address vector corresponds to a candidate address, and among the K candidate addresses, (N*M-1) reference addresses are selected according to the previously calculated address vectors.

[0097] Optionally, based on the address vector, the step of selecting (N*M-1) reference addresses from K candidate addresses includes:

[0098] Sort K address vectors according to the vector modulus to obtain an address vector sequence; in the address vector sequence, traverse and calculate the vector angle between the first address vector and the second address vector, and the first address vector is sorted before the second address vector; if the vector angle is less than a preset angle threshold, the second address vector is discarded, and the next address vector after the second address vector is determined as the second address vector; if the vector angle is greater than or equal to the preset angle threshold, the second address vector is determined as the reference address; if (N*M-1) reference addresses are obtained, stop traversing; if the traversal is completed, if T reference addresses are obtained, sort the remaining address vectors according to the vector modulus to obtain a remaining address vector sequence, where T is less than (N*M-1); in the address vectors corresponding to the T reference addresses, the first second address vector after the first address vector is determined as a new first address vector; in the remaining address vector sequence, iteratively calculate and determine R reference addresses, where R+T=(N*M-1).

[0099] In the embodiment of the present invention, the above-mentioned sorting by vector modulus may be understood as a process of sorting the K address vectors according to the modulus of each address vector. The above-mentioned modulus may be a distance.

[0100] Furthermore, the sorted address vectors are combined into a new address vector sequence in the order of their modulus lengths. This address vector sequence can reflect the importance or influence of each address vector.

[0101] The above vector angle can be understood as the angle between two vectors. The vector angle can be used to describe the relative position or motion state between two objects, such as the relative direction, speed or acceleration between two objects. At the same time, the vector angle can also be used to describe the relationship between two variables, such as the correlation between two variables or the degree of mutual influence.

[0102] The above T is a quantifier, which can be 1, 2, 3, 4, etc.

[0103] The above-mentioned iterative calculation is a method of gradually approaching the solution of a problem by repeated calculation. Iterative calculation is to decompose a complex problem into several simple sub-problems, and finally obtain the solution of the problem by repeatedly calculating these sub-problems.

[0104] Specifically, K address vectors are sorted by vector modulus to obtain an address vector sequence. Then, in the address vector sequence, the vector angle between each pair of adjacent address vectors is traversed and calculated, and the second vector in the vector pair with an angle less than a preset threshold is retained as a reference address. When N*M-1 reference addresses are obtained, the traversal is stopped. If the number of reference addresses obtained is less than NM-1, the remaining address vectors are sorted by vector modulus to obtain a new address vector sequence, and T reference addresses are determined in the new address vector sequence, where T is less than NM-1. Finally, in the address vectors corresponding to the T reference addresses, the first second address vector after the first address vector is determined as the new first address vector; in the remaining address vector sequence, iterative calculation is performed, and R reference addresses are determined, so that R+T=(N*M-1).

[0105] Optionally, the step of selecting (N*M-1) reference addresses from the K candidate addresses with the target address as the center includes:

[0106] Based on the target address and K candidate addresses, a Delaunay triangulation is constructed, where the Delaunay triangulation includes multiple Delaunay triangles, each of which does not include any other addresses except the three vertex addresses constituting the Delaunay triangle; in the Delaunay triangulation, K1 first target Delaunay triangles whose vertex addresses include the target address are determined, where K1 is less than K; when the number of vertices other than the target address vertex of the K1 first target Delaunay triangles is less than 8, the first target Delaunay triangle with the largest first circumscribed circle area is determined among the K1 first target Delaunay triangles, the target address is connected to the center point of the first target Delaunay triangle with the largest first circumscribed circle area, and the target address is extended as the starting point to obtain a discrimination ray; the first non-first target Delaunay triangle that intersects the discrimination ray is determined as the first target Delaunay triangle, until The number of vertices of the first target Delaunay triangle other than the target address vertex reaches 8; when the number of vertices of K1 first target Delaunay triangles other than the target address vertex is greater than 8, determine the triangle area of ​​each first target Delaunay triangle in the K1 first target Delaunay triangles, and determine the second circumscribed circle area of ​​each first target Delaunay triangle; delete the vertices other than the target address vertex in the first target Delaunay triangle with the smallest triangle area and second circumscribed area until the number of vertices of the first target Delaunay triangle other than the target address vertex reaches 8; with the 8 vertices other than the target address vertex, search outward for new vertices of adjacent Delaunay triangles until (N*M-1) vertices are obtained; in the Delaunay triangulation, determine the corresponding (N*M-1) reference addresses based on the (N*M-1) vertices.

