Deep learning-based environmental pollution risk assessment method and system, computing device and storage medium

Abstract

The invention provides an environmental pollution risk assessment method and system based on deep learning, a computing device and a storage medium, and relates to the technical field of environmental pollution risk assessment, and the method comprises the steps: determining a preliminary environmental pollution condition of a target region through employing a multi-scale multi-factor coupling analysis method based on the parameter values of physicochemical property parameters and background information; verifying and correcting the initial environmental pollution condition to obtain a target environmental pollution condition; according to the target environmental pollution condition, the historical environmental pollution risk assessment result, the public health data and the parameter values of the organism response parameters, the target environmental pollution risk assessment result is obtained; and an environmental pollution risk assessment model including a deep learning-based multilayer perceptron, an attention mechanism and a generative adversarial network is adopted to assess the overall influence degree of the environmental pollution condition of the target area on public health and an ecological system, so that a target environmental pollution risk assessment result can be obtained. According to the technical scheme provided by the invention, the accuracy of the target environmental pollution risk assessment result is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of environmental pollution risk assessment, and in particular, to a method, system, computing device, and storage medium for environmental pollution risk assessment based on deep learning. Background Art

[0002] Environmental pollution risk assessment is an important part of environmental protection and management, and is used to evaluate the impact of environmental pollution on public health and ecosystems. Traditional environmental pollution risk assessment methods are usually based on statistics and empirical models, which have limitations when dealing with complex and multi-factor environmental pollution problems.

[0003] In recent years, with the development of deep learning technology, environmental pollution risk assessment methods based on deep learning have gradually become a research hotspot. These methods have shown certain advantages in dealing with complex and multi-factor environmental pollution problems, but there are still some significant defects: for example, the models used in these methods are usually static, which limits the application of the models in dynamic environments. And these methods usually use fixed weights to process various key parameters, resulting in poor adaptability of the models under different environmental conditions, resulting in poor accuracy of environmental pollution risk assessment results. Summary of the invention

[0004] The embodiments of the present application provide an environmental pollution risk assessment method, system, computing device and storage medium based on deep learning, which are used to solve the problem of poor accuracy of environmental pollution risk assessment results caused by unreasonable model design in the prior art.

[0005] In a first aspect, an embodiment of the present application provides an environmental pollution risk assessment method based on deep learning, comprising: Obtaining parameter values, background information, historical environmental pollution risk assessment results, and public health data of multiple environmental parameters in the target area; the multiple environmental parameters include physical and chemical property parameters and organism response parameters; Based on the parameter values ​​of the physical and chemical property parameters and the background information, a multi-scale multi-factor coupling analysis method is used to determine the preliminary environmental pollution situation of the target area; Verifying and correcting the preliminary environmental pollution situation to obtain the target environmental pollution situation; According to the target environmental pollution situation, the historical environmental pollution risk assessment results, the public health data and the parameter values ​​of the organism response parameters, an environmental pollution risk assessment model is used to assess the overall impact of the environmental pollution situation in the target area on public health and the ecosystem, so as to obtain the target environmental pollution risk assessment results; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a double-layer Bayesian network.

[0006] Optionally, according to the target environmental pollution situation, the historical environmental pollution risk assessment results, the public health data and the parameter value of the organism response parameter, an environmental pollution risk assessment model is used to assess the overall impact of the environmental pollution situation in the target area on public health and the ecosystem to obtain a target environmental pollution risk assessment result; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a two-layer Bayesian network, including: Extracting characteristic information of key parameters from the target environmental pollution situation, the historical environmental pollution risk assessment results, the public health data, and the parameter values ​​of the organism response parameters respectively; A multi-layer perceptron is used to process the characteristic information of all key parameters to obtain preliminary risk assessment results; Calculating the importance weight of each key parameter using an attention mechanism based on the preliminary risk assessment result, and applying the importance weight to the feature information of the corresponding key parameter to generate weighted feature information of the key parameter; Based on the weighted feature information of all key parameters, a new data sample is generated by using a generative adversarial network, the new data sample is used to optimize a multilayer perceptron, and the weighted feature information of all key parameters is processed by using the optimized multilayer perceptron to obtain an optimized risk assessment result; Define a first node structure and a first edge structure corresponding to the public health layer and a second node structure and a second edge structure corresponding to the ecosystem layer in a two-layer Bayesian network; perform an environmental pollution risk assessment on the public health layer based on the first node structure and the first edge structure to obtain a first reference risk assessment result; perform an environmental pollution risk assessment on the ecosystem layer based on the second node structure and the second edge structure to obtain a second reference risk assessment result; The optimized risk assessment result is adjusted in combination with the first reference risk assessment result and the second reference risk assessment result to obtain a target environmental pollution risk assessment result.

[0007] Optionally, the multi-layer perceptron is used to process the characteristic information of all key parameters to obtain a preliminary risk assessment result, including: Performing nonlinear mapping processing on the characteristic information of the key parameters to obtain a preliminary score for each key parameter; A normalization processing algorithm is introduced to standardize the preliminary scores of all key parameters to obtain standardized preliminary scores; A weighted average algorithm is introduced to comprehensively calculate all standardized preliminary scores to obtain preliminary risk assessment results.

