Method and system for determining degree of NAPL contamination in groundwater

Abstract

Disclosed in the present application are a method and system for determining the degree of NAPL contamination in groundwater on the basis of image learning. The method comprises: acquiring water body image information in a groundwater monitoring well, preprocessing the water body image information, and extracting image features from the water body image information; acquiring substance types and concentration information of NAPL substances in groundwater according to monitoring data of the groundwater monitoring well within a preset time; matching the image features with the substance types and the concentration information, and constructing a feature data set, which is conductive to constructing a data information base for NAPL contamination in the groundwater by combining the feature data set with local hydrogeological information; and on the basis of deep learning, constructing a model for determining the degree of NAPL contamination in the groundwater, using data in the data information base to perform training, determining the degree of NAPL contamination in the groundwater, and generating a contamination early warning according to the degree of contamination. In the present application, a correlation between water body image information and substance types and concentration levels is established, such that the degree of NAPL contamination is rapidly determined, thereby providing data support for contamination early warning and contamination treatment in preset areas.

Description

A method and system for determining the degree of NAPL pollution in groundwater

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to a Chinese patent application filed with the Patent Office of China on June 2, 2023, with application number 202310645177.6 and invention name “A method and system for determining the degree of NAPL pollution in groundwater based on image learning”, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present application relates to the field of groundwater pollution technology, and more specifically, to a method and system for determining the degree of NAPL pollution in groundwater based on image learning. Background Art

[0004] Groundwater pollution must be resolved to ensure water safety for urban residents. Groundwater pollution assessment is an important part of groundwater pollution investigation and remediation, providing a basis for groundwater pollution prevention and control in the study area.

[0005] After entering the soil aquifer, non-aqueous phase liquids (NAPLs) cannot mix with water and instead exist in the soil and groundwater in liquid and gaseous forms. The main NAPL substances include light non-aqueous phase liquids (LNAPLs) such as petroleum hydrocarbons and benzene series, and heavy non-aqueous phase liquids (DNAPLs) such as chlorinated hydrocarbons and polychlorinated biphenyls. After NAPL substances leak into the soil, they will migrate vertically into the groundwater and diffuse radially with the flow of groundwater, forming a contamination plume. The hydrophobicity of NAPLs in contaminated sites, the differences in the migration characteristics of complex pollutants, the heterogeneity of the strata, and the frequent interaction between surface water and groundwater all increase the judgment of the degree of NAPL contamination in groundwater. Therefore, how to accurately determine and provide early warning of the degree of NAPL contamination in groundwater is an urgent problem that needs to be solved.

[0006] Summary of the Invention

[0007] In order to solve the above technical problems, this application proposes a method and system for determining the degree of NAPL contamination in groundwater based on image learning.

[0008] The first aspect of the present application provides a method for determining the degree of NAPL contamination in groundwater based on image learning, comprising:

[0009] Obtaining water body image information from groundwater monitoring wells in a preset area, preprocessing the water body image information, and extracting image features of the water body image information;

[0010] Obtain the type and concentration information of NAPL substances in groundwater based on the monitoring data of groundwater monitoring wells within a preset time;

[0011] Matching the image features with the substance type and concentration information to construct a feature data set, which is conducive to combining the feature data set with local hydrogeological information to construct a data information database of groundwater NAPL pollution;

[0012] A groundwater NAPL pollution degree determination model is constructed based on deep learning. The model is trained using data in the database to determine the pollution degree of groundwater NAPL and generate pollution warnings based on the pollution degree.

[0013] In this solution, the image features of water body image information are extracted, specifically:

[0014] Acquire water body image information according to preset time intervals through a preset image sensor device in a groundwater monitoring well, perform filtering, denoising and image enhancement on the water body image information, and divide the filtered, denoising and image enhancement into a plurality of image blocks;

[0015] Calculating the information entropy of the image block, generating a mask image of the water body image information based on the information entropy, and performing image segmentation using the mask image to obtain preprocessed water body image information;

[0016] The preprocessed water body image information is imported into the CSPNET network for feature representation to obtain the feature map. The hole feature pyramid module is introduced and hole convolution with different hole ratios is used to extract multi-scale features and obtain the multi-scale feature map.

[0017] The multi-scale feature map is converted into a channel description through average pooling, the channel attention weight of the multi-scale feature map is obtained through two convolutional layers and sigmoid and ReLU functions, and the channel attention weight is multiplied by the multi-scale feature map to obtain an attention map;

[0018] The image features of the water body image information are obtained by element-wise multiplication of the attention map and the feature map.