[0107] In an embodiment of the present invention, the above-mentioned Delaunay triangulation can be understood as a set of a series of connected but non-overlapping triangles, and the circumscribed circles of these triangles do not contain any other points in the face domain. The Delaunay triangulation forms a series of triangles by triangulating a discrete point set. The circumscribed circle of each triangle of the Delaunay triangulation does not contain any other points in the face.

[0108] The above-mentioned determination ray can be understood as a ray extending from a target point to a certain direction, which is used to determine whether a point is inside, on the boundary or outside a certain polygon.

[0109] Optionally, the step of extracting spatiotemporal features from the adjacency matrix N*M to obtain the spatiotemporal features of the adjacency matrix includes:

[0110] The spatiotemporal features of the adjacency matrix N*M are extracted through the trained capsule network to obtain the spatiotemporal features of the adjacency matrix.

[0111] In an embodiment of the present invention, the capsule network is a neural network structure that simulates the working principle of human neurons and can capture the spatiotemporal dependencies and dynamic changes in data.

[0112] Furthermore, the capsule network is trained through a data set, which includes a sample adjacency matrix and a label matrix. The sample adjacency matrix is ​​obtained in the same way as the adjacency matrix N*M. Each sample matrix unit in the sample adjacency matrix corresponds to a historical environmental data C from time point T1 to time point T2. i∈[T1，T2],j∈N*M , C i∈[T1，T2],j∈N*M Indicates the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T1 time point and the T2 time point. Each label matrix unit in the label matrix corresponds to a historical environment data D from the T3 time point to the T4 time point. i∈[T3，T4],j∈N*M , D i∈[T3，T4],j∈N*M Represents the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T3 time point and the T4 time point, where the T1 time point is earlier than the T2 time point, the T3 time point is earlier than the T4 time point, and the T2 time point is earlier than the T3 time point.

[0113] Furthermore, the untrained capsule network is supervisedly trained through the data set, and after the training is completed, a trained capsule network is obtained. The untrained capsule network refers to an untrained capsule network.

[0114] Specifically, the trained capsule network extracts spatiotemporal features from the adjacency matrix N*M. The capsule network learns the connection relationship between nodes and the attribute information of edges from the adjacency matrix N*M, thereby extracting a vector representation with spatiotemporal features.

[0115] Optionally, based on the spatiotemporal characteristics, the environmental risk of the target address is predicted, and the step of obtaining the environmental risk identification result of the target address includes:

[0116] The trained capsule network is used to predict and process the spatiotemporal features to obtain the environmental risk identification results of the target address. The capsule network is trained through a data set, which includes a sample adjacency matrix and a label matrix. The sample adjacency matrix is ​​obtained in the same way as the adjacency matrix N*M. Each sample matrix unit in the sample adjacency matrix corresponds to a historical environmental data C from time point T1 to time point T2. i∈[T1，T2],j∈N*M , C i∈[T1，T2],j∈N*M Indicates the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T1 time point and the T2 time point. Each label matrix unit in the label matrix corresponds to a historical environment data D from the T3 time point to the T4 time point. i∈[T3，T4],j∈N*M , D i∈[T3，T4],j∈N*M Represents the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T3 time point and the T4 time point, where the T1 time point is earlier than the T2 time point, the T3 time point is earlier than the T4 time point, and the T2 time point is earlier than the T3 time point.

[0117] In the embodiment of the present invention, the capsule network is a neural network structure that simulates the working principle of human neurons and can capture the spatiotemporal dependencies and dynamic changes in data. During the training process, the capsule network automatically learns how to predict environmental risks based on the input spatiotemporal features.

[0118] It should be noted that the present invention uses capsule network to predict the environmental risk of the target address's spatiotemporal characteristics to obtain the environmental risk identification result of the target address. The environmental risk identification result can help users understand the environmental risks that the target address may face and provide strong support for the environmental risk identification of the target address.