[0008] Optionally, the using an attention mechanism to calculate the importance weight of each key parameter based on the preliminary risk assessment result includes: Conduct contextual relevance analysis on each key parameter and determine the change trend of each key parameter; A dynamic weight allocation algorithm is introduced to dynamically adjust the initial weight of each key parameter according to the change trend and context relevance of the key parameter to generate a dynamically adjusted preliminary weight; the preliminary risk assessment result is divided into multiple subspaces, the local importance score of each key parameter in the subspace is calculated, and the local importance scores of the key parameter in all the subspaces are aggregated to obtain the global importance score of the key parameter; Merging the preliminary weight with the global importance score to obtain a comprehensive importance score of the key parameter; The comprehensive importance scores of all key parameters are standardized to obtain the importance weights of all key parameters.

[0009] Optionally, aggregating the local importance scores of the key parameter in all the subspaces to obtain a global importance score of the key parameter includes: Evaluate the importance of each subspace to obtain a subspace evaluation result, and determine the subspace initial weight according to the subspace evaluation result; Adopting an adaptive adjustment algorithm to dynamically adjust the weight of each subspace, and obtaining the adjusted weight of the subspace; The product of the local importance score of the key parameter in each subspace and the adjusted weight of the corresponding subspace is used as the weighted local importance score of the subspace; The weighted local importance scores of all subspaces are summed to obtain the global importance score of the key parameters.

[0010] Optionally, the combining the first reference risk assessment result and the second reference risk assessment result to adjust the optimized risk assessment result to obtain a target environmental pollution risk assessment result includes: extracting a shared feature representation of the first reference risk assessment result and the second reference risk assessment result; fusing the shared feature representation, the first reference risk assessment result, and the second reference risk assessment result to obtain a fusion result; The environmental restoration potential index of the target area is obtained, and the environmental restoration potential index and the fusion result are used as residuals to adjust the optimized risk assessment result to obtain the target environmental pollution risk assessment result.

[0011] Optionally, the multi-scale multi-factor coupling analysis method includes a geographically weighted regression method, a multi-scale geometric analysis method and a multi-factor coupling analysis method; The method of determining the preliminary environmental pollution situation of the target area by using a multi-scale multi-factor coupling analysis method based on the parameter values ​​of the physical and chemical property parameters and the background information includes: Based on the physical and chemical property parameters and background information, a geographically weighted regression model is constructed; in each sub-region, a local regression analysis is performed using the geographically weighted regression model to generate a preliminary environmental pollution situation in the sub-region, and the preliminary environmental pollution situations of all sub-regions are integrated to generate the first overall environmental pollution situation of the target region; Combined with the geographical features in the background information, the target area is divided into multiple sub-areas of different scales; a multi-scale geometric analysis method is used to extract geometric features in each sub-area; a multi-modal fusion technology is used to perform multi-factor fusion on the physical and chemical property parameters, the background information and the geometric features to obtain multi-factor features after fusion; a comprehensive analysis method is used to perform a comprehensive analysis on the multi-factor features to generate a second overall environmental pollution situation of the target area; The preliminary environmental pollution situation of the target area is determined according to the first overall environmental pollution situation of the target area and the second overall environmental pollution situation of the target area.

[0012] In a second aspect, the embodiment of the present application provides an environmental pollution risk assessment system based on deep learning, comprising: An acquisition module, used to acquire parameter values, background information, historical environmental pollution risk assessment results and public health data of multiple environmental parameters in the target area; the multiple environmental parameters include physical and chemical property parameters and organism response parameters; A determination module, for determining the preliminary environmental pollution situation of the target area by adopting a multi-scale multi-factor coupling analysis method based on the parameter values ​​of the physicochemical property parameters and the background information; A processing module, used to verify and correct the preliminary environmental pollution situation to obtain the target environmental pollution situation; The evaluation module is used to evaluate the overall impact of the environmental pollution situation in the target area on public health and the ecosystem based on the target environmental pollution situation, the historical environmental pollution risk assessment results, the public health data and the parameter values ​​of the organism response parameters using an environmental pollution risk assessment model to obtain a target environmental pollution risk assessment result; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a double-layer Bayesian network.

[0013] In a third aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a deep learning-based environmental pollution risk assessment method as described in any one of the first aspects.

[0014] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements an environmental pollution risk assessment method based on deep learning as described in any one of the first aspects.

[0015] In an embodiment of the present application, a method for environmental pollution risk assessment based on degree learning is provided, the method comprising: obtaining parameter values, background information, historical environmental pollution risk assessment results and public health data of multivariate environmental parameters in a target area; the multivariate environmental parameters include physicochemical property parameters and organism response parameters; based on the parameter values ​​and background information of the physicochemical property parameters, a multi-scale multi-factor coupling analysis method is used to determine the preliminary environmental pollution situation of the target area; the preliminary environmental pollution situation is verified and corrected to obtain the target environmental pollution situation; according to the target environmental pollution situation, historical environmental pollution risk assessment results, public health data and parameter values ​​of the organism response parameters, an environmental pollution risk assessment model is used to assess the overall impact of the environmental pollution situation in the target area on public health and the ecosystem to obtain the target environmental pollution risk assessment result; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a double-layer Bayesian network.