[0019] In this solution, the characteristic dataset is combined with local hydrogeological information to construct a NAPL-contaminated groundwater data database. Specifically:

[0020] Read the historical monitoring data of groundwater monitoring wells in the preset area, judge the effectiveness of groundwater monitoring wells based on the historical monitoring data, and eliminate invalid groundwater monitoring wells;

[0021] Acquire monitoring data from groundwater monitoring wells valid within a preset time, clean the monitoring data, obtain the type and concentration information of NAPL-like substances in the groundwater, intersect image features with the type and concentration information of NAPL-contaminated substances, and construct a feature data set;

[0022] Obtain the corresponding groundwater flow direction, flow velocity, surface water-groundwater interaction frequency, and stratigraphic structure information in the preset area to construct a hydrogeological dataset and combine it with the characteristic dataset;

[0023] An information database on NAPL pollution in groundwater reservoirs is established, and the combined characteristic data is structured and stored in the information database. The information database is used to manage the data on NAPL pollution in groundwater in a preset area and to update and optimize the model for determining the degree of NAPL pollution in groundwater.

[0024] In this solution, a groundwater NAPL pollution degree determination model is constructed based on deep learning and trained using data from the database. Specifically:

[0025] A groundwater NAPL pollution degree determination model was constructed based on deep learning methods, and feature datasets were extracted from the information database.

[0026] Using big data, we obtain the pollution degree judgment criteria for each substance type corresponding to NAPL pollution in the preset search space, and select the judgment criteria with the highest frequency in the test data to build a pollution degree judgment system for each substance type.

[0027] Determine the pollution degree based on the pollution degree determination system for each substance type and the data in the corresponding feature data set of the groundwater monitoring well monitoring data, and set the pollution degree data label for the corresponding image feature;

[0028] The groundwater NAPL pollution degree determination model is trained using the data-labeled feature dataset. When the number of training times meets the preset number of times, the trained groundwater NAPL pollution degree determination model is output;

[0029] The image features of the current water body image information of each groundwater monitoring well and the geographical hydrological information of the preset area are used as the input of the groundwater NAPL pollution degree determination model to identify the substance type and concentration information corresponding to the image features of each groundwater monitoring well;

[0030] Obtain the current spatial distribution of NAPL-like substances in groundwater in a preset area, and obtain the pollution degree of NAPL-like pollution in groundwater in each area of ​​the preset area.

[0031] In this solution, a pollution warning is generated according to the pollution level, which includes:

[0032] Divide the preset area into several sub-areas based on the distribution of groundwater monitoring wells, obtain the substance type and historical concentration information of NAPL substances in the sub-areas within the preset time, and generate a concentration time series corresponding to the substance type based on the historical concentration information;

[0033] Trace the pollution sources of various substances based on the concentration time series corresponding to the substance types in different sub-areas, mark the sub-areas where the pollution sources are located, and set labels for the sub-areas where the pollution sources are located based on the substance types;

[0034] Extracting concentration changes at preset time intervals based on the concentration time series, and setting different change thresholds for labeled sub-regions and unlabeled sub-regions;

[0035] Determine whether the sub-region is tagged, read the change threshold value based on the determination result, and when the concentration change is greater than the threshold value, if the sub-region is not tagged, obtain the environmental characteristics and climate characteristics of the sub-region as the pollution impact characteristics of the sub-region; if the sub-region is tagged, obtain the environmental characteristics, climate characteristics, and pollution source operation characteristics of the sub-region as the pollution impact characteristics of the sub-region;

[0036] At the same time, in different sub-areas, the average value of the concentration change in the sub-area corresponding to the preset time interval is obtained, and the average value of the concentration change is used as the pollution change benchmark value in the sub-area within the preset time.

[0037] In this solution, a pollution warning is generated according to the pollution level, specifically:

[0038] Based on the current spatial distribution of NAPL substances in groundwater in the preset area, the current substance type and current concentration information of different sub-areas are obtained and compared with the historical substance type and concentration information corresponding to the previous time step;

[0039] When the type of NAPL substance in groundwater in a sub-region changes or the concentration change is greater than the pollution change benchmark value, early warning information in the sub-region is generated, and the early warning information is visualized in combination with the spatial distribution;

[0040] Based on whether the sub-region is labeled, the environmental characteristics and climate characteristics, or the environmental characteristics, climate characteristics, and pollution source operation characteristics of the sub-region within a preset time period in the future are obtained, and the obtained characteristics are compared with the pollution impact characteristics of the sub-region using Manhattan distance for similarity calculation;

[0041] When the Manhattan distance is less than or equal to a preset distance threshold, it is proved that the similarity meets the preset requirements, and a pollution warning for the sub-region is generated. Based on the pollution warning, emergency measures for the pollution of NAPL substances in groundwater in the sub-region are obtained using big data means.

[0042] A second aspect of the present application further provides a system for determining the degree of NAPL contamination in groundwater based on image learning. The system includes: a memory and a processor. The memory includes a program for determining the degree of NAPL contamination in groundwater based on image learning. When the program is executed by the processor, the following steps are implemented:

[0043] Obtaining water body image information from groundwater monitoring wells in a preset area, preprocessing the water body image information, and extracting image features of the water body image information;

[0044] Obtain the type and concentration information of NAPL substances in groundwater based on the monitoring data of groundwater monitoring wells within a preset time;

[0045] Matching the image features with the substance type and concentration information to construct a feature data set, which is conducive to combining the feature data set with local hydrogeological information to construct a data information database of groundwater NAPL pollution;

[0046] A groundwater NAPL pollution degree determination model is constructed based on deep learning. The model is trained using data in the database to determine the pollution degree of groundwater NAPL and generate pollution warnings based on the pollution degree.

