A method and platform for intelligent safety risk management and control in industrial parks

Abstract

The present application discloses a method and platform for intelligent control of safety risks in industrial parks, which relates to the field of industrial park safety management. The method includes: obtaining the real-time environmental information of the industrial park at the current moment and the regional meteorological environmental information of the area where the industrial park is located at the current moment; obtaining the real-time comprehensive environmental information of the park through the real-time environmental information and the regional meteorological environmental information; establishing a three-dimensional model of the industrial park; obtaining the image information of the industrial park through an image acquisition device; obtaining the raw material storage information of chemical raw materials through the three-dimensional model of the park and image recognition technology; combining the raw material storage information and the real-time comprehensive environmental information of the park to determine the danger level of the park; if the danger level of the park is greater than the preset danger level threshold of the park, then generating the optimal evacuation path and sending it to the mobile terminal held by the staff in the industrial park. The present application has the effect of improving the safety and efficiency of park safety management.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of industrial park safety management, and in particular, to a method and platform for intelligent safety risk management and control in industrial parks. Background Art

[0002] Chemical parks are important carriers for the development of the chemical industry, with advantages such as standardization and clustering. However, due to the wide variety of chemical raw materials in chemical parks, and many of them are dangerous, chemical parks face complex safety challenges. Once an accident occurs, it is very likely to cause serious casualties and property losses. Traditional industrial parks usually use decentralized monitoring systems, such as gas detectors, temperature sensors, pressure monitors, etc., to monitor different environmental parameters. On the other hand, the safety management and risk assessment of traditional industrial parks mainly rely on manual regular inspections and preset emergency plans. Manual inspections are not only inefficient, but also easily affected by subjective factors, resulting in misjudgments, and preset emergency plans are difficult to flexibly adjust according to specific circumstances.

[0003] The management model of traditional industrial parks may be effective in relatively stable and simple environments, but with the development of science and technology, its limitations are becoming increasingly prominent in the face of increasingly complex environments and changing risk factors in modern industrial parks. Especially in some large-scale comprehensive industrial parks, different types of chemicals, changing weather conditions, complex geographical environments and other factors interact with each other, making safety management more challenging. Existing technologies can no longer meet the requirements of high safety and efficiency in the park's safety management process. Summary of the invention

[0004] The embodiments of the present application provide a method and platform for intelligent management and control of security risks in an industrial park, which are used to solve the problems of insufficient security and low efficiency in the existing park security management process.

[0005] To achieve the above objectives, the embodiments of the present application adopt the following technical solutions:

[0006] In a first aspect, a method for intelligent safety risk management and control of an industrial park is provided, which is applied to an industrial park storing different types of chemical raw materials, and the method comprises:

[0007] The real-time environmental information of the industrial park at the current moment is obtained in real time through multiple types of sensors preset in different areas of the industrial park, and the regional meteorological environmental information of the area where the industrial park is located at the current moment is obtained through a weather forecast system;

[0008] Establishing a park meteorological environment generation model based on a neural network model, combining the real-time environmental information and the regional meteorological environment information and using the park meteorological environment generation model to generate park real-time comprehensive environmental information containing comprehensive environmental information of all areas of the industrial park;

[0009] Acquire the park plane map information of the industrial park, and establish the park three-dimensional model of the industrial park based on the park plane map information;

[0010] Acquire image information of the industrial park through image acquisition devices preset in different areas of the industrial park;

[0011] Using the three-dimensional model of the industrial park as a carrier, image recognition technology is used to identify and analyze the image information to obtain raw material storage information of the chemical raw materials currently stored in the industrial park;

[0012] Combining the raw material storage information and the real-time comprehensive environmental information of the industrial park, a multi-dimensional analysis is performed on the storage situation of chemical raw materials in the industrial park by region, and the danger level of the industrial park is determined according to the analysis results;

[0013] If the danger level of the industrial park is greater than a preset danger level threshold, an optimal evacuation path is generated based on the three-dimensional model of the industrial park and using a path optimization algorithm, and the optimal evacuation path is sent to a mobile terminal held by a staff member in the industrial park.

[0014] Optionally, the park meteorological environment generation model is established based on the neural network model, and the real-time comprehensive environmental information of the park containing the comprehensive environmental information of all areas of the industrial park is generated by combining the real-time environmental information and the regional meteorological environmental information and using the park meteorological environment generation model as follows:

[0015] A spatial feature extraction module is constructed based on a convolutional neural network model, and a temporal feature extraction module is constructed based on a long short-term memory neural network model. The spatial feature extraction module, the temporal feature extraction module and a preset multi-scale feature fusion module are combined to construct a park meteorological environment generation model;

[0016] Extracting historical environmental information and historical comprehensive environmental information of the industrial park in the same historical period from a preset park environment database, wherein the historical environmental information is environmental information acquired by the multi-type sensors in different areas during the historical period, and both the historical environmental information and the real-time environmental information include sensor location information of the multi-type sensors in the industrial park;

[0017] Using the historical period as a query index, retrieve the regional historical meteorological environment information of the area where the industrial park is located during the historical period from the weather forecast system;

[0018] Inputting the historical environmental information, the historical comprehensive environmental information of the industrial park and the historical meteorological environmental information of the region into the industrial park meteorological environment generation model, training the industrial park meteorological environment generation model to learn the spatial characteristic change relationship and the temporal characteristic change relationship of the environmental information in the industrial park, and obtaining the trained industrial park meteorological environment generation model;

[0019] The real-time environmental information and the regional meteorological environmental information are input into the trained park meteorological environment generation model, and the trained park meteorological environment generation model outputs the park real-time comprehensive environmental information containing the comprehensive environmental information of all areas of the industrial park.

[0020] Optionally, the historical environmental information, the industrial park's historical comprehensive environmental information, and the regional historical meteorological environmental information are input into the industrial park's meteorological environment generation model, and the industrial park's meteorological environment generation model is trained to learn the spatial feature change relationship and the temporal feature change relationship of the environmental information in the industrial park, so as to obtain the trained industrial park's meteorological environment generation model, including the following steps:

[0021] Inputting the historical environmental information into the spatial feature extraction module in the park meteorological environment generation model, and respectively extracting global environmental spatial correlation features and local environmental spatial correlation features from the historical environmental information through the global spatial feature extraction layer and the local spatial feature extraction layer in the spatial feature extraction module;

[0022] The global environment space correlation feature and the local environment space correlation feature are merged into an environment space correlation feature matrix, and a space feature loss function of the space feature extraction module is constructed based on the environment space correlation feature matrix;

[0023] Respectively performing time series processing on the historical environmental information and the regional historical meteorological environmental information to obtain historical environmental time series data and regional historical meteorological environmental time series data, and inputting the historical environmental time series data and the regional historical meteorological environmental time series data into the time series feature extraction module in the park meteorological environment generation model;

[0024] Utilizing the time series feature extraction module to extract historical environment time series features and regional historical meteorological environment time series features from the historical environment time series data and the regional historical meteorological environment time series data respectively;

[0025] The historical environmental time series features and the regional historical meteorological environmental time series features are merged into an environmental time series correlation feature matrix, and a time series feature loss function of the time series feature extraction module is constructed based on the environmental time series correlation feature matrix;

[0026] Inputting the environment spatial correlation feature matrix and the environment temporal correlation feature matrix into the multi-scale feature fusion module, using the multi-scale feature fusion module to output the environment multi-dimensional feature correlation matrix, and constructing a feature fusion loss function of the multi-scale feature fusion module based on the multi-dimensional feature correlation matrix;

[0027] The spatial feature loss function, the temporal feature loss function and the feature fusion loss function are combined to obtain a comprehensive loss function of the park meteorological environment generation model, and a comprehensive loss function value of the comprehensive loss function is calculated according to the historical comprehensive environmental information of the park;

[0028] The module parameters of the spatial feature extraction module, the temporal feature extraction module and the multi-scale feature fusion module are iteratively updated based on the comprehensive loss function value until the comprehensive loss function value reaches a minimum value, thereby obtaining the trained park meteorological environment generation model.

[0029] Optionally, the raw material storage information includes raw material hazard feature information and raw material location information, and the three-dimensional model of the industrial park is used as a carrier, and image recognition technology is used to identify and analyze the image information to obtain the raw material storage information of the chemical raw materials currently stored in the industrial park, including the following steps:

[0030] grayscale all the image information to obtain multiple grayscale images;

[0031] Extracting edge point sets of the grayscale image using an edge enhancement operator, and integrating the edge point sets into grayscale image edge features;

[0032] Based on the edge features of the grayscale image and using a threshold segmentation algorithm, a target image region in the grayscale image containing the chemical raw material is separated from the image background;

[0033] Identify the hazardous material logo on the surface of the chemical raw material in the target image area by image recognition technology, classify the hazardous material logo, and obtain the raw material hazard characteristic information corresponding to the chemical raw material according to the classification result;

[0034] Marking the image information containing the target image area as target image information;

[0035] The device position information of the target image acquisition device that photographs the target image information is acquired through the three-dimensional model of the park, and the device position information is used as the raw material position information of the chemical raw material corresponding to the target image information.