[0016] The multi-scale multi-factor coupling analysis method in the embodiment of the present application can realize analysis at different time and space scales through multi-scale analysis, capture the multi-scale change trend of environmental pollution, and enable the environmental pollution risk assessment model to better simulate the complexity and uncertainty of the real world by coupling multiple factors, and improve the reliability of the target environmental pollution risk assessment results. The multi-layer perceptron can capture the complex relationship in the input data such as the target environmental pollution situation, historical environmental pollution risk assessment results, public health data and parameter values ​​of the biological response parameters through multi-layer nonlinear transformation, improve the expression ability of the environmental pollution risk assessment model, and enable the environmental pollution risk assessment model to more accurately assess the environmental pollution risk. The adversarial generative network can generate new data samples, increase the diversity and quantity of training data, prevent overfitting, and improve the generalization ability of the environmental pollution risk assessment model. The double-layer Bayesian network can evaluate the risks at two levels respectively by defining the node structure and edge structure of the public health layer and the ecosystem layer, provide a more comprehensive target environmental pollution risk assessment result, and quantify the uncertainty of the assessment result. The attention mechanism dynamically adjusts the importance weights of key parameters, so that the environmental pollution risk assessment model can flexibly adjust the weights according to changes in environmental conditions, and improve the adaptability of the environmental pollution risk assessment model.

[0017] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 A flowchart of an environmental pollution risk assessment method based on deep learning provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of an environmental pollution risk assessment system based on deep learning provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution 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.

[0021] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The sequence numbers of the operations, such as 11, 12, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to different types.

[0022] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0023] Figure 1 A flowchart of a method for environmental pollution risk assessment based on deep learning is provided in an embodiment of the present application, such as Figure 1 As shown, the method includes: S11. Obtain parameter values, background information, historical environmental pollution risk assessment results and public health data of multiple environmental parameters in the target area; the multiple environmental parameters include physical and chemical property parameters and organism response parameters.

[0024] Among them, the physical and chemical property parameters include the concentration, emission, emission frequency, solubility, volatility, etc. of pollutants. These parameters reflect the physical and chemical properties of pollutants. Organism response parameters: including the physiological response, growth and development status, survival rate, etc. of organisms. These parameters reflect the impact of pollutants on organisms. Background information includes geographic information, meteorological information, land use type and population density. Among them, geographic information includes latitude and longitude, altitude, topography, etc. Meteorological information includes temperature, humidity, wind speed, rainfall, etc. Land use types include industrial areas, residential areas, agricultural areas, etc. Population density includes the population size and distribution of the target area. The historical environmental pollution risk assessment results may include historical data and historical assessment results, among which historical data is the record of environmental pollution in the target area over a period of time, such as historical data of pollutant concentrations, historical risk assessment reports, etc. Historical assessment results include past risk assessment results, such as historical environmental pollution events and their impact assessments. Public health data include disease incidence, medical records and health survey data. Among them, disease incidence includes the incidence of various diseases in the target area, especially those related to environmental pollution. Medical records include medical records of hospitals and clinics, such as the number of patients, diagnosis results, etc. Health survey data include public health status data collected through questionnaires and other means.

[0025] This embodiment does not specifically limit the manner of acquiring the above various information. For example, this embodiment can use environmental monitoring equipment to collect parameter values ​​of multiple environmental parameters in the target area.

[0026] S12. Based on the parameter values ​​and background information of the physical and chemical property parameters, a multi-scale and multi-factor coupling analysis method is used to determine the preliminary environmental pollution situation in the target area.

[0027] Among them, this embodiment can capture the multi-scale change trend of environmental pollution through analysis of different time and space scales. For example, hourly scale, daily scale, monthly scale, etc. Multi-factor coupling can comprehensively consider the interaction of multiple factors such as physical and chemical property parameters, background information, historical data and public health data to improve the comprehensiveness and accuracy of the assessment. The preliminary environmental pollution situation includes the distribution of pollutants, concentration change trends, and the impact range of pollutants in the target area.

[0028] S13, verifying and correcting the preliminary environmental pollution situation to obtain the target environmental pollution situation. The verification is to verify the accuracy of the preliminary environmental pollution situation by comparing the preliminary environmental pollution situation with the known data or standard range in the target area. The preliminary environmental pollution situation is corrected according to the verification result to obtain the target environmental pollution situation to improve the accuracy of the assessment.

[0029] S14. Based on the target environmental pollution situation, historical environmental pollution risk assessment results, public health data and parameter values ​​of organism response parameters, an environmental pollution risk assessment model is used to assess the overall impact of environmental pollution in the target area on public health and the ecosystem, so as to obtain the target environmental pollution risk assessment results; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a double-layer Bayesian network.

