A full-process data management method and system for groundwater environment monitoring wells

Abstract

The present invention discloses a full-process data management method and system for groundwater environment monitoring wells, which relates to the technical field of groundwater environment monitoring well investigations and includes: obtaining multi-source investigation data of groundwater environment monitoring wells and performing preprocessing, extracting defect semantic features based on the multi-source investigation data of defective groundwater environment monitoring wells, performing multi-label identification of groundwater environment monitoring wells according to the defect semantic features, realizing comprehensive status evaluation and generating corresponding operation and maintenance information; selecting groundwater environment monitoring wells that meet the requirements to read water quality detection data, extracting the spatio-temporal features of the water quality detection data for predicting the groundwater environment quality within the region, and generating monitoring early warnings through the prediction results of the groundwater environment quality within the region. The present invention ensures the efficient implementation of groundwater monitoring well investigation and evaluation projects, improves the daily management ability of groundwater monitoring wells, and enhances the groundwater environment supervision ability.

Description

Technical Field

[0001] The present invention relates to the technical field of groundwater environment monitoring well surveys, and more specifically, to a full-process data management method and system for groundwater environment monitoring wells. Background Art

[0002] As an important reliance for carrying out groundwater environment monitoring work, the "health status" of groundwater environment monitoring wells directly affects the accuracy of monitoring results. Therefore, the orderly development of groundwater monitoring well surveys → groundwater monitoring well assessments → groundwater monitoring well management and maintenance → groundwater monitoring network construction → long-sequence analysis and display of groundwater monitoring data → groundwater monitoring early warning work can solidly promote groundwater monitoring work and improve the groundwater environment supervision ability.

[0003] Characteristics such as a large number of monitoring well survey categories, a large quantity, rich survey content (including information entry, photographing, and signing), and a large number of survey and review personnel. When using traditional questionnaires for surveys, problems such as long on-site filling times in the field, large digital workloads, high error rates in survey information entry, and complex management may occur. To avoid the above problems, a full-process intelligent data management system for multi-source data including groundwater environment monitoring well surveys, assessments, management, maintenance, and automatic evaluation of groundwater quality data is required, which can efficiently carry out the implementation of groundwater monitoring well survey and assessment projects, greatly improve work efficiency, and ensure the quality of projects. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a full-process data management method and system for groundwater environment monitoring wells, realizing full-process intelligent data management of multi-source data such as groundwater environment monitoring well surveys, assessments, management, maintenance, and automatic evaluation of groundwater quality data.

[0005] The first aspect of the present invention provides a full-process data management method for groundwater environment monitoring wells, including the following steps:

[0006] Obtain multi-source survey data of groundwater environment monitoring wells and perform preprocessing, and conduct a preliminary status assessment of the groundwater environment monitoring wells based on the preprocessed multi-source survey data;

[0007] Read the substandard groundwater environment monitoring wells through the preliminary status assessment, extract defect semantic features based on the multi-source survey data of the substandard groundwater environment monitoring wells, construct a comprehensive status assessment model, perform multi-label identification of the groundwater environment monitoring wells based on the defect semantic features, and conduct a comprehensive status assessment based on the defect category labels;

[0008] Generate operation and maintenance information based on the evaluation results of the comprehensive status assessment, select the groundwater environment monitoring wells that meet the requirements to read water quality detection data, and extract the spatio-temporal features of the water quality detection data;

[0009] Predict the groundwater environmental quality within the region based on the spatio-temporal characteristics, and generate monitoring and early warning through the prediction results of the groundwater environmental quality within the region.

[0010] In this solution, obtain the multi-source survey data of the groundwater environmental monitoring wells and perform preprocessing. Based on the preprocessed multi-source survey data, conduct a preliminary status assessment of the groundwater environmental monitoring wells, specifically:

[0011] Obtain the basic information, basic conditions, structural conditions, and connectivity conditions of the groundwater environmental monitoring wells within the region as the multi-source survey data of the groundwater monitoring wells. Merge the multi-source survey data and use a unique identifier for matching;

[0012] Perform data cleaning on the matched multi-source survey data, process duplicate data, outliers, and missing values, and perform dimensionality reduction and standardization processing on the multi-source survey data after data cleaning to obtain the preprocessed multi-source survey data;

[0013] Read the historical multi-source survey data according to the unique identifier corresponding to the groundwater environmental monitoring well, and obtain the status assessment label corresponding to the historical multi-source survey data. Select the historical multi-source survey data with qualified monitoring functions based on the status assessment label;

[0014] Train and generate an adversarial network based on the obtained historical multi-source survey data to obtain a generator model and a discriminator model. Use the obtained generator model as a decoder, and train an autoencoder network through the obtained historical multi-source survey data to obtain an encoder that maps the preprocessed multi-source survey data to the latent space;

[0015] Obtain the latent feature sample corresponding to the historical multi-source survey data that is closest to the preprocessed multi-source survey data in the latent space, and decode it through the decoder to obtain a reconstructed data that is similar but not the same as the input multi-source survey data;

[0016] Obtain the residual between the preprocessed multi-source survey data and the reconstructed data, and judge whether there are defects in the groundwater environmental monitoring well according to the comparison result between the residual and the preset difference threshold, so as to achieve a preliminary status assessment.