[0036] Optionally, the method further includes:

[0037] Performing resolution enhancement on the target image region using an LDSR algorithm;

[0038] Using OCR technology to identify the size text features from the target image area after resolution enhancement, and obtaining the actual size information corresponding to the chemical raw material by analyzing the size text features;

[0039] Calculating image size information of the chemical raw material in the image information according to the edge features of the grayscale image;

[0040] The actual distance between the target image acquisition device and the chemical raw material in the three-dimensional model of the park is calculated in combination with the actual size information, the image size information and the preset device parameters of the target image acquisition device, and the raw material position information is corrected according to the actual distance.

[0041] Optionally, the step of performing a regional multi-dimensional analysis on the storage situation of chemical raw materials in the industrial park in combination with the raw material storage information and the real-time comprehensive environmental information of the industrial park, and determining the danger level of the industrial park according to the analysis results comprises the following steps:

[0042] For any stored chemical raw materials in the industrial park, judging whether the stored chemical raw materials are toxic gas raw materials or flammable and explosive raw materials according to the chemical raw material types in the raw material hazard characteristic information of the stored chemical raw materials;

[0043] If the stored chemical raw material is the toxic gas raw material, the toxic gas leakage data of the target area in the industrial park is obtained from the real-time comprehensive environmental information of the park, and the toxic gas concentration data of the target area is obtained based on the toxic gas leakage data, and the target area is the area where the raw material location information of the stored chemical raw material is located;

[0044] If the toxic gas concentration data of the target area is 0, the toxic gas danger level value of the target area is set to zero;

[0045] If the toxic gas concentration data of the target area is greater than 0, the toxic gas diffusion range of the target area is obtained by comprehensive calculation based on the toxic gas concentration data and the real-time comprehensive environmental information of the park in the target area, and the toxic gas danger level value of the target area is determined according to the toxic gas diffusion range;

[0046] If the stored chemical raw materials are the flammable and explosive raw materials, the probability of static electricity occurrence in the target area is predicted based on the real-time temperature data and real-time humidity data of the real-time comprehensive environmental information of the park in the target area, and the explosion hazard value of the target area is calculated using the static electricity occurrence probability;

[0047] Allocating different regional weights to each of the target regions according to the area of ​​the target region and the amount of the stored chemical raw materials;

[0048] Based on the regional weight and in combination with the toxic gas hazard level values ​​and / or the explosion hazard level values ​​of all the target areas, a weighted calculation is performed to obtain a comprehensive hazard level value of the industrial park, and the comprehensive hazard level value is divided into corresponding park hazard level levels using a preset hazard level threshold.

[0049] Optionally, the step of comprehensively calculating the toxic gas diffusion range of the target area by combining the toxic gas concentration data and the real-time comprehensive environmental information of the park in the target area, and determining the toxic gas danger level value of the target area according to the toxic gas diffusion range includes the following steps:

[0050] Constructing a three-dimensional coordinate system of the industrial park with the center of the industrial park as the origin in the three-dimensional model of the industrial park;

[0051] Extracting wind direction characteristic data, wind speed characteristic data and real-time temperature data of the target area from the real-time comprehensive environmental information of the park;

[0052] Importing the wind direction characteristic data and the wind speed characteristic data into the park three-dimensional model, determining the wind direction point coordinates of the strongest wind direction point in the target area through the park three-dimensional coordinate system, and calculating the wind direction angle in the target area based on the wind direction point coordinates;

[0053] Calculate the toxic gas diffusion concentration in the target area by combining the wind direction angle, the toxic gas concentration data and the real-time temperature data;

[0054] Determine the main wind direction axis of the target area according to the wind direction angle, combine the wind speed characteristic data, the toxic gas diffusion concentration and the toxic gas concentration data, and use a Gaussian puff model to calculate the toxic gas diffusion distance on the main wind direction axis in the target area, and estimate the toxic gas diffusion range of the target area based on the toxic gas diffusion distance;

[0055] The toxic gas danger level value of the target area is calculated based on the area ratio of the toxic gas diffusion range in the target area.

[0056] Optionally, the calculation formula for the toxic gas diffusion concentration is as follows:

[0057] ,

[0058] Where: is the toxic gas diffusion concentration in the target area n, is the gas diffusion fixed coefficient, is the real-time temperature data of the target area n, is the wind direction angle of the target area n, is the toxic gas concentration data of the target area n, and are respectively the horizontal and vertical coordinates of the chemical raw materials stored in the target area n in the three-dimensional coordinate system of the park;

[0059] The calculation formula of the toxic gas diffusion distance is as follows:

[0060] ,

[0061] Where: represents the diffusion distance of the toxic gas in the target area n, represents the gas diffusion time, represents the exponential function, Indicates the effective discharge height, Represents the wind speed characteristic data of the target area n.

[0062] Optionally, generating an optimal evacuation path based on the park three-dimensional model and using a path optimization algorithm comprises the following steps:

[0063] Using the center point of the area where the workers in the industrial park gather as the starting point coordinates and the center point of the preset safe area in the industrial park as the end point coordinates, a plurality of initial evacuation paths of the industrial park are generated using the three-dimensional model of the industrial park;

[0064] Encoding all the initial evacuation paths as an initial population of a path optimization algorithm, and setting a maximum number of iterations of the path optimization algorithm;

[0065] Importing the toxic gas diffusion range into the park three-dimensional coordinate system to generate a range coordinate set corresponding to the toxic gas diffusion range;

[0066] Integrate the raw material location information of all the stored chemical raw materials into a raw material location coordinate set;

[0067] Constructing the constraint conditions of the path optimization algorithm in combination with the range coordinate set and the raw material position coordinate set;

[0068] Iteratively updating the initial population based on the constraint conditions until the number of iterative updates reaches the maximum number of iterations to obtain an optimal population;

[0069] The fitness of each individual in the optimal population is calculated according to a preset fitness function, and the individual with the highest fitness in the optimal population is decoded as an optimal evacuation path.

[0070] In the second aspect, the present application provides a security risk intelligent management and control platform for an industrial park, including:

[0071] a memory configured to store instructions; and

[0072] The processor is configured to call the instructions from the memory and to implement the method for intelligent management and control of safety risks in an industrial park according to the first aspect when executing the instructions.

[0073] Through the above technical scheme, the real-time environmental information of the industrial park at the current moment is obtained by using multiple types of sensors preset in different areas of the industrial park, and the regional meteorological environmental information of the area where the industrial park is located at the previous moment is obtained through the weather forecast system; a park meteorological environment generation model is established by using a neural network model, and the real-time environmental information and the regional meteorological environmental information are input into the trained park meteorological environment generation model to obtain the park real-time comprehensive environmental information of all regional comprehensive environmental information of the industrial park; using the three-dimensional model of the park as a carrier, the image information is identified and analyzed by using image recognition technology to obtain the raw material storage information of chemical raw materials; the raw material location information contained in the raw material storage information is corrected by using an algorithm; the storage situation of chemical raw materials in the industrial park is analyzed in multiple dimensions by region in combination with the raw material storage information and the real-time comprehensive environmental information of the park, and the park danger level of the industrial park is determined according to the analysis results; if the park danger level is greater than the preset park danger level threshold, an optimal evacuation path is generated and sent to the mobile terminal held by the staff in the industrial park.

[0074] As described above, the present invention obtains real-time environmental information through multi-type sensors and weather forecast systems, generates real-time comprehensive environmental information of the park using a neural network model, and obtains chemical raw material storage information in combination with a three-dimensional model and image recognition technology, thereby realizing comprehensive monitoring of the industrial park environment and raw material storage conditions. The danger level of the park is determined through multi-dimensional analysis, and the optimal evacuation path is generated when the danger level exceeds the threshold, which can timely warn and guide personnel to evacuate. This method integrates multi-source data, improves the accuracy of environmental monitoring and risk assessment, and realizes real-time dynamic monitoring of the safety status of the park. At the same time, the automatically generated evacuation path can effectively guide emergency evacuation and improve the efficiency and safety of responding to emergencies. In summary, this application comprehensively considers factors such as the various types of chemicals in some large comprehensive industrial parks, the changeable weather in the park, and the complex geographical environment of the park, improves the efficiency of park safety warning and park safety, reduces the risk of safety accidents, and provides strong protection for the safety of life and property of park personnel.