[0030] Among them, the multi-layer perceptron can capture the complex relationships in the input data such as the target environmental pollution situation, historical environmental pollution risk assessment results, public health data and parameter values ​​of biological response parameters through multi-layer nonlinear transformation, improve the expressiveness of the environmental pollution risk assessment model, and enable the environmental pollution risk assessment model to more accurately assess environmental pollution risks. The attention mechanism dynamically adjusts the importance weights of key parameters, so that the environmental pollution risk assessment model can flexibly adjust the weights according to changes in environmental conditions, thereby improving the adaptability of the environmental pollution risk assessment model. The adversarial generative network can generate new data samples, increase the diversity and quantity of training data, prevent overfitting, and improve the generalization ability of the environmental pollution risk assessment model. The two-layer Bayesian network can evaluate the risks at the two levels respectively by defining the node structure and edge structure of the public health layer and the ecosystem layer, provide more comprehensive target environmental pollution risk assessment results, and quantify the uncertainty of the assessment results.

[0031] Exemplarily, when this embodiment uses the above-mentioned deep learning-based environmental pollution risk assessment method to assess the environmental pollution risk of a city industrial park, especially the impact on the health of surrounding residents and the ecological environment, the following steps are performed: The concentration data of pollutants were obtained from environmental monitoring stations. The response data of plants were obtained from the plant park management system around the industrial park. The geographical information of the industrial park was obtained from the geographic information system, the meteorological data from the meteorological station, the land use type from the land use database, and the population density data from the statistical publication platform. The records of environmental pollution events and risk assessment reports in the past few years were obtained from the environmental protection department. The disease incidence data were obtained from hospitals and clinics, and the health survey data were obtained from the health department. The geographical weighted regression method was used to analyze the changing trends of pollutant concentrations at different time scales (hours, days, and months). The multi-scale geometric analysis method was used to extract geographical features at different spatial scales, and the preliminary environmental pollution situation was generated through multimodal fusion technology in combination with physical and chemical property parameters and background information. The preliminary environmental pollution situation was verified based on the known historical data, and the preliminary environmental pollution situation was revised based on the verification results. The environmental pollution risk assessment model including multi-layer perceptron based on deep learning, attention mechanism and adversarial generation network, and double-layer Bayesian network was used to evaluate the overall impact of environmental pollution in the target area on public health and the ecosystem.

[0032] Through the above steps, this embodiment can systematically evaluate the environmental pollution risk of a city's industrial park, especially its impact on the health of surrounding residents and the ecological environment. This method combines multiple technologies such as multi-scale multi-factor coupling analysis, multi-layer perceptron, attention mechanism, adversarial generative network and double-layer Bayesian network to ensure the accuracy, robustness and comprehensiveness of the target environmental pollution risk assessment results.

[0033] In some possible embodiments, the multi-scale multi-factor coupling analysis method includes a geographically weighted regression method, a multi-scale geometric analysis method, and a multi-factor coupling analysis method. In environmental pollution risk assessment, the geographically weighted regression method can be used to analyze the relationship between pollutant concentration and geographical location, thereby identifying pollution hotspots. By performing local regression analysis in each sub-region, the geographically weighted regression method can reveal local spatial heterogeneity and generate preliminary environmental pollution conditions in the sub-region. The multi-scale geometric analysis method can be used to extract geographical features from different scales, such as rivers, roads, buildings, etc., which may affect the diffusion and accumulation of pollutants. By extracting geometric features in each sub-region and performing multi-factor fusion, the multi-scale geometric analysis method can provide a more detailed description of environmental pollution. Multimodal fusion technology can fuse different types of data such as physical and chemical property parameters, background information, and geometric features to generate more comprehensive and accurate multi-factor features. These features can be used for further comprehensive analysis to generate overall environmental pollution conditions.

[0034] Accordingly, S12, based on the parameter values ​​of the physical and chemical property parameters and background information, a multi-scale multi-factor coupling analysis method is used to determine the preliminary environmental pollution situation of the target area, including: Step 121: construct a geographically weighted regression model based on the physicochemical property parameters and background information; in each sub-region, use the geographically weighted regression model to perform local regression analysis to generate a preliminary environmental pollution situation in the sub-region, and integrate the preliminary environmental pollution situations of all sub-regions to generate the first overall environmental pollution situation of the target area.

[0035] Step 122, combining the geographic features in the background information, dividing the target area into multiple sub-areas of different scales; using a multi-scale geometric analysis method to extract geometric features in each sub-area; using multi-modal fusion technology to perform multi-factor fusion of physical and chemical property parameters, background information and geometric features, and generating multi-factor features after fusion; using a comprehensive analysis method to perform a comprehensive analysis of the multi-factor features, and generate a second overall environmental pollution situation of the target area. The comprehensive analysis method can be used to perform a comprehensive analysis of the multi-factor features to generate the overall environmental pollution situation of the target area. These methods may include statistical analysis, machine learning models, expert systems, etc.

[0036] Step 123: Determine the preliminary environmental pollution situation of the target area according to the first overall environmental pollution situation of the target area and the second overall environmental pollution situation of the target area.

[0037] Through these steps, this embodiment can directly generate the preliminary environmental pollution situation of the target area, providing data support for providing a comprehensive and accurate environmental pollution assessment result. This method can not only capture the local spatial heterogeneity, but also comprehensively consider the interaction of multiple factors, thereby providing technical support for subsequently improving the reliability and accuracy of the target environmental pollution assessment results.

