Environmental pollution monitoring method and device based on evaluation model

Abstract

The invention discloses an environmental pollution monitoring method and device based on an evaluation model, and relates to the related field of environmental pollution monitoring, and the method comprises the steps: analyzing the environmental pollution evaluation features of the evaluation model, and building a multi-dimensional feature list of the evaluation model; carrying out existence form, diffusion mode, pollution target and pollution analysis on the environmental pollution poison monitoring target to obtain pollution poison monitoring characteristics, and carrying out monitoring evaluation target response analysis to obtain response evaluation demand parameters; performing traversal matching on the response evaluation demand parameters and the model evaluation features to obtain missing demand parameters; performing new training on the evaluation model to obtain an enhanced evaluation model; and carrying out pollutant monitoring evaluation on the environmental monitoring data by using the enhanced evaluation model to obtain an environmental pollution monitoring evaluation result, and carrying out pollution monitoring feedback. The technical problem that existing environmental pollution monitoring is insufficient in evaluation result accuracy and reliability is solved, and the technical effect of improving the evaluation result accuracy and reliability is achieved.

Description

Technical Field

[0001] The present application relates to the field of environmental pollution monitoring, and in particular to an environmental pollution monitoring method and device based on an evaluation model. Background Art

[0002] Environmental pollution is one of the major challenges facing the world today. The accuracy of its monitoring and evaluation is directly related to the effective implementation of environmental protection measures and the maintenance of ecological balance. At present, the monitoring of environmental pollution mainly relies on some statistical prediction models. These methods evaluate the environmental pollution status by establishing prediction models based on historical data. However, although the statistical prediction models can process a large amount of data, they are often limited by the fixed structure of the model and the limitations of the training data. They are difficult to adapt to the complex and changeable environmental pollution conditions, resulting in insufficient accuracy and reliability of the evaluation results.

[0003] Among the current related technologies, environmental pollution monitoring based on assessment models has technical problems such as insufficient accuracy and reliability of assessment results. Summary of the invention

[0004] The present application provides an environmental pollution monitoring method and device based on an assessment model, adopts an analytical assessment model, establishes a multidimensional feature list, analyzes the characteristics of pollutants, obtains response assessment demand parameters, matches the demand parameters with model characteristics, determines missing parameters, conducts additional training on the assessment model, obtains an enhanced model, uses the enhanced model to evaluate environmental monitoring data, obtains monitoring results and provides feedback, and other technical means to achieve the technical effect of improving the accuracy and reliability of the assessment results.

[0005] The present application provides an environmental pollution monitoring method based on an assessment model, including: parsing the environmental pollution assessment characteristics of the assessment model, establishing a multidimensional feature list of the assessment model; performing existence form, diffusion mode, pollution target, and pollution analysis on environmental pollution toxic monitoring targets to obtain pollution toxic monitoring characteristics, and performing monitoring assessment target response analysis based on the pollution toxic monitoring characteristics to obtain response assessment demand parameters; using the response assessment demand parameters to traverse and match model assessment characteristics in the multidimensional feature list of the assessment model to obtain missing demand parameters; performing additional training on the assessment model based on the missing demand parameters to obtain an enhanced assessment model; using the enhanced assessment model to perform pollutant monitoring and assessment on environmental monitoring data to obtain environmental pollution monitoring and assessment results, and performing pollution monitoring feedback on the environmental pollution monitoring and assessment results.

[0006] In a possible implementation, the environmental pollution assessment features of the assessment model are analyzed, a multidimensional feature list of the assessment model is established, and the following processing is performed: according to the multi-dimensional characteristics of training sample features, model architecture features, training methods, output forms, and model performance features, the features of the assessment model are analyzed to obtain multi-dimensional assessment features; according to the dimensional feature labels of the multi-dimensional assessment features, the multi-dimensional assessment features are integrated to construct a multi-dimensional feature list of the assessment model.

[0007] In a possible implementation, the evaluation model is feature analyzed according to the multiple dimensions of training sample features, model architecture features, training methods, output forms, and model performance features to obtain multi-dimensional evaluation features, and the following processing is performed: feature analysis of the training samples in terms of data source, sample contamination toxicant type, data features, sample quantity distribution, and data labels to obtain contaminated training sample features; logical feature analysis of pollutant evaluation is performed according to the model type and model structure of the model architecture to obtain model architecture features; training method feature analysis is performed according to the training type and training parameters to obtain training method features; output form feature analysis is performed according to the type, dimension, and range of the model output to obtain output form features; performance feature analysis is performed according to the accuracy, real-time, and robustness of the model output results to obtain model performance features; the contaminated training sample features, model architecture features, training method features, output form features, and model performance features are combined to obtain the multi-dimensional evaluation features.

[0008] In a possible implementation, before additional training is performed on the assessment model based on the missing requirement parameters, the following processing is performed: determine whether the missing requirement parameters exist; when they do not exist, use the assessment model to perform pollutant monitoring and assessment on environmental monitoring data; when they exist, perform training method analysis based on the missing requirement parameters and the assessment model to obtain an additional training method; and perform additional training on the assessment model based on the missing requirement parameters and the additional training method.

[0009] In a possible implementation, the following processing is performed: the newly added training method includes: transfer learning, incremental learning or integrated learning.

[0010] In a possible implementation, the environmental pollution and toxic substance monitoring target is analyzed in terms of existence form, diffusion mode, pollution target, and pollution to obtain pollution and toxic substance monitoring characteristics, and the following processing is performed: based on the environmental pollution and toxic substance monitoring target, sample data of the pollution target is collected, including historical pollution record data and experimental data, to record the monitoring concentration, existence form, propagation change characteristics, pollution target impact site, and pollution impact data of environmental pollution and toxic substances; based on the existence form and pollution target impact site, the monitoring characteristics of the collected data are determined; based on the sample data of the pollution target, the spatiotemporal influence relationship between the propagation change characteristics and the monitoring concentration and pollution impact data is fitted to determine the toxic substance diffusion monitoring characteristics; based on the collected data monitoring characteristics and the toxic substance diffusion monitoring characteristics, the pollution and toxic substance monitoring characteristics are obtained.

[0011] In a possible implementation, a monitoring and evaluation target response analysis is performed based on the pollution and toxicant monitoring characteristics to obtain response evaluation requirement parameters, and the following processing is performed: based on the monitoring constraints of pollution and toxicant, the accuracy relationship of the collected data monitoring characteristics and the diffusion path range of the toxicant diffusion monitoring characteristics are fitted for the pollution and toxicant monitoring characteristics to obtain the monitoring response constraints of the pollution and toxicant monitoring characteristics; with the monitoring response constraints of the pollution and toxicant monitoring characteristics as the target, the pollution identification process of the pollution and toxicant monitoring characteristics is analyzed to obtain the response evaluation requirement parameters.

[0012] In a possible implementation, after obtaining the enhanced assessment model, the following processing is performed: establishing the identification label and input layer screening and processing rules of the enhanced assessment model; obtaining environmental monitoring data and target identification labels; matching and identifying the target identification labels with the enhanced assessment model to obtain a matching model or a matching integrated sub-model; screening and processing the environmental monitoring data according to the input layer screening and processing rules, and inputting the screened environmental monitoring data into the matching model or the matching integrated sub-model.

