Environmental pollutant toxicity detection and evaluation method

Abstract

The invention discloses an environmental pollutant toxicity detection and evaluation method, which comprises the following steps: deploying a multi-parameter sensor array in a target monitoring environment, and obtaining initial monitoring data; performing deep mining and pollutant toxicity feature extraction on the initial monitoring data to generate a pollutant feature data set; forming a customized evaluation model framework according to the target environment monitoring requirement, the preset monitoring parameters and the configuration of the evaluation model; inputting the generated pollutant characteristic data set into a preset customized evaluation model, and performing toxicity evaluation on the environmental pollutants to obtain a preliminary evaluation result; and automatically adjusting parameters of the customized evaluation model based on the obtained preliminary evaluation result by adopting an adaptive learning algorithm to obtain an optimized evaluation model, carrying out toxicity evaluation again to obtain an optimized evaluation result, and the like. The method adapts to monitoring requirements under different environmental conditions, and comprehensively evaluates the toxicity risk of pollutants.

Description

Technical Field

[0001] The invention belongs to the field of environmental detection, and in particular relates to a method for detecting and evaluating the toxicity of environmental pollutants. Background Art

[0002] With the rapid development of industrialization and urbanization, environmental pollution problems are becoming increasingly serious, especially the impact of pollutants on the ecological environment and human health has attracted widespread attention. Traditional methods for assessing the toxicity of environmental pollutants mainly rely on laboratory analysis, which is not only time-consuming and costly, but also unable to achieve real-time and dynamic monitoring of the toxicity of pollutants. In addition, laboratory analysis can usually only provide static data at a specific time point, which is difficult to reflect the dynamic characteristics of pollutant toxicity changing with time and environmental conditions.

[0003] In the field of environmental monitoring, some automated and online monitoring technologies have been developed, but these technologies often focus on monitoring pollutant concentrations while neglecting the comprehensive assessment of pollutant toxicity. The toxicity of pollutants is not only related to concentration, but also closely related to factors such as the chemical properties of pollutants, environmental conditions, and the sensitivity of organisms. Therefore, existing monitoring technologies cannot comprehensively assess the toxicity risks of pollutants, nor can they provide sufficient scientific basis for environmental management and decision-making.

[0004] In addition, the toxicity assessment of environmental pollutants is a complex process involving multi-parameter monitoring, data analysis, and model prediction. Existing assessment methods often use a single assessment model, which is difficult to adapt to monitoring needs under different environmental conditions. Moreover, these methods lack depth in data processing and analysis, and cannot fully utilize the information in a large amount of monitoring data, resulting in insufficient accuracy and reliability of the assessment results. Summary of the invention

[0005] The present invention is proposed to solve the problems existing in the prior art such as the inability to realize real-time monitoring of pollutant toxicity, lack of depth in data processing and analysis, inability to predict the changing trend of pollutant toxicity, inability of the evaluation model to adapt to different environmental conditions and inability to be optimized according to specific environmental conditions. Its purpose is to provide a method for detecting and evaluating the toxicity of environmental pollutants.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for detecting and evaluating the toxicity of environmental pollutants comprises the following steps:

[0008] S1. Deploy a multi-parameter sensor array in the target monitoring environment to obtain initial monitoring data;

[0009] S2. Use big data analysis and cloud computing technology to perform in-depth mining of the initial monitoring data obtained in step S1 and extract toxicity characteristics of pollutants to generate a pollutant characteristic data set;

[0010] S3. According to the monitoring requirements of the target environment, preset the configuration of monitoring parameters and evaluation models to form a customized evaluation model framework;

[0011] S4, inputting the pollutant characteristic data set generated in step S2 into a preset customized assessment model to conduct a toxicity assessment of environmental pollutants and obtain a preliminary assessment result;

[0012] S5. Adopt an adaptive learning algorithm, based on the preliminary evaluation results obtained in step S4, automatically adjust the parameters of the customized evaluation model to obtain an optimized evaluation model, use the optimized evaluation model to perform toxicity evaluation on the pollutant characteristic data set again, and obtain an optimized evaluation result.