[0047] This application discloses a method and system for determining the degree of NAPL contamination in groundwater based on image learning, including: obtaining water image information from groundwater monitoring wells, performing preprocessing, and extracting image features of the water image information; obtaining the substance type and concentration information of NAPL-like substances in the groundwater based on monitoring data from the groundwater monitoring wells within a preset time; matching the image features with the substance type and concentration information to construct a feature dataset, which facilitates the construction of a data information database of NAPL-contaminated groundwater by combining the historical feature dataset with local hydrogeological information; constructing a groundwater NAPL contamination degree determination model based on deep learning, using data from the data information database for training, determining the degree of groundwater NAPL contamination, and generating a pollution warning based on the degree of contamination. By establishing a correlation between water image information and substance type and concentration level, this application achieves rapid determination of the degree of NAPL contamination, providing data support for pollution warning and pollution control within a preset area. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] FIG1 shows a flow chart of a method for determining the degree of NAPL contamination in groundwater based on image learning according to the present application;

[0049] FIG2 shows a flow chart of the method for constructing a groundwater NAPL pollution degree determination model in the present application;

[0050] FIG3 shows a flow chart of a method for generating a pollution warning based on pollution levels in the present application;

[0051] FIG4 shows a block diagram of a groundwater NAPL pollution degree determination system based on image learning in the present application. DETAILED DESCRIPTION

[0052] In order to more clearly understand the above-mentioned objects, features and advantages of the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features therein can be combined with each other in the absence of conflict.

[0053] In the following description, many specific details are set forth to facilitate a full understanding of the present application. However, the present application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present application is not limited to the specific embodiments disclosed below.

[0054] FIG1 shows a flow chart of a method for determining the degree of NAPL contamination in groundwater based on image learning in the present application.

[0055] As shown in FIG1 , the first aspect of the present application provides a method for determining the degree of NAPL contamination in groundwater based on image learning, comprising:

[0056] S102, obtaining water body image information from groundwater monitoring wells in a preset area, preprocessing the water body image information, and extracting image features of the water body image information;

[0057] S104, obtaining the type and concentration information of NAPL substances in groundwater based on the monitoring data of the groundwater monitoring well within a preset time;

[0058] S106, matching the image features with the substance type and concentration information to construct a feature data set, which is conducive to combining the feature data set with local hydrogeological information to construct a data information database of groundwater NAPL pollution;

[0059] S108, constructing a groundwater NAPL pollution degree determination model based on deep learning, using data in the data information database for training, determining the groundwater NAPL pollution degree, and generating a pollution warning based on the pollution degree.

[0060] It should be noted that water body image information is obtained according to preset time intervals through a preset image sensor device in a groundwater monitoring well, and the water body image information is filtered, denoised and image enhanced. Optionally, the water body image information is enhanced through a generative adversarial network. In the generative adversarial network, the original water body image information is input into the generator, and after a series of operations, the enhanced image is output, the difference between the output image and the reference image is obtained, and the parameters of the generator are adjusted. The results are fed back to the generator to continuously improve the generator, and the corresponding discriminator further adjusts the discrimination ability to eventually achieve a balance, and the water body image information is enhanced through the generative adversarial network.

[0061] The filtering denoising and image enhancement are divided into several image blocks; the information entropy of the image blocks is calculated, the feature abundance in the image is represented by the information entropy, a mask image of the water body image information is generated according to the information entropy, and the mask image is used to perform image segmentation to obtain the preprocessed water body image information; the preprocessed water body image information is imported into the CSPNET network for feature representation to obtain a feature map; the CSPNET network can obtain richer and more powerful feature representations while reducing the amount of calculation, introduce a void feature pyramid module, and use void convolution with different void rates to perform multi-scale feature extraction to obtain a multi-scale feature map. The introduction of void convolution can effectively increase the receptive field to include more context information without increasing the amount of calculation; the multi-scale feature map is converted into a channel description f through average pooling. MN represents the size information of the feature map, T c (i,j) represents the cth channel T c At the value of point (i, j), the channel attention weight q of the multi-scale feature map is obtained through two convolutional layers and sigmoid and Relu functions, q = Sigmoid(conv(Relu(conv(f)))), conv represents convolution calculation, and the channel attention weight is multiplied by the multi-scale feature map to obtain the attention map; the image features of the water body image information are obtained by element-wise multiplication of the attention map and the feature map.

[0062] FIG2 shows a flow chart of the method for constructing a groundwater NAPL pollution degree determination model in the present application.