[0013] In a possible implementation, the missing requirement parameters are obtained by performing the following processing: performing data standardization and feature dimension alignment processing on the response evaluation requirement parameters and the model evaluation features in the multidimensional feature list; configuring matching rules, including data matching, range matching, and logical matching, and matching the missing requirement parameters with the model evaluation features in the multidimensional feature list item by item based on the matching rules to obtain the missing parameters and the missing types; and obtaining the missing requirement parameters based on the missing parameters and the missing types.

[0014] The present application also provides an environmental pollution monitoring device based on an assessment model, including: an environmental pollution assessment feature parsing module, which is used to parse the environmental pollution assessment features of the assessment model and establish a multidimensional feature list of the assessment model; a response assessment demand parameter acquisition module, which is used to analyze the existence form, diffusion mode, pollution target, and pollution of the environmental pollution toxic monitoring target to obtain the pollution toxic monitoring features, and perform monitoring and assessment target response analysis based on the pollution toxic monitoring features to obtain response assessment demand parameters; a missing demand parameter acquisition module, which is used to use the response assessment demand parameters to traverse and match the model assessment features in the multidimensional feature list of the assessment model to obtain missing demand parameters; a model additional training module, which is used to perform additional training on the assessment model according to the missing demand parameters to obtain an enhanced assessment model; a pollutant monitoring and assessment module, which is used to use the enhanced assessment model to perform pollutant monitoring and assessment on environmental monitoring data, obtain environmental pollution monitoring and assessment results, and use the environmental pollution monitoring and assessment results for pollution monitoring feedback.

[0015] The environmental pollution monitoring method and device based on the evaluation model proposed in this application first analyzes the environmental pollution evaluation characteristics of the evaluation model, establishes a multidimensional feature list of the evaluation model, and then analyzes the existence form, diffusion mode, pollution target, and pollution of the environmental pollution toxic monitoring target to obtain the pollution toxic monitoring characteristics, and performs monitoring evaluation target response analysis based on the pollution toxic monitoring characteristics to obtain response evaluation demand parameters, and then uses the response evaluation demand parameters to traverse and match the model evaluation characteristics in the multidimensional feature list of the evaluation model to obtain missing demand parameters, and then performs additional training on the evaluation model based on the missing demand parameters to obtain an enhanced evaluation model, and finally uses the enhanced evaluation model to perform pollutant monitoring and evaluation on environmental monitoring data to obtain environmental pollution monitoring and evaluation results, and the environmental pollution monitoring and evaluation results are used for pollution monitoring feedback. The technical effect of improving the accuracy and reliability of the evaluation results is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the device according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0017] Figure 1 A schematic diagram of the process of an environmental pollution monitoring method based on an assessment model provided in an embodiment of the present application.

[0018] Figure 2 A schematic diagram of the structure of an environmental pollution monitoring device based on an evaluation model provided in an embodiment of the present application.

[0019] Explanation of the accompanying drawings: environmental pollution assessment feature analysis module 10, response assessment requirement parameter acquisition module 20, missing requirement parameter acquisition module 30, model additional training module 40, pollutant monitoring and assessment module 50. DETAILED DESCRIPTION

[0020] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0021] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0022] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions. For example, a process, method, device, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0023] The present application embodiment provides an environmental pollution monitoring method based on an evaluation model, such as Figure 1 As shown, the method includes: Step S100, analyzing the environmental pollution assessment features of the assessment model and establishing a multi-dimensional feature list of the assessment model.

[0024] Specifically, obtain an existing environmental pollution assessment model, which is an algorithm model based on machine learning and is used to assess the extent or impact of environmental pollution. Identify the key features of the model used to assess environmental pollution by reading the model's documentation or source code, or through the feature engineering steps during model training. These features include the type and concentration of pollutants, the location of pollution sources, wind direction and speed, etc. Organize the identified assessment features into a multidimensional feature list, with each feature corresponding to a dimension, for subsequent matching with the monitoring target features.

[0025] In a possible implementation, the environmental pollution assessment features of the analytical assessment model are established to establish a multidimensional feature list of the assessment model. Step S100 further includes step S110, which performs feature analysis on the assessment model according to the training sample features, model architecture features, training methods, output forms, and model performance features to obtain multidimensional assessment features. Specifically, the original data set used to train the assessment model is obtained from the model developer or data provider. The original data is preprocessed, including data cleaning (removing duplicates, missing or outliers), data conversion (normalization, standardization), etc. Key features are extracted from the preprocessed data, which are part of the model input, such as pollutant concentration, timestamp, geographic location, etc. Statistical methods (such as mean, variance, correlation analysis) or visualization tools (such as scatter plots, histograms) are used to analyze the distribution and correlation of these features in the training samples.

[0026] Consult the model's documentation or source code to understand the model's architecture, including the structure of the input layer, hidden layer, and output layer, and how the layers are connected. Use tools (such as TensorBoard, Netron) to visualize the model architecture, analyze the number, type, and initialization methods of the model's parameters (such as weights and biases), and how they affect the model's performance.

[0027] Identify the algorithm used to train the model from the model documentation or source code, such as gradient descent, Adam optimizer, etc. Analyze the key steps in the training process, such as the adjustment of learning rate, the selection of batch size, the application of regularization methods, etc. Review the logs during the training process to understand the performance and performance changes of the model at different training stages.

[0028] Determine the data type of the model output, such as continuous values ​​(pollutant concentration predictions), categorical labels (pollution level classification), etc. Analyze the format of the model output, such as JSON, CSV, API response, etc.

[0029] According to the application scenario and objectives of the model, select appropriate evaluation indicators, such as accuracy, recall, F1 score, mean square error, etc. Evaluate the performance of the model on the validation set or test set, and use the selected evaluation indicators to measure the accuracy and generalization ability of the model. Compare the performance of the model with the baseline model or other related models to understand the advantages and disadvantages of the model and the room for improvement.

[0030] Step S120, according to the dimensional feature labels of the multidimensional evaluation features, the multidimensional evaluation features are integrated to construct a multidimensional feature list of the evaluation model. Specifically, a unique label is assigned to each multidimensional evaluation feature, which can be a descriptive name or code for reference and identification. By manually sorting or using automated tools (such as Excel, Python scripts), features with the same or similar attributes are classified together to form a structured multidimensional feature list, which contains a detailed description of each feature, data type, value range, source (such as training data, model architecture, training method, etc.) and other information. This implementation method provides an accurate and comprehensive information basis for subsequent steps by performing comprehensive feature analysis and integration of the evaluation model, which helps to quickly identify and utilize these features in subsequent steps, thereby improving the accuracy and efficiency of environmental pollution monitoring methods.