[0119] like Figure 2 As shown, an embodiment of the present invention provides an environmental risk identification device for site selection of a coating plant, and the environmental risk identification device for site selection of a coating plant includes:

[0120] The acquisition module 201 is used to acquire an adjacency matrix N*M of a target area, wherein the target area includes a target address and a reference address, and the adjacency matrix includes a central matrix unit corresponding to the target address and an adjacency matrix unit corresponding to the reference address, and each matrix unit in the adjacency matrix corresponds to a historical environment data B i,j∈N*M , the B i,j∈N*M The table is the environment data of the address corresponding to the jth matrix unit among the N*M matrix units at the i-th time point;

[0121] An extraction module 202 is used to extract the spatiotemporal features of the adjacency matrix N*M to obtain the spatiotemporal features of the adjacency matrix;

[0122] The prediction module 203 is used to predict the environmental risk of the target address based on the space-time matrix to obtain an environmental risk identification result of the target address.

[0123] Optionally, the acquisition module 201 is also used to determine the target address in the target area, and determine the built-up scale of the target plant; based on the built-up scale of the target plant, determine the environmental impact area of ​​the target site; within the environmental impact area, determine K candidate addresses, K is greater than N*M; among the K candidate addresses, select (N*M-1) reference addresses with the target address as the center; obtain the first historical environment data corresponding to the (N*M-1) reference addresses, and the second historical environment data corresponding to the target address; based on the (N*M-1) reference addresses, the target address, the first historical environment data and the second historical environment data, construct an adjacency matrix N*M of the target area.

[0124] Optionally, the acquisition module 201 is also used to determine the intersection area of ​​the target area and the environmental impact area; if the area ratio of the intersection area to the target area is greater than or equal to a preset ratio threshold, K candidate addresses are determined in the intersection area; if the area ratio of the intersection area to the target area is less than a preset ratio threshold, based on the area ratio, K candidate addresses are determined in proportion in the intersection area, the non-intersection area of ​​the target area, and the non-intersection area of ​​the environmental impact area.

[0125] Optionally, the acquisition module 201 is further used to calculate the vectors between the target address and each candidate address to obtain K address vectors, each address vector corresponding to one candidate address; based on the address vectors, (N*M-1) reference addresses are selected from the K candidate addresses.

[0126] Optionally, the acquisition module 201 is also used to sort the K address vectors according to the vector modulus to obtain an address vector sequence; in the address vector sequence, traverse and calculate the vector angle between the first address vector and the second address vector, and the first address vector is sorted before the second address vector; if the vector angle is less than a preset angle threshold, the second address vector is discarded, and the next address vector after the second address vector is determined as the second address vector; if the vector angle is greater than or equal to the preset angle threshold, the second address vector is determined as the reference address; if (N*M-1) reference addresses are obtained, the traversal is stopped; if the traversal is completed, if T reference addresses are obtained, the remaining address vectors are sorted according to the vector modulus to obtain a remaining address vector sequence, where T is less than (N*M-1); in the address vectors corresponding to the T reference addresses, the first second address vector after the first address vector is determined as a new first address vector; in the remaining address vector sequence, iterative calculation is performed, and R reference addresses are determined, where R+T=(N*M-1).

[0127] Optionally, the acquisition module 201 is also used to construct a Delaunay triangulation network based on the target address and the K candidate addresses, the Delaunay triangulation network including multiple Delaunay triangles, each of the Delaunay triangles not including any other addresses other than the three vertex addresses constituting the Delaunay triangle; in the Delaunay triangulation network, K1 first target Delaunay triangles whose vertex addresses include the target address are determined, K1 is less than K; when the number of vertices of the K1 first target Delaunay triangles other than the target address vertex is less than 8, among the K1 first target Delaunay triangles, the first target Delaunay triangle with the largest first circumscribed circle area is determined, the target address is connected to the center point of the first target Delaunay triangle with the largest first circumscribed circle area, and the target address is extended as the starting point to obtain a discrimination ray; the first non-first target Delaunay triangle that intersects the discrimination ray is determined as the the first target Delaunay triangle, until the number of vertices of the first target Delaunay triangle other than the target address vertex reaches 8; when the number of vertices of the K1 first target Delaunay triangles other than the target address vertex is greater than 8, determine the triangle area of ​​each of the K1 first target Delaunay triangles, and determine the second circumscribed circle area of ​​each of the first target Delaunay triangles; delete the vertices other than the target address vertex in the first target Delaunay triangle with the smallest triangle area and the second circumscribed circle, until the number of vertices of the first target Delaunay triangle other than the target address vertex reaches 8; with the 8 vertices other than the target address vertices, search outward for new vertices of adjacent Delaunay triangles until (N*M-1) vertices are obtained; in the Delaunay triangulation, determine the corresponding (N*M-1) reference addresses according to the (N*M-1) vertices.