[0063] According to an embodiment of the present application, a groundwater NAPL pollution degree determination model is constructed based on deep learning and trained using data in a data information database. Specifically:

[0064] S202, constructing a groundwater NAPL pollution degree determination model based on a deep learning method and extracting a feature data set from an information database;

[0065] S204, using big data to obtain the pollution degree determination criteria for each substance type corresponding to NAPL pollution in a preset search space, and selecting the determination criteria with the highest frequency in the test data to construct a pollution degree determination system for each substance type;

[0066] S206, determining the pollution degree based on the pollution degree determination system for each substance type and the data in the feature data set corresponding to the monitoring data of the groundwater monitoring well, and setting a data label of the pollution degree for the corresponding image feature;

[0067] S208, using the labeled feature data set to train a groundwater NAPL pollution degree determination model, and when the number of training times meets the preset number of times, outputting the trained groundwater NAPL pollution degree determination model;

[0068] S210, using the image features of the current water body image information of each groundwater monitoring well and the geographical and hydrological information of the preset area as inputs to the groundwater NAPL pollution degree determination model, and identifying the substance type and concentration information corresponding to the image features of each groundwater monitoring well;

[0069] S212, obtaining the current spatial distribution of NAPL-like substances in groundwater in a preset area, and obtaining the pollution degree of NAPL-like pollution in groundwater in each area of ​​the preset area.

[0070] It should be noted that the deep learning method includes a BP neural network or a support vector machine, and the preset search space includes a relevant literature database and an expert experience database. The historical monitoring data of groundwater monitoring wells in a preset area are read, the effectiveness of the groundwater monitoring wells is judged based on the historical monitoring data, and invalid groundwater monitoring wells are eliminated; the monitoring data of groundwater monitoring wells that are valid within a preset time are obtained, the monitoring data are cleaned, the material type and concentration information of NAPL-like substances in the groundwater are obtained, the image features are combined with the material type and concentration information of NAPL pollution, and a feature data set is constructed; the corresponding groundwater flow direction, flow rate, surface water-groundwater interaction frequency, and stratigraphic structure information in the preset area are obtained to construct a hydrogeological data set, which is combined with the feature data set; an information database of NAPL pollution in groundwater reservoirs is established, the combined feature data is structured and stored in the information database, data management of NAPL pollution in the preset area is performed through the information database, and the groundwater NAPL pollution degree determination model is updated and optimized through the continuously updated data set.

[0071] It should be noted that the image features in the feature data set are clustered according to the material category, and the scatter distribution of the image features in the feature space is obtained under the same material category; statistical analysis is performed on the image features under the same material category to obtain the image features with the highest number of occurrences, and similarity calculation is performed based on the image features to obtain a preset number of image features that meet the preset similarity standard, and an image feature set is constructed; the image features in the image feature set are obtained at the corresponding positions of the scatter distribution, and the cluster center points of the image features of the material category are obtained based on the corresponding positions; the distance threshold corresponding to the material category is obtained by the average distance from other points in the scatter distribution to the cluster center point, and the distance between the image features of each image block in the current water body image information and the cluster center point of the image features of a certain material category is compared with the distance threshold corresponding to the material category. If it is less than the distance threshold, it is determined that the water body contains NAPL pollutants of the material category.

[0072] FIG3 shows a flow chart of the method for generating a pollution warning according to the pollution degree in the present application.

[0073] According to an embodiment of the present application, a pollution warning is generated according to the pollution level, specifically:

[0074] S302, based on the current spatial distribution of NAPL substances in groundwater in a preset area, obtain current substance type and current concentration information in different sub-areas, and compare them with historical substance type and concentration information corresponding to the previous time step;

[0075] S304: When the type of NAPL substance in the groundwater in the sub-region changes or the concentration change is greater than the pollution change reference value, early warning information is generated in the sub-region, and the early warning information is visualized in combination with the spatial distribution;

[0076] S306, obtaining environmental characteristics and climate characteristics, or environmental characteristics, climate characteristics, and pollution source operation characteristics of the sub-region within a preset time period in the future, based on whether the sub-region is labeled, and calculating similarity between the obtained characteristics and the pollution impact characteristics of the sub-region using Manhattan distance;

[0077] S308: When the Manhattan distance is less than or equal to a preset distance threshold, it is proved that the similarity meets the preset requirements, and a pollution warning for the sub-region is generated. Based on the pollution warning, emergency measures for the pollution of NAPL substances in groundwater in the sub-region are obtained using big data means.

[0078] It should be noted that the preset area is divided into several sub-areas according to the distribution of groundwater monitoring wells, the substance type and historical concentration information of NAPL substances in the groundwater in the sub-areas within a preset time are obtained, and a concentration time series corresponding to the substance type is generated based on the historical concentration information; the pollution source of each substance is traced based on the concentration time series corresponding to the substance type in different sub-areas, and the sub-area where the pollution source is located is marked, and a label is set for the sub-area where the pollution source is located according to the substance type; based on the concentration time series, the concentration change at a preset time interval is extracted, and different change thresholds are set for the labeled sub-areas and the unlabeled sub-areas; Determine whether the sub-area is labeled, and read the change threshold based on the judgment result. When the concentration change is greater than the threshold, if the sub-area is not labeled, obtain the environmental characteristics and climate characteristics of the sub-area as the pollution impact characteristics of the sub-area. If the sub-area is labeled, obtain the environmental characteristics, climate characteristics and pollution source operation characteristics of the sub-area as the pollution impact characteristics of the sub-area. The pollution source operation characteristics include the factory's production plan and production quantity, etc. At the same time, in different sub-areas, obtain the average value of the concentration change of the sub-area corresponding to the preset time interval, and use the average concentration change as the pollution change baseline value within the preset time of the sub-area.