[0031] In a possible implementation, the evaluation model is feature parsed according to the training sample features, model architecture features, training methods, output forms, and model performance features to obtain multi-dimensional evaluation features. Step S110 further includes step S111, which performs feature parsing of the training sample data source, sample contamination type, data features, sample quantity distribution, and data labels to obtain contaminated training sample features. Specifically, the metadata or documents of the training sample are reviewed to confirm the source channel of the data (such as monitoring stations, laboratories, public databases, etc.). The training samples are traversed to identify different types of contamination based on the chemical identification or classification labels in the samples. Key features are extracted from the raw data using feature engineering methods (such as statistical calculation, feature selection, feature transformation, etc.). The number of samples of various types of contamination and toxic substances is counted, and a sample quantity distribution diagram is drawn. Use a text editor or data visualization tool (such as Excel, Python's Pandas library, etc.) to open the file containing the sample data and its labels, view the contents of the label column, and determine what the label corresponding to each sample is. Alternatively, you can write a script (such as a Python script) to read the data file and extract the contents of the label column. You can use the script to perform statistical analysis on the labels, such as calculating the number and frequency of each label, and output the types and distribution of labels.

[0032] Step S112, perform pollutant assessment logical feature analysis based on the model type and model structure of the model architecture to obtain model architecture features. Specifically, consult the model documentation or source code to confirm which type of model it is (such as neural network, decision tree, support vector machine, etc.). Use the model visualization tools or APIs provided by machine learning frameworks (such as TensorFlow, PyTorch, etc.) to generate a structural diagram of the model, showing the input layer, hidden layer, output layer and their connection methods. Based on the model structure and training sample characteristics, deduce how the model evaluates pollutants based on the input data, that is, determine the rules and processes for the model to evaluate pollutants based on the input data.

[0033] Step S113, perform training method feature analysis based on training type and training parameters to obtain training method features. Specifically, consult the model training document or code to confirm the training type (such as supervised learning, unsupervised learning, reinforcement learning, etc.). List and review the key parameters used in the training process (such as learning rate, batch size, number of iterations, etc.).

[0034] Step S114: Analyze the output form characteristics according to the type, dimension, and range of the model output to obtain the output form characteristics. Specifically, consult the model document or code, confirm the data type of the model output (such as numerical value, classification label, probability distribution, etc.), count the dimensions of the model output (such as single numerical value, vector, matrix, etc.), and determine the data range or limit of the model output (such as a pollution index of 0-100).

[0035] Step S115, analyze the performance characteristics according to the accuracy, real-time performance, and robustness of the model output results to obtain the model performance characteristics. Specifically, use the test set or validation set to evaluate the accuracy of the model (such as classification accuracy, regression error, etc.). Measure the response time of the model from input to output to evaluate its real-time performance. Analyze the stability and robustness of the model by introducing noise, outliers, or changing the distribution of input data.

[0036] Step S116, combining the contaminated training sample features, model architecture features, training method features, output form features and model performance features to obtain the multi-dimensional evaluation features. Specifically, all the features obtained in steps S111 to S115 are sorted, summarized and combined to form a comprehensive multi-dimensional evaluation feature set. This implementation method helps to quickly identify and utilize these features in subsequent steps by combining these features into a comprehensive multi-dimensional evaluation feature set. This structured feature representation method helps to integrate and interoperate with other systems or methods.

[0037] Step S200, analyze the existence form, diffusion mode, pollution target, and contamination of the environmental pollution and toxicant monitoring target to obtain the pollution and toxicant monitoring characteristics, and perform monitoring and evaluation target response analysis based on the pollution and toxicant monitoring characteristics to obtain response evaluation demand parameters.

[0038] Specifically, laboratory analysis, such as microscopic observation, X-ray diffraction (XRD) or scanning electron microscopy (SEM) and other techniques, can be used to determine the physical form of the pollutant (such as gas, liquid, solid). For gases, gas chromatography (GC) or mass spectrometry (MS) can be used for analysis; for liquids and solids, infrared spectroscopy (IR), nuclear magnetic resonance (NMR) or Raman spectroscopy can be used. Chemical analysis techniques, such as titration, spectrophotometry or electrochemical methods, can be used to determine the volatility, solubility and other chemical properties of the pollutant. In addition, by measuring its melting point, boiling point, density and other physical properties, its chemical stability can also be indirectly reflected.

[0039] Study the diffusion modes of pollutants in the environment, such as air transmission, water flow transmission, soil penetration, etc. For air transmission, use gas chromatography-mass spectrometry (GC-MS) technology to monitor the concentration of pollutants in the atmosphere, and combine meteorological data (such as wind speed and wind direction) and geographic information system (GIS) to simulate the diffusion path of pollutants. For water flow transmission, collect water samples through water quality monitoring stations, and use liquid chromatography-mass spectrometry (LC-MS) or inductively coupled plasma mass spectrometry (ICP-MS) and other technologies to analyze the types and concentrations of pollutants in the water. Combine hydrological models (such as SWAT, MIKE, etc.) to simulate the water flow transmission process of pollutants. For soil penetration, soil sampling and laboratory analysis (such as soil gas extraction-gas chromatography) are used to evaluate the penetration of pollutants in the soil. At the same time, soil erosion and deposition models (such as RUSLE, WEPP, etc.) are used to predict the migration and accumulation of pollutants in the soil.

[0040] Identify the objects that pollutants may affect, such as human health, ecological environment, water resources, etc. For human health, evaluate the impact of pollutants on human health through epidemiological surveys, biological monitoring (such as the content of pollutants in blood and urine) and risk assessment models (such as USEPA's IRIS database). For the ecological environment, use ecotoxicology experiments (such as acute toxicity tests, chronic toxicity tests) and ecological risk assessment methods (such as OECD's ecological risk assessment framework) to assess the potential impact of pollutants on ecosystems (such as plants, animals, and microorganisms). For water resources, identify the impact of pollutants on water resources through water quality monitoring and ecological health assessment (such as biodiversity index, comprehensive water quality index).

[0041] Evaluate the toxicity, persistence, bioaccumulation, etc. of pollutants to determine their potential harm to the environment and humans. Use toxicity tests (such as acute toxicity tests, genotoxicity tests, carcinogenicity tests) and bioaccumulation experiments (such as determination of bioconcentration factor BCF and biomagnification factor BMF) to evaluate the toxicity, persistence and bioaccumulation of pollutants. Combine environmental persistence assessment models (such as PBTK models) and biogeochemical cycle models to predict the long-term behavior and impact of pollutants in the environment.

[0042] Based on the above analysis, determine the objectives of monitoring and evaluation, such as monitoring changes in pollutant concentrations, evaluating the spread of pollutants, predicting the long-term impact of pollutants, etc. According to the monitoring and evaluation objectives, determine the required monitoring parameters, such as monitoring frequency (regular monitoring, real-time monitoring), monitoring location (near pollution sources, sensitive areas), monitoring methods (on-site monitoring, laboratory analysis), etc.