[0038] In some possible embodiments, S14, based on the target environmental pollution situation, historical environmental pollution risk assessment results, public health data and parameter values ​​of the organism response parameters, an environmental pollution risk assessment model is used to assess the overall impact of the environmental pollution situation in the target area on public health and the ecosystem to obtain the target environmental pollution risk assessment result; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a two-layer Bayesian network, including: Step 141: extract characteristic information of key parameters from the target environmental pollution situation, historical environmental pollution risk assessment results, public health data, and parameter values ​​of organism response parameters. Such characteristic information may include time series of pollutant concentrations, key indicators of historical risk assessment, changing trends of disease incidence, physiological responses of organisms, etc.

[0039] Step 142: Use a multi-layer perceptron to process the characteristic information of all key parameters to obtain a preliminary risk assessment result. The preliminary risk assessment result provides a preliminary estimate of the impact of environmental pollution on public health and the ecosystem, providing a basis for further optimization and adjustment.

[0040] Step 143: Based on the preliminary risk assessment results, the importance weights of each key parameter are calculated using the attention mechanism, and the importance weights are applied to the feature information of the corresponding key parameters to generate weighted feature information of the key parameters. By calculating the importance weights of each key parameter through the attention mechanism, the weights of these parameters can be dynamically adjusted so that the environmental pollution risk assessment model can better focus on the parameters that have a greater impact on the assessment results. The weighted feature information of the key parameters can more accurately reflect the impact of each parameter on the environmental pollution risk assessment results, thereby improving the accuracy of the environmental pollution risk assessment model.

[0041] Step 144: Generate new data samples using a generative adversarial network based on the weighted feature information of all key parameters, optimize the multilayer perceptron using the new data samples, and process the weighted feature information of all key parameters using the optimized multilayer perceptron to obtain an optimized risk assessment result. It should be understood that the optimized multilayer perceptron can better learn the distribution characteristics of the data and improve the accuracy of environmental pollution risk assessment.

[0042] Step 145, define a first node structure and a first edge structure corresponding to the public health layer and a second node structure and a second edge structure corresponding to the ecosystem layer in the double-layer Bayesian network; perform an environmental pollution risk assessment on the public health layer based on the first node structure and the first edge structure to obtain a first reference risk assessment result; perform an environmental pollution risk assessment on the ecosystem layer based on the second node structure and the second edge structure to obtain a second reference risk assessment result.

[0043] It should be understood that the first node structure and the first edge structure are used to define the nodes (such as disease incidence, medical records, etc.) and edges (such as causality, correlation, etc.) of the public health layer to conduct risk assessment at the public health layer; the second node structure and the second edge structure are used to define the nodes (such as organism response, ecological diversity, etc.) and edges (such as food chain relationships, ecological balance, etc.) of the ecosystem layer to conduct risk assessment at the ecosystem layer. The first reference risk assessment results provide a detailed assessment of the impact of environmental pollution on public health, providing a reference for comprehensive assessment. The second reference risk assessment results provide a detailed assessment of the impact of environmental pollution on the ecosystem, providing a reference for comprehensive assessment.

[0044] Step 146: Combine the first reference risk assessment result and the second reference risk assessment result to adjust the optimized risk assessment result to obtain a target environmental pollution risk assessment result.

[0045] Through the above steps, this embodiment can comprehensively and accurately assess the overall impact of environmental pollution in the target area on public health and the ecosystem, and generate reliable target environmental pollution risk assessment results. This method combines multiple technologies such as multi-layer perceptron, attention mechanism, adversarial generative network and double-layer Bayesian network, which significantly improves the accuracy and reliability of target environmental pollution risk assessment results.

[0046] In the above embodiment, as a possible implementation, step 142, using a multi-layer perceptron to process the characteristic information of all key parameters to obtain a preliminary risk assessment result, includes: Step a1: Perform nonlinear mapping on the feature information of the key parameters to obtain a preliminary score for each key parameter. The nonlinear mapping can convert the feature information of the key parameters into a more expressive form, so that the model can better capture the nonlinear relationship between these features.

[0047] Step a2: introduce a normalization processing algorithm to normalize the preliminary scores of all key parameters to obtain standardized preliminary scores. The normalization processing can standardize the preliminary scores of key parameters of different dimensions to ensure that each parameter has the same weight and influence in subsequent calculations.

[0048] Step a3: Introduce a weighted average algorithm to comprehensively calculate all standardized preliminary scores to obtain preliminary risk assessment results. The weighted average algorithm can be used to comprehensively calculate all standardized preliminary scores to generate preliminary risk assessment results. By assigning different weights to different parameters, the impact of each parameter on the risk assessment results can be more accurately reflected.

[0049] Through the steps in the above process, this embodiment can generate a comprehensive and accurate preliminary risk assessment result, providing a reliable basis for environmental pollution risk assessment. This method can not only capture the nonlinear relationship between key parameters, but also improve the stability and accuracy of the assessment results through standardization and weighted average processing.

[0050] In the above embodiment, as a possible implementation, step 143, using the attention mechanism to calculate the importance weight of each key parameter based on the preliminary risk assessment result, includes: Step b1: Analyze the contextual relevance of each key parameter and determine the change trend of each key parameter. Contextual relevance analysis can be used to identify the change trend of key parameters under different time and space conditions, so as to better understand the impact of these parameters on environmental pollution risks.