[0075] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 A schematic diagram of a process for intelligent management and control of safety risks in an industrial park provided in an embodiment of the present application;

[0077] Figure 2 A schematic diagram of the structure of a park meteorological environment generation model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0079] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back...), such directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0080] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0081] Figure 1 The following schematically shows a flow chart of a method for intelligently managing and controlling security risks in an industrial park according to an embodiment of the present application. Figure 1 As shown, the embodiment of the present application provides a method for intelligent management and control of safety risks in an industrial park, which may include the following steps:

[0082] S101. Real-time environmental information of the industrial park at the current moment is obtained in real time through multiple types of sensors preset in different areas of the industrial park, and regional meteorological environmental information of the area where the industrial park is located at the current moment is obtained through a weather forecast system.

[0083] In this embodiment, the multiple types of sensors mainly include temperature sensors, air pressure sensors, humidity sensors, wind speed and direction sensors, gas sensors and other sensors. The weather forecast system is a weather station in the area where the industrial park is located. Various types of sensors are installed in advance in different areas of the industrial park, and real-time environmental information of the industrial park is obtained in real time through these sensors, including real-time temperature information, real-time humidity information, etc., and regional meteorological environment information is obtained through the weather station in the area where the industrial park is located.

[0084] S102. Establish a park meteorological environment generation model based on a neural network model, combine real-time environmental information and regional meteorological environmental information, and use the park meteorological environment generation model to generate real-time comprehensive environmental information of the park that includes comprehensive environmental information of all areas of the industrial park.

[0085] In this embodiment, a park meteorological environment generation model is established by using a neural network model, and the historical environmental information of the industrial park in the same historical period and the historical comprehensive environmental information of the park are obtained from a preset park environment database. At the same time, the regional historical meteorological environment information of the same historical period in the area where the industrial park is located is retrieved from the weather forecast system. The park meteorological environment generation model is trained by the historical environmental information, the historical comprehensive environmental information of the park and the historical meteorological environment information of the region to obtain the park meteorological environment generation model after training. The real-time environmental information and the regional meteorological environment information are input into the trained park meteorological environment generation model to obtain the park real-time comprehensive environmental information containing the comprehensive environmental information of all regions of the industrial park. The main function of this step is to obtain the park real-time comprehensive environmental information of all regions of the industrial park. Due to the large scope of the industrial park and the complex geographical conditions, the local regional meteorological environment information cannot accurately describe the meteorological environment information of all regions in the industrial park. At the same time, these meteorological environment information are relatively lagging, which will cause errors in the judgment of the subsequent industrial park safety warning. Therefore, it is necessary to combine the real-time environmental information obtained by multiple types of sensors distributed in various regions in the park and the park meteorological environment generation model to obtain the park real-time comprehensive environmental information containing the comprehensive environmental information of all regions of the industrial park. For example, if the collected temperature data of one area is lower than the real-time temperature data, then if there is a toxic gas leak in the area, the calculated toxic gas diffusion concentration will be smaller, causing an error in the danger level value of the area, affecting the classification of the danger level of the park, and failing to issue a safety warning in time, which may lead to serious safety accidents, causing casualties and property losses.

[0086] S103: Acquire the plan map information of the industrial park, and establish a three-dimensional model of the industrial park based on the plan map information.

[0087] In this embodiment, the obtained industrial park plan map information includes a scale map of the industrial park and a building plan of the industrial park. The building plan of the industrial park contains information such as the plan layout (division of production areas, processing areas, etc.), size (for example, a large factory building is 100 meters long, 60 meters wide, and 10 meters high), structure (the location of walls, columns, piers, etc.) of all buildings in the entire area. The establishment of a three-dimensional model of the park requires the use of a variety of modeling software, such as Autodesk Revit (building information modeling software), AutoCAD (drawing tool software), and 3DS Max (three-dimensional animation rendering software). According to the plan map of the industrial park, the overall building distribution, building proportions, and road distribution of the industrial park are constructed. After continuously optimizing the modeling of the industrial park, the three-dimensional model of the industrial park is finally obtained. The three-dimensional model of the industrial park includes the proportional structures of all buildings in the industrial park, as well as the distribution of all roads, greening conditions, terrain, and entrances and exits of the industrial park. At the same time, the positions of various types of sensors and image acquisition devices (high-definition cameras) pre-installed in the park are also imported into the three-dimensional model of the park. The installation positions of various types of sensors and image acquisition devices can be directly obtained from the three-dimensional model of the park.

[0088] S104. Acquire image information of the industrial park through image acquisition devices preset in different areas of the industrial park.

[0089] In this embodiment, the image acquisition device is a high-definition camera. By arranging high-definition cameras in different areas of the industrial park, it is possible to collect real-time image information such as the transportation status of chemical raw materials in the park, the production status of the park, and the hazardous raw material identification on the surface of chemical raw materials. The collected image information is helpful to obtain the raw material location information and raw material hazard feature information of chemical raw materials.

[0090] S105. Using the three-dimensional model of the industrial park as a carrier, image recognition technology is used to identify and analyze image information to obtain raw material storage information of chemical raw materials currently stored in the industrial park.

[0091] In this embodiment, the raw material storage information of the chemical raw material includes raw material hazard characteristic information and raw material location information. The image information is collected by a high-definition camera. The warehouse location for storing chemical raw materials can be obtained through the three-dimensional model of the park. The transportation status of chemical raw materials can be observed through the image information collected by the high-definition camera. The chemical raw material location information can be preliminarily obtained by combining these two ways. The chemical raw material location information preliminarily obtained is corrected by the algorithm to obtain the raw material location information. At the same time, the dangerous raw material logo on the surface of the chemical raw material can be collected by the high-definition camera inside the warehouse. For example, the dangerous raw material logo on the surface of the chemical raw material is a picture with a red background and black flames, and the digital text "0.84 / G / 4 / 1.2", which means that the chemical raw material is a flammable and explosive liquid with a relative density of 0.8 (water = 1), a flash point of 4 degrees Celsius, and an explosion limit data of 1.2% (V / V) (percentage concentration). The collected dangerous raw material logos on the surface of the chemical raw materials are analyzed and classified to obtain the raw material hazard characteristic information.

[0092] S106. Combine the raw material storage information and the real-time comprehensive environmental information of the industrial park to conduct a multi-dimensional analysis of the storage situation of chemical raw materials in each region, and determine the park hazard level of the industrial park based on the analysis results.

[0093] In this embodiment, if the raw material hazard characteristic information shows that toxic gas raw materials or flammable and explosive raw materials are stored in the current area, the toxic gas hazard level value and the combustion and explosion hazard level value of these two hazardous raw materials are calculated respectively in combination with the real-time comprehensive environmental information of the park in the current area. The regional weight assigned to the current area is determined based on the area and storage quantity of the area where the hazardous raw materials are stored. Generally, the larger the area and the more storage quantity, the larger the corresponding regional weight. The same method is used to assign regional weights to other areas where toxic gas raw materials or flammable and explosive raw materials are stored. According to the regional weight and in combination with the toxic gas hazard level value and the combustion and explosion hazard level value of all areas, the comprehensive hazard level value of the industrial park is obtained, and then the comprehensive hazard level value is graded to obtain the park hazard level.

[0094] S107. If the danger level of the industrial park is greater than a preset danger level threshold, an optimal evacuation path is generated based on the three-dimensional model of the industrial park and using a path optimization algorithm. The optimal evacuation path is sent to a mobile terminal held by a staff member in the industrial park.

[0095] In this embodiment, the path optimization algorithm may be a gray wolf optimization algorithm, which is an optimization algorithm inspired by the behavior of gray wolves in nature. Through this algorithm, when a dangerous accident is about to occur in the park, the optimal evacuation route can be generated for the park staff. The mobile terminal is a communication device (mobile phone) held by the park staff. If the danger level of the park is greater than the preset park danger level threshold, the path optimization algorithm is used to generate the optimal evacuation path according to the three-dimensional model of the park, and the optimal evacuation route is sent to the communication device held by the park staff in the form of a plane map, and an alarm instruction is also sent to the park alarm system.