[0079] FIG4 shows a block diagram of a groundwater NAPL pollution degree determination system based on image learning in the present application.

[0080] A second aspect of the present application further provides a groundwater NAPL pollution degree determination system 4 based on image learning. The system includes: a memory 41 and a processor 42. The memory includes a groundwater NAPL pollution degree determination method program based on image learning. When the groundwater NAPL pollution degree determination method program based on image learning is executed by the processor, the following steps are implemented:

[0081] Obtaining water body image information from groundwater monitoring wells in a preset area, preprocessing the water body image information, and extracting image features of the water body image information;

[0082] Obtain the type and concentration information of NAPL substances in groundwater based on the monitoring data of groundwater monitoring wells within a preset time;

[0083] Matching the image features with the substance type and concentration information to construct a feature data set, which is conducive to combining the feature data set with local hydrogeological information to construct a data information database of groundwater NAPL pollution;

[0084] A groundwater NAPL pollution degree determination model is constructed based on deep learning. The model is trained using data in the database to determine the pollution degree of groundwater NAPL and generate pollution warnings based on the pollution degree.

[0085] It should be noted that water body image information is obtained according to preset time intervals through a preset image sensor device in a groundwater monitoring well, and the water body image information is filtered, denoised and image enhanced. Optionally, the water body image information is enhanced through a generative adversarial network. In the generative adversarial network, the original water body image information is input into the generator, and after a series of operations, the enhanced image is output, the difference between the output image and the reference image is obtained, and the parameters of the generator are adjusted. The results are fed back to the generator to continuously improve the generator, and the corresponding discriminator further adjusts the discrimination ability to eventually achieve a balance, and the water body image information is enhanced through the generative adversarial network.

[0086] The filtering denoising and image enhancement are divided into several image blocks; the information entropy of the image blocks is calculated, the feature abundance in the image is represented by the information entropy, a mask image of the water body image information is generated according to the information entropy, and the mask image is used to perform image segmentation to obtain the preprocessed water body image information; the preprocessed water body image information is imported into the CSPNET network for feature representation to obtain a feature map; the CSPNET network can obtain richer and more powerful feature representations while reducing the amount of calculation, introduce a void feature pyramid module, and use void convolution with different void rates to perform multi-scale feature extraction to obtain a multi-scale feature map. The introduction of void convolution can effectively increase the receptive field to include more context information without increasing the amount of calculation; the multi-scale feature map is converted into a channel description f through average pooling. MN represents the size information of the feature map, T c (i,j) represents the cth channel T c At the value of point (i, j), the channel attention weight q of the multi-scale feature map is obtained through two convolutional layers and sigmoid and Relu functions, q = Sigmoid(conv(Relu(conv(f)))), conv represents convolution calculation, and the channel attention weight is multiplied by the multi-scale feature map to obtain the attention map; the image features of the water body image information are obtained by element-wise multiplication of the attention map and the feature map.

[0087] According to an embodiment of the present application, a groundwater NAPL pollution degree determination model is constructed based on deep learning and trained using data in a data information database. Specifically:

[0088] A groundwater NAPL pollution degree determination model was constructed based on deep learning methods, and feature datasets were extracted from the information database.

[0089] Using big data, we obtain the pollution degree judgment criteria for each substance type corresponding to NAPL pollution in the preset search space, and select the judgment criteria with the highest frequency in the test data to build a pollution degree judgment system for each substance type.

[0090] Determine the pollution degree based on the pollution degree determination system for each substance type and the data in the corresponding feature data set of the groundwater monitoring well monitoring data, and set the pollution degree data label for the corresponding image feature;

[0091] The groundwater NAPL pollution degree determination model is trained using the data-labeled feature dataset. When the number of training times meets the preset number of times, the trained groundwater NAPL pollution degree determination model is output;

[0092] The image features of the current water body image information of each groundwater monitoring well and the geographical hydrological information of the preset area are used as the input of the groundwater NAPL pollution degree determination model to identify the substance type and concentration information corresponding to the image features of each groundwater monitoring well;

[0093] Obtain the current spatial distribution of NAPL-like substances in groundwater in a preset area, and obtain the pollution degree of NAPL-like pollution in groundwater in each area of ​​the preset area.

