CNN-based underground water pollution source identification method and system under strong heterogeneous condition

Abstract

The invention discloses a convolutional neural network CNN-based underground water pollution source identification method and system under a strong heterogeneous condition, and relates to the technical field of underground water environmental protection. Reasonably generalizing hydrogeological conditions, and constructing a groundwater solute transport conceptual model; generating a plurality of groups of working conditions representing the emission intensity of the pollution source by adopting a random sampling method, and generating an aquifer permeability coefficient field through sequential Gaussian simulation; based on the aquifer permeability coefficient field, constructing an underground water solute transport-double-domain reference model; inputting a plurality of groups of working conditions into the groundwater solute transport-double-domain reference model one by one so as to form samples of'model input-model response 'and form a training sample set and a test sample set; the underground water pollution source identification model is constructed based on the training sample set by adopting the deep learning-CNN algorithm, and the position and the emission intensity of the pollution source can be accurately identified by using the model, so that the pollution source identification efficiency and precision are effectively improved, and a reliable solution can be provided for underground water pollution treatment.

Description

Technical Field

[0001] The present invention relates to the technical field of groundwater environmental protection, and in particular to a method and system for identifying groundwater pollution sources under strong heterogeneity conditions based on a convolutional neural network (CNN). Background Art

[0002] Due to the concealment and complexity of groundwater pollution, the use of traditional monitoring and investigation methods to obtain information on groundwater pollution sources is not only inefficient, but also has high engineering costs. Accurately identifying the number, location coordinates and release history of groundwater pollution sources is the primary task of groundwater pollution control and is crucial for subsequent groundwater environmental remediation. However, on the other hand, the natural underground aquifer system is a highly heterogeneous and complex system. When pollutants invade the underground aquifer, their migration patterns and spatiotemporal variation characteristics are deeply influenced by the heterogeneity of the aquifer medium. When dealing with such problems, traditional pollution source identification methods often fail to fully consider the impact of spatial heterogeneity, resulting in inaccurate simulation results that are difficult to match with actual conditions.

[0003] Groundwater pollution source identification is a process that reversely derives key details describing groundwater pollution sources, such as the number of pollution sources, precise location coordinates, and release history, based on existing prior data (covering information such as pollutant concentration and distribution, water level, and water quality parameters). However, the current traditional methods frequently and repeatedly call the groundwater solute transport simulation model when identifying groundwater pollution sources. This process is not only computationally intensive, but also time-consuming and cumbersome, making it difficult to ensure the efficiency of the identification process and the accuracy of the final results. Summary of the invention

[0004] The purpose of the present invention is to propose a method and system for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN, so as to cope with the complexity and heterogeneity of the underground aquifer system and realize accurate identification of the location coordinates of the groundwater pollution sources and their emission intensity.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN is provided, comprising the following steps:

[0006] After analyzing the hydrogeological and groundwater pollution related data, the hydrogeological conditions are reasonably generalized and a conceptual model of groundwater solute transport is constructed to accurately describe the migration and movement of pollutants in groundwater;

[0007] In the conceptual model of groundwater solute transport, according to the pollution source characteristics and the prior interval of the aquifer permeability coefficient, a random sampling method is used to generate several groups of conditions representing the emission intensity of the pollution source, and aquifer permeability coefficient field is generated by sequential Gaussian simulation to represent the contaminated underground aquifer system under strong heterogeneity conditions;

[0008] Based on the aquifer permeability field, a dual-domain benchmark model of groundwater solute transport is constructed to accurately reflect the physical and chemical processes of pollutants in the underground aquifer system;

[0009] Input several groups of working conditions into the groundwater solute transport-dual domain benchmark model one by one to obtain the model response of the observation well observation data. The model response is the location coordinates of the pollution source and its emission intensity, thereby forming a sample of "model input-model response" and constituting a training sample set and a test sample set;

[0010] The deep learning-CNN algorithm is used to build a groundwater pollution source identification model based on the training sample set, and the model is used to predict the location coordinates and emission intensity of the pollution source.

[0011] Furthermore, the simulated annealing method is used to optimize the Latin hypercube sampling algorithm SA-LHS and randomly sample several groups of working conditions.

[0012] Furthermore, Modflow and MT3DMS programs were used to construct a groundwater solute transport-dual domain benchmark model.

[0013] Furthermore, the groundwater pollution source identification model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer.