[0017] In this solution, read the unqualified groundwater environmental monitoring wells through the preliminary status assessment, and extract the defect semantic features according to the multi-source survey data of the unqualified groundwater environmental monitoring wells, specifically:

[0018] Obtain the preliminary status estimation results of the groundwater environmental monitoring wells within the region, screen the groundwater environmental monitoring wells with defects according to the preliminary status estimation results, and generate unqualified labels;

[0019] Extract the multi-source survey data corresponding to the groundwater environmental monitoring wells with non-compliant labels, and use a preset shallow convolutional neural network to extract and compress the defect features of the multi-source survey data. The shallow convolutional neural network consists of three convolutional layers, three max pooling layers, and two fully connected layers. Obtain the one-dimensional feature vector of the non-compliant groundwater environmental monitoring wells in the fully connected layer;

[0020] Obtain the monitoring well damage instances with severity annotations according to the historical multi-source survey data, perform clustering analysis through the monitoring well damage instances, obtain the clustering clusters corresponding to the last clustering result after iterative clustering, generate a subset of defect data, and determine the defect types corresponding to each subset of defect data;

[0021] Extract defect attributes according to the monitoring well damage instance samples in each subset of defect data, integrate the defect attribute vectors to construct a defect attribute set, use similarity calculation to select the n attribute vectors with the highest similarity in the defect attribute set using the one-dimensional feature vector, and splice the one-dimensional feature vector with the selected attribute vectors to generate defect semantic features.

[0022] In this solution, construct a comprehensive status evaluation model, perform multi-label recognition of groundwater environmental monitoring wells according to the defect semantic features, and perform comprehensive status evaluation according to the defect category labels. Specifically:

[0023] Construct a comprehensive status evaluation model according to the defect semantic feature extraction module and the relationship network module. The relationship network module consists of two fully connected layers. Use the monitoring well damage instance samples with severity annotations in each subset of defect data to construct training samples, obtain the relationship scores of each training sample, and perform model training according to the minimum mean square error between the relationship scores and the one-hot encoding of the training samples;

[0024] In the first fully connected layer, activate the defect semantic features obtained by the defect semantic feature extraction module using the ReLU activation function and output them. In the second fully connected layer, use the Sigmoid function to map the output of the first fully connected layer to the interval (0,1) to generate relationship scores;

[0025] Perform multi-label recognition of the corresponding defects of the groundwater environmental monitoring wells according to the relationship scores, determine the defect category labels and corresponding severities of the non-compliant groundwater environmental monitoring wells, and generate a comprehensive status evaluation result based on the preset evaluation threshold interval according to the defect category labels and severities.

[0026] In this solution, generate operation and maintenance information according to the evaluation results of the comprehensive status evaluation. Specifically:

[0027] Obtain the comprehensive status evaluation results of non-compliant groundwater environment monitoring wells within the area, and select the groundwater environment monitoring wells with poor monitoring functions and general monitoring functions for suspension according to the comprehensive status evaluation results;

[0028] Obtain the defect category label with the highest severity corresponding to each groundwater environment monitoring well among the suspended groundwater environment monitoring wells with general monitoring functions, and obtain the severity trend change corresponding to the defect category label according to historical multi-source survey data;

[0029] Compare the severity trend change with the trend change threshold, cancel the suspension of the groundwater environment monitoring wells with general monitoring functions that are less than the trend change threshold, and finally generate operation and maintenance information according to the unique identifiers corresponding to the suspended groundwater environment monitoring wells with poor monitoring functions and general monitoring functions, and send and display it according to the preset method.

[0030] In this solution, select the groundwater environment monitoring wells that meet the requirements to read the water quality detection data, and extract the spatio-temporal characteristics of the water quality detection data. Specifically:

[0031] Obtain the groundwater environment monitoring wells with good monitoring functions in the area, extract the water quality detection data sequence within the preset time, and calculate the Pearson correlation coefficient based on the water quality detection data sequence to obtain the correlation of monitoring data between groundwater environment monitoring wells;

[0032] Determine the connection relationship between groundwater monitoring wells according to the distance and monitoring data correlation between groundwater environment monitoring wells, generate the corresponding topological structure and represent it graphically. Take the groundwater environment monitoring wells with good monitoring functions in the area as graph nodes, and set the edge structure between nodes according to the connection relationship;

[0033] Construct an adjacency matrix based on the edge structure, use a graph attention network to learn the graph representation, take the adjacency matrix as the input, obtain the attention weights between any two nodes in the neighborhood according to the multi-head self-attention mechanism, use the attention weights to perform weighted aggregation on the adjacency matrix, update the feature representation, and obtain the spatial characteristics of the groundwater environment quality in the area;

[0034] Import the updated feature representation into a gated recurrent unit, use the hidden state update to obtain the temporal characteristics of the groundwater environment quality in the area, introduce a temporal attention mechanism to perform attention weighting on the temporal characteristics at different times, and fuse the weighted temporal characteristics with the spatial characteristics after pooling operation to obtain the spatio-temporal characteristics of the water quality detection data in the area.