[0043] In a possible implementation, the environmental pollution toxic monitoring target is analyzed for existence form, diffusion mode, pollution target, and pollution to obtain pollution toxic monitoring characteristics. Step S200 further includes step S210, according to the environmental pollution toxic monitoring target, collects sample data of the pollution target, including historical pollution record data and experimental data, which are used to record the monitoring concentration, existence form, propagation change characteristics, pollution target impact part, and pollution impact data of environmental pollution toxic. Specifically, by consulting the historical records of the environmental protection department, the database of the scientific research institution or the public environmental pollution report, collect data of similar environmental pollution events that occurred in the past. These data contain information such as the type, concentration, occurrence time, location, and impact range of the pollution toxic. Simulate the environmental pollution scene under laboratory conditions and record the behavior characteristics of the pollution toxic under different conditions, such as the existence form (gaseous, liquid, solid), propagation speed, etc. The experimental data is obtained by means of chemical analysis, biological testing, etc. The historical pollution record data provides pollution cases in the real world, and the experimental data provides the behavior characteristics of pollutants under controlled conditions. The two are combined to reflect the actual situation of the pollutants.

[0044] Step S220, according to the existence form and the affected part of the pollution target, determine the monitoring characteristics of the collected data. Specifically, according to the collected sample data, the existence form of the pollutant is analyzed by chemical analysis, such as gas, liquid or solid. By comparing the environmental changes before and after the pollution, the main parts affected by the pollution, such as soil, water source, air or organisms, are identified. Based on this, the monitoring characteristics of the collected data are determined, that is, the focus and direction of the monitoring are determined. Different existence forms and affected parts require different monitoring methods and means.

[0045] Step S230, according to the sample data of the pollution target, fit the transmission change characteristics and the time-space change influence relationship between the monitoring concentration and the pollution impact data, and determine the poison diffusion monitoring characteristics. Specifically, mathematical models or simulation software are used to fit the propagation change characteristics of pollutants (the characteristics of the propagation speed and influence range of pollutants in the environment changing over time) according to the propagation speed, influence range and other information in the sample data. This can be a time-space dynamic model used to describe the concentration distribution of pollutants at different times and places. By analyzing the toxic effects and bioaccumulation of pollutants at different concentrations, a quantitative relationship between the monitoring concentration (the actual concentration of pollutants in the environment) and the pollution impact is established. According to the diffusion law and influence range of pollutants, the monitoring characteristics of poison diffusion are determined to formulate effective monitoring strategies.

[0046] Step S240, obtaining the pollution and poison monitoring characteristics according to the collected data monitoring characteristics and the poison diffusion monitoring characteristics. Specifically, the collected data monitoring characteristics and the poison diffusion monitoring characteristics determined in steps S220 and S230 are comprehensively analyzed and integrated to form a complete pollution and poison monitoring feature set. This feature set should include information such as the existence form, impact site, propagation change characteristics, and relationship with monitoring concentration and pollution impact of pollutants. This implementation method integrates monitoring characteristics from different aspects to fully understand the behavioral characteristics and potential risks of pollutants, providing a basis for formulating targeted monitoring and prevention and control measures.

[0047] In a possible implementation, a monitoring and evaluation target response analysis is performed according to the pollution and poison monitoring characteristics to obtain response evaluation demand parameters, and step S200 further includes step S250, according to the monitoring constraints of the pollution and poison, the pollution and poison monitoring characteristics are fitted with the accuracy relationship of the collected data monitoring characteristics and the diffusion path range of the poison diffusion monitoring characteristics, so as to obtain the monitoring response constraints of the pollution and poison monitoring characteristics. Specifically, the data sources required for the pollution and poison monitoring targets are analyzed, such as historical pollution records, real-time monitoring data, laboratory analysis results, etc. The requirements for monitoring accuracy are determined according to the toxicity level and potential hazards of the pollution and poison. For example, for highly toxic substances, even small concentrations of pollution require high-precision monitoring. Based on historical data and experimental data, a relationship model between monitoring concentration and data accuracy is established using statistical methods or machine learning algorithms. This model is a regression model used to predict the required monitoring accuracy at different concentrations. According to the requirements of toxicity level and monitoring accuracy, the accuracy settings of the monitoring system are adjusted and optimized to ensure that the required monitoring accuracy is achieved within the critical concentration range.

[0048] Using geographic information systems (GIS) and diffusion models in environmental science, we simulate the diffusion path and range of pollutants in the environment. These models perform dynamic simulations based on environmental factors such as wind fields and water flows. Based on the simulation results, we define the possible regional range of pollutants and adjust and optimize the layout of monitoring points according to actual needs to ensure that all key areas are covered.

[0049] Step S260, taking the monitoring response constraint of the pollution and poison monitoring feature as the target, the pollution identification process analysis of the pollution and poison monitoring feature is performed to obtain the response evaluation requirement parameters. Specifically, based on the monitoring data and the results of the diffusion model, the main pollution sources and potential pollution paths are identified. The propagation process, transformation mechanism and influencing factors of pollutants are analyzed, including physical, chemical and biological processes. According to the results of the pollution identification process, key monitoring indicators such as pollutant concentration, diffusion rate, and impact range are determined. Based on the key indicators and monitoring constraints, the required monitoring frequency, number of sampling points, monitoring methods and other technical parameters are calculated. This implementation method ensures that the monitoring system achieves the required monitoring accuracy and coverage within the key concentration range and diffusion area through precision relationship fitting and diffusion path range fitting. At the same time, through the analysis of the pollution identification process and the extraction of response evaluation requirement parameters, a scientific basis is provided for the design and optimization of the monitoring system, ensuring that the monitoring system can accurately identify pollution sources, assess the degree of pollution and formulate effective response measures. Doing so not only improves the pertinence and effectiveness of the monitoring system, but also enhances the accuracy and reliability of environmental pollution monitoring.

[0050] Step S300: traverse and match the response evaluation requirement parameters with the model evaluation features in the multidimensional feature list of the evaluation model to obtain the missing requirement parameters.

[0051] Specifically, the response assessment requirement parameters are compared with each feature in the multidimensional feature list of the assessment model one by one to determine which requirement parameters match the model features and which requirement parameters are missing from the model. By traversing the matching, the parameters that are not in the model feature list but are required to achieve the monitoring and assessment objectives are identified, namely the missing requirement parameters.

[0052] In a possible implementation, the step S300 of obtaining the missing requirement parameter further includes a step S310 of performing data standardization and feature dimension alignment processing on the response evaluation requirement parameter and the model evaluation feature in the multidimensional feature list. Specifically, data standardization refers to converting data from different sources and different formats into a unified format and unit for comparison and analysis. Feature dimension alignment refers to ensuring that the feature dimensions in the multidimensional feature list and the response evaluation requirement parameter have the same format, unit and quantity for matching and comparison. For numerical parameters such as detection limit and response time, the units are first unified. For example, the unit of the detection limit is converted from ppm to μg / m³, and the response time is converted from seconds to milliseconds to ensure that all numerical parameters have the same measurement unit. For categorical parameters such as sensor type, enumerated values ​​are used for encoding. For example, "electrochemical" sensor is encoded as 0 and "optical" sensor is encoded as 1 for computer processing. For composite parameters containing multiple dimensions, such as the input dimension "time series + spatial coordinate", it is split into atomic features. For example, the time series feature is marked as 1, and the spatial coordinate feature is also marked as 1 as two independent feature dimensions.