[0096] In one embodiment, a park meteorological environment generation model is established based on a neural network model, and the park meteorological environment generation model is combined with real-time environmental information and regional meteorological environmental information to generate park real-time comprehensive environmental information containing comprehensive environmental information of all areas of the industrial park as follows:

[0097] A spatial feature extraction module is constructed based on the convolutional neural network model, and a temporal feature extraction module is constructed based on the long short-term memory neural network model. The spatial feature extraction module, the temporal feature extraction module and the preset multi-scale feature fusion module are combined to construct a park meteorological environment generation model.

[0098] The historical environmental information and the historical comprehensive environmental information of the industrial park in the same historical period are extracted from the preset park environmental database. The historical environmental information is the environmental information obtained by multiple types of sensors in different areas in the historical period. Both the historical environmental information and the real-time environmental information contain the sensor location information of multiple types of sensors in the industrial park.

[0099] Using the historical time period as the query index, retrieve the historical meteorological environment information of the area where the industrial park is located in the historical time period from the weather forecast system;

[0100] Input historical environmental information, historical comprehensive environmental information of the industrial park, and historical meteorological environmental information of the region into the industrial park meteorological environment generation model, train the industrial park meteorological environment generation model to learn the spatial characteristic change relationship and temporal characteristic change relationship of environmental information in the industrial park, and obtain the trained industrial park meteorological environment generation model;

[0101] The real-time environmental information and regional meteorological environmental information are input into the trained industrial park meteorological environment generation model, and the trained industrial park meteorological environment generation model outputs the real-time comprehensive environmental information of the industrial park, which contains the comprehensive environmental information of all areas of the industrial park.

[0102] In this embodiment, refer to Figure 2First, a spatial feature extraction module and a temporal feature extraction module are constructed. The spatial feature extraction module is based on a convolutional neural network (CNN) model, which can effectively capture the spatial distribution characteristics of environmental data in industrial parks. In specific implementation, a combination of multi-layer convolutional layers, pooling layers, and fully connected layers can be used. The temporal feature extraction module is based on a long short-term memory neural network (LSTM) model, which can learn the time series characteristics of environmental data. The LSTM module can contain multiple LSTM layers, such as using a 2-layer LSTM, each layer containing 128 neurons. These two modules are combined with the preset multi-scale feature fusion module to form a complete park meteorological environment generation model. The multi-scale feature fusion module can use an attention mechanism or a gating mechanism to effectively fuse spatial features and temporal features. For example, a gated recurrent unit (GRU) can be used to achieve feature fusion, and the hidden state dimension of the GRU can be set to 256. The model constructed in this way can simultaneously consider the spatial distribution and temporal evolution of environmental data, thereby generating park meteorological environment information more accurately.

[0103] Next, the historical environmental information and the historical comprehensive environmental information of the industrial park in the same historical period are extracted from the preset park environment database. The historical environmental information contains the environmental data obtained by multiple types of sensors in different regions during the historical period, such as temperature, humidity, wind speed, wind direction, toxic gas concentration, etc. The historical comprehensive environmental information of the park is the comprehensive information obtained after integrating and processing these raw data. Both the historical environmental information and the real-time environmental information contain the location information of the sensor in the industrial park. The purpose of this step is to provide sufficient historical data for model training so that the model can learn the long-term patterns and periodic characteristics of environmental changes. Then, using the historical period as the query index, the regional historical meteorological environmental information of the area where the industrial park is located in the historical period is retrieved from the weather forecast system. The purpose of this step is to obtain larger-scale meteorological information to supplement the possible deficiencies in local sensor data. The retrieved information may include macro meteorological data such as temperature, precipitation, wind direction, wind speed, and atmospheric pressure.

[0104] Next, the historical environmental information, the park's historical comprehensive environmental information, and the region's historical meteorological environmental information are input into the park's meteorological environment generation model for model training. During the training process, the model learns the spatial feature change relationship and the temporal feature change relationship of environmental information in the industrial park. Specifically, the CNN part learns the spatial distribution pattern of environmental data, such as the temperature difference between different regions and the spatial distribution of pollutant concentrations. The LSTM part learns the time series characteristics of environmental data, such as the daily change cycle of temperature and seasonal changes. The multi-scale feature fusion module learns how to effectively combine these spatial and temporal features. During the training process, the back propagation algorithm and gradient descent method can be used to optimize the model parameters. The loss function can be selected as mean square error (MSE) or mean absolute error (MAE). Mini-batch gradient descent can be used for training, and the batch size can be set to 64 or 128. The learning rate can be initially set to 0.001, and a learning rate decay strategy is used. The number of training iterations can be set to 100 rounds, and an early stopping strategy is used to avoid overfitting.

[0105] Finally, the real-time environmental information and regional meteorological environmental information are input into the trained park meteorological environment generation model to generate the park's real-time comprehensive environmental information containing comprehensive environmental information of all areas of the industrial park. The real-time environmental information comes from the sensor network in the park, including real-time readings of various environmental parameters. The regional meteorological environmental information is the current and short-term forecast data obtained in real time from the weather forecast system. The model first processes the spatial features of the input data through the CNN module to capture the environmental differences between different regions. Then, the LSTM module processes the time series data to predict the environmental change trend in the short term. Finally, the multi-scale feature fusion module combines these features to generate the final real-time comprehensive environmental information of the park. The output comprehensive environmental information includes parameters such as temperature, humidity, wind speed, air quality, etc. in each area of ​​the park, as well as the short-term predicted values ​​of these parameters. For example, the model may output a data structure of the form {Region 1:{Temperature:25°C,Humidity:60%,PM2.5:50μg / m³,...},Region 2:{...},...}.

[0106] In one embodiment, historical environmental information, park historical comprehensive environmental information and regional historical meteorological environmental information are input into a park meteorological environment generation model, and the park meteorological environment generation model is trained to learn the spatial characteristic change relationship and the temporal characteristic change relationship of environmental information in the industrial park, and the trained park meteorological environment generation model is obtained, which includes the following steps:

[0107] The historical environmental information is input into the spatial feature extraction module in the park meteorological environment generation model, and the global environmental spatial correlation features and the local environmental spatial correlation features are respectively extracted from the historical environmental information through the global spatial feature extraction layer and the local spatial feature extraction layer in the spatial feature extraction module;

[0108] The global environment spatial correlation features and the local environment spatial correlation features are integrated into an environment spatial correlation feature matrix, and a spatial feature loss function of a spatial feature extraction module is constructed based on the environment spatial correlation feature matrix;

[0109] The historical environment information and the regional historical meteorological environment information are respectively processed in time series to obtain historical environment time series data and regional historical meteorological environment time series data, and the historical environment time series data and the regional historical meteorological environment time series data are input into the time series feature extraction module in the park meteorological environment generation model;

[0110] The time series feature extraction module is used to extract the historical environment time series features and the regional historical meteorological environment time series features from the historical environment time series data and the regional historical meteorological environment time series data respectively;

[0111] The historical environmental time series characteristics and the regional historical meteorological environmental time series characteristics are integrated into the environmental time series correlation feature matrix, and the time series feature loss function of the time series feature extraction module is constructed based on the environmental time series correlation feature matrix;

[0112] Input the environment spatial correlation feature matrix and the environment temporal correlation feature matrix into the multi-scale feature fusion module, use the multi-scale feature fusion module to output the environment multi-dimensional feature correlation matrix, and construct the feature fusion loss function of the multi-scale feature fusion module based on the multi-dimensional feature correlation matrix;

[0113] The spatial feature loss function, the temporal feature loss function and the feature fusion loss function are combined to obtain the comprehensive loss function of the park's meteorological environment generation model, and the comprehensive loss function value of the comprehensive loss function is calculated based on the park's historical comprehensive environmental information;

[0114] The module parameters of the spatial feature extraction module, the temporal feature extraction module and the multi-scale feature fusion module are iteratively updated based on the comprehensive loss function value until the comprehensive loss function value reaches the minimum value, and the trained park meteorological environment generation model is obtained.

[0115] In this embodiment, first, the historical environmental information is input into the spatial feature extraction module in the park meteorological environment generation model. The module includes a global spatial feature extraction layer and a local spatial feature extraction layer, which are used to extract the global environmental spatial association features and the local environmental spatial association features, respectively. The global spatial feature extraction layer can use a convolutional neural network with a large-size convolution kernel to capture the environmental feature associations within the entire park. In specific implementation, three convolution layers can be used, each using 64, 128 and 256 convolution kernels, and ReLU is selected as the activation function. The local spatial feature extraction layer uses a small-size convolution kernel to capture the environmental feature details in the local area. Similarly, three convolution layers can be used, each using 32, 64 and 128 convolution kernels. Both feature extraction layers can be used in conjunction with a maximum pooling layer to reduce the feature dimension and improve the translation invariance of the model. Next, the global environmental spatial association features and the local environmental spatial association features are fused into an environmental spatial association feature matrix. The fusion process can be done by feature concatenation or weighted summation. Based on this environment spatial correlation feature matrix, the spatial feature loss function of the spatial feature extraction module is constructed. The loss function can be selected as mean square error (MSE) or mean absolute error (MAE). The spatial feature loss function is used to evaluate the difference between the spatial features extracted by the model and the real spatial features, and guide the model to optimize the spatial feature extraction capability.