[0035] In this solution, predict the groundwater environment quality in the area according to the spatio-temporal characteristics, and generate a monitoring warning through the prediction results of the groundwater environment quality in the area. Specifically:

[0036] Construct a groundwater environmental quality prediction model based on a graph attention network and a gated recurrent unit. Obtain the spatio-temporal characteristics of the water quality detection quantity in the region through the graph attention network and the gated recurrent unit, and obtain the prediction result of the groundwater environmental quality in the region through the fully connected layer for the spatio-temporal characteristics.

[0037] Conduct dynamic analysis based on the prediction result of the groundwater environmental quality within a preset time period to obtain the trend change. Set the absolute threshold and the trend change threshold. When the obtained prediction result of the groundwater environmental quality in the region or the trend change is greater than the corresponding absolute threshold or trend change threshold, generate a monitoring warning and send and display it in a preset manner.

[0038] The second aspect of the present invention provides a full-process data management system for groundwater environmental monitoring wells. The system includes: a multi-source survey data integration module, a monitoring well status evaluation module, a monitoring well operation and maintenance warning module, a groundwater environmental quality analysis module, and a groundwater monitoring warning module.

[0039] The multi-source survey data integration module is responsible for obtaining the multi-source survey data of the groundwater environmental monitoring wells and performing preprocessing.

[0040] The monitoring well status evaluation module is responsible for judging whether there are defects in the groundwater environmental monitoring wells according to the preprocessed multi-source survey data, realizing the preliminary status evaluation, extracting the defect semantic features from the multi-source survey data of the groundwater environmental monitoring wells that do not meet the standard in the preliminary status evaluation, constructing a comprehensive status evaluation model, performing multi-label recognition of the groundwater environmental monitoring wells according to the defect semantic features, and performing comprehensive status evaluation according to the defect category labels.

[0041] The monitoring well operation and maintenance warning module is responsible for generating operation and maintenance information according to the evaluation result of the comprehensive status evaluation and sending and displaying it in a preset manner.

[0042] The groundwater environmental quality analysis module is responsible for selecting the groundwater environmental monitoring wells that meet the requirements in the region, reading the water quality detection data, and extracting the spatio-temporal characteristics of the water quality detection data to predict the groundwater environmental quality in the region.

[0043] The groundwater monitoring warning module is responsible for generating a water environment monitoring warning through the prediction result of the groundwater environmental quality in the region and sending and displaying it in a preset manner.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] Through the integration, preprocessing, and status evaluation of multi-source survey data corresponding to groundwater environmental monitoring wells, combined with the extraction and prediction of spatio-temporal characteristics of water quality detection data, the full-process data management and monitoring early warning of groundwater environmental monitoring wells are realized. Through the extraction and prediction of spatio-temporal characteristics of water quality detection data, the changing trend of groundwater environmental quality can be accurately predicted. Finally, through the monitoring early warning mechanism, groundwater environmental problems can be discovered and handled in a timely manner to ensure the safety of the groundwater environment.

[0046] The present invention provides a full-process intelligent data management system for multi-source data such as the investigation, evaluation, management, maintenance of groundwater environmental monitoring wells, and the automatic evaluation of groundwater quality data, which can efficiently carry out the implementation of investigation and evaluation projects for groundwater monitoring wells, greatly improving work efficiency and work quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments or exemplifications of the present invention, the following will briefly introduce the drawings required for use in the description of the embodiments or exemplifications. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the drawings shown.

[0048] Figure 1 Shows the flowchart of the full-process data management method for groundwater environmental monitoring wells;

[0049] Figure 2 Shows the flowchart of extracting the defect semantic features of non-compliant groundwater environmental monitoring wells;

[0050] Figure 3 Shows the flowchart of extracting the spatio-temporal characteristics of water quality detection data within the region;

[0051] Figure 4 Shows the block diagram of the full-process data management system for groundwater environmental monitoring wells. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] In order to more clearly understand the above-mentioned objects, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0053] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0054] Figure 1The flowchart of the full-process data management method for groundwater environment monitoring wells is shown.

[0055] As Figure 1 shown, in the first embodiment of the present invention, a full-process data management method for groundwater environment monitoring wells is provided, including:

[0056] S102, obtaining multi-source survey data of groundwater environment monitoring wells and performing preprocessing, and performing a preliminary status assessment on the groundwater environment monitoring wells according to the preprocessed multi-source survey data;

[0057] S104, reading the non-compliant groundwater environment monitoring wells through the preliminary status assessment, extracting defect semantic features according to the multi-source survey data of the non-compliant groundwater environment monitoring wells, constructing a comprehensive status assessment model, performing multi-label identification of the groundwater environment monitoring wells according to the defect semantic features, and performing a comprehensive status assessment according to the defect category labels;

[0058] S106, generating operation and maintenance information according to the evaluation results of the comprehensive status assessment, selecting the groundwater environment monitoring wells that meet the requirements to read the water quality detection data, and extracting the spatio-temporal features of the water quality detection data;

[0059] S108, predicting the groundwater environment quality in the region according to the spatio-temporal features, and generating a monitoring warning through the prediction results of the groundwater environment quality in the region.