[0051] Step b2: Introduce a dynamic weight allocation algorithm to dynamically adjust the initial weight of each key parameter according to the changing trend and contextual relevance of the key parameters, and generate a dynamically adjusted preliminary weight; divide the preliminary risk assessment results into multiple subspaces, calculate the local importance score of each key parameter in the subspace, aggregate the local importance scores of the key parameters in all subspaces, and obtain the global importance score of the key parameters.

[0052] Among them, the dynamic weight allocation algorithm can adopt the following formula: ; in, For the The initial weights after adjustment of key parameters; For the The initial weights of the key parameters; For the The changing trend of key parameters; For the contextual relevance of key parameters; is a balancing factor, which is used to adjust the relative weights of change trends and contextual relevance; is the total number of key parameters.

[0053] The dynamic weight allocation algorithm can dynamically adjust the initial weight of each key parameter according to the changing trend and contextual relevance of the key parameters, so that its importance under different environmental conditions can be reasonably reflected. Dividing the preliminary risk assessment results into multiple subspaces can more carefully analyze the local importance of each key parameter in different subspaces, thereby improving the accuracy of the assessment results. By calculating the local importance score of each key parameter in different subspaces, the importance of these parameters under different environmental conditions can be more accurately reflected. By aggregating the local importance scores of all subspaces, the global importance scores of key parameters can be obtained, thereby more comprehensively evaluating the importance of these parameters.

[0054] Step b3: Fusing the preliminary weights with the global importance scores to obtain the comprehensive importance scores of the key parameters. By fusing the preliminary weights with the global importance scores, a more accurate comprehensive importance score of the key parameters can be obtained, thereby better reflecting the importance of these parameters in risk assessment.

[0055] Step b4: Standardize the comprehensive importance scores of all key parameters to obtain the importance weights of all key parameters.

[0056] Through the above steps, this embodiment can generate a comprehensive and accurate key parameter importance weight, providing a reliable basis for environmental pollution risk assessment. This method can not only dynamically adjust the weight of key parameters, but also improve the accuracy and robustness of environmental pollution assessment results through multi-scale analysis and local importance scoring.

[0057] In the above embodiment, as a possible implementation manner, in step b2, the local importance scores of the key parameters in all subspaces are aggregated to obtain the global importance score of the key parameters, including: Step b21, evaluate the importance of each subspace, obtain the subspace evaluation result, and determine the subspace initial weight according to the subspace evaluation result. The subspace evaluation result reflects the importance of each subspace in the overall evaluation. The subspace evaluation result can be used to determine the initial weight of each subspace, so as to give different weights to different subspaces in the subsequent aggregation process. The subspace initial weight can ensure that the importance of each subspace is reasonably reflected in the aggregation process.

[0058] Step b22: Adopt an adaptive adjustment algorithm to dynamically adjust the weight of each subspace to obtain the adjusted weight of the subspace. The adaptive adjustment algorithm dynamically adjusts the weight of each subspace to improve the accuracy and robustness of the evaluation result.

[0059] Among them, the adaptive adjustment algorithm can adopt the following formula: ; in, For the The weights of the subspaces after adjustment; For the The initial weights of the subspaces; is the adjustment factor, which is used to adjust the effect of the coefficient of variation on the weight; For the The coefficient of variation of the subspace; is the total number of subspaces.

[0060] Step b23: multiply the local importance score of the key parameter in each subspace by the weight adjusted for the corresponding subspace as the weighted local importance score of the subspace. The weighted local importance score can more accurately reflect the importance of the key parameter in different subspaces, thereby improving the accuracy of the global importance score.

[0061] Step b24: sum the weighted local importance scores of all subspaces to obtain the global importance score of the key parameter. The global importance score can fully reflect the importance of the key parameter in the overall assessment and provide a reliable basis for the final risk assessment.

[0062] Through these steps, this embodiment can generate a comprehensive and accurate global importance score of key parameters, providing a reliable basis for environmental pollution risk assessment. This method can not only dynamically adjust the weight of the subspace, but also improve the accuracy and robustness of environmental pollution risk assessment results through multi-scale analysis and local importance scoring.

[0063] In the above embodiment, as a possible implementation method, step 146, combining the first reference risk assessment result and the second reference risk assessment result, adjusting the optimized risk assessment result to obtain the target environmental pollution risk assessment result, includes: Step c1: extracting shared feature representations of the first reference risk assessment result and the second reference risk assessment result. The shared feature representations can be extracted from the first reference risk assessment result and the second reference risk assessment result to discover common features between the two results, thereby improving the accuracy of the fusion result.

[0064] Step c2: Fusing the shared feature representation, the first reference risk assessment result, and the second reference risk assessment result to obtain a fusion result. The fusion result can integrate the first reference risk assessment result and the second reference risk assessment result to generate a more comprehensive risk assessment result.

[0065] Step c3, obtain the environmental recovery potential index of the target area, and use the environmental recovery potential index and the fusion result as residuals to adjust the optimized risk assessment results to obtain the target environmental pollution risk assessment results. The environmental recovery potential index refers to an indicator that assesses the ability of the environment to return to its normal state through natural recovery or human intervention measures after being polluted. The environmental recovery potential index can be used to evaluate the self-repair ability of the environment and the effectiveness of management measures, providing additional information for risk assessment. The residuals can be used to measure the prediction errors of the model, and these errors can be reduced by adjusting the model to improve the accuracy of the environmental pollution risk assessment results.