[0116] Then, the historical environmental information and the regional historical meteorological environmental information are processed in time series to obtain the historical environmental time series data and the regional historical meteorological environmental time series data. Time series processing can include steps such as data alignment, normalization, and sliding window segmentation. Data alignment ensures that data from different sources correspond to each other in the time dimension. Normalization can use minimum-maximum normalization or Z-score normalization to make data of different scales comparable. Sliding window segmentation is to divide continuous time series data into segments of fixed length. The processed data is input into the time series feature extraction module in the park meteorological environment generation model. The time series feature extraction module is used to extract the historical environmental time series features and the regional historical meteorological environment time series features from the historical environmental time series data and the regional historical meteorological environment time series data respectively. Each time step output of LSTM can be regarded as the time series feature of the corresponding moment. For example, if the input sequence length is 24 (corresponding to 24 hours of data), the output shape of LSTM is (batch_size, 24, 128), where 128 is the number of LSTM hidden units. You can choose to use the output of the last time step as the feature representation of the entire sequence, or use the average of the outputs of all time steps. In this way, both the historical environmental time series features and the regional historical meteorological environment time series features can be represented as vectors of fixed dimensions.

[0117] The historical environmental time series features and the regional historical meteorological environmental time series features are fused into an environmental time series correlation feature matrix. The fusion method can use an attention mechanism or a gating mechanism. Taking the attention mechanism as an example, the correlation score between the two features can be calculated, and then weighted fusion is performed according to the score. The fused feature matrix contains comprehensive information about local environmental changes and large-scale meteorological changes. The time series feature loss function of the time series feature extraction module is constructed based on the environmental time series correlation feature matrix. The time series feature loss function can use the time series prediction error. This loss function evaluates the model's ability to capture time series features and guides the model to optimize the time series feature extraction process. Next, the environmental spatial correlation feature matrix and the environmental time series correlation feature matrix are input into the multi-scale feature fusion module. The goal of the multi-scale feature fusion module is to organically combine spatial features and temporal features to generate a comprehensive environmental representation. Feature fusion can be achieved by using a self-attention mechanism or a graph neural network. Taking the self-attention mechanism as an example, spatial features and temporal features can be regarded as different representations, and fusion is achieved by calculating the attention weights between them. The specific implementation can use a multi-head attention mechanism, such as using 8 attention heads, each with a dimension of 64. After fusion, the multi-dimensional feature correlation matrix of the environment is output, which integrates the environmental information in spatial and temporal dimensions. The feature fusion loss function of the multi-scale feature fusion module is constructed based on the multi-dimensional feature correlation matrix. The loss function can use cross entropy or KL divergence to evaluate the matching degree between the fused features and the real environment state.

[0118] Finally, the comprehensive loss function of the park meteorological environment generation model is obtained by combining the spatial feature loss function, the temporal feature loss function and the feature fusion loss function. The comprehensive loss function can be expressed as the weighted sum of these three losses: , where α, β, and γ are weight coefficients used to balance the importance of different loss terms. The weights can be determined to the optimal value by methods such as cross-validation or grid search. The comprehensive loss function value of the comprehensive loss function is calculated. The calculation process involves forward propagating the historical data through the model, then comparing it with the historical comprehensive environmental information of the park to obtain each loss term, and finally weighted summing it. The module parameters of the spatial feature extraction module, the temporal feature extraction module, and the multi-scale feature fusion module are iteratively updated based on the comprehensive loss function value. The update process can use optimization algorithms such as stochastic gradient descent (SGD) or Adam. The iterative process continues until the comprehensive loss function value reaches the minimum value or reaches the preset number of iterations, and finally the trained park meteorological environment generation model is obtained.

[0119] In one embodiment, the raw material storage information includes raw material hazard feature information and raw material location information. The three-dimensional model of the industrial park is used as a carrier, and image recognition technology is used to identify and analyze the image information. The raw material storage information of the chemical raw materials currently stored in the industrial park is obtained, which includes the following steps:

[0120] Grayscale all image information to obtain multiple grayscale images;

[0121] The edge point set of the grayscale image is extracted by using the edge enhancement operator, and the edge point set is integrated into the edge feature of the grayscale image;

[0122] Based on the edge features of the grayscale image and using a threshold segmentation algorithm, the target image area containing chemical raw materials in the grayscale image is separated from the image background;

[0123] Identify hazardous material marks on the surface of chemical raw materials in the target image area through image recognition technology, classify the hazardous material marks, and obtain the raw material hazard characteristic information of the corresponding chemical raw materials according to the classification results;

[0124] Marking the image information containing the target image area as the target image information;

[0125] The device position information of the target image acquisition device that shoots the target image information is obtained through the three-dimensional model of the park, and the device position information is used as the raw material position information of the chemical raw material corresponding to the target image information.

[0126] In this embodiment, the image information is grayed, which can make the object contours in the image information clearer and the details more obvious. After obtaining multiple grayscale images, the edge point set of the grayscale image is extracted using an edge enhancement operator, and the obtained edge point set is integrated into the grayscale image edge feature, which is helpful to identify the outline and shape of the chemical raw material in the image, and at the same time, the color distribution in the image can be understood, which is convenient for image segmentation and subsequent extraction of image size information of the chemical raw material. Among them, the commonly used edge enhancement operators include Sobel operator, Laplace operator and Roberts operator. After obtaining the grayscale image edge feature, the target image area of ​​the chemical raw material is separated from the image background using a threshold segmentation algorithm, which can remove the image background that interferes with the target image area, and facilitate more accurate identification of hazardous raw material logos on the surface of the chemical raw material.

[0127] Specifically, the main principle of the threshold segmentation algorithm is to divide the pixel set according to the gray level, so that each subset forms an area corresponding to the real scene, each area has consistent attributes, and the adjacent areas do not have such consistent attributes. The threshold segmentation algorithm can accurately separate the target image area of ​​chemical raw materials from the image background, thereby avoiding the interference of the image background on the identification of the target image area.

[0128] By using image recognition technology to analyze the hazardous characteristics of chemical raw materials, chemical raw materials that may evaporate or leak toxic gases are classified as toxic gas raw materials, such as chlorine, ammonia, etc., which are toxic gases that can harm the human body. Chemical raw materials that may burn or explode are classified as flammable and explosive raw materials, such as gasoline, ethanol, methane, aluminum powder, magnesium powder, nitrocellulose, etc., which are all flammable and explosive chemical raw materials.

[0129] After marking the image information containing the target image area as the target image information, the location of the high-definition camera that captured the target image information is obtained through the three-dimensional model of the park, so that the raw material location information of the corresponding chemical raw material can be preliminarily determined.

[0130] In one embodiment, it also includes:

[0131] Use LDSR algorithm to enhance the resolution of target image area;

[0132] Use OCR technology to identify the size text features from the target image area after resolution enhancement, and obtain the actual size information of the corresponding chemical raw materials by analyzing the size text features;

[0133] Calculate the image size information of the chemical raw material in the image information according to the edge features of the grayscale image;

[0134] The actual distance between the target image acquisition device and the chemical raw material in the three-dimensional model of the park is calculated by combining the actual size information, image size information and the preset device parameters of the target image acquisition device, and the raw material position information is corrected according to the actual distance.

[0135] In this embodiment, the target image area refers to an image with the image background removed and containing chemical raw materials. The LDSR algorithm is an image super-resolution algorithm based on a deep convolutional neural network (DCNN). The algorithm builds a deep convolutional neural network model to learn the mapping relationship from low-resolution images to high-resolution images, while retaining the texture and detail information of the image, to achieve super-resolution enhancement of the image. This embodiment uses the LDSR algorithm to enhance the resolution of the target image area, so that the size and text features of the target image area can be identified using OCR technology later.