[0060] It should be noted that the basic information, basic conditions, structural conditions, and connectivity conditions of the groundwater environment monitoring wells in the region are obtained as the multi-source survey data of the groundwater monitoring wells. The basic information includes the name number, monitoring well type, monitoring well address, affiliated basin, location relationship with the site, etc. The basic conditions include the monitoring well type, wellhead protection device, well pipe type, preservation status of the upper well pipe, outer diameter and wall thickness of the well pipe, well depth, etc. The structural conditions include the number of water intake sections, length of the well wall pipe, length of the filter pipe, length of the sedimentation pipe, monitored layer, etc. The connectivity conditions include the data of the permeability sensitivity experiment work carried out, oscillator details, sampling frequency, excitation head, water level recovery situation, etc. The multi-source survey data are merged and matched using a unique identifier (such as the groundwater environment monitoring well number) to ensure that the data of each monitoring well can be accurately associated; the matched multi-source survey data are subjected to data cleaning to process duplicate data, outliers, and missing values, and the dimensionality of the data-cleaned multi-source survey data is reduced. According to the status assessment requirements of the groundwater environment monitoring wells, key features related to the monitoring well status are selected, redundant or irrelevant features are removed, the data dimension is reduced, and the format and unit of the data are unified through standardization processing for subsequent analysis. After the processing is completed, the preprocessed multi-source survey data are obtained.

[0061] Read historical multi-source survey data according to the unique identifier corresponding to the groundwater environment monitoring well, and obtain the status evaluation label corresponding to the historical multi-source survey data, which is divided into good monitoring function, general monitoring function, and poor monitoring function; extract the historical multi-source survey data corresponding to the good monitoring function and general monitoring function labels as the historical multi-source survey data with qualified monitoring function; train a generative adversarial network according to the obtained historical multi-source survey data, and the generative adversarial network can model the latent space corresponding to the multi-source survey data under the condition of qualified monitoring function, obtain the generator model and the discriminator model, use the generator model to learn the latent distribution of the multi-source survey data, and use the discriminator module to fit the latent distribution of the multi-source survey data. Take the obtained generator model as the decoder, and train the autoencoder network with the obtained historical multi-source survey data, and train with the goal of minimizing the reconstruction error, and obtain the encoder to map the preprocessed multi-source survey data to the latent space; the vector in the latent space can represent the important information of the sample, can reduce the influence of randomness, obtain the latent feature sample corresponding to the historical multi-source survey data closest to the preprocessed multi-source survey data in the latent space, and decode through the decoder to obtain a reconstructed data similar but different from the input multi-source survey data; obtain the residual between the preprocessed multi-source survey data and the reconstructed data, and judge whether there are defects in the groundwater environment monitoring well according to the comparison result between the residual and the preset difference threshold, so as to realize the preliminary status evaluation.

[0062] Figure 2 The flowchart of extracting the defect semantic features of the unqualified groundwater environment monitoring well is shown.

[0063] According to the embodiments of the present invention, read the unqualified groundwater environment monitoring wells through the preliminary status evaluation, and extract the defect semantic features from the multi-source survey data of the unqualified groundwater environment monitoring wells, specifically:

[0064] S202, obtain the preliminary status estimation result of the groundwater environment monitoring wells in the region, screen the groundwater environment monitoring wells with defects according to the preliminary status estimation result, and generate an unqualified label;

[0065] S204, extract the multi-source survey data corresponding to the groundwater environment monitoring wells with unqualified labels, and use the preset shallow convolutional neural network to extract and compress the defect features of the multi-source survey data. The shallow convolutional neural network consists of three convolutional layers, three max-pooling layers and two fully connected layers, and obtain the one-dimensional feature vector of the unqualified groundwater environment monitoring well in the fully connected layer;

[0066] S206. Obtain the monitoring well damage instances with severity annotations based on historical multi-source survey data, perform clustering analysis through the monitoring well damage instances, obtain the clustering clusters corresponding to the last clustering result after iterative clustering, generate subsets of defect data, and determine the defect types corresponding to each subset of defect data;

[0067] S208. Extract defect attributes from the monitoring well damage instance samples in each subset of defect data, integrate the defect attribute vectors to construct a defect attribute set, use similarity calculation to select the n attribute vectors with the highest similarity in the defect attribute set using the one-dimensional feature vector, and splice the one-dimensional feature vector with the selected attribute vectors to generate defect semantic features.

[0068] It should be noted that a preset shallow convolutional neural network learns task-specific information and extracts defect features corresponding to groundwater environment monitoring wells with non-compliance labels. The shallow convolutional neural network consists of three convolutional layers, three max-pooling layers, and two fully connected layers. A max-pooling layer is provided after each convolutional layer to prevent overfitting of the shallow convolutional neural network and reduce computational costs. The output of the fully connected layer of the shallow convolutional neural network is used as a one-dimensional feature vector as the defect feature. Each type of defect of the groundwater environment monitoring well can be described by using multiple attributes. Defect attributes are obtained based on the monitoring well damage instances with severity annotations, such as defect location, defect size, and defect structure. Describing the defect semantic attributes through attribute vectors can enhance the internal connection of the defect semantic attributes, better reflect the occurrence process of the defects, use similarity calculation to select the n attribute vectors with the highest similarity in the defect attribute set using the one-dimensional feature vector, reveal the association between features, and splice the obtained one-dimensional feature vector with the attribute vectors to generate defect semantic features.