[0066] Through these steps, this embodiment can generate a comprehensive and accurate target environmental pollution risk assessment result. This method can not only integrate information from multiple data sources, but also provide additional assessment information through the environmental restoration potential index, thereby improving the accuracy and reliability of the target environmental pollution risk assessment result.

[0067] Figure 2A schematic diagram of the structure of an environmental pollution risk assessment system based on deep learning provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system includes: The acquisition module 21 is used to obtain parameter values, background information, historical environmental pollution risk assessment results and public health data of multiple environmental parameters in the target area; the multiple environmental parameters include physical and chemical property parameters and organism response parameters.

[0068] The determination module 22 is used to determine the preliminary environmental pollution situation of the target area by adopting a multi-scale multi-factor coupling analysis method based on the parameter values ​​of the physical and chemical property parameters and background information.

[0069] The processing module 23 is used to verify and correct the preliminary environmental pollution situation to obtain the target environmental pollution situation.

[0070] The evaluation module 24 is used to evaluate the overall impact of the environmental pollution situation in the target area on public health and the ecosystem based on the target environmental pollution situation, historical environmental pollution risk assessment results, public health data and parameter values ​​of the organism response parameters, using an environmental pollution risk assessment model to obtain the target environmental pollution risk assessment results; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a double-layer Bayesian network.

[0071] Figure 2 The environmental pollution risk assessment system based on deep learning can perform Figure 1 The implementation principle and technical effect of the deep learning-based environmental pollution risk assessment method described in the illustrated embodiment will not be repeated. The specific manner in which each module and unit performs operations in the deep learning-based environmental pollution risk assessment system in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0072] In one possible design, Figure 2 The deep learning-based environmental pollution risk assessment system of the illustrated embodiment can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32 .

[0073] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0074] The processing component 32 is used to: obtain parameter values, background information, historical environmental pollution risk assessment results and public health data of multivariate environmental parameters in the target area; the multivariate environmental parameters include physicochemical property parameters and organism response parameters; based on the parameter values ​​and background information of the physicochemical property parameters, a multi-scale multi-factor coupling analysis method is used to determine the preliminary environmental pollution situation in the target area; the preliminary environmental pollution situation is verified and corrected to obtain the target environmental pollution situation; based on the target environmental pollution situation, historical environmental pollution risk assessment results, public health data and parameter values ​​of organism response parameters, an environmental pollution risk assessment model is used to assess the overall impact of the environmental pollution situation in the target area on public health and the ecosystem to obtain the target environmental pollution risk assessment result; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a double-layer Bayesian network.

[0075] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0076] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0077] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0078] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0079] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0080] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0081] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The environmental pollution risk assessment method based on deep learning in the illustrated embodiment.

[0082] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0083] The system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0084] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for environmental pollution risk assessment based on deep learning, characterized in that: include: Obtain parameter values, background information, historical environmental pollution risk assessment results and public health data of multiple environmental parameters in the target area; The multivariate environmental parameters include physical and chemical property parameters and organism response parameters; Based on the parameter values ​​of the physical and chemical property parameters and the background information, a multi-scale multi-factor coupling analysis method is used to determine the preliminary environmental pollution situation of the target area; Verifying and correcting the preliminary environmental pollution situation to obtain the target environmental pollution situation; According to the target environmental pollution situation, the historical environmental pollution risk assessment results, the public health data and the parameter values ​​of the organism response parameters, an environmental pollution risk assessment model is used to assess the overall impact of the environmental pollution situation in the target area on public health and the ecosystem, so as to obtain the target environmental pollution risk assessment results; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a double-layer Bayesian network.

2. The method according to claim 1, characterized in that The environmental pollution risk assessment model is used to assess the overall impact of the environmental pollution in the target area on public health and the ecosystem based on the target environmental pollution situation, the historical environmental pollution risk assessment results, the public health data and the parameter values ​​of the organism response parameters, so as to obtain the target environmental pollution risk assessment results; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a two-layer Bayesian network, including: Extracting characteristic information of key parameters from the target environmental pollution situation, the historical environmental pollution risk assessment results, the public health data, and the parameter values ​​of the organism response parameters respectively; A multi-layer perceptron is used to process the characteristic information of all key parameters to obtain preliminary risk assessment results; Calculating the importance weight of each key parameter using an attention mechanism based on the preliminary risk assessment result, and applying the importance weight to the feature information of the corresponding key parameter to generate weighted feature information of the key parameter; Based on the weighted feature information of all key parameters, a new data sample is generated by using a generative adversarial network, the new data sample is used to optimize a multilayer perceptron, and the weighted feature information of all key parameters is processed by using the optimized multilayer perceptron to obtain an optimized risk assessment result; Define a first node structure and a first edge structure corresponding to the public health layer and a second node structure and a second edge structure corresponding to the ecosystem layer in a two-layer Bayesian network; perform an environmental pollution risk assessment on the public health layer based on the first node structure and the first edge structure to obtain a first reference risk assessment result; perform an environmental pollution risk assessment on the ecosystem layer based on the second node structure and the second edge structure to obtain a second reference risk assessment result; The optimized risk assessment result is adjusted in combination with the first reference risk assessment result and the second reference risk assessment result to obtain a target environmental pollution risk assessment result.