[0014] Furthermore, the convolution layer scans and performs convolution operations on the output results of the previous layer through the convolution kernel, extracts local features, and performs dimensionality reduction processing. The specific expression is:

[0015]

[0016] In the formula, is the jth feature output of the lth layer, M j is the jth convolutional region of the lth layer, is the i-th feature output of the l-1th layer, is the jth weight value of the i-th convolution kernel in the l-th layer, is the i-th bias of the l-th layer, and * represents the convolution operation.

[0017] After the convolution operation, an activation function is used to enhance the nonlinearity of the extracted local features, improve the learning speed and applicability of the convolutional neural network, further optimize the model and increase the calculation speed. The ReLU function is generally used for nonlinear calculations:

[0018] ReLU = max(0, x)

[0019] Where x is a time series.

[0020] Furthermore, the pooling layer compresses the extracted features to reduce the dimension and increase the calculation speed. The commonly used pooling method is maximum pooling, and its operation model is:

[0021]

[0022] In the formula, is the value of the t-th neuron in the ith feature of the l-th layer (t∈[(j-1)W,jW]), W is the width of the pooling region, is the value of the neuron in the l+1th layer.

[0023] Furthermore, the fully connected layer acts as a classifier to integrate and classify the output results of the pooling layer and output the results after mapping. The mathematical model of the specific process is:

[0024] O=f(b o +f v w o )

[0025] Where b o is the deviation vector, f v is the characteristic vector, w o is the weight matrix; through the step of the fully connected layer, the process from feature extraction to target recognition of data is realized.

[0026] According to a second aspect of an embodiment of the present disclosure, a groundwater pollution source identification system under strong heterogeneity conditions based on CNN is provided, comprising:

[0027] The conceptual model building module analyzes the hydrogeological and groundwater pollution related data, reasonably generalizes the hydrogeological conditions, and constructs a groundwater solute transport conceptual model to accurately describe the migration and movement of pollutants in groundwater;

[0028] A working condition acquisition module, in the groundwater solute transport conceptual model, generates several groups of working conditions representing the emission intensity of the pollution source by using a random sampling method according to the pollution source characteristics and the prior interval of the aquifer permeability coefficient, and generates the aquifer permeability coefficient field by sequential Gaussian simulation to represent the polluted underground aquifer system under strong heterogeneity conditions;

[0029] The benchmark model building module builds a dual-domain benchmark model of groundwater solute transport based on the aquifer permeability field to accurately reflect the physical and chemical processes of pollutants in the underground aquifer system;

[0030] The sample set acquisition module inputs several groups of working conditions into the groundwater solute transport-dual domain benchmark model one by one to obtain the model response of the observation well observation data. The model response is the location coordinates of the pollution source and its emission intensity, thereby forming a sample of "model input-model response" and constituting a training sample set and a test sample set;

[0031] The prediction module adopts the deep learning-CNN algorithm to build a groundwater pollution source identification model based on the training sample set, and uses this model to predict the location coordinates and emission intensity of the pollution source.

[0032] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored and running on the memory, wherein when the processor executes the program, the method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN is implemented.

[0033] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the CNN-based method for identifying groundwater pollution sources under strong heterogeneity conditions is implemented.

[0034] Compared with the prior art, the above technical solution adopted by the present invention has the following advantages: 1) The present invention establishes a groundwater pollution source identification model under strong heterogeneity conditions based on CNN. Compared with the traditional groundwater pollution source identification method, this model can accurately identify groundwater pollution sources under strong heterogeneity conditions, effectively solve complex high-dimensional nonlinear problems, and reduce the use of the conceptual model of groundwater solute transport, avoiding cumbersome calculation processes, shortening the calculation cycle, and improving calculation efficiency.

[0035] 2) The present invention solves the accuracy and stability problems of groundwater pollution source identification. It uses the stronger learning ability of CNN to grasp the complex input-output relationship in the groundwater pollution source identification model. It has good robustness and generalization ability to cope with pollution sources of different types and intensities and groundwater pollution source identification problems under complex and strongly heterogeneous hydrogeological conditions, thereby ensuring the stability of the model.

[0036] 3) Based on the prior data, the present invention obtains the observation data of 7 observation wells in any year, and can accurately identify the location of the pollution source and its emission intensity. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings in the specification, which constitute a part of the present application, are used to provide further understanding of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute improper limitations on the present application.