[0069] Construct a comprehensive status evaluation model based on the defect semantic feature extraction module and the relationship network module. The relationship network module consists of two fully connected layers. Use the monitoring well damage instance samples with severity annotations in each subset of defect data to construct training samples, obtain the relationship scores of each training sample, and train the model according to the minimum mean square error between the relationship scores and the one-hot encoding of the training samples; in the first fully connected layer, activate the defect semantic features obtained by the defect semantic feature extraction module using the ReLU activation function and output them, and in the second fully connected layer, use the Sigmoid function to map the output of the first fully connected layer to the interval (0,1) to generate relationship scores; perform multi-label recognition of the corresponding defects of the groundwater environment monitoring wells according to the relationship scores, determine the defect category labels and corresponding severities of the non-compliant groundwater environment monitoring wells, and generate a comprehensive status evaluation result based on the defect category labels and severities based on a preset evaluation threshold interval.

[0070] It should be noted that the comprehensive status evaluation results of non-compliant groundwater environmental monitoring wells in the acquisition area are obtained, and the groundwater environmental monitoring wells with poor monitoring functions and general monitoring functions are selected for suspension according to the comprehensive status evaluation results; the defect category label with the highest severity corresponding to each groundwater environmental monitoring well is obtained from the suspended groundwater environmental monitoring wells with general monitoring functions, and the severity trend change corresponding to the defect category label is obtained according to historical multi-source survey data; the severity trend change is compared with the trend change threshold, and the suspended groundwater environmental monitoring wells with general monitoring functions with a value less than the trend change threshold are cancelled from suspension. Finally, the operation and maintenance information is generated based on the unique identifiers corresponding to the suspended groundwater environmental monitoring wells with poor monitoring functions and general monitoring functions, and is sent and displayed according to a preset method.

[0071] Figure 3 The flowchart shows the spatio-temporal characteristics of water quality detection data in the extraction area.

[0072] According to the embodiments of the present invention, groundwater environmental monitoring wells that meet the requirements are selected to read water quality detection data, and the spatio-temporal characteristics of the water quality detection data are extracted, specifically:

[0073] S302, obtain the groundwater environmental monitoring wells in the area with a good monitoring function in the comprehensive status evaluation results, extract the water quality detection data sequence within a preset time, and calculate the Pearson correlation coefficient based on the water quality detection data sequence to obtain the correlation of monitoring data between groundwater environmental monitoring wells;

[0074] S304, determine the connection relationship between groundwater monitoring wells according to the distance and monitoring data correlation between groundwater environmental monitoring wells, generate the corresponding topological structure and represent it graphically, use the groundwater environmental monitoring wells with good monitoring functions in the area as graph nodes, and set the edge structure between nodes according to the connection relationship;

[0075] S306, construct an adjacency matrix based on the edge structure, use a graph attention network to learn the graph representation, take the adjacency matrix as the input, obtain the attention weights between any two nodes in the neighborhood according to the multi-head self-attention mechanism, use the attention weights to perform weighted aggregation on the adjacency matrix, update the feature representation, and obtain the spatial characteristics of the groundwater environmental quality in the area;

[0076] S308, import the updated feature representation into a gated recurrent unit, use the hidden state update to obtain the temporal characteristics of the groundwater environmental quality in the area, introduce a temporal attention mechanism to perform attention weighting on the temporal characteristics at different times, and fuse the weighted temporal characteristics with the spatial characteristics after pooling operation to obtain the spatio-temporal characteristics of the water quality detection data in the area.

[0077] It should be noted that the attention weights between any two nodes in the neighborhood are obtained according to the multi-head self-attention mechanism. The multi-head self-attention mechanism accelerates the calculation speed of the model through parallel computing. The multi-head self-attention mechanism is used to scale the features corresponding to each monitoring well node according to the attention scores, and aggregate the relevant information of the neighboring monitoring well nodes. In the multi-layer structure, the splicing and integration of the last layer uses the average operation method to obtain the feature representation corresponding to the monitoring well node , specifically:

[0078] ,

[0079] where is the activation function, is the total number of attention heads, is the number of items in the attention head, is the parameter vector between nodes in the attention head, is the weight matrix between nodes in the attention head, is the feature representation of the neighboring monitoring well nodes.

[0080] A gated recurrent unit is added after the graph attention network. The gated recurrent unit overcomes the problems of vanishing gradients and exploding gradients in complex networks, and updates the feature representation corresponding to the monitoring well node by using the hidden state, captures the time dependence, introduces a time attention mechanism in the update of the hidden state to dynamically adjust the weights between the hidden states, performs attention weighting on the time features at different times, and highlights the more important time features. The weighted time features are fused with the spatial features after the pooling operation to obtain the spatio-temporal features of the water quality detection data in the region.