[0128] Optionally, the extraction module 202 is further used to extract spatiotemporal features of the adjacency matrix N*M through a trained capsule network to obtain the spatiotemporal features of the adjacency matrix;

[0129] The prediction module 203 is also used to predict the spatiotemporal features through the trained capsule network to obtain the environmental risk identification result of the target address; wherein the capsule network is trained through a data set, and the data set includes a sample adjacency matrix and a label matrix. The sample adjacency matrix is ​​obtained in the same way as the adjacency matrix N*M. Each sample matrix unit in the sample adjacency matrix corresponds to a historical environmental data C from time point T1 to time point T2. i∈[T1，T2],j∈N*M , the C i∈[T1，T2],j∈N*MIndicates the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T1 time point and the T2 time point. Each label matrix unit in the label matrix corresponds to a historical environment data D between the T3 time point and the T4 time point. i∈[T3，T4],j∈N*M , the D i∈[T3，T4],j∈N*M Represents the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T3 time point and the T4 time point, wherein the T1 time point is earlier than the T2 time point, the T3 time point is earlier than the T4 time point, and the T2 time point is earlier than the T3 time point.

[0130] The environmental risk identification device for site selection of a coating plant provided by the embodiment of the present invention can implement each process implemented by the environmental risk identification method for site selection of a coating plant in the above method embodiment, and can achieve the same beneficial effects. To avoid repetition, it will not be described here.

[0131] See also Figure 3 , Figure 3 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, it includes: a memory 302, a processor 301, and a computer program of an environmental risk identification method for coating plant site selection stored in the memory 302 and executable on the processor 301, wherein:

[0132] The processor 301 is used to call the computer program stored in the memory 302 and execute the following steps:

[0133] Obtain an adjacency matrix N*M of the target area, wherein the target area includes a target address and a reference address, the adjacency matrix includes a central matrix unit corresponding to the target address and an adjacency matrix unit corresponding to the reference address, and each matrix unit in the adjacency matrix corresponds to a historical environment data B i,j∈N*M , the B i,j∈N*M The table is the environment data of the address corresponding to the jth matrix unit among the N*M matrix units at the i-th time point;

[0134] Extracting spatiotemporal features from the adjacency matrix N*M to obtain spatiotemporal features of the adjacency matrix;

[0135] Based on the space-time matrix, the environmental risk of the target address is predicted to obtain an environmental risk identification result of the target address.

[0136] Optionally, the step of obtaining the adjacency matrix N*M of the target area performed by the processor 301 includes:

[0137] Determine a target address in the target area, and determine the construction scale of the target factory building;

[0138] Determine the environmental impact area of ​​the target plant site based on the built-up scale of the target plant;

[0139] In the environmental impact area, K candidate addresses are determined, where K is greater than N*M;

[0140] Among the K candidate addresses, (N*M-1) reference addresses are selected with the target address as the center;

[0141] Acquire the first historical environment data corresponding to the (N*M-1) reference addresses and the second historical environment data corresponding to the target address;

[0142] Based on the (N*M-1) reference addresses, the target address, the first historical environment data and the second historical environment data, an adjacency matrix N*M of the target area is constructed.

[0143] Optionally, the step of determining K candidate addresses in the environmental impact area executed by the processor 301 includes:

[0144] determining an intersection area between the target area and the environmental impact area;

[0145] If the area ratio of the intersection region to the target region is greater than or equal to a preset ratio threshold, K candidate addresses are determined in the intersection region;

[0146] If the area ratio of the intersection area to the target area is less than a preset ratio threshold, K candidate addresses are determined proportionally in the intersection area, the non-intersection area of ​​the target area, and the non-intersection area of ​​the environmental impact area based on the area ratio.