[0094] It should be noted that the historical monitoring data of groundwater monitoring wells in the preset area are read, the effectiveness of the groundwater monitoring wells is judged based on the historical monitoring data, and invalid groundwater monitoring wells are eliminated; the monitoring data of the groundwater monitoring wells that are valid within the preset time are obtained, the monitoring data are cleaned, the material type and concentration information of NAPL-like substances in the groundwater are obtained, the image features are intersected with the material type and concentration information of NAPL pollution, and a feature data set is constructed; the corresponding groundwater flow direction, flow velocity, surface water-groundwater interaction frequency, and stratum structure information in the preset area are obtained to construct a hydrogeological data set, which is combined with the feature data set; an information database of NAPL pollution in groundwater reservoirs is established, the combined feature data is structured and stored in the information database, data management of NAPL pollution in the preset area is carried out through the information database, and the groundwater NAPL pollution degree determination model is updated and optimized through the continuously updated data set.

[0095] According to an embodiment of the present application, a pollution warning is generated according to the pollution level, specifically:

[0096] Based on the current spatial distribution of NAPL substances in groundwater in the preset area, the current substance type and current concentration information of different sub-areas are obtained and compared with the historical substance type and concentration information corresponding to the previous time step;

[0097] When the type of NAPL substance in groundwater in a sub-region changes or the concentration change is greater than the pollution change benchmark value, early warning information in the sub-region is generated, and the early warning information is visualized in combination with the spatial distribution;

[0098] Based on whether the sub-region is labeled, the environmental characteristics and climate characteristics, or the environmental characteristics, climate characteristics, and pollution source operation characteristics of the sub-region within a preset time period in the future are obtained, and the obtained characteristics are compared with the pollution impact characteristics of the sub-region using Manhattan distance for similarity calculation;

[0099] When the Manhattan distance is less than or equal to a preset distance threshold, it is proved that the similarity meets the preset requirements, and a pollution warning for the sub-region is generated. Based on the pollution warning, emergency measures for the pollution of NAPL substances in groundwater in the sub-region are obtained using big data means.

[0100] It should be noted that the preset area is divided into several sub-areas according to the distribution of groundwater monitoring wells, the substance type and historical concentration information of NAPL substances in the groundwater in the sub-areas within a preset time are obtained, and a concentration time series corresponding to the substance type is generated based on the historical concentration information; the pollution source of each substance is traced based on the concentration time series corresponding to the substance type in different sub-areas, and the sub-area where the pollution source is located is marked, and a label is set for the sub-area where the pollution source is located according to the substance type; based on the concentration time series, the concentration change at a preset time interval is extracted, and different change thresholds are set for the labeled sub-areas and the unlabeled sub-areas; Determine whether the sub-area is labeled, and read the change threshold based on the judgment result. When the concentration change is greater than the threshold, if the sub-area is not labeled, obtain the environmental characteristics and climate characteristics of the sub-area as the pollution impact characteristics of the sub-area. If the sub-area is labeled, obtain the environmental characteristics, climate characteristics and pollution source operation characteristics of the sub-area as the pollution impact characteristics of the sub-area. The pollution source operation characteristics include the factory's production plan and production quantity, etc. At the same time, in different sub-areas, obtain the average value of the concentration change of the sub-area corresponding to the preset time interval, and use the average concentration change as the pollution change baseline value within the preset time of the sub-area.

[0101] The third aspect of the present application also provides a computer-readable storage medium, which includes a method program for determining the degree of groundwater NAPL pollution based on image learning. When the method program for determining the degree of groundwater NAPL pollution based on image learning is executed by a processor, the steps of the method for determining the degree of groundwater NAPL pollution based on image learning as described in any one of the above items are implemented.

[0102] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0103] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0104] In addition, all functional units in the embodiments of the present application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the above-mentioned integrated units can be implemented in the form of hardware or in the form of hardware plus software functional units.

[0105] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0106] Alternatively, if the above-mentioned integrated unit of the present application is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0107] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for determining the degree of NAPL pollution in groundwater based on image learning, characterized in that: The following steps are involved: Acquire water body image information in groundwater monitoring wells in a preset area, preprocess the water body image information, and extract image features of the water body image information; Obtain the type and concentration information of NAPL substances in groundwater based on the monitoring data of groundwater monitoring wells within a preset time; Matching the image features with the substance type and concentration information to construct a feature data set, which is conducive to combining the feature data set with local hydrogeological information to construct a data information database of groundwater NAPL pollution; A groundwater NAPL pollution degree determination model is constructed based on deep learning, and the data in the data information database is used for training to determine the pollution degree of groundwater NAPL, and a pollution warning is generated according to the pollution degree.