[0038] Figure 1 This is an overview map of the example area;

[0039] Figure 2 The schematic diagram of the dual-domain model applicable to strong heterogeneity conditions;

[0040] Figure 3 This is the schematic diagram of the CNN algorithm. DETAILED DESCRIPTION

[0041] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0042] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0043] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0044] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and systems according to various embodiments of the present disclosure. It should be noted that each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of a code may include one or more executable instructions for implementing the logical functions specified in each embodiment. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and / or block diagram, and the combination of boxes in the flowchart and / or block diagram can be implemented using a dedicated hardware-based system that performs a specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0045] Embodiment 1:

[0046] This embodiment provides a method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN, including the following steps:

[0047] S1. After analyzing the hydrogeological and groundwater pollution related data, reasonably generalize the hydrogeological conditions and construct a conceptual model of groundwater solute transport to accurately describe the migration and movement of pollutants in groundwater;

[0048] S2. In the conceptual model of groundwater solute transport, according to the prior intervals of pollution source characteristics (pollutant types, pollutant concentration distribution, etc.) and aquifer permeability coefficients (porosity, permeability coefficient, etc.) (as shown in Table 1), a random sampling method is used to generate several groups of conditions representing the emission intensity of pollution sources, and aquifer permeability coefficient fields are generated by sequential Gaussian simulation to represent the contaminated underground aquifer system under strong heterogeneity conditions;

[0049] Combined with the specific circumstances of the embodiment, such as Figure 1 As shown in the figure, it is assumed that the model boundary is set at a position where the influence of pollutant migration can be approximately ignored. Among them, the blue line AB and the blue line CD are generalized as the constant head boundary, and the black line AC and the black line BD are generalized as the water-proof boundary. It is assumed that the permeability coefficient value is constant within its corresponding grid to simplify the calculation.

[0050] Table 1 Prior intervals of pollution source parameters

[0051]

[0052] S3. Based on the aquifer permeability field, a dual-domain benchmark model of groundwater solute transport is constructed, such as Figure 2 to accurately reflect the physicochemical processes of pollutants in underground aquifer systems;

[0053] The physicochemical parameters of the benchmark model are shown in Table 2:

[0054] Table 2 Physicochemical parameters

[0055]

[0056] Based on the above information, Modflow and MT3DMS programs were used to construct a groundwater solute transport-dual domain benchmark model and initialize the model.

[0057] S4. Input several groups of working conditions into the groundwater solute transport-dual domain benchmark model one by one to obtain the model response of the observation well observation data, and the model response is the location coordinates of the pollution source and its emission intensity, thereby forming a sample of "model input-model response" and constituting a training sample set and a test sample set;

[0058] Among them, the number of training sample sampling can be 350, which is used for model training, and another 150 groups of test samples are formed for model performance testing.

[0059] S5. Using deep learning-CNN algorithm, a groundwater pollution source identification model is constructed based on the training sample set, such as Figure 3 As shown, the model is used to predict the location coordinates and emission intensity of pollution sources.

[0060] It should be noted that the accuracy and reliability of the model were verified by using training samples and test samples for inversion and forecasting calculations.

[0061] The groundwater pollution source identification model constructed by the CNN algorithm shows high accuracy in predicting the location coordinates of the pollution source and its emission intensity. According to the test results of the test samples, the groundwater pollution source identification model constructed by the CNN algorithm has high accuracy in fitting the emission history (Table 3) and identifying the pollution source coordinates (Table 4). This shows that the groundwater pollution source identification model constructed by the CNN algorithm can still accurately learn the complex input-output relationship in the groundwater pollution source identification model under strong heterogeneity conditions, thereby obtaining important information about the pollution source.

[0062] Table 3. Comparison between the 10 average emission values ​​of the pollution identification model and the actual emission values

[0063]

[0064] Table 4. Comparison of 10 average coordinates of the pollution identification model and the actual position coordinates

[0065]

[0066] In summary, the present invention effectively solves the shortcomings of traditional groundwater pollution source identification models when processing complex data sets. The present invention not only reduces the calculation time and amount of calculation, but also realizes the accurate identification of the coordinate position of groundwater pollution sources and their emission intensity under strong heterogeneous conditions while ensuring efficient calculation. This innovative achievement not only provides a scientific basis for the precise control of groundwater pollution, but also provides a reference for the identification and detection of pollution sources in other environmental fields. At the same time, computer algorithms and models are applied interdisciplinaryly to solve complex environmental problems in groundwater pollution, providing strong technical support for the prevention and control of groundwater pollution and promoting technological innovation in this field.