[0136] The target image area after resolution enhancement is first corrected and denoised using OCR technology. This is because the dimension text in the target image area may be tilted or stained. The dimension text is then segmented into columns to segment each character in the text. Then, the dimension text features are identified and extracted to obtain the dimension text features. The obtained dimension text features are parsed to obtain the actual size information of the corresponding chemical raw materials, such as 2mx2mx2.5m. The image size information of the chemical raw materials in the image information is calculated using the pre-obtained grayscale image edge features, such as 20cmx20cmx25cm. The grayscale image edge features can reflect the length, width and height of the chemical raw materials in the image information, that is, the image size information. The size reduction ratio of the chemical raw materials in the image can be calculated based on the actual size information and the image size information. Then, the actual distance between the high-definition camera and the chemical raw materials can be calculated by combining the installation height, lens focal length and shooting rotation angle in the high-definition camera device parameters. The raw material position information is corrected based on the actual distance obtained, so that the obtained raw material position information is more accurate.

[0137] In one embodiment, combining the raw material storage information and the real-time comprehensive environmental information of the industrial park to perform a multi-dimensional analysis of the storage of chemical raw materials in the industrial park by region, and determining the park danger level of the industrial park according to the analysis results includes the following steps:

[0138] For any stored chemical raw materials in the industrial park, determine whether the stored chemical raw materials are toxic gas raw materials or flammable and explosive raw materials according to the chemical raw material types in the raw material hazard characteristic information of the stored chemical raw materials;

[0139] If the stored chemical raw materials are toxic gas raw materials, the toxic gas leakage data of the target area in the industrial park is obtained from the real-time comprehensive environmental information of the industrial park, and the toxic gas concentration data of the target area is obtained based on the toxic gas leakage data. The target area is the area where the raw material location information of the stored chemical raw materials is located;

[0140] If the toxic gas concentration data of the target area is 0, the toxic gas danger level value of the target area is set to zero;

[0141] If the toxic gas concentration data of the target area is greater than 0, the toxic gas diffusion range of the target area is calculated by combining the toxic gas concentration data and the real-time comprehensive environmental information of the park in the target area, and the toxic gas danger level value of the target area is determined according to the toxic gas diffusion range;

[0142] If the stored chemical raw materials are flammable and explosive raw materials, the probability of static electricity occurrence in the target area is predicted based on the real-time temperature data and real-time humidity data of the real-time comprehensive environmental information of the park in the target area, and the explosion hazard value of the target area is calculated using the static electricity occurrence probability;

[0143] Different regional weights are assigned to each target area according to the area of ​​the target area and the amount of chemical raw materials stored;

[0144] Based on the regional weights and combined with the toxic gas hazard level values ​​and / or explosion hazard level values ​​of all target areas, a weighted calculation is performed to obtain the comprehensive hazard level value of the industrial park. The comprehensive hazard level value is divided into corresponding park hazard level levels using preset hazard level thresholds.

[0145] In this embodiment, whether any chemical raw material stored in the industrial park is a toxic gas raw material or a flammable and explosive raw material is determined based on the type of chemical raw material in the raw material hazard characteristic information. If the stored chemical raw material is a toxic gas raw material, the toxic gas leakage data (obtained through a gas sensor) in the real-time comprehensive environmental information of the industrial park is directly retrieved. The toxic gas leakage data includes the toxic gas concentration data where the toxic gas leakage occurs. The toxic gas concentration data of the target area can be obtained through the toxic gas leakage data. The target area refers to the area where the raw material location information of the chemical raw material is currently stored.

[0146] If the toxic gas concentration data of the target area is 0, it means that there is no toxic gas leakage in the target area, and the toxic gas danger level value of the target area is immediately set to zero. If the toxic gas concentration data of the target area is greater than 0, it means that there is a toxic gas leakage in the target area, and the toxic gas diffusion range of the target area is calculated by combining the toxic gas concentration data and the wind direction angle, toxic gas concentration data and real-time temperature data contained in the real-time comprehensive environmental information of the park in the target area, and the toxic gas danger level value of the target area is determined according to the toxic gas diffusion range.

[0147] If the stored chemical raw materials are flammable and explosive raw materials, the real-time temperature data and real-time humidity data in the real-time comprehensive environmental information of the park in the target area can be used to establish a regression equation for static electricity occurrence probability prediction, and the static electricity occurrence probability in the target area can be obtained. Then, the real-time temperature data, real-time humidity data, static electricity occurrence probability data, and the raw material flash point data and the raw material explosion lower limit data contained in the raw material hazard characteristic information are input into the preset combustion and explosion hazard prediction model to obtain the combustion and explosion hazard value. The preset combustion and explosion hazard prediction model is constructed using a neural network model (such as a BP neural network model). For flammable and explosive chemical raw materials, once static electricity occurs, it is very likely to cause the chemical raw materials to burn or explode, so it is necessary to use the probability of static electricity occurrence as one of the factors constituting the combustion and explosion hazard value.

[0148] Different regional weights are assigned to each target area according to the area of ​​the target area and the amount of stored chemical raw materials. If the area of ​​the target area is large and the amount of stored chemical raw materials is large, the regional weight assigned to the target area is also larger. According to the assigned regional weights, combined with the toxic gas hazard level value and / or the combustion and explosion hazard level value of the target area, the target area may have both the toxic gas hazard level value and the combustion and explosion hazard level value, or may have only one of them. After weighted calculation, the comprehensive hazard level value of the industrial park is obtained, and the comprehensive hazard level value is graded using the preset hazard level threshold to obtain the corresponding park hazard level.

[0149] If the danger level of the park is greater than the preset danger level threshold, the optimal evacuation route is generated based on the three-dimensional model of the park using a path optimization algorithm. The optimal evacuation route is sent to the communication equipment held by the park staff in the form of a plane map, and an alarm command is also sent to the park alarm system.

[0150] If the park danger level is less than or equal to the preset park danger level threshold, the corresponding staff protection plan is generated according to the toxic gas danger level value and the explosion danger level value of the target area. For example, when the toxic gas danger level value is greater than 0, the park staff must wear chemical safety protective glasses and mask-type adhesive tape anti-gas clothing, and when the explosion danger level value is greater than 0, they must wear anti-static work clothes.

[0151] In one embodiment, the toxic gas diffusion range of the target area is obtained by comprehensive calculation based on the toxic gas concentration data and the real-time comprehensive environmental information of the park in the target area, and the toxic gas danger level value of the target area is determined according to the toxic gas diffusion range, including the following steps:

[0152] In the three-dimensional model of the industrial park, a three-dimensional coordinate system of the industrial park is constructed with the center of the industrial park as the origin;

[0153] Extract wind direction characteristic data, wind speed characteristic data and real-time temperature data of the target area from the real-time comprehensive environmental information of the park;

[0154] Import the wind direction characteristic data and wind speed characteristic data into the park three-dimensional model, determine the wind direction point coordinates of the strongest wind direction point in the target area through the park three-dimensional coordinate system, and calculate the wind direction angle in the target area based on the wind direction point coordinates;

[0155] The toxic gas diffusion concentration in the target area is calculated by combining wind direction angle, toxic gas concentration data and real-time temperature data;

[0156] Determine the main wind direction axis of the target area according to the wind direction angle, combine the wind speed characteristic data, toxic gas diffusion concentration and toxic gas concentration data, and use the Gaussian puff model to calculate the toxic gas diffusion distance on the main wind direction axis in the target area, and estimate the toxic gas diffusion range in the target area based on the toxic gas diffusion distance;

[0157] The toxic gas danger level value of the target area is calculated based on the area ratio of the toxic gas diffusion range in the target area.

[0158] In this embodiment, a three-dimensional coordinate system of the industrial park is constructed with the center of the industrial park as the origin in the three-dimensional model of the park. The three-dimensional coordinate system takes the center of the industrial park as the origin, the east of the ground plane of the industrial park as the x-axis, the north of the ground plane of the industrial park as the y-axis, and the z-axis directly above the center of the industrial park. The wind direction characteristic data (such as east wind), wind speed characteristic data (such as real-time wind speed, average wind speed, etc.) and real-time temperature data of the target area are extracted from the real-time comprehensive environmental information of the park, and the wind direction characteristic data and wind speed characteristic data are imported into the three-dimensional model of the park. The strongest wind direction point in the wind direction characteristic data is found by using professional equipment such as anemometers, and then the wind direction point coordinates of the strongest wind direction point in the target area are determined by the three-dimensional coordinate system of the park, for example, the wind direction point coordinates are (76, 0, 7), and the wind direction angle in the target area is calculated by using trigonometric functions according to the wind direction point coordinates of the strongest wind direction point, for example, an angle of 30 degrees with the x-axis as the base.