[0081] A groundwater environmental quality prediction model is constructed based on the graph attention network and the gated recurrent unit. The relevant standards such as the "Groundwater Quality Standard" (GB / T 14848-2017) are flexibly written into the groundwater environmental quality prediction model, and quality prediction is carried out on the water quality detection data. The groundwater environmental quality prediction levels are Class I, Class II, Class III, Class IV, and Class V, and the quality prediction results automatically form the corresponding evaluation result charts. The spatio-temporal features of the water quality detection quantity in the region are obtained through the graph attention network and the gated recurrent unit, and the spatio-temporal features are used to obtain the groundwater environmental quality prediction results in the region through the fully connected layer; dynamic analysis is carried out according to the groundwater environmental quality prediction results within the preset time period to obtain the trend changes, and the absolute threshold and the trend change threshold are set. When the obtained groundwater environmental quality prediction results or trend changes in the region are greater than the corresponding absolute threshold or trend change threshold, a monitoring warning is generated and sent and displayed in a preset manner.

[0082] Obtain the water quality detection data of each groundwater environment monitoring well in the area and the prediction results of the groundwater environment quality. Based on the water quality detection data, determine whether there is pollution in the corresponding groundwater environment. Obtain the groundwater environment monitoring wells with pollution. Generate a pollution portrait of the monitoring wells according to their groundwater environment quality prediction results, pollutant categories, and pollutant concentrations. Cluster the groundwater environment monitoring wells with pollution in the area based on the pollution portrait. Divide similar pollution areas according to the clustering results. Sort the pollution areas using the average groundwater environment quality prediction results. Select the area with the highest pollution degree based on the sorting results. Trace the pollution source in the selected area in combination with the pollutant categories. After tracing the pollution source, conduct a dynamic analysis of groundwater pollution according to the distribution of the pollution area. Retrieve historical pollution monitoring information according to the pollutant categories and the geographical and hydrological characteristics in the area. Use the historical pollution monitoring information to supplement the training of the groundwater environment quality prediction model. Generate a predicted pollution path according to the groundwater environment quality prediction results of each groundwater environment monitoring well. Obtain the next pollution area through the predicted pollution path, and generate a groundwater environmental pollution warning.

[0083] Figure 4 Fig. shows a block diagram of a full-process data management system for groundwater environment monitoring wells.

[0084] In a second aspect of the present invention, there is provided a full-process data management system 4 for groundwater environment monitoring wells, which includes: a multi-source survey data integration module 401, a monitoring well status evaluation module 402, a monitoring well operation and maintenance warning module 403, a groundwater environment quality analysis module 404, and a groundwater monitoring warning module 405;

[0085] The multi-source survey data integration module 401 is responsible for obtaining multi-source survey data of groundwater environment monitoring wells and performing preprocessing;

[0086] The monitoring well status evaluation module 402 is responsible for judging whether there are defects in the groundwater environment monitoring wells according to the preprocessed multi-source survey data, realizing preliminary status evaluation, extracting defect semantic features from the multi-source survey data of the groundwater environment monitoring wells that do not meet the standard in the preliminary status evaluation, constructing a comprehensive status evaluation model, performing multi-label identification of the groundwater environment monitoring wells according to the defect semantic features, and performing comprehensive status evaluation according to the defect category labels;

[0087] The monitoring well operation and maintenance warning module 403 is responsible for generating operation and maintenance information according to the evaluation results of the comprehensive status evaluation, and sending and displaying it through a preset method;

[0088] The groundwater environment quality analysis module 404 is responsible for selecting the groundwater environment monitoring wells that meet the requirements in the area, reading the water quality detection data, and extracting the spatio-temporal characteristics of the water quality detection data to predict the groundwater environment quality in the area;

[0089] The groundwater monitoring and early warning module 405 is responsible for generating water environment monitoring and early warning based on the prediction results of the groundwater environmental quality within the region, and sending and displaying them in a preset manner.

[0090] Preferably, the full-process data management system for groundwater environmental monitoring wells further includes a data visualization module, which is used to visually display multi-source survey data, status evaluation results, operation and maintenance information, water quality detection data, and monitoring and early warning results, facilitating users to intuitively understand the status of groundwater environmental monitoring wells and the trend of water quality changes.

[0091] The third embodiment of the present invention provides a computer-readable storage medium, which includes a program for the full-process data management method of groundwater environmental monitoring wells. When the program for the full-process data management method of groundwater environmental monitoring wells is executed by a processor, the steps of the full-process data management method of groundwater environmental monitoring wells are implemented.

[0092] In several embodiments provided in the present application, it should be understood that the disclosed method can be implemented in other ways. The above-described embodiments are merely illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as: multiple modules 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 various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or modules can be electrical, mechanical, or other forms.

[0093] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium includes: removable storage devices, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks, or optical discs and other various media that can store program codes.

[0094] Alternatively, if the above integrated modules of the present invention are implemented in the form of software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as removable storage devices, ROM, RAM, magnetic disks, or optical discs that can store program codes.