[0147] Optionally, the step of selecting (N*M-1) reference addresses from the K candidate addresses with the target address as the center, performed by the processor 301, includes:

[0148] Calculate the vectors between the target address and each candidate address to obtain K address vectors, each address vector corresponding to one candidate address;

[0149] Based on the address vector, (N*M-1) reference addresses are selected from the K candidate addresses.

[0150] Optionally, the step of selecting (N*M-1) reference addresses from the K candidate addresses based on the address vector, performed by the processor 301, includes:

[0151] Sort the K address vectors according to the vector modulus length to obtain an address vector sequence;

[0152] In the address vector sequence, traverse and calculate the vector angle between the first address vector and the second address vector, the first address vector is sorted before the second address vector;

[0153] If the vector angle is less than a preset angle threshold, discarding the second address vector, and determining the next address vector after the second address vector as the second address vector;

[0154] If the vector angle is greater than or equal to a preset angle threshold, determining the second address vector as a reference address;

[0155] If (N*M-1) reference addresses are obtained, stop traversing;

[0156] If the traversal is completed and T reference addresses are obtained, the remaining address vectors are sorted according to the vector modulus length to obtain a remaining address vector sequence, where T is less than (N*M-1);

[0157] Among the address vectors corresponding to the T reference addresses, determining the first second address vector after the first address vector as a new first address vector;

[0158] In the remaining address vector sequence, iterative calculation is performed to determine R reference addresses, R+T=(N*M-1).

[0159] Optionally, the step of selecting (N*M-1) reference addresses from the K candidate addresses with the target address as the center, performed by the processor 301, includes:

[0160] Based on the target address and the K candidate addresses, construct a Delaunay triangulation network, wherein the Delaunay triangulation network includes a plurality of Delaunay triangles, and each of the Delaunay triangles does not include any other addresses except the three vertex addresses constituting the Delaunay triangle;

[0161] In the Delaunay triangulation, determining K1 first target Delaunay triangles whose vertex addresses include the target address, where K1 is less than K;

[0162] When the number of vertices of the K1 first target Delaunay triangles other than the target address vertex is less than 8, determine the first target Delaunay triangle with the largest first circumscribed circle area among the K1 first target Delaunay triangles, connect the target address with the center point of the first target Delaunay triangle with the largest first circumscribed circle area, and extend the ray starting from the target address to obtain a discrimination ray;

[0163] Determine the first non-first target Delaunay triangle intersecting the discrimination ray as the first target Delaunay triangle, until the number of vertices of the first target Delaunay triangle, excluding the target address vertex, reaches 8;

[0164] When the number of vertices of the K1 first target Delaunay triangles other than the target address vertex is greater than 8, determining the triangle area of ​​each of the K1 first target Delaunay triangles, and determining the second circumscribed circle area of ​​each of the first target Delaunay triangles;

[0165] Deleting vertices other than the target address vertex in the first target Delaunay triangle whose triangle area and the second circumscribed area are the smallest, until the number of vertices other than the target address vertex of the first target Delaunay triangle reaches 8;

[0166] Using the 8 vertices outside the target address vertex, search outward for new vertices of the adjacent Delaunay triangle until (N*M-1) vertices are obtained;

[0167] In the Delaunay triangulation, corresponding (N*M-1) reference addresses are determined based on (N*M-1) vertices.

[0168] Optionally, the step of extracting spatiotemporal features from the adjacency matrix N*M to obtain spatiotemporal features of the adjacency matrix performed by the processor 301 includes:

[0169] Extracting spatiotemporal features of the adjacency matrix N*M through the trained capsule network to obtain the spatiotemporal features of the adjacency matrix;

[0170] The step of predicting the environmental risk of the target address based on the spatiotemporal characteristics and obtaining the environmental risk identification result of the target address performed by the processor 301 includes:

[0171] The temporal and spatial features are predicted and processed by a trained capsule network to obtain an environmental risk identification result of the target address;

[0172] The capsule network is trained by a data set, and the data set includes a sample adjacency matrix and a label matrix. The sample adjacency matrix is ​​obtained in the same way as the adjacency matrix N*M. Each sample matrix unit in the sample adjacency matrix corresponds to a historical environmental data C from time point T1 to time point T2. i∈[T1，T2],j∈N*M , the C i∈[T1，T2],j∈N*MIndicates the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T1 time point and the T2 time point. Each label matrix unit in the label matrix corresponds to a historical environment data D between the T3 time point and the T4 time point. i∈[T3，T4],j∈N*M , the D i∈[T3，T4],j∈N*M Represents the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T3 time point and the T4 time point, wherein the T1 time point is earlier than the T2 time point, the T3 time point is earlier than the T4 time point, and the T2 time point is earlier than the T3 time point.