2. The method for determining the degree of NAPL contamination in groundwater based on image learning according to claim 1 is characterized in that: The image features of water body image information are extracted, specifically: Acquire water body image information according to preset time intervals through a preset image sensor device in a groundwater monitoring well, filter and de-noise the water body image information and perform image enhancement, and divide the filtering and de-noising and image enhancement into a plurality of image blocks; Calculating the information entropy of the image block, generating a mask image of the water body image information according to the information entropy, and performing image segmentation using the mask image to obtain preprocessed water body image information; The preprocessed water body image information is imported into the CSPNET network for feature characterization to obtain feature maps. The hole feature pyramid module is introduced, and hole convolutions with different hole rates are used to extract multi-scale features to obtain multi-scale feature maps. The multi-scale feature map is converted into a channel description through average pooling, the channel attention weight of the multi-scale feature map is obtained through two convolutional layers and sigmoid and Relu functions, and the channel attention weight is multiplied by the multi-scale feature map to obtain an attention map; The image features of the water body image information are obtained by element-wise multiplication of the attention map and the feature map.

3. The method for determining the degree of NAPL contamination in groundwater based on image learning according to claim 1 is characterized in that: It is beneficial to combine the characteristic data set with the local hydrogeological information to build a data information database of NAPL contaminated groundwater, specifically: Read the historical monitoring data of groundwater monitoring wells in the preset area, judge the effectiveness of groundwater monitoring wells based on the historical monitoring data, and eliminate invalid groundwater monitoring wells; Acquire monitoring data of groundwater monitoring wells valid within a preset time, clean the monitoring data, obtain the material type and concentration information of NAPL-like substances in groundwater, intersect image features with the material type and concentration information of NAPL pollution, and construct a feature data set; Obtain the corresponding groundwater flow direction, flow velocity, surface water-groundwater interaction frequency, and stratigraphic structure information in the preset area to construct a hydrogeological data set and combine it with the characteristic data set; An information database of NAPL pollution in groundwater reservoirs is established, and the combined characteristic data is structured and stored in the information database. The data management of NAPL pollution in groundwater in a preset area and the update and optimization of the groundwater NAPL pollution degree determination model are performed through the information database.

4. The method for determining the degree of NAPL contamination in groundwater based on image learning according to claim 1 is characterized in that: A groundwater NAPL pollution degree determination model was constructed based on deep learning, and the data in the data information database was used for training. Specifically: A model for determining the degree of groundwater NAPL pollution was constructed based on deep learning methods, and feature data sets were extracted from the information database. By using big data, the pollution degree judgment standards corresponding to each substance type of NAPL pollution are obtained in the preset search space, and the judgment standards with the highest frequency in the test data are selected to construct the pollution degree judgment system for each substance type; Determine the pollution degree based on the pollution degree determination system for each substance type and the data in the corresponding feature data set of the monitoring data of the groundwater monitoring well, and set the pollution degree data annotation for the corresponding image feature; The groundwater NAPL pollution degree determination model is trained using the feature data set after data annotation. When the number of training times meets the preset number of times, the trained groundwater NAPL pollution degree determination model is output; The image features of the current water body image information of each groundwater monitoring well and the geographical hydrological information of the preset area are used as the input of the groundwater NAPL pollution degree determination model to identify the substance type and concentration information corresponding to the image features of each groundwater monitoring well; Obtain the current spatial distribution of NAPL-like substances in groundwater in a preset area, and obtain the pollution degree of NAPL-like pollution in groundwater in each area of ​​the preset area.

5. The method for determining the degree of NAPL contamination in groundwater based on image learning according to claim 1 is characterized in that: Generate pollution warning according to the pollution level, which previously included: Divide the preset area into several sub-areas according to the distribution of groundwater monitoring wells, obtain the material type and historical concentration information of groundwater NAPL substances in the sub-areas within a preset time, and generate a concentration time series corresponding to the material type according to the historical concentration information; According to the concentration time series corresponding to the substance types in different sub-areas, the pollution sources of various substances are traced, and the sub-areas where the pollution sources are located are marked, and the labels of the sub-areas where the pollution sources are located are set according to the substance types; Extracting the concentration variation at a preset time interval based on the concentration time series, and setting different variation thresholds for the labeled sub-region and the unlabeled sub-region; Determine whether the sub-region is tagged, read the change threshold value according to the determination result, and when the concentration change is greater than the threshold value, if the sub-region is not tagged, obtain the environmental characteristics and climate characteristics of the sub-region as the pollution impact characteristics of the sub-region; if the sub-region is tagged, obtain the environmental characteristics, climate characteristics and pollution source operation characteristics of the sub-region as the pollution impact characteristics of the sub-region; At the same time, in different sub-areas, the average value of the concentration change in the sub-area corresponding to the preset time interval is obtained, and the average value of the concentration change is used as the pollution change reference value of the sub-area within the preset time.