[0053] If the response assessment requirement parameter contains dimensions that are not covered in the multidimensional feature list, such as "support for satellite data fusion", a new corresponding column needs to be added to the multidimensional feature list to record this feature. For feature dimensions that already exist in the multidimensional feature list and the response assessment requirement parameter, ensure that their values ​​have the same format and unit. For example, if the "sensor detection limit" in the multidimensional feature list is in μg / m³, the detection limit in the response assessment requirement parameter should also be converted to the same unit.

[0054] Step S320, configure matching rules, including data matching, range matching, and logical matching, and match the missing requirement parameters with the model evaluation features in the multidimensional feature list item by item based on the matching rules to obtain the missing parameters and missing types. Specifically, configure matching rules, and matching rules refer to a set of rules used to determine whether the response evaluation requirement parameters match the feature values ​​in the multidimensional feature list. Data matching means that when the response evaluation requirement parameter is exactly the same as a feature value in the multidimensional feature list, it is considered to be a complete match; range matching means that when the response evaluation requirement parameter is a range (such as response time <10 seconds), and the feature value in the multidimensional feature list is a specific value, it is determined whether the value falls within the range specified by the requirement parameter; logical matching means that when the response evaluation requirement parameter is a functional requirement (such as "support dynamic warning threshold"), check whether there is a corresponding functional mark (such as "yes / no") in the multidimensional feature list to indicate whether the model has this function.

[0055] Traverse each item in the response evaluation requirement parameter list, find the corresponding feature dimension in the multi-dimensional feature list according to the configured matching rules, and match it. Record the matching results, including whether it matches, the matching type (complete match, range match, logical match), and any reasons for mismatch.

[0056] Step S330, according to the missing parameters and the missing types, obtain the missing requirement parameters. Specifically, according to the matching results of step 320, all unmatched response evaluation requirement parameters are identified, and these parameters are missing parameters. Each missing parameter is classified to determine its missing type, including data missing (such as lack of training data for a specific scenario), feature missing (such as the model input does not contain key features) and performance missing (such as the model or hardware performance does not meet the standards). Each missing parameter and its missing type are recorded in the missing requirement parameter record table. This implementation ensures an accurate match between the response evaluation requirement parameters and the multidimensional feature list of the evaluation model, thereby identifying the requirements that are not currently covered by the model (i.e., missing requirement parameters), providing a clear guiding direction for subsequent model optimization, and ensuring that the model can be continuously iterated and improved to meet more specific needs.

[0057] Step S400: additionally train the evaluation model according to the missing requirement parameters to obtain an enhanced evaluation model.

[0058] Specifically, for missing demand parameters, relevant environmental monitoring data are collected, such as new pollutant concentration data, new pollution source location data, etc. Feature engineering is performed on the collected new data to extract features related to the missing demand parameters. The processed new data and its features are added to the training set of the evaluation model, and the model is retrained to obtain an enhanced evaluation model that includes the missing demand parameters.

[0059] In a possible implementation, before the evaluation model is additionally trained according to the missing requirement parameters, step S400 further includes step S410, determining whether the missing requirement parameters exist. Specifically, after obtaining the missing requirement parameters, the system first checks whether these parameters are empty or whether they meet the preset "existence" conditions (for example, non-zero values, non-empty strings, etc.). This is achieved through conditional judgment statements (such as if statements) in programming languages, and each missing requirement parameter is checked one by one.

[0060] Step S420, when it does not exist, the evaluation model is used to perform pollutant monitoring and evaluation on the environmental monitoring data. Specifically, if the judgment result shows that there are no missing required parameters, that is, all necessary evaluation features have been met, the system directly calls the existing evaluation model. Using environmental monitoring data (such as air quality data, water quality data, etc.) as input, through the processing flow of the model, the environmental pollution monitoring and evaluation results are output, that is, step S500 is no longer executed.

[0061] Step S430: When there are missing required parameters, the training method analysis is performed based on the missing required parameters and the evaluation model to obtain a new training method. Specifically, when there are missing required parameters, the properties of these parameters (such as data type, value range, etc.) and their relationship with the existing features of the evaluation model are analyzed. According to the analysis results, a suitable training method is selected to supplement these missing parameters.

[0062] Step S440, additional training is performed on the evaluation model based on the missing demand parameters and the newly added training method. Specifically, according to the newly added training method determined in step 430, training data containing the missing demand parameters is collected or generated. Use these data to train or fine-tune the evaluation model to supplement the missing feature parameters. During the training process, it may be necessary to adjust the hyperparameters of the model (such as learning rate, batch size, etc.) to optimize the training effect. After the training is completed, the performance of the model is evaluated using the validation data set to ensure that the newly added parameters do not damage the overall performance of the model. Environmental pollution monitoring is a complex and changeable task. The evaluation model needs to constantly adapt to new environmental conditions and pollutant types. This implementation method can ensure that the evaluation model is comprehensive and accurate enough to cope with various environmental pollution monitoring scenarios by identifying and supplementing the missing demand parameters. At the same time, the use of a flexible training method can efficiently update the model, reduce the cost and time of retraining, and thus quickly respond to the challenges of environmental changes.

[0063] In a possible implementation, step S430 further includes step S431, and the newly added training method includes: transfer learning, incremental learning or integrated learning.

[0064] Specifically, analyze the nature of the missing required parameters. First, determine the data type (such as numeric type, categorical type, etc.) and data structure (such as vector, matrix, etc.) of the missing required parameters to select the appropriate training algorithm. Different algorithms have different adaptability to data types. Evaluate the importance of the missing required parameters to the performance improvement of the model and whether these parameters need to be supplemented urgently. If a parameter is critical and urgent to the performance of the model, then a training method that can quickly adapt to new data is needed.

[0065] Analyze and evaluate the current state of the model. The architecture of the model (such as the number of neural network layers and nodes) will affect the choice of training method. For example, for deep neural networks, incremental learning is more suitable, which can gradually update the model without changing the network structure. The performance of the current model (such as accuracy, recall, etc.) is also an important factor in choosing a training method. If the model performance is already very high, then a more sophisticated training method (such as fine-tuning in transfer learning) is needed to avoid performance degradation.

[0066] Evaluate available training resources and check if there is enough new data to train the model. If data is scarce, transfer learning is a better choice, which can use knowledge in related fields to accelerate the training process. Analyze available computing resources (such as CPU, GPU, memory, etc.). Some training methods (such as incremental learning of deep learning) require more computing resources to gradually update the model.

[0067] Based on the above analysis, choose the appropriate additional training method. If the missing requirement parameters have some similarity with the existing data or tasks, transfer learning is an effective choice, which allows the model to use the knowledge learned on related tasks to accelerate the learning process of new tasks. When new data continues to arrive and the model needs to be updated in real time, incremental learning is a suitable choice, which allows the model to gradually learn new data without retraining the entire model. If the missing requirement parameters involve multiple different features or models, an integrated learning method can be used to combine the prediction results of multiple models to improve the overall performance. This implementation method provides strong support and guidance for new model training by comprehensively considering the nature of the missing requirement parameters, evaluating the current state of the model, and the available training resources.