3. The method according to claim 2, characterized in that The multi-layer perceptron is used to process the characteristic information of all key parameters to obtain preliminary risk assessment results, including: Performing nonlinear mapping processing on the characteristic information of the key parameters to obtain a preliminary score for each key parameter; A normalization processing algorithm is introduced to standardize the preliminary scores of all key parameters to obtain standardized preliminary scores; A weighted average algorithm is introduced to comprehensively calculate all standardized preliminary scores to obtain preliminary risk assessment results.

4. The method according to claim 2, characterized in that: The attention mechanism is used to calculate the importance weight of each key parameter based on the preliminary risk assessment result, including: Conduct contextual relevance analysis on each key parameter and determine the change trend of each key parameter; A dynamic weight allocation algorithm is introduced to dynamically adjust the initial weight of each key parameter according to the change trend and context relevance of the key parameter to generate a dynamically adjusted preliminary weight; the preliminary risk assessment result is divided into multiple subspaces, the local importance score of each key parameter in the subspace is calculated, and the local importance scores of the key parameter in all the subspaces are aggregated to obtain the global importance score of the key parameter; Merging the preliminary weight with the global importance score to obtain a comprehensive importance score of the key parameter; The comprehensive importance scores of all key parameters are standardized to obtain the importance weights of all key parameters.

5. The method according to claim 4, characterized in that The step of aggregating the local importance scores of the key parameters in all the subspaces to obtain the global importance scores of the key parameters includes: Evaluate the importance of each subspace to obtain a subspace evaluation result, and determine the subspace initial weight according to the subspace evaluation result; Adopting an adaptive adjustment algorithm to dynamically adjust the weight of each subspace, and obtaining the adjusted weight of the subspace; The product of the local importance score of the key parameter in each subspace and the adjusted weight of the corresponding subspace is used as the weighted local importance score of the subspace; The weighted local importance scores of all subspaces are summed to obtain the global importance score of the key parameters.

6. The method according to claim 2, characterized in that The step of combining the first reference risk assessment result and the second reference risk assessment result to adjust the optimized risk assessment result to obtain a target environmental pollution risk assessment result includes: extracting a shared feature representation of the first reference risk assessment result and the second reference risk assessment result; fusing the shared feature representation, the first reference risk assessment result, and the second reference risk assessment result to obtain a fusion result; The environmental restoration potential index of the target area is obtained, and the environmental restoration potential index and the fusion result are used as residuals to adjust the optimized risk assessment result to obtain the target environmental pollution risk assessment result.

7. The method according to claim 1, characterized in that The multi-scale multi-factor coupling analysis method includes a geographically weighted regression method, a multi-scale geometric analysis method and a multi-factor coupling analysis method; The method of determining the preliminary environmental pollution situation of the target area by using a multi-scale multi-factor coupling analysis method based on the parameter values ​​of the physical and chemical property parameters and the background information includes: Based on the physical and chemical property parameters and background information, a geographically weighted regression model is constructed; in each sub-region, a local regression analysis is performed using the geographically weighted regression model to generate a preliminary environmental pollution situation in the sub-region, and the preliminary environmental pollution situations of all sub-regions are integrated to generate the first overall environmental pollution situation of the target region; Combined with the geographical features in the background information, the target area is divided into multiple sub-areas of different scales; a multi-scale geometric analysis method is used to extract geometric features in each sub-area; a multi-modal fusion technology is used to perform multi-factor fusion on the physical and chemical property parameters, the background information and the geometric features to obtain multi-factor features after fusion; a comprehensive analysis method is used to perform a comprehensive analysis on the multi-factor features to generate a second overall environmental pollution situation of the target area; The preliminary environmental pollution situation of the target area is determined according to the first overall environmental pollution situation of the target area and the second overall environmental pollution situation of the target area.

8. A deep learning-based environmental pollution risk assessment system, characterized in that: include: An acquisition module, used to obtain parameter values, background information, historical environmental pollution risk assessment results and public health data of multiple environmental parameters in the target area; The multivariate environmental parameters include physical and chemical property parameters and organism response parameters; A determination module, for determining the preliminary environmental pollution situation of the target area by adopting a multi-scale multi-factor coupling analysis method based on the parameter values ​​of the physicochemical property parameters and the background information; A processing module, used to verify and correct the preliminary environmental pollution situation to obtain the target environmental pollution situation; The evaluation module is used to evaluate the overall impact of the environmental pollution situation in the target area on public health and the ecosystem based on the target environmental pollution situation, the historical environmental pollution risk assessment results, the public health data and the parameter values ​​of the organism response parameters using an environmental pollution risk assessment model to obtain a target environmental pollution risk assessment result; the environmental pollution risk assessment model is an integrated model including a multi-layer perceptron based on deep learning, an attention mechanism and a generative adversarial network, and a double-layer Bayesian network.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an environmental pollution risk assessment method based on deep learning as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, an environmental pollution risk assessment method based on deep learning as described in any one of claims 1 to 7 is implemented.

Images