[0067] Embodiment 2:

[0068] This embodiment provides a groundwater pollution source identification system under strong heterogeneity conditions based on CNN, including:

[0069] The conceptual model building module analyzes the hydrogeological and groundwater pollution related data, reasonably generalizes the hydrogeological conditions, and constructs a groundwater solute transport conceptual model to accurately describe the migration and movement of pollutants in groundwater;

[0070] A working condition acquisition module, in the groundwater solute transport conceptual model, generates several groups of working conditions representing the emission intensity of the pollution source by using a random sampling method according to the pollution source characteristics and the prior interval of the aquifer permeability coefficient, and generates the aquifer permeability coefficient field by sequential Gaussian simulation to represent the polluted underground aquifer system under strong heterogeneity conditions;

[0071] The benchmark model building module builds a dual-domain benchmark model of groundwater solute transport based on the aquifer permeability field to accurately reflect the physical and chemical processes of pollutants in the underground aquifer system;

[0072] The sample set acquisition module inputs several groups of working conditions into the groundwater solute transport-dual domain benchmark model one by one to obtain the model response of the observation well observation data. The model response is the location coordinates of the pollution source and its emission intensity, thereby forming a sample of "model input-model response" and constituting a training sample set and a test sample set;

[0073] The prediction module adopts the deep learning-CNN algorithm to build a groundwater pollution source identification model based on the training sample set, and uses this model to predict the location coordinates and emission intensity of the pollution source.

[0074] Embodiment three:

[0075] An electronic device includes a memory, a processor, and a computer program stored and running on the memory, wherein the processor implements the above-mentioned method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN when executing the program, including:

[0076] After analyzing the hydrogeological and groundwater pollution related data, the hydrogeological conditions are reasonably generalized and a conceptual model of groundwater solute transport is constructed to accurately describe the migration and movement of pollutants in groundwater;

[0077] In the conceptual model of groundwater solute transport, according to the pollution source characteristics and the prior interval of the aquifer permeability coefficient, a random sampling method is used to generate several groups of conditions representing the emission intensity of the pollution source, and aquifer permeability coefficient field is generated by sequential Gaussian simulation to represent the contaminated underground aquifer system under strong heterogeneity conditions;

[0078] Based on the aquifer permeability field, a dual-domain benchmark model of groundwater solute transport is constructed to accurately reflect the physical and chemical processes of pollutants in the underground aquifer system;

[0079] Input several groups of working conditions into the groundwater solute transport-dual domain benchmark model one by one to obtain the model response of the observation well observation data. The model response is the location coordinates of the pollution source and its emission intensity, thereby forming a sample of "model input-model response" and constituting a training sample set and a test sample set;

[0080] The deep learning-CNN algorithm is used to build a groundwater pollution source identification model based on the training sample set, and the model is used to predict the location coordinates and emission intensity of the pollution source.

[0081] Embodiment 4:

[0082] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN, comprising:

[0083] After analyzing the hydrogeological and groundwater pollution related data, the hydrogeological conditions are reasonably generalized and a conceptual model of groundwater solute transport is constructed to accurately describe the migration and movement of pollutants in groundwater;

[0084] In the conceptual model of groundwater solute transport, according to the pollution source characteristics and the prior interval of the aquifer permeability coefficient, a random sampling method is used to generate several groups of conditions representing the emission intensity of the pollution source, and aquifer permeability coefficient field is generated by sequential Gaussian simulation to represent the contaminated underground aquifer system under strong heterogeneity conditions;

[0085] Based on the aquifer permeability field, a dual-domain benchmark model of groundwater solute transport is constructed to accurately reflect the physical and chemical processes of pollutants in the underground aquifer system;

[0086] Input several groups of working conditions into the groundwater solute transport-dual domain benchmark model one by one to obtain the model response of the observation well observation data. The model response is the location coordinates of the pollution source and its emission intensity, thereby forming a sample of "model input-model response" and constituting a training sample set and a test sample set;

[0087] The deep learning-CNN algorithm is used to build a groundwater pollution source identification model based on the training sample set, and the model is used to predict the location coordinates and emission intensity of the pollution source.

[0088] Those skilled in the art should understand that the modules or steps of the present disclosure can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present disclosure is not limited to any specific combination of hardware and software.