6. The method for determining the degree of NAPL contamination in groundwater based on image learning according to claim 5 is characterized in that: A pollution warning is generated according to the pollution degree, specifically: Based on the current spatial distribution of NAPL substances in groundwater in the preset area, the current substance types and current The previous concentration information is compared with the historical substance type and concentration information corresponding to the previous time step; When the type of NAPL substance in groundwater in a sub-region changes or the concentration change is greater than the pollution change reference value, early warning information in the sub-region is generated, and the early warning information is visualized in combination with the spatial distribution; According to whether the sub-region is labeled, the environmental characteristics and climate characteristics or the environmental characteristics, climate characteristics and pollution source operation characteristics of the sub-region in the future preset time are obtained, and the obtained characteristics are used to calculate the similarity with the pollution impact characteristics of the sub-region using Manhattan distance; When the Manhattan distance is less than or equal to the preset distance threshold, it is proved that the similarity meets the preset requirements, and a pollution warning for the sub-region is generated. Based on the pollution warning, big data means are used to obtain pollution emergency measures for groundwater NAPL substances in the sub-region.

7. A groundwater NAPL pollution degree determination system based on image learning, characterized in that: The system includes: a memory and a processor, wherein the memory includes a method program for determining the degree of NAPL pollution in groundwater based on image learning, and when the method program for determining the degree of NAPL pollution in groundwater based on image learning is executed by the processor, the following steps are implemented: Acquire water body image information in groundwater monitoring wells in a preset area, preprocess the water body image information, and extract image features of the water body image information; Obtain the type and concentration information of NAPL substances in groundwater based on the monitoring data of groundwater monitoring wells within a preset time; Matching the image features with the substance type and concentration information to construct a feature data set, which is conducive to combining the feature data set with local hydrogeological information to construct a data information database of groundwater NAPL pollution; A groundwater NAPL pollution degree determination model is constructed based on deep learning, and the data in the data information database is used for training to determine the pollution degree of groundwater NAPL, and a pollution warning is generated according to the pollution degree.

8. The groundwater NAPL pollution degree determination system based on image learning according to claim 7 is characterized in that: The image features of water body image information are extracted, specifically: Acquire water body image information according to preset time intervals through a preset image sensor device in a groundwater monitoring well, filter and de-noise the water body image information and perform image enhancement, and divide the filtering and de-noising and image enhancement into a plurality of image blocks; Calculating the information entropy of the image block, generating a mask image of the water body image information according to the information entropy, and performing image segmentation using the mask image to obtain preprocessed water body image information; The preprocessed water body image information is imported into the CSPNET network for feature characterization to obtain feature maps. The hole feature pyramid module is introduced, and hole convolutions with different hole rates are used to extract multi-scale features to obtain multi-scale feature maps. The multi-scale feature map is converted into a channel description through average pooling, the channel attention weight of the multi-scale feature map is obtained through two convolutional layers and sigmoid and Relu functions, and the channel attention weight is multiplied by the multi-scale feature map to obtain an attention map; The image features of the water body image information are obtained by element-wise multiplication of the attention map and the feature map.

9. The groundwater NAPL pollution degree determination system based on image learning according to claim 7 is characterized in that: It is beneficial to combine the characteristic data set with the local hydrogeological information to build a data information database of NAPL contaminated groundwater, specifically: Read the historical monitoring data of groundwater monitoring wells in the preset area, judge the effectiveness of groundwater monitoring wells based on the historical monitoring data, and eliminate invalid groundwater monitoring wells; Acquire monitoring data of groundwater monitoring wells valid within a preset time, clean the monitoring data, obtain the material type and concentration information of NAPL-like substances in groundwater, intersect image features with the material type and concentration information of NAPL pollution, and construct a feature data set; Obtain the corresponding groundwater flow direction, flow velocity, surface water-groundwater interaction frequency, and stratigraphic structure information in the preset area to construct a hydrogeological data set and combine it with the characteristic data set; An information database of NAPL pollution in groundwater reservoirs is established, and the combined characteristic data is structured and stored in the information database. The data management of NAPL pollution in groundwater in a preset area and the update and optimization of the groundwater NAPL pollution degree determination model are performed through the information database.

10. The groundwater NAPL pollution degree determination system based on image learning according to claim 7 is characterized in that: A groundwater NAPL pollution degree determination model was constructed based on deep learning, and the data in the data information database was used for training. Specifically: A model for determining the degree of groundwater NAPL pollution was constructed based on deep learning methods, and feature data sets were extracted from the information database. By using big data, the pollution degree judgment standards corresponding to each substance type of NAPL pollution are obtained in the preset search space, and the judgment standards with the highest frequency in the test data are selected to construct the pollution degree judgment system for each substance type; Determine the pollution degree based on the pollution degree determination system for each substance type and the data in the corresponding feature data set of the monitoring data of the groundwater monitoring well, and set the pollution degree data annotation for the corresponding image feature; The groundwater NAPL pollution degree determination model is trained using the feature data set after data annotation. When the number of training times meets the preset number of times, the trained groundwater NAPL pollution degree determination model is output; The image features of the current water body image information of each groundwater monitoring well and the geographical hydrological information of the preset area are used as the input of the groundwater NAPL pollution degree determination model to identify the substance type and concentration information corresponding to the image features of each groundwater monitoring well; Obtain the current spatial distribution of NAPL-like substances in groundwater in a preset area, and obtain the pollution degree of NAPL-like pollution in groundwater in each area of ​​the preset area.

Images