[0173] The electronic device provided in the embodiment of the present invention can implement each process implemented by the environmental risk identification method for site selection of a coating plant in the above method embodiment, and can achieve the same beneficial effect. To avoid repetition, it will not be described here.

[0174] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the various processes of the environmental risk identification method for site selection of a paint plant provided in an embodiment of the present invention or the environmental risk identification method for site selection of a paint plant on the application side, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0175] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The computer-readable storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0176] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for identifying environmental risks in the site selection of a coating plant, characterized in that: include: Obtain an adjacency matrix N*M of the target area, wherein the target area includes a target address and a reference address, the adjacency matrix includes a central matrix unit corresponding to the target address and an adjacency matrix unit corresponding to the reference address, and each matrix unit in the adjacency matrix corresponds to a historical environment data B i,j∈N*M , the B i,j∈N*M Represents the environment data of the address corresponding to the jth matrix unit among the N*M matrix units at the i-th time point; Extracting spatiotemporal features from the adjacency matrix N*M to obtain spatiotemporal features of the adjacency matrix; Based on the spatiotemporal characteristics, the environmental risk of the target address is predicted to obtain an environmental risk identification result of the target address.

2. The method according to claim 1, characterized in that The step of obtaining the adjacency matrix N*M of the target area includes: Determine a target address in the target area, and determine the construction scale of the target factory building; Determine the environmental impact area of ​​the target plant based on the built-up scale of the target plant; In the environmental impact area, K candidate addresses are determined, where K is greater than N*M; Among the K candidate addresses, (N*M-1) reference addresses are selected with the target address as the center; Acquire the first historical environment data corresponding to the (N*M-1) reference addresses and the second historical environment data corresponding to the target address; Based on the (N*M-1) reference addresses, the target address, the first historical environment data and the second historical environment data, an adjacency matrix N*M of the target area is constructed.

3. The method according to claim 2, characterized in that The step of determining K candidate addresses within the environmental impact area includes: determining an intersection area between the target area and the environmental impact area; If the area ratio of the intersection region to the target region is greater than or equal to a preset ratio threshold, K candidate addresses are determined in the intersection region; If the area ratio of the intersection area to the target area is less than a preset ratio threshold, K candidate addresses are determined proportionally in the intersection area, the non-intersection area of ​​the target area, and the non-intersection area of ​​the environmental impact area based on the area ratio.

4. The method according to claim 2, characterized in that The step of selecting (N*M-1) reference addresses from the K candidate addresses with the target address as the center comprises: Calculate the vectors between the target address and each candidate address to obtain K address vectors, each address vector corresponding to one candidate address; Based on the address vector, (N*M-1) reference addresses are selected from the K candidate addresses.

5. The method according to claim 4, characterized in that The step of selecting (N*M-1) reference addresses from the K candidate addresses based on the address vector includes: Sort the K address vectors according to the vector modulus length to obtain an address vector sequence; In the address vector sequence, traverse and calculate the vector angle between the first address vector and the second address vector, the first address vector is sorted before the second address vector; If the vector angle is less than a preset angle threshold, discarding the second address vector, and determining the next address vector after the second address vector as the second address vector; If the vector angle is greater than or equal to a preset angle threshold, determining the second address vector as a reference address; If (N*M-1) reference addresses are obtained, stop traversing; If the traversal is completed and T reference addresses are obtained, the remaining address vectors are sorted according to the vector modulus length to obtain a remaining address vector sequence, where T is less than (N*M-1); Among the address vectors corresponding to the T reference addresses, determining the first second address vector after the first address vector as a new first address vector; In the remaining address vector sequence, iterative calculation is performed to determine R reference addresses, R+T=(N*M-1).