[0095] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention.

Claims

1. A method for managing the whole process data of groundwater environment monitoring wells, characterized in that: The following steps are involved: Acquire multi-source survey data of groundwater environment monitoring wells and perform preprocessing, and perform preliminary status assessment on the groundwater environment monitoring wells based on the preprocessed multi-source survey data; Read the substandard groundwater environment monitoring wells through the preliminary status assessment, extract the defect semantic features based on the multi-source survey data of the substandard groundwater environment monitoring wells, build a comprehensive status assessment model, perform multi-label recognition of the groundwater environment monitoring wells based on the defect semantic features, and perform comprehensive status assessment based on the defect category labels; Generate operation and maintenance information according to the evaluation results of the comprehensive status evaluation, select groundwater environment monitoring wells that meet the requirements to read water quality detection data, and extract the spatiotemporal characteristics of the water quality detection data; Predict the groundwater environment quality in the region based on the temporal and spatial characteristics, and generate monitoring and early warning based on the groundwater environment quality prediction results in the region; The substandard groundwater environment monitoring wells are read through the preliminary status assessment, and the defect semantic features are extracted based on the multi-source survey data of the substandard groundwater environment monitoring wells, specifically: Obtain preliminary status estimation results of groundwater environment monitoring wells in the region, screen defective groundwater environment monitoring wells according to the preliminary status estimation results, and generate non-compliance labels; Extract multi-source survey data corresponding to groundwater environment monitoring wells with substandard labels, and use a preset shallow convolutional neural network to extract and compress defect features of the multi-source survey data. The shallow convolutional neural network consists of three convolutional layers, three maximum pooling layers and two fully connected layers. The one-dimensional feature vector of the substandard groundwater environment monitoring wells is obtained in the fully connected layer; Acquire monitoring well damage instances with severity labels based on historical multi-source survey data, perform cluster analysis on the monitoring well damage instances, obtain cluster clusters corresponding to the last clustering result after iterative clustering, generate defect data subsets, and determine the defect type corresponding to each defect data subset; In each defect data subset, defect attributes are extracted according to the monitoring well damage instance samples, and the defect attribute vectors are integrated to construct a defect attribute set. The one-dimensional feature vector is used for similarity calculation to select the n attribute vectors with the highest similarity in the defect attribute set, and the one-dimensional feature vector is concatenated with the selected attribute vector to generate defect semantic features.

2. A method for managing the full process data of groundwater environment monitoring wells according to claim 1, characterized in that: Obtain multi-source survey data of groundwater environment monitoring wells and perform preprocessing, and conduct preliminary status assessment of the groundwater environment monitoring wells based on the preprocessed multi-source survey data, specifically: Obtaining basic information, foundation conditions, structural conditions, and connectivity conditions of groundwater environment monitoring wells in the region as multi-source survey data of groundwater monitoring wells, merging the multi-source survey data, and matching them using unique identifiers; Perform data cleaning on the matched multi-source survey data, process duplicate data, outliers and missing values, and perform dimensionality reduction and standardization on the cleaned multi-source survey data to obtain pre-processed multi-source survey data; Read historical multi-source survey data according to the unique identifier corresponding to the groundwater environment monitoring well, obtain the status evaluation label corresponding to the historical multi-source survey data, and select the historical multi-source survey data that meets the monitoring function standard based on the status evaluation label; A generative adversarial network is trained according to the acquired historical multi-source survey data to obtain a generator model and a discriminator model, the acquired generator model is used as a decoder, and an autoencoder network is trained with the acquired historical multi-source survey data to obtain an encoder to map the pre-processed multi-source survey data to a latent space; Acquire latent feature samples corresponding to historical multi-source survey data closest to the pre-processed multi-source survey data in the latent space, and decode through a decoder to acquire reconstructed data that is similar to but not identical to the input multi-source survey data; The residuals of the preprocessed multi-source survey data and the reconstructed data are obtained, and the presence of defects in the groundwater environment monitoring well is determined based on the comparison results of the residuals with a preset difference threshold, so as to achieve a preliminary status assessment.

3. A method for managing the full process data of groundwater environment monitoring wells according to claim 1, characterized in that: A comprehensive status assessment model is constructed to perform multi-label identification of groundwater environment monitoring wells based on the defect semantic features, and to perform comprehensive status assessment based on the defect category labels, specifically: A comprehensive state assessment model is constructed based on a defect semantic feature extraction module and a relationship network module. The relationship network module is composed of two fully connected layers. A training sample is constructed using monitoring well damage instance samples with severity annotations in each defect data subset. A relationship score of each training sample is obtained. The minimum mean square error of the unique hot encoding of the training sample is calculated based on the relationship score to perform model training. In the first fully connected layer, the defect semantic features obtained by the defect semantic feature extraction module are activated and output using the ReLU activation function, and in the second fully connected layer, the output of the first fully connected layer is mapped to the (0,1) interval using the Sigmoid function to generate a relationship score; Multi-label identification of defects corresponding to groundwater environment monitoring wells is performed based on the relationship scores, the defect category labels and corresponding severity of substandard groundwater environment monitoring wells are determined, and a comprehensive status assessment result is generated based on the defect category labels and severity based on a preset assessment threshold interval.