[0159] The toxic gas diffusion concentration in the target area is calculated based on the wind direction angle, toxic gas concentration data and real-time temperature data. The calculation formula is as follows:

[0160] ,

[0161] in, is the toxic gas diffusion concentration in the target area n, is the gas diffusion fixed coefficient, is the real-time temperature data of target area n, is the wind direction angle of target area n, is the toxic gas concentration data of the target area n, and are the horizontal and vertical coordinates of the chemical raw materials stored in the target area n in the three-dimensional coordinate system of the park. For example, taking the toxic gas ammonia as an example, assuming =26℃, =30°, =40, =30 (obtained from the ratio of the three-dimensional park model to the industrial park), =40mg / m³, =2.81×10 -5(The diffusion coefficient of ammonia in the air can be obtained by looking up the table), and the toxic gas diffusion concentration of ammonia is calculated to be 3.375×10 -4 mg / m³. In addition, the toxic gas diffusion concentration is real-time.

[0162] The main wind direction axis of the target area, such as the x-axis, is determined according to the wind direction angle. The wind speed characteristic data, toxic gas diffusion concentration and toxic gas concentration data are combined, and the toxic gas diffusion distance of the toxic gas in the target area on the main wind direction axis is calculated using the Gaussian puff model. The toxic gas diffusion distance calculation formula is as follows:

[0163] ,

[0164] Where: Indicates the diffusion distance of toxic gas in target area n, represents the gas diffusion time, represents the exponential function, Indicates the effective discharge height, Represents the wind speed characteristic data of target area n.

[0165] The Gaussian puff model is a model used to describe the diffusion of atmospheric pollution, especially in sudden leakage accidents, when the pollution source suddenly releases a large amount of harmful gases in a short period of time, the model is used to solve the ground pollution concentration. This model assumes that the volume of the polluted gas cloud grows horizontally and vertically, simulating the changes in the polluted gas cloud in time and space. Applied in this embodiment, after further improvement of the Gaussian puff model, the diffusion distance of the leaked toxic gas can be accurately calculated, and the diffusion range of the toxic gas in the target area can be estimated based on the diffusion distance of the toxic gas. The toxic gas hazard level value of the target area is calculated based on the regional proportion of the toxic gas diffusion range in the target area (for example, 65%). (For example, the toxic gas hazard level value is 6.5).

[0166] In one embodiment, generating an optimal evacuation path based on a three-dimensional model of a park and using a path optimization algorithm includes the following steps:

[0167] Using the center point of the area where the staff in the industrial park gather as the starting point coordinate and the center point of the preset safe area in the industrial park as the end point coordinate, the three-dimensional model of the industrial park is used to generate multiple initial evacuation routes of the industrial park;

[0168] Encode all initial evacuation paths as the initial population of the path optimization algorithm and set the maximum number of iterations of the path optimization algorithm;

[0169] Import the diffusion range of toxic gases into the three-dimensional coordinate system of the park to generate a range coordinate set corresponding to the diffusion range of toxic gases;

[0170] Integrate the raw material location information of all stored chemical raw materials into a raw material location coordinate set;

[0171] The constraints of the path optimization algorithm are constructed by combining the range coordinate set and the raw material location coordinate set;

[0172] Iteratively update the initial population based on the constraint conditions until the number of iterative updates reaches the maximum number of iterations to obtain the optimal population;

[0173] The fitness of each individual in the optimal population is calculated according to the preset fitness function, and the individual with the highest fitness in the optimal population is decoded as the optimal evacuation path.

[0174] In this embodiment, the center point of the area where the staff in the industrial park gathers is used as the starting point coordinate, and the center point of the preset safety area in the industrial park is used as the end point coordinate. Generally, an open place such as a square in the park is used as a safety area. Multiple evacuation paths are generated in the three-dimensional model of the park, and all evacuation paths are encoded, for example: {path 1, path 2, path 3...}. At the same time, the maximum number of iterations of the path optimization algorithm (such as the gray wolf optimization algorithm) is set, and the toxic gas diffusion range is imported into the three-dimensional coordinate system of the park. The boundary point coordinates of multiple boundary points on the range boundary of the toxic gas diffusion range are combined to generate a range coordinate set containing the toxic gas diffusion range. At the same time, based on the three-dimensional coordinate system of the park, the raw material location information of all stored chemical raw materials is integrated into a raw material location coordinate set, and the constraint conditions of the path optimization algorithm are constructed according to the range coordinate set and the raw material location coordinate set. The role of the constraint conditions is to exclude the evacuation paths within the toxic gas range and the evacuation paths with chemical raw materials stored on the paths, so as to prevent dangerous accidents during the evacuation of personnel. The evacuation path is iterated multiple times according to the constraints until the number of iterations reaches the maximum number of iterations, and then the fitness of each evacuation path is calculated using the preset fitness function, and the evacuation path with the highest fitness is taken as the optimal evacuation path. The optimal evacuation path is then marked in the 3D model of the park. At the same time, in order to facilitate the review of the park staff, the 3D model of the park marked with the optimal evacuation path is converted into a flat map, and then sent to the mobile devices held by the park staff (usually mobile phones or electronic watches). At the same time, the alarm facilities in the park will also receive alarm instructions to remind the park staff to evacuate to a safe area in an orderly manner.

[0175] The present application also discloses an intelligent safety risk management and control platform for industrial parks, including:

[0176] a memory configured to store instructions; and

[0177] The processor is configured to call instructions from the memory and implement the above-mentioned method of intelligent management and control of security risks in the industrial park when executing the instructions.

[0178] An embodiment of the present application also provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the above-mentioned intelligent management and control of security risks in an industrial park.

[0179] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0180] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0181] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0182] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0183] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0184] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0185] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0186] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0187] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for intelligent management and control of safety risks in an industrial park, characterized in that: Applicable to industrial parks storing different types of chemical raw materials, the method comprises the following steps: The real-time environmental information of the industrial park at the current moment is obtained in real time through multiple types of sensors preset in different areas of the industrial park, and the regional meteorological environmental information of the area where the industrial park is located at the current moment is obtained through a weather forecast system; Establishing a park meteorological environment generation model based on a neural network model, combining the real-time environmental information and the regional meteorological environment information and using the park meteorological environment generation model to generate park real-time comprehensive environmental information containing comprehensive environmental information of all areas of the industrial park; Acquire the park plane map information of the industrial park, and establish the park three-dimensional model of the industrial park based on the park plane map information; Acquire image information of the industrial park through image acquisition devices preset in different areas of the industrial park; Using the three-dimensional model of the industrial park as a carrier, image recognition technology is used to identify and analyze the image information to obtain raw material storage information of the chemical raw materials currently stored in the industrial park, wherein the raw material storage information includes raw material hazard feature information and raw material location information; For any stored chemical raw materials in the industrial park, judging whether the stored chemical raw materials are toxic gas raw materials or flammable and explosive raw materials according to the chemical raw material types in the raw material hazard characteristic information of the stored chemical raw materials; If the stored chemical raw material is the toxic gas raw material, the toxic gas leakage data of the target area in the industrial park is obtained from the real-time comprehensive environmental information of the park, and the toxic gas concentration data of the target area is obtained based on the toxic gas leakage data, and the target area is the area where the raw material location information of the stored chemical raw material is located; If the toxic gas concentration data of the target area is 0, the toxic gas danger level value of the target area is set to zero; If the toxic gas concentration data of the target area is greater than 0, the toxic gas diffusion range of the target area is obtained by comprehensive calculation based on the toxic gas concentration data and the real-time comprehensive environmental information of the park in the target area, and the toxic gas danger level value of the target area is determined according to the toxic gas diffusion range; If the stored chemical raw materials are the flammable and explosive raw materials, the probability of static electricity occurrence in the target area is predicted based on the real-time temperature data and real-time humidity data of the real-time comprehensive environmental information of the park in the target area, and the explosion hazard value of the target area is calculated using the static electricity occurrence probability; Allocating different regional weights to each of the target regions according to the area of ​​the target region and the amount of the stored chemical raw materials; Based on the regional weights and in combination with the toxic gas hazard level values ​​and / or the explosion hazard level values ​​of all the target areas, a weighted calculation is performed to obtain a comprehensive hazard level value of the industrial park, and the comprehensive hazard level value is divided into corresponding park hazard levels using a preset hazard level threshold; If the danger level of the industrial park is greater than a preset danger level threshold, an optimal evacuation path is generated based on the three-dimensional model of the industrial park and using a path optimization algorithm, and the optimal evacuation path is sent to a mobile terminal held by a staff member in the industrial park.