[0068] In one possible implementation, after obtaining the enhanced assessment model, step S400 further includes step S400a, establishing identification tags and input layer screening and processing rules for the enhanced assessment model. Specifically, specific identification tags are established for the enhanced assessment model based on its application scenarios and monitoring objectives. These tags are used to identify key information such as the types of pollutants that the model can handle and the monitoring environment. For example, labels such as "PM2.5 monitoring" and "water quality heavy metal monitoring" can be set for the model. According to the input requirements of the model, rules for screening and processing input data are formulated. These rules include requirements for data types, data formats, data ranges, and other aspects. For example, for a water quality monitoring model, key data such as dissolved oxygen, pH value, and heavy metal content in the water need to be screened out and converted into a format that the model can recognize.

[0069] Step S400b, obtain environmental monitoring data and target identification tags. Specifically, collect or obtain historical monitoring data in real time through environmental monitoring equipment (such as sensors, drones, etc.). These data cover all aspects of the monitoring target, such as the concentration, existence form, and propagation change characteristics of pollutants. According to the requirements of the monitoring task, set corresponding identification tags for the acquired environmental monitoring data. These tags match the model identification tags established in step S400a so that the model can be correctly selected and used.

[0070] Step S400c, matching and identifying the target identification tag with the enhanced assessment model to obtain a matching model or a matching integrated sub-model. Specifically, according to the target identification tag of the environmental monitoring data, a matching model or integrated sub-model is selected from the established enhanced assessment model library. The matching process can be achieved by comparing the similarity of the tags or keyword matching. In some cases, multiple models need to be integrated to improve the accuracy and robustness of the prediction. These integrated sub-models can be models based on different training data, training methods or model architectures.

[0071] Step S400d, screen and process the environmental monitoring data according to the input layer screening and processing rules, and input the screened environmental monitoring data into the matching model or matching integrated sub-model. Specifically, according to the input layer screening and processing rules formulated in step S400a, screen and process the environmental monitoring data, including data cleaning, format conversion, data range adjustment and other operations. Input the screened environmental monitoring data into the model or integrated sub-model obtained by matching identification to conduct pollutant monitoring and evaluation. This implementation method ensures the quality and accuracy of the input data by establishing identification tags and input layer screening and processing rules; through matching identification and the selection of integrated sub-models, it ensures that the selected model matches the requirements of the monitoring task, thereby improving the accuracy and reliability of the prediction.

[0072] Step S500, using the enhanced assessment model to perform pollutant monitoring and assessment on the environmental monitoring data, obtain environmental pollution monitoring and assessment results, and use the environmental pollution monitoring and assessment results for pollution monitoring feedback.

[0073] Specifically, the actual environmental monitoring data, such as pollutant concentration, wind direction and wind speed, etc., are input into the enhanced assessment model (including the assessment model with missing demand parameters), the enhanced assessment model is started, and the input environmental monitoring data is processed and analyzed. After the model processes the data, it outputs the environmental pollution monitoring and assessment results, including information such as the concentration, diffusion range, and potential hazards of pollutants. The assessment results are fed back to relevant departments or personnel so that corresponding environmental protection measures can be taken. The embodiment of the present application adopts an analytical assessment model, establishes a multidimensional feature list, analyzes the characteristics of pollutants, obtains response assessment demand parameters, matches demand parameters with model features, determines missing parameters, performs additional training on the assessment model, obtains an enhanced model, uses the enhanced model to evaluate environmental monitoring data, obtains monitoring results and feedbacks, and other technical means, thereby achieving the technical effect of improving the accuracy and reliability of the assessment results.

[0074] The following is an example of an enhanced assessment model for monitoring sudden heavy metal pollution in water bodies. The existing assessment model is a time series prediction model based on LSTM (Long Short-Term Memory Network) to predict the concentration changes of mercury ions (Hg²⁺) in rivers.

[0075] The characteristics of the training samples are as follows: Data source: water quality monitoring data of a river basin in the past five years (sampling frequency: once per hour). Features include: mercury ion concentration (μg / L), dissolved oxygen (DO, mg / L), water temperature (℃), pH value, flow rate (m / s), and upstream industrial wastewater discharge (tons / hour).

[0076] The model architecture features are as follows: Input layer: contains 7 time series features (mercury concentration, DO, water temperature, etc.). Hidden layer: 2 layers of LSTM units, 128 neurons in each layer. Output layer: predicts the mercury concentration in the next 2 hours.

[0077] The training method features are as follows: Loss function: mean square error (MSE). Optimizer: Adam (learning rate = 0.001). Training parameters: batch size = 32, iterations = 200, early stopping to prevent overfitting.

[0078] The monitoring characteristics of pollutants are as follows: Existence form: Mercury ions are mainly dissolved (Hg²⁺), and some are adsorbed on suspended particles. Diffusion mode: Diffusion through river flow, significantly affected by water flow velocity and river topography. Pollution target: 5 key monitoring points along the river (A1-A5), covering drinking water sources and ecologically sensitive areas. Toxicity analysis: Mercury ions are highly neurotoxic and low concentration changes need to be monitored in real time (threshold: 0.01 μg / L).

[0079] The response assessment requirement parameters are as follows: Monitoring frequency: real-time (sampling once every minute). Accuracy requirement: ±5% (±10% error is allowed when the concentration is lower than 0.01 μg / L). Diffusion path prediction: Combined with the hydrological model (SWAT), the pollutant migration trajectory is output.

[0080] The characteristics of existing evaluation models are shown in Table 1.

[0081] Table 1: Characteristics of existing models

[0082] The missing required parameters are as follows: Insufficient real-time performance: the model has a delay of 30 seconds and cannot meet the real-time warning requirements. Insufficient low-concentration detection capability: the detection limit is 0.1 μg / L, which is higher than the safety threshold of 0.01 μg / L. Lack of diffusion path prediction function: the hydrological model is not integrated.

[0083] A new training method is added, which selects transfer learning, fine-tunes the existing LSTM model, and adds a diffusion path prediction branch. Specifically, trainable=False is used to force the layer parameters to be read-only to protect the weight parameters of the first two layers of the LSTM model from being modified; a new diffusion prediction head is added, and the fully connected layer is used as a transition layer. The ReLU activation function is used to enhance the nonlinear expression ability. The output layer corresponds to the pollution probability distribution of the five monitoring points, which corresponds one-to-one to the five key monitoring points in the river basin. The output is converted into a probability distribution. The optimizer selects the RMSprop adaptive learning rate algorithm (more suitable for small sample fine-tuning than Adam), and the learning rate is 0.0005, which is used to fine-tune the parameter update step size and balance the convergence speed and accuracy. The loss function selects categorical_crossentropy, which is used to accurately match the probability distribution of monitoring points.

[0084] In the above, refer to Figure 1 The environmental pollution monitoring method based on the evaluation model according to the embodiment of the present invention is described in detail. Figure 2 An environmental pollution monitoring device based on an evaluation model according to an embodiment of the present invention is described.