4. A method for managing the full process data of groundwater environment monitoring wells according to claim 1, characterized in that: Generate operation and maintenance information according to the evaluation result of the comprehensive status evaluation, specifically: Obtain comprehensive status evaluation results of substandard groundwater environment monitoring wells in the region, and select groundwater environment monitoring wells with poor monitoring functions and general monitoring functions for suspension based on the comprehensive status evaluation results; Obtain the defect category label of the highest severity corresponding to each underground environmental monitoring well in the suspended groundwater environmental monitoring wells with general monitoring functions, and obtain the severity trend change corresponding to the defect category label based on historical multi-source survey data; The severity trend change is compared with the trend change threshold, and the groundwater environment monitoring with general monitoring function that is less than the trend change threshold is unsuspended. Finally, operation and maintenance information is generated based on the unique identifiers corresponding to the suspended groundwater environment monitoring wells with poor monitoring functions and general monitoring functions, and sent and displayed according to a preset method.

5. A method for managing the full process data of groundwater environment monitoring wells according to claim 1, characterized in that: Select groundwater environment monitoring wells that meet the requirements to read water quality test data and extract the spatiotemporal characteristics of the water quality test data, specifically: Obtain groundwater environment monitoring wells with good monitoring functions as shown in the comprehensive status assessment results in the region, extract water quality detection data sequences within a preset time, and calculate the Pearson correlation coefficient based on the water quality detection data sequences to obtain the correlation of monitoring data between groundwater environment monitoring wells; Determine the connection relationship between groundwater monitoring wells according to the distance between them and the correlation of monitoring data, generate the corresponding topological structure and represent it in a graph, take the groundwater monitoring wells with good monitoring functions in the area as graph nodes, and set the edge structure between nodes according to the connection relationship; An adjacency matrix is ​​constructed based on the edge structure, and the graph representation is learned using a graph attention network. The adjacency matrix is ​​used as input, and the attention weight between any two nodes in the neighborhood is obtained according to a multi-head self-attention mechanism. The adjacency matrix is ​​weightedly aggregated using the attention weight, and the feature representation is updated to obtain the spatial characteristics of the groundwater environmental quality in the region. The updated feature representation is imported into the gated recurrent unit, and the hidden state is used to update the temporal characteristics of the groundwater environmental quality in the region. The temporal attention mechanism is introduced to weight the attention of the temporal features at different moments. The weighted temporal features are fused with the spatial features after pooling operation to obtain the spatiotemporal characteristics of water quality detection data in the region.

6. A method for managing the full process data of groundwater environment monitoring wells according to claim 1, characterized in that: The groundwater environment quality in the region is predicted based on the spatiotemporal characteristics, and monitoring and early warning are generated based on the groundwater environment quality prediction results in the region, specifically: A groundwater environmental quality prediction model is constructed based on a graph attention network and a gated recurrent unit, and the spatiotemporal characteristics of the number of water quality detections in the region are obtained through the graph attention network and the gated recurrent unit, and the spatiotemporal characteristics are passed through a fully connected layer to obtain the groundwater environmental quality prediction results in the region; Dynamic analysis is performed based on the groundwater environmental quality prediction results within the preset time period to obtain trend changes, set absolute thresholds and trend change thresholds, and when the obtained groundwater environmental quality prediction results or trend changes in the area are greater than the corresponding absolute thresholds or trend change thresholds, a monitoring warning is generated and sent and displayed in the preset manner.

7. A full-process data management system for groundwater environment monitoring wells, characterized in that: A method for managing the full-process data of a groundwater environment monitoring well as described in any one of claims 1 to 6 is implemented, wherein the system comprises: a multi-source survey data integration module, a monitoring well status assessment module, a monitoring well operation and maintenance early warning module, a groundwater environment quality analysis module, and a groundwater monitoring early warning module; The multi-source survey data integration module is responsible for acquiring multi-source survey data of groundwater environment monitoring wells and performing pre-processing; The monitoring well status assessment module is responsible for judging whether the groundwater environment monitoring well has defects based on the preprocessed multi-source survey data, realizing preliminary status assessment, extracting defect semantic features based on the multi-source survey data of the groundwater environment monitoring well that does not meet the preliminary status assessment standard, building a comprehensive status assessment model, performing multi-label recognition of the groundwater environment monitoring well based on the defect semantic features, and performing comprehensive status assessment based on the defect category labels; The monitoring well operation and maintenance early warning module is responsible for generating operation and maintenance information according to the evaluation results of the comprehensive status evaluation, and sending and displaying it in a preset manner; The groundwater environment quality analysis module is responsible for selecting groundwater environment monitoring wells that meet the requirements in the region to read water quality detection data, extracting the spatiotemporal characteristics of the water quality detection data to predict the groundwater environment quality in the region; The groundwater monitoring and early warning module is responsible for generating water environment monitoring and early warning through the groundwater environment quality prediction results in the region, and sending and displaying them in a preset manner.

Images