6. The method according to claim 2, characterized in that The step of selecting (N*M-1) reference addresses from the K candidate addresses with the target address as the center comprises: Based on the target address and the K candidate addresses, construct a Delaunay triangulation network, wherein the Delaunay triangulation network includes a plurality of Delaunay triangles, and each of the Delaunay triangles does not include any other addresses except for the three vertex addresses constituting the Delaunay triangle; In the Delaunay triangulation, determining K1 first target Delaunay triangles whose vertex addresses include the target address, where K1 is less than K; When the number of vertices of the K1 first target Delaunay triangles other than the target address vertex is less than 8, determine the first target Delaunay triangle with the largest first circumscribed circle area among the K1 first target Delaunay triangles, connect the target address with the center point of the first target Delaunay triangle with the largest first circumscribed circle area, and extend the ray starting from the target address to obtain a discrimination ray; Determine the first non-first target Delaunay triangle intersecting the discrimination ray as the first target Delaunay triangle, until the number of vertices of the first target Delaunay triangle, excluding the target address vertex, reaches 8; When the number of vertices of the K1 first target Delaunay triangles other than the target address vertex is greater than 8, determining the triangle area of ​​each of the K1 first target Delaunay triangles, and determining the second circumscribed circle area of ​​each of the first target Delaunay triangles; Deleting vertices other than the target address vertex in the first target Delaunay triangle whose triangle area and the second circumscribed area are the smallest, until the number of vertices other than the target address vertex in the first target Delaunay triangle reaches 8; Using 8 vertices other than the target address vertex, search outward for new vertices of the adjacent Delaunay triangle until (N*M-1) vertices are obtained; In the Delaunay triangulation, corresponding (N*M-1) reference addresses are determined based on (N*M-1) vertices.

7. The method according to any one of claims 1 to 6, characterized in that The step of extracting spatiotemporal features from the adjacency matrix N*M to obtain the spatiotemporal features of the adjacency matrix includes: Extracting spatiotemporal features of the adjacency matrix N*M through the trained capsule network to obtain the spatiotemporal features of the adjacency matrix; The step of predicting the environmental risk of the target address based on the spatiotemporal characteristics to obtain the environmental risk identification result of the target address includes: The temporal and spatial features are predicted and processed by a trained capsule network to obtain an environmental risk identification result of the target address; The capsule network is trained by a data set, and the data set includes a sample adjacency matrix and a label matrix. The sample adjacency matrix is ​​obtained in the same way as the adjacency matrix N*M. Each sample matrix unit in the sample adjacency matrix corresponds to a historical environmental data C from time point T1 to time point T2. i∈[T1，T2],j∈N*M , the C i∈[T1，T2],j∈N*M Indicates the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T1 time point and the T2 time point; each label matrix unit in the label matrix corresponds to a historical environment data D from the T3 time point to the T4 time point i∈[T3，T4],j∈N*M , the D i∈[T3，T4],j∈N*M Represents the historical environment data of the address corresponding to the jth matrix unit in the N*M matrix units at the i-th time point between the T3 time point and the T4 time point, wherein the T1 time point is earlier than the T2 time point, the T3 time point is earlier than the T4 time point, and the T2 time point is earlier than the T3 time point.

8. An environmental risk identification device for site selection of a coating plant, characterized in that: The environmental risk identification device for the site selection of the coating plant includes: The acquisition module is used to acquire the adjacency matrix N*M of the target area, wherein the target area includes the target address and the reference address, and the adjacency matrix includes the central matrix unit corresponding to the target address and the adjacency matrix unit corresponding to the reference address, and each matrix unit in the adjacency matrix corresponds to a historical environment data B i,j∈N*M , the B i,j∈N*M represents the corresponding environmental data of the jth matrix unit among the N*M matrix units at the i-th time point; An extraction module, used for performing spatiotemporal feature extraction on the adjacency matrix N*M to obtain the spatiotemporal features of the adjacency matrix; The prediction module is used to predict the environmental risk of the target address based on the spatiotemporal characteristics to obtain an environmental risk identification result of the target address.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for identifying environmental risks in the site selection of a coating plant as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for identifying environmental risks in the site selection of a paint shop as described in any one of claims 1 to 7.

Images