[0085] The environmental pollution monitoring device based on the evaluation model according to the embodiment of the present invention is used to solve the technical problem of insufficient accuracy and reliability of the evaluation results in the prior art, and achieve the technical effect of improving the accuracy and reliability of the evaluation results. The environmental pollution monitoring device based on the evaluation model includes: an environmental pollution evaluation feature analysis module 10, a response evaluation requirement parameter acquisition module 20, a missing requirement parameter acquisition module 30, a model additional training module 40, and a pollutant monitoring and evaluation module 50.

[0086] The environmental pollution assessment feature parsing module 10 is used to parse the environmental pollution assessment features of the assessment model and establish a multi-dimensional feature list of the assessment model; the response assessment demand parameter acquisition module 20 is used to analyze the existence form, diffusion mode, pollution target, and pollution of the environmental pollution toxic monitoring target to obtain the pollution toxic monitoring feature, and perform monitoring assessment target response analysis based on the pollution toxic monitoring feature to obtain the response assessment demand parameter; the missing demand parameter acquisition module 30 is used to use the response assessment demand parameter to traverse and match the model assessment features in the multi-dimensional feature list of the assessment model to obtain the missing demand parameter; the model additional training module 40 is used to perform additional training on the assessment model according to the missing demand parameter to obtain an enhanced assessment model; the pollutant monitoring and assessment module 50 is used to use the enhanced assessment model to perform pollutant monitoring and assessment on the environmental monitoring data, obtain the environmental pollution monitoring and assessment result, and use the environmental pollution monitoring and assessment result for pollution monitoring feedback.

[0087] The specific configuration of the environmental pollution assessment feature analysis module 10 will be described in detail below. As described above, the environmental pollution assessment features of the assessment model are analyzed, and a multidimensional feature list of the assessment model is established. The environmental pollution assessment feature analysis module 10 may further include: a model feature analysis unit is used to perform feature analysis on the assessment model according to the training sample features, model architecture features, training methods, output forms, and model performance features to obtain multidimensional assessment features; a multidimensional feature list construction unit is used to integrate the multidimensional assessment features according to the dimensional feature labels of the multidimensional assessment features to construct a multidimensional feature list of the assessment model.

[0088] Among them, according to the multi-dimensional characteristics of training sample characteristics, model architecture characteristics, training methods, output forms, and model performance characteristics, the evaluation model is feature analyzed to obtain multi-dimensional evaluation characteristics. The model feature analysis unit may further include: a pollution training sample feature acquisition subunit is used to perform feature analysis on the training samples in terms of data source, sample pollution toxic type, data characteristics, sample quantity distribution, and data labels to obtain pollution training sample characteristics; a model architecture feature acquisition subunit is used to perform pollutant assessment logic feature analysis according to the model type and model structure of the model architecture to obtain model architecture characteristics; a training method feature acquisition subunit is used to perform training method feature analysis according to the training type and training parameters to obtain training method characteristics; an output form feature acquisition subunit is used to perform output form feature analysis according to the type, dimension, and range of the model output to obtain output form characteristics; a model performance feature acquisition subunit is used to perform performance feature analysis according to the accuracy, real-time, and robustness of the model output results to obtain model performance characteristics; a feature combination subunit is used to combine the pollution training sample characteristics, model architecture characteristics, training method characteristics, output form characteristics, and model performance characteristics to obtain the multi-dimensional evaluation characteristics.

[0089] The specific configuration of the model additional training module 40 will be described in detail below. As described above, before the evaluation model is additionally trained according to the missing demand parameters, the model additional training module 40 may further include: a judgment unit for judging whether the missing demand parameters exist; a pollutant monitoring and evaluation unit for, when the missing demand parameters do not exist, using the evaluation model to perform pollutant monitoring and evaluation on environmental monitoring data; a training method analysis unit for, when the missing demand parameters exist, performing training method analysis based on the evaluation model to obtain an additional training method; and an additional training unit for performing additional training on the evaluation model based on the missing demand parameters and the additional training method.

[0090] Among them, the training method analysis unit may further include: a new training method acquisition subunit for the new training method including: transfer learning, incremental learning or integrated learning.

[0091] The specific configuration of the response assessment demand parameter acquisition module 20 will be described in detail below. As described above, the existence form, diffusion mode, pollution target, and contamination analysis of the environmental pollution and poison monitoring target are performed to obtain the pollution and poison monitoring characteristics. The response assessment demand parameter acquisition module 20 may further include: a sample data collection unit is used to collect sample data of the pollution target according to the environmental pollution and poison monitoring target, including historical pollution record data and experimental data, which are used to record the monitoring concentration, existence form, propagation change characteristics, pollution target impact site, and pollution impact data of environmental pollution and poison; a collection data monitoring feature determination unit is used to determine the collection data monitoring characteristics according to the existence form and pollution target impact site; a poison diffusion monitoring feature determination unit is used to fit the propagation change characteristics and the spatiotemporal influence relationship between the monitoring concentration and the pollution impact data according to the sample data of the pollution target, and determine the poison diffusion monitoring characteristics; a pollution and poison monitoring feature acquisition unit is used to obtain the pollution and poison monitoring characteristics according to the collection data monitoring characteristics and the poison diffusion monitoring characteristics.

[0092] Among them, monitoring and evaluation target response analysis is performed according to the pollution and toxicant monitoring characteristics to obtain response evaluation requirement parameters, and the response evaluation requirement parameter acquisition module 20 may further include: a monitoring response constraint acquisition unit is used to perform precision relationship fitting of the collected data monitoring characteristics and diffusion path range fitting of the toxicant diffusion monitoring characteristics on the pollution and toxicant monitoring characteristics according to the monitoring constraint conditions of the pollution and toxicant, so as to obtain the monitoring response constraints of the pollution and toxicant monitoring characteristics; a pollution identification process analysis unit is used to perform pollution identification process analysis on the pollution and toxicant monitoring characteristics with the monitoring response constraints of the pollution and toxicant monitoring characteristics as the target, so as to obtain the response evaluation requirement parameters.

[0093] Among them, after obtaining the enhanced assessment model, the device may further include: an identification label and input layer screening and processing rule establishment module for establishing the identification label and input layer screening and processing rules of the enhanced assessment model; an environmental monitoring data and target identification label acquisition module for acquiring environmental monitoring data and target identification labels; a matching identification module for matching and identifying the enhanced assessment model according to the target identification label to obtain a matching model or a matching integrated sub-model; a screening and processing module for screening and processing the environmental monitoring data according to the input layer screening and processing rules, and inputting the screened environmental monitoring data into the matching model or the matching integrated sub-model.

[0094] The specific configuration of the missing requirement parameter acquisition module 30 will be described in detail below. As described above, to obtain the missing requirement parameters, the missing requirement parameter acquisition module 30 may further include: a data standardization and feature dimension alignment processing unit for performing data standardization and feature dimension alignment processing on the response evaluation requirement parameters and the model evaluation features in the multidimensional feature list; an item-by-item matching unit for configuring matching rules, including data matching, range matching, and logical matching, and matching the missing requirement parameters and the model evaluation features in the multidimensional feature list item by item based on the matching rules to obtain the missing parameters and the missing type; a missing requirement parameter acquisition unit for obtaining the missing requirement parameters according to the missing parameters and the missing type.