[0013] In the above technical solution, the step S1 deploys a multi-parameter sensor array in the target monitoring environment to obtain initial monitoring data, which specifically includes the following steps:

[0014] S11. Install and calibrate the multi-parameter sensor array;

[0015] The multi-parameter sensor array includes chemical sensors, biological sensors and physical sensors;

[0016] S12, collecting pollutant data in the target monitoring environment in real time through a multi-parameter sensor array;

[0017] S13, preprocessing the collected pollutant data, wherein the preprocessing includes data cleaning, denoising and normalization;

[0018] S14. Integrate the preprocessed data into an initial monitoring data set.

[0019] In the above technical solution, step S2 uses big data analysis and cloud computing technology to perform deep mining and toxicity feature extraction of the initial monitoring data set obtained in step S1 to generate a pollutant feature data set, which specifically includes the following steps:

[0020] S21, performing data cleaning on the initial monitoring data set obtained in step S1 to form a cleaned monitoring data set;

[0021] The data cleaning includes removing outliers, filling missing data and data standardization;

[0022] S22. Apply big data analysis techniques to conduct statistical analysis and pattern recognition on the cleaned monitoring data sets to identify potential toxicity patterns of pollutants;

[0023] S23. Use cloud computing resources to conduct in-depth analysis of the identified potential toxicity patterns and extract the toxicity characteristics of pollutants;

[0024] Said in-depth analysis includes machine learning classification and regression analysis;

[0025] S24, screening the extracted pollutant toxicity characteristics through a feature selection algorithm to determine the most representative pollutant toxicity characteristic set;

[0026] S25. The most representative pollutant toxicity characteristics are collected after screening to generate a pollutant characteristic data set.

[0027] In the above technical solution, step S3 presets the configuration of monitoring parameters and evaluation model according to the target environment monitoring requirements to form a customized evaluation model framework, which specifically includes the following steps:

[0028] S31. Analyze the specific needs of the target monitoring environment and determine the preliminary settings of monitoring parameters;

[0029] The specific requirements of the target monitoring environment include pollutant types, monitoring frequency, data accuracy and environmental conditions;

[0030] S32, selecting parameters of the multi-parameter sensor array according to the specific requirements analysis results of the target detection environment;

[0031] S33. Design the architecture of the evaluation model based on the specific requirements analysis results of the target detection environment;

[0032] The framework of the evaluation model includes determining the input parameters, output results and processing algorithms of the model;

[0033] S34. Use historical monitoring data and known pollutant toxicity information to conduct preliminary training and parameter adjustment of the assessment model to form a basic assessment model;

[0034] S35, applying the basic evaluation model to the actual monitoring data, performing model verification, and obtaining performance indicators of the evaluation model;

[0035] The performance indicators of the evaluation model include precision, recall and F1 score;

[0036] S36, further adjusting and optimizing the parameters of the evaluation model according to the performance indicators of the evaluation model to obtain an optimized evaluation model;

[0037] S37. Integrate the optimized assessment model with monitoring parameters and data processing procedures to form a customized assessment model framework.

[0038] In the above technical solution, the step S4 inputs the generated pollutant characteristic data set into a preset customized evaluation model to perform toxicity evaluation of environmental pollutants and obtain preliminary evaluation results, which specifically includes the following steps:

[0039] S41, inputting the pollutant characteristic data set generated in step S2 into the optimized evaluation model obtained in step S36 according to the preset format and requirements;

[0040] S42, in the evaluation model obtained in step S3, executing a toxicity evaluation algorithm according to the input pollutant characteristic data set, the toxicity evaluation algorithm performs calculations according to the monitoring parameters and evaluation model configuration preset in step S37 to evaluate the toxicity of the pollutant;

[0041] S43. The assessment model processes the input pollutant characteristic data set and generates preliminary toxicity assessment results based on built-in logic and parameters.

[0042] In the above technical solution, the toxicity assessment algorithm specifically includes the following steps:

[0043] S421. Define a loss function for evaluating the model;

[0044] The loss function is used to measure the difference between the predicted value of the evaluation model and the actual pollutant toxicity value;

[0045] S422, based on the defined loss function, calculate the initial loss value of the evaluation model; by inputting the pollutant characteristic data set obtained in step S2 into the evaluation model, and comparing the predicted output of the model with the actual toxicity value, the formula is expressed as:

[0046]

[0047] Where: L represents the overall loss, N is the number of samples, l is the loss function, y i is the actual toxicity value of the i-th sample, is the toxicity value predicted by the model;

[0048] S423, using a gradient descent algorithm to calculate the gradient of the evaluation model parameters according to the initial loss value;

[0049] The gradient calculation formula is as follows:

[0050]

[0051] Where: θ represents the model parameters, represents the loss function, L represents the gradient with respect to the model parameter θ;

[0052] S424. Update the parameters of the evaluation model according to the calculated gradient. The parameter update rule is determined according to the adaptive learning algorithm. The update formula is:

[0053]

[0054] Among them, θ t represents the model parameters at time step t, η t represents the learning rate, represents the gradient at time step t;

[0055] S425. Repeat steps S422 to S424 until the loss value of the evaluation model converges or reaches a predetermined number of iterations.

[0056] In the above technical solution, the step S5 adopts an adaptive learning algorithm to automatically adjust the parameters of the customized evaluation model based on the preliminary evaluation results to obtain an optimized evaluation model, and uses the optimized evaluation model to perform toxicity evaluation on the pollutant characteristic data set again to obtain an optimized evaluation result, which specifically includes the following steps:

[0057] S51, analyzing the accuracy and reliability of the preliminary evaluation results obtained in step S4 to determine whether there are systematic deviations or parameters to be optimized;

[0058] The systematic method includes calculating the accuracy, recall and F1 score of the predictions of the evaluation model;

[0059] S52, identifying the evaluation model parameters to be adjusted according to the analysis result of step S51;

[0060] The evaluation model parameters include evaluation model weights, threshold settings, or algorithm-specific hyperparameters;

[0061] S53, applying an adaptive learning algorithm to dynamically adjust the identified parameters according to changes in the preliminary evaluation results obtained in step S4, wherein the adaptive learning algorithm uses historical gradient information to automatically adjust the learning rate and the parameter update step size to obtain new parameter values;

[0062] S54, updating the evaluation model according to the new parameter value provided by the adaptive learning algorithm;

[0063] S55. Use the updated assessment model to conduct toxicity assessment on the pollutant characteristic data set again to obtain optimized assessment results.

[0064] In the above technical solution, step S6 uses a pre-trained deep learning algorithm to perform trend analysis and prediction on the optimized evaluation results obtained in step S5 to generate a prediction report on the toxicity of pollutants, which specifically includes the following steps:

[0065] S61, inputting the optimized evaluation result obtained in step S5 into the pre-trained deep learning model;

[0066] S62, executing the forward propagation process of the deep learning model, and calculating the toxicity assessment result of the deep learning model for each sample;

[0067] S63. In the output layer of the deep learning model, a softmax function is applied to convert the output of the deep learning model into a probability distribution, representing the probabilities of different toxicity categories;

[0068] S64. Determine the prediction category of pollutant toxicity based on the probability distribution output by the deep learning model;

[0069] S65. Analyze and predict the toxicity trend of pollutants using the prediction results of the deep learning model in combination with time series analysis technology, and generate the predicted change trend of pollutant toxicity in the future time period;

[0070] S66. Deep learning model generates prediction reports on pollutant toxicity;

[0071] The forecast report includes predicted toxicity levels, changing trends and expected impacts.

[0072] The beneficial effects of the present invention are:

[0073] The invention provides a method for detecting and evaluating the toxicity of environmental pollutants, which is adapted to the monitoring needs under different environmental conditions and comprehensively evaluates the toxicity risks of pollutants. By deploying a multi-parameter sensor array, real-time monitoring of environmental pollutants is achieved, so that environmental managers can quickly respond to environmental changes, take effective measures in time, and reduce the potential risks of environmental pollution to ecology and human health. Big data analysis and cloud computing technology are used to perform deep mining and feature extraction on initial monitoring data to generate a pollutant feature data set, achieve comprehensive and deep analysis, improve the accuracy and reliability of toxicity evaluation, and provide more scientific decision support for environmental management. An adaptive learning algorithm is used to automatically adjust the parameters of the evaluation model according to real-time data and historical data. The adaptive and optimized evaluation capabilities ensure that the evaluation model can continuously provide accurate toxicity evaluation results, and improve the applicability and effectiveness of the evaluation model. A pre-trained deep learning algorithm is used to perform trend analysis and prediction on the optimized evaluation results, and a prediction report on the toxicity of pollutants is generated. The predictive and decision support capabilities provide forward-looking guidance for environmental management, which is helpful to prevent and mitigate potential environmental pollution events. According to the target environmental monitoring needs, the configuration of the monitoring parameters and the evaluation model is preset to form a customized evaluation model framework. The customized capability enables the evaluation model to better adapt to different monitoring needs and environmental conditions, and improves the pertinence and accuracy of the evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 It is a flow chart of the method of the present invention.

[0075] For ordinary technicians in this field, other relevant drawings can be obtained based on the above drawings without any creative work. DETAILED DESCRIPTION

[0076] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.

[0077] like Figure 1 As shown, a method for detecting and evaluating the toxicity of environmental pollutants comprises the following steps:

[0078] S1. Deploy a multi-parameter sensor array in the target monitoring environment to obtain initial monitoring data, specifically including the following steps:

[0079] S11. Install and calibrate the multi-parameter sensor array;

[0080] The multi-parameter sensor array includes chemical sensors, biological sensors and physical sensors;

[0081] The multi-parameter sensor array is installed in key areas to ensure comprehensive monitoring of pollutants in the environment. During the installation process, each sensor is precisely positioned to maximize monitoring efficiency. After installation, the sensor array is calibrated to ensure data accuracy and consistency. The calibration process may include the use of known standard materials or calibration equipment to set parameters such as sensor sensitivity, linearity, repeatability, and drift; the deployment location, quantity, and type of the sensor array are optimized according to monitoring requirements and environmental characteristics to ensure comprehensive coverage of the monitoring area and obtain representative data;

[0082] S12, collecting pollutant data in the target monitoring environment in real time through a multi-parameter sensor array;

[0083] The collected data includes multi-dimensional information such as chemical composition, biological activity and physical state. The collection frequency is optimized according to monitoring needs and sensor performance to ensure the real-time and accuracy of the data;

[0084] S13, preprocessing the collected pollutant data, wherein the preprocessing includes data cleaning, denoising and normalization;

[0085] The collected raw data usually contains noise and outliers, so preprocessing is required. The preprocessing steps include data cleaning, denoising and normalization. The data cleaning step uses filtering algorithms to remove noise and outliers, such as limiting filtering, median filtering or Gaussian filtering. The denoising step further reduces random fluctuations in the data and ensures data stability. The normalization step converts the data into a unified dimension and range for subsequent analysis;

[0086] S14, integrating the preprocessed data into an initial monitoring data set;

[0087] The integration steps include data synchronization, timestamp correction, and data formatting to ensure the temporal continuity and spatial consistency of the data set. These steps provide high-quality input data for subsequent deep mining and feature extraction, and provide a solid foundation for the toxicity assessment of environmental pollutants.

[0088] S2. Using big data analysis and cloud computing technology to perform deep mining on the initial monitoring data set obtained in step S1 and extract the toxicity characteristics of pollutants to generate a pollutant characteristic data set, specifically including the following steps:

[0089] S21, performing data cleaning on the initial monitoring data set obtained in step S1 to form a cleaned monitoring data set; the data cleaning includes removing outliers, filling missing data and data standardization;

[0090] Real-time collection of pollutant data in the environment is done through multi-parameter sensor arrays that can monitor the concentration of multiple pollutants, including chemical, biological and physical parameters. The collected data will be transmitted to the central data processing unit, which has big data analysis and cloud computing technology to process and analyze the collected data. The initial monitoring data obtained is transmitted to the central data processing unit via wireless or wired communication. This step ensures the real-time and reliability of data transmission. The central data processing unit receives the data stream from the sensor array and performs preliminary data synchronization and format unification processing.

[0091] Data cleaning is necessary for the initial monitoring data received, as the raw data may contain noise, outliers, or missing values. The data cleaning process includes using filtering algorithms to remove noise, using interpolation methods to fill in missing data, and performing data standardization to ensure data quality and consistency;

[0092] S22. Apply big data analysis techniques to conduct statistical analysis and pattern recognition on the cleaned monitoring data sets to identify potential toxicity patterns of pollutants;

[0093] This step involves applying big data analysis techniques, such as cluster analysis, correlation analysis, and principal component analysis, to identify potential toxicity patterns of pollutants; these analyses help understand the relationship between pollutant characteristics and toxicity and identify key factors that may affect toxicity;

[0094] S23. Use cloud computing resources to conduct in-depth analysis of the identified potential toxicity patterns and extract the toxicity characteristics of pollutants;

[0095] Said in-depth analysis includes machine learning classification and regression analysis;

[0096] This step may include machine learning algorithms such as support vector machines, random forests, or neural networks for classification and regression analysis, which are able to learn from large amounts of data and extract toxicity signatures of pollutants;

[0097] S24, screening the extracted pollutant toxicity characteristics through a feature selection algorithm to determine the most representative pollutant toxicity characteristic set;

[0098] Feature selection algorithms, such as recursive feature elimination or model-based feature selection, help reduce the dimensionality of the feature space while retaining the most valuable information;

[0099] S25, collecting the most representative pollutant toxicity characteristics after screening to generate a pollutant characteristic data set;

[0100] This dataset will be used for subsequent toxicity assessment and model optimization. To ensure the accuracy and reliability of the dataset, validation and testing will be performed. This includes comparative analysis with known toxicity data and the use of cross-validation methods to evaluate the performance of the model.

[0101] S26. Verify and test the generated pollutant characteristic data set;

[0102] The validation process includes comparative analysis with known toxicity data and cross-validation methods;

[0103] Step S2 uses big data analysis and cloud computing technology to conduct in-depth mining of the transmitted initial monitoring data. This process includes data cleaning to remove noise and outliers; feature selection to identify the most representative data features; pattern recognition to discover potential laws in the data; and association rule learning to establish the relationship between pollutant characteristics and toxicity. Through these steps, a detailed pollutant characteristic data set is generated to provide accurate input data for subsequent toxicity assessment.

[0104] S3. According to the target environmental monitoring requirements, preset monitoring parameters and configuration of the assessment model to form a customized assessment model framework for toxicity assessment and model optimization, which specifically includes the following steps:

[0105] S31. Analyze the specific needs of the target monitoring environment and determine the preliminary settings of monitoring parameters;

[0106] The specific requirements of the target monitoring environment include, but are not limited to, types of pollutants, monitoring frequency, data accuracy and environmental conditions;

[0107] This analysis includes an assessment of pollutant types, monitoring frequency, data accuracy, and environmental conditions. For example, if the target environment is an industrial area, the pollutants that may need to be monitored include sulfur dioxide, nitrogen oxides, and particulate matter. The monitoring frequency may be once an hour, and the data accuracy requirements may be very high to ensure that abnormal changes in pollutant concentrations can be detected in a timely manner;

[0108] S32, selecting parameters of the multi-parameter sensor array according to the specific requirements analysis results of the target detection environment;

[0109] Based on these requirements, select the appropriate sensor type and configuration. For example, chemical sensors may be used to monitor gaseous pollutants, biosensors are used to monitor microbial activity in water bodies, and physical sensors are used to monitor particulate matter. The configuration of sensors will be optimized according to the type of pollutants and the monitoring frequency to ensure that the monitoring area can be fully covered and representative data can be obtained. Depending on the specific needs of the target detection environment, it may be a single requirement or a requirement with regional characteristics. For example, only gaseous pollutants need to be detected, or different regions or environments have different assessment standards for the pollutants. At this time, the parameters of the multi-parameter sensor array need to be adjusted in a targeted manner;

[0110] S33. Design the architecture of the evaluation model based on the specific needs analysis results of the target detection environment, which includes determining the model input parameters, such as sensor data, environmental condition parameters (such as temperature, humidity), etc., output results, such as pollutant toxicity level, and processing algorithms, such as data fusion algorithms, machine learning classification algorithms, etc. The design of the evaluation model will take into account the complexity and real-time requirements of data processing;

[0111] S34. Use historical monitoring data and known pollutant toxicity information to conduct preliminary training and parameter adjustment of the assessment model to form a basic assessment model;

[0112] This step involves using statistical learning methods to determine model parameters, or using machine learning methods to train the model. For example, a support vector machine (SVM) or random forest algorithm can be used to train the model to identify the toxicity characteristics of different pollutants;

[0113] S35, applying the basic evaluation model to the actual monitoring data, performing model verification, and obtaining performance indicators of the evaluation model;

[0114] The performance indicators of the evaluation model include precision, recall and F1 score;

[0115] This step involves comparing the model’s predictions with actual pollutant toxicity data to assess the model’s accuracy and reliability;

[0116] S36, further adjusting and optimizing the parameters of the evaluation model according to the performance indicators of the evaluation model to obtain an optimized evaluation model;

[0117] This step may involve using cross-validation methods to optimize model parameters, or using techniques such as grid search to find the best combination of parameters;

[0118] S37, integrating the optimized assessment model with monitoring parameters and data processing flow to form a customized assessment model framework for automatically receiving pollutant characteristic data sets, performing processing and toxicity assessment, and outputting results that meet the target environmental monitoring requirements;

[0119] This framework will be able to automatically receive pollutant feature data sets from sensor arrays, process and evaluate toxicity, and output results that meet the target environmental monitoring needs. For example, if the monitoring requirements are real-time monitoring and rapid response, the assessment model framework will include real-time data processing and rapid assessment algorithms to ensure that toxicity assessment results can be provided in a timely manner.

[0120] Step S3 involves optimizing the structural design of the evaluation model, setting the learning rate, selecting the loss function, and defining the evaluation index. The customized evaluation model framework can be adjusted according to different environmental monitoring requirements to adapt to different monitoring scenarios and pollutant characteristics.

[0121] S4. Input the generated pollutant characteristic data set into a preset customized assessment model to conduct toxicity assessment of environmental pollutants and obtain preliminary assessment results, which specifically includes the following steps:

[0122] S41, inputting the pollutant characteristic data set generated in step S2 into the optimized evaluation model obtained in step S36 according to the preset format and requirements;

[0123] Extract pollutant characteristic data sets from the central data processing unit. These data sets are processed through big data analysis and cloud computing technology and contain detailed toxicity characteristics of pollutants. This step is the basis of the evaluation process and ensures the accuracy and reliability of subsequent evaluations;

[0124] Input the pollutant characteristic data set into the assessment model according to the preset format and requirements. This step ensures that the data set is compatible with the model's input interface and provides the necessary input data for the model;

[0125] The preset formats and requirements refer to the requirements in the data processing flow in the customized evaluation model framework;

[0126] S42, in the evaluation model obtained in step S3, executing a toxicity evaluation algorithm according to the input pollutant characteristic data set, the toxicity evaluation algorithm performs calculations according to the monitoring parameters and evaluation model configuration preset in step S37 to evaluate the toxicity of the pollutant;

[0127] The algorithm calculates and evaluates the toxicity of pollutants based on the preset monitoring parameters and assessment model configuration. This step is the core of the assessment process. Through the calculation of the algorithm, the toxicity assessment results of pollutants can be obtained;

[0128] The toxicity assessment algorithm specifically includes the following steps:

[0129] S421. Define a loss function for evaluating the model;

[0130] The loss function is used to measure the difference between the predicted value of the evaluation model and the actual pollutant toxicity value;

[0131] Define the loss function of the evaluation model, which is the key to evaluating the performance of the model. Usually, the mean square error or cross entropy loss function is selected to ensure that the model can accurately predict the toxicity value of the pollutant.

[0132] S422, based on the defined loss function, calculate the initial loss value of the evaluation model; by inputting the pollutant characteristic data set obtained in step S2 into the evaluation model, and comparing the predicted output of the model with the actual toxicity value, the formula is expressed as:

[0133]

[0134] Where: L represents the overall loss, N is the number of samples, l is the loss function, y i is the actual toxicity value of the i-th sample, is the toxicity value predicted by the model;

[0135] This step is achieved by inputting the pollutant characteristic data set into the evaluation model and comparing the model's predicted output with the actual toxicity value. The calculation of the loss value provides a benchmark for subsequent model optimization.

[0136] S423, using a gradient descent algorithm to calculate the gradient of the evaluation model parameters according to the initial loss value;

[0137] The gradient calculation formula is as follows:

[0138]

[0139] Where: θ represents the model parameters, represents the loss function, L represents the gradient with respect to the model parameter θ;

[0140] The gradient descent algorithm is used to calculate the gradient of the evaluation model parameters. This step is achieved by taking the partial derivatives of the loss function with respect to the model parameters. The direction of the gradient indicates the direction of parameter adjustment to reduce the loss.

[0141] S424. Update the parameters of the evaluation model according to the calculated gradient. The parameter update rule is determined according to the adaptive learning algorithm. The update formula is:

[0142]

[0143] Among them, θ t represents the model parameters at time step t, η t represents the learning rate, represents the gradient at time step t;

[0144] This step is achieved through an adaptive learning algorithm, which automatically adjusts the learning rate based on the gradient information to dynamically adjust the parameter update step size.

[0145] S425, repeating steps S422 to S424 until the loss value of the evaluation model converges or reaches a predetermined number of iterations, obtaining;

[0146] This process ensures the optimal performance of the evaluation model in toxicity assessment tasks and improves the predictive accuracy of the model.

[0147] S426. Use the final optimized evaluation model to perform toxicity evaluation on the pollutant characteristic data set, obtain preliminary toxicity evaluation results, and output the results.

[0148] This step is to apply the optimized model to actual toxicity assessment and provide a scientific basis for environmental management and decision-making.

[0149] S43. The assessment model processes the input pollutant characteristic data set and generates preliminary toxicity assessment results based on built-in logic and parameters, expressing the toxicity level of pollutants in the form of numerical values, categories or probabilities, providing important information for environmental management and decision-making;

[0150] S44. Perform quality control on the preliminary toxicity assessment results;

[0151] The quality control of the toxicity assessment results includes verifying the integrity of the data, the rationality of the assessment results, and the consistency with historical data;

[0152] This step ensures the accuracy and reliability of the evaluation results and provides a solid foundation for subsequent evaluation model optimization and toxicity prediction.

[0153] S45. Output the preliminary evaluation results that have undergone quality control to the result database to provide basic data for optimization of the evaluation model and toxicity prediction;

[0154] This step provides basic data for evaluation model optimization and toxicity prediction, allowing the evaluation process to be continuously improved and optimized.

[0155] In step S4, the evaluation model may include a machine learning model, a deep learning model, a statistical model, etc. The selection and configuration of the model are determined based on the characteristics of the pollutant characteristic data set and the evaluation requirements.

[0156] S5. Adopting an adaptive learning algorithm, based on the obtained preliminary evaluation results, automatically adjusting the parameters of the customized evaluation model to obtain an optimized evaluation model, and using the optimized evaluation model to re-evaluate the toxicity of the pollutant characteristic data set to obtain an optimized evaluation result, specifically comprising the following steps:

[0157] S51, analyzing the accuracy and reliability of the preliminary evaluation results obtained in step S4, and determining whether there are systematic deviations or parameters to be optimized by calculating performance indicators such as the accuracy, recall rate, and F1 score of the evaluation model prediction;

[0158] The preliminary evaluation results are generated by a customized evaluation model. The preliminary evaluation results include the toxicity characteristics of the pollutants