2. The method according to claim 1, characterized in that The steps of establishing a park meteorological environment generation model based on a neural network model, combining the real-time environmental information and the regional meteorological environment information and using the park meteorological environment generation model to generate the park real-time comprehensive environmental information containing the comprehensive environmental information of all areas of the industrial park are as follows: A spatial feature extraction module is constructed based on a convolutional neural network model, and a temporal feature extraction module is constructed based on a long short-term memory neural network model. The spatial feature extraction module, the temporal feature extraction module and a preset multi-scale feature fusion module are combined to construct a park meteorological environment generation model; Extracting historical environmental information and historical comprehensive environmental information of the industrial park in the same historical period from a preset park environment database, wherein the historical environmental information is environmental information acquired by the multi-type sensors in different areas during the historical period, and both the historical environmental information and the real-time environmental information include sensor location information of the multi-type sensors in the industrial park; Using the historical period as a query index, retrieve the regional historical meteorological environment information of the area where the industrial park is located during the historical period from the weather forecast system; Inputting the historical environmental information, the historical comprehensive environmental information of the industrial park and the historical meteorological environmental information of the region into the industrial park meteorological environment generation model, training the industrial park meteorological environment generation model to learn the spatial characteristic change relationship and the temporal characteristic change relationship of the environmental information in the industrial park, and obtaining the trained industrial park meteorological environment generation model; The real-time environmental information and the regional meteorological environmental information are input into the trained park meteorological environment generation model, and the trained park meteorological environment generation model outputs the park real-time comprehensive environmental information containing the comprehensive environmental information of all areas of the industrial park.

3. The method according to claim 2, characterized in that The step of inputting the historical environmental information, the industrial park historical comprehensive environmental information and the regional historical meteorological environmental information into the industrial park meteorological environment generation model, training the industrial park meteorological environment generation model to learn the spatial feature change relationship and the temporal feature change relationship of the environmental information in the industrial park, and obtaining the trained industrial park meteorological environment generation model comprises the following steps: Inputting the historical environmental information into the spatial feature extraction module in the park meteorological environment generation model, and respectively extracting global environmental spatial correlation features and local environmental spatial correlation features from the historical environmental information through the global spatial feature extraction layer and the local spatial feature extraction layer in the spatial feature extraction module; The global environment space correlation feature and the local environment space correlation feature are merged into an environment space correlation feature matrix, and a space feature loss function of the space feature extraction module is constructed based on the environment space correlation feature matrix; Respectively performing time series processing on the historical environmental information and the regional historical meteorological environmental information to obtain historical environmental time series data and regional historical meteorological environmental time series data, and inputting the historical environmental time series data and the regional historical meteorological environmental time series data into the time series feature extraction module in the park meteorological environment generation model; Utilizing the time series feature extraction module to extract historical environment time series features and regional historical meteorological environment time series features from the historical environment time series data and the regional historical meteorological environment time series data respectively; The historical environmental time series features and the regional historical meteorological environmental time series features are merged into an environmental time series correlation feature matrix, and a time series feature loss function of the time series feature extraction module is constructed based on the environmental time series correlation feature matrix; Inputting the environment spatial correlation feature matrix and the environment temporal correlation feature matrix into the multi-scale feature fusion module, using the multi-scale feature fusion module to output the environment multi-dimensional feature correlation matrix, and constructing a feature fusion loss function of the multi-scale feature fusion module based on the multi-dimensional feature correlation matrix; The spatial feature loss function, the temporal feature loss function and the feature fusion loss function are combined to obtain a comprehensive loss function of the park meteorological environment generation model, and a comprehensive loss function value of the comprehensive loss function is calculated according to the historical comprehensive environmental information of the park; The module parameters of the spatial feature extraction module, the temporal feature extraction module and the multi-scale feature fusion module are iteratively updated based on the comprehensive loss function value until the comprehensive loss function value reaches a minimum value, thereby obtaining the trained park meteorological environment generation model.

4. The method according to claim 1, characterized in that The method of using the three-dimensional model of the industrial park as a carrier and using image recognition technology to identify and analyze the image information to obtain the raw material storage information of the chemical raw materials currently stored in the industrial park includes the following steps: grayscale all the image information to obtain multiple grayscale images; Extracting edge point sets of the grayscale image using an edge enhancement operator, and integrating the edge point sets into grayscale image edge features; Based on the edge features of the grayscale image and using a threshold segmentation algorithm, a target image region in the grayscale image containing the chemical raw material is separated from the image background; Identify the hazardous material logo on the surface of the chemical raw material in the target image area by image recognition technology, classify the hazardous material logo, and obtain the raw material hazard characteristic information corresponding to the chemical raw material according to the classification result; Marking the image information containing the target image area as target image information; The device position information of the target image acquisition device that photographs the target image information is acquired through the three-dimensional model of the park, and the device position information is used as the raw material position information of the chemical raw material corresponding to the target image information.

5. The method according to claim 4, characterized in that The method further comprises: Performing resolution enhancement on the target image region using an LDSR algorithm; Using OCR technology to identify the size text features from the target image area after resolution enhancement, and obtaining the actual size information corresponding to the chemical raw material by analyzing the size text features; Calculating image size information of the chemical raw material in the image information according to the edge features of the grayscale image; The actual distance between the target image acquisition device and the chemical raw material in the three-dimensional model of the park is calculated in combination with the actual size information, the image size information and the preset device parameters of the target image acquisition device, and the raw material position information is corrected according to the actual distance.

6. The method according to claim 1, characterized in that The step of comprehensively calculating the toxic gas diffusion range of the target area by combining the toxic gas concentration data and the real-time comprehensive environmental information of the park in the target area, and determining the toxic gas danger level value of the target area according to the toxic gas diffusion range comprises the following steps: Constructing a three-dimensional coordinate system of the industrial park with the center of the industrial park as the origin in the three-dimensional model of the industrial park; Extracting wind direction characteristic data, wind speed characteristic data and real-time temperature data of the target area from the real-time comprehensive environmental information of the park; Importing the wind direction characteristic data and the wind speed characteristic data into the park three-dimensional model, determining the wind direction point coordinates of the strongest wind direction point in the target area through the park three-dimensional coordinate system, and calculating the wind direction angle in the target area based on the wind direction point coordinates; Calculate the toxic gas diffusion concentration in the target area by combining the wind direction angle, the toxic gas concentration data and the real-time temperature data; Determine the main wind direction axis of the target area according to the wind direction angle, combine the wind speed characteristic data, the toxic gas diffusion concentration and the toxic gas concentration data, and use a Gaussian puff model to calculate the toxic gas diffusion distance on the main wind direction axis in the target area, and estimate the toxic gas diffusion range of the target area based on the toxic gas diffusion distance; The toxic gas danger level value of the target area is calculated based on the area ratio of the toxic gas diffusion range in the target area.

7. The method according to claim 6, characterized in that The calculation formula for the toxic gas diffusion concentration is as follows: ， Where: is the toxic gas diffusion concentration in the target area n, is the gas diffusion fixed coefficient, is the real-time temperature data of the target area n, is the wind direction angle of the target area n, is the toxic gas concentration data of the target area n, and are respectively the horizontal and vertical coordinates of the chemical raw materials stored in the target area n in the three-dimensional coordinate system of the park; The calculation formula of the toxic gas diffusion distance is as follows: ， Where: represents the diffusion distance of the toxic gas in the target area n, represents the gas diffusion time, represents the exponential function, Indicates the effective discharge height, Represents the wind speed characteristic data of the target area n.

8. The method according to claim 1, characterized in that: The generating of the optimal evacuation path based on the park three-dimensional model and using a path optimization algorithm comprises the following steps: Using the center point of the area where the workers in the industrial park gather as the starting point coordinates and the center point of the preset safe area in the industrial park as the end point coordinates, a plurality of initial evacuation paths of the industrial park are generated using the three-dimensional model of the industrial park; Encoding all the initial evacuation paths as an initial population of a path optimization algorithm, and setting a maximum number of iterations of the path optimization algorithm; Importing the toxic gas diffusion range into the park three-dimensional coordinate system to generate a range coordinate set corresponding to the toxic gas diffusion range; Integrate the raw material location information of all the stored chemical raw materials into a raw material location coordinate set; Constructing the constraint conditions of the path optimization algorithm in combination with the range coordinate set and the raw material position coordinate set; Iteratively updating the initial population based on the constraint conditions until the number of iterative updates reaches the maximum number of iterations to obtain an optimal population; The fitness of each individual in the optimal population is calculated according to a preset fitness function, and the individual with the highest fitness in the optimal population is decoded as an optimal evacuation path.

9. An intelligent safety risk management and control platform for industrial parks, characterized in that: include: a memory configured to store instructions; as well as The processor is configured to call the instructions from the memory and to implement the method for intelligent management and control of safety risks of an industrial park according to any one of claims 1 to 8 when executing the instructions.

Images