[0095] The environmental pollution monitoring device based on the evaluation model provided in the embodiment of the present invention can execute the environmental pollution monitoring method based on the evaluation model provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0096] Although the present application makes various references to certain modules in the device according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0097] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An environmental pollution monitoring method based on an assessment model, characterized in that: include: Analyze the environmental pollution assessment characteristics of the assessment model and establish a multi-dimensional feature list of the assessment model; Conducting an analysis of the existence form, diffusion mode, pollution target, and contamination of environmental pollution and toxic substance monitoring targets to obtain pollution and toxic substance monitoring characteristics, and conducting a monitoring and evaluation target response analysis based on the pollution and toxic substance monitoring characteristics to obtain response evaluation demand parameters; The response evaluation requirement parameters are used to perform traversal matching with the model evaluation features in the multidimensional feature list of the evaluation model to obtain the missing requirement parameters; Performing additional training on the evaluation model according to the missing requirement parameters to obtain an enhanced evaluation model; The enhanced assessment model is used to perform pollutant monitoring and assessment on environmental monitoring data to obtain environmental pollution monitoring and assessment results, and the environmental pollution monitoring and assessment results are used for pollution monitoring feedback.

2. The environmental pollution monitoring method based on the evaluation model according to claim 1 is characterized in that: The environmental pollution assessment features of the analytical assessment model are used to establish a multi-dimensional feature list of the assessment model, including: According to the multi-dimensional characteristics of training samples, model architecture, training methods, output forms, and model performance, the evaluation model is characterized to obtain multi-dimensional evaluation characteristics; The multidimensional evaluation features are integrated according to their dimensional feature labels to construct a multidimensional feature list of the evaluation model.

3. The environmental pollution monitoring method based on the evaluation model according to claim 2 is characterized in that: According to the multi-dimensional characteristics of training samples, model architecture, training methods, output forms, and model performance, the evaluation model is feature parsed to obtain multi-dimensional evaluation features, including: Perform feature analysis on the training samples, including data source, sample contamination type, data features, sample quantity distribution and data labels, to obtain the contaminated training sample features; Conduct pollutant assessment logic feature analysis based on the model type and model structure of the model architecture to obtain model architecture features; Perform training method feature analysis according to training type and training parameters to obtain training method features; Analyze the output form characteristics according to the type, dimension and range of the model output to obtain the output form characteristics; Perform performance characteristic analysis based on the accuracy, real-time performance, and robustness of the model output results to obtain model performance characteristics; The contaminated training sample characteristics, model architecture characteristics, training method characteristics, output form characteristics and model performance characteristics are combined to obtain the multi-dimensional evaluation characteristics.

4. The environmental pollution monitoring method based on the evaluation model according to claim 1 is characterized in that: Additional training is performed on the evaluation model according to the missing requirement parameters, which also includes: Determine whether the missing requirement parameter exists; When it does not exist, using the assessment model to conduct pollutant monitoring assessment on the environmental monitoring data; If it exists, a training method analysis is performed according to the missing requirement parameters and the evaluation model to obtain a new training method; Additional training is performed on the evaluation model based on the missing requirement parameters and the additional training method.

5. The environmental pollution monitoring method based on the evaluation model according to claim 4 is characterized in that: The newly added training methods include: transfer learning, incremental learning or integrated learning.

6. The environmental pollution monitoring method based on the evaluation model according to claim 1 is characterized in that: The monitoring characteristics of environmental toxicants are obtained by analyzing the existence form, diffusion mode, pollution target, and pollution of the environmental toxicants monitoring target, including: According to the environmental pollution toxic monitoring target, collect sample data of the pollution target, including historical pollution record data and experimental data, which are used to record the monitoring concentration, existence form, propagation change characteristics, pollution target impact parts and pollution impact data of environmental pollution toxic substances; Determine the monitoring characteristics of the collected data according to the existing form and the affected parts of the pollution target; According to the sample data of the pollution target, the temporal and spatial influence relationship between the propagation change characteristics and the monitoring concentration and pollution impact data is fitted to determine the monitoring characteristics of the spread of poisons; The pollution and poison monitoring characteristics are obtained according to the collected data monitoring characteristics and the poison diffusion monitoring characteristics.

7. The environmental pollution monitoring method based on the evaluation model according to claim 6 is characterized in that: According to the monitoring characteristics of the pollution and toxic substances, the monitoring and evaluation target response analysis is carried out to obtain the response evaluation demand parameters, including: According to the monitoring constraint conditions of the pollution and poison, the monitoring characteristics of the pollution and poison are fitted with the accuracy relationship of the collected data monitoring characteristics and the diffusion path range of the poison diffusion monitoring characteristics to obtain the monitoring response constraints of the pollution and poison monitoring characteristics; Taking the monitoring response constraint of the pollution and poison monitoring feature as the target, the pollution identification process of the pollution and poison monitoring feature is analyzed to obtain the response evaluation requirement parameter.

8. The environmental pollution monitoring method based on the evaluation model according to claim 5 is characterized in that: Get an enhanced evaluation model, which also includes: Establishing identification tags and input layer screening and processing rules of the enhanced evaluation model; Obtain environmental monitoring data and target identification tags; Perform matching and identification with the enhanced evaluation model according to the target identification label to obtain a matching model or a matching integrated sub-model; The environmental monitoring data is screened and processed according to the input layer screening and processing rules, and the screened environmental monitoring data is input into the matching model or the matching integrated sub-model.

9. The environmental pollution monitoring method based on the evaluation model according to claim 7 is characterized in that: The obtaining of missing requirement parameters includes: Performing data standardization and feature dimension alignment processing on the response evaluation requirement parameters and the model evaluation features in the multidimensional feature list; Configure matching rules, including data matching, range matching, and logic matching, and match the missing requirement parameters with the model evaluation features in the multidimensional feature list item by item based on the matching rules to obtain the missing parameters and missing types; The missing requirement parameters are obtained according to the missing parameters and the missing type.

10. An environmental pollution monitoring device based on an evaluation model, characterized in that: The device is used to implement the environmental pollution monitoring method based on the assessment model according to any one of claims 1 to 9, and the device comprises: Environmental pollution assessment feature analysis module, used to analyze the environmental pollution assessment features of the assessment model and establish a multi-dimensional feature list of the assessment model; The response assessment demand parameter acquisition module is used to analyze the existence form, diffusion mode, pollution target, and pollution of the environmental pollution and toxic monitoring target, obtain the pollution and toxic monitoring characteristics, and perform monitoring and assessment target response analysis based on the pollution and toxic monitoring characteristics to obtain the response assessment demand parameters; A missing requirement parameter acquisition module is used to perform traversal matching using the response evaluation requirement parameters and the model evaluation features in the multidimensional feature list of the evaluation model to obtain the missing requirement parameters; A model additional training module is used to perform additional training on the evaluation model according to the missing requirement parameters to obtain an enhanced evaluation model; The pollutant monitoring and evaluation module is used to use the enhanced evaluation model to perform pollutant monitoring and evaluation on the environmental monitoring data, obtain environmental pollution monitoring and evaluation results, and use the environmental pollution monitoring and evaluation results for pollution monitoring feedback.

Images