[0089] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0090] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN, characterized in that: The following steps are involved: After analyzing the hydrogeological and groundwater pollution related data, the hydrogeological conditions are reasonably generalized and a conceptual model of groundwater solute transport is constructed to accurately describe the migration and movement of pollutants in groundwater; In the conceptual model of groundwater solute transport, according to the pollution source characteristics and the prior interval of the aquifer permeability coefficient, a random sampling method is used to generate several groups of conditions representing the emission intensity of the pollution source, and aquifer permeability coefficient field is generated by sequential Gaussian simulation to represent the contaminated underground aquifer system under strong heterogeneity conditions; Based on the aquifer permeability field, a dual-domain benchmark model of groundwater solute transport is constructed to accurately reflect the physical and chemical processes of pollutants in the underground aquifer system; Input several groups of working conditions into the groundwater solute transport-dual domain benchmark model one by one to obtain the model response of the observation well observation data. The model response is the location coordinates of the pollution source and its emission intensity, thereby forming a sample of "model input-model response" and constituting a training sample set and a test sample set; The deep learning-CNN algorithm is used to build a groundwater pollution source identification model based on the training sample set, and the model is used to predict the location coordinates and emission intensity of the pollution source.

2. According to claim 1, the method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN is characterized in that: The simulated annealing method is used to optimize the Latin hypercube sampling algorithm SA-LHS and randomly sample several groups of working conditions.

3. According to claim 1, the method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN is characterized in that: Modflow and MT3DMS programs were used to construct a dual-domain benchmark model of groundwater solute transport.

4. According to claim 1, the method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN is characterized in that: The groundwater pollution source identification model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer.

5. According to claim 4, the method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN is characterized in that: The convolution layer scans and performs convolution operations on the output results of the previous layer through the convolution kernel, extracts local features, and performs dimensionality reduction processing. The specific expression is: In the formula, is the jth feature output of the lth layer, M j is the jth convolutional region of the lth layer, is the i-th feature output of the l-1th layer, is the jth weight value of the i-th convolution kernel in the l-th layer, is the i-th bias of the l-th layer, and * represents the convolution operation; After the convolution operation, an activation function is used to enhance the nonlinearity of the extracted local features, so the ReLU function is used for nonlinear calculation: ReLU = max(0, x) Where x is a time series.

6. According to claim 4, the method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN is characterized in that: The pooling layer compresses the extracted features to reduce the dimension, and its operation model is: In the formula, is the value of the t-th neuron in the ith feature of the l-th layer (t∈[(j-1)W,jW]), W is the width of the pooling region, P i l+1 (j) is the value of the neuron in the l+1th layer.

7. The method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN according to claim 4 is characterized in that: The fully connected layer acts as a classifier, integrating and classifying the output results of the pooling layer and outputting the results after mapping. The mathematical model of the specific process is: O=f(b o +f v w o ) Where b o is the deviation vector, f v is the characteristic vector, w o is the weight matrix; through the step of the fully connected layer, the process from feature extraction to target recognition of data is realized.

8. A groundwater pollution source identification system under strong heterogeneity conditions based on CNN, characterized by: include: The conceptual model building module analyzes the hydrogeological and groundwater pollution related data, reasonably generalizes the hydrogeological conditions, and constructs a groundwater solute transport conceptual model to accurately describe the migration and movement of pollutants in groundwater; A working condition acquisition module, in the groundwater solute transport conceptual model, generates several groups of working conditions representing the emission intensity of the pollution source by using a random sampling method according to the pollution source characteristics and the prior interval of the aquifer permeability coefficient, and generates the aquifer permeability coefficient field by sequential Gaussian simulation to represent the polluted underground aquifer system under strong heterogeneity conditions; The benchmark model building module builds a dual-domain benchmark model of groundwater solute transport based on the aquifer permeability field to accurately reflect the physical and chemical processes of pollutants in the underground aquifer system; The sample set acquisition module inputs several groups of working conditions into the groundwater solute transport-dual domain benchmark model one by one to obtain the model response of the observation well observation data. The model response is the location coordinates of the pollution source and its emission intensity, thereby forming a sample of "model input-model response" and constituting a training sample set and a test sample set; The prediction module adopts the deep learning-CNN algorithm to build a groundwater pollution source identification model based on the training sample set, and uses this model to predict the location coordinates and emission intensity of the pollution source.

9. An electronic device comprising a memory, a processor and a computer program stored and running on the memory, wherein when the processor executes the program, the method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for identifying groundwater pollution sources under strong heterogeneity conditions based on CNN as described in any one of claims 1 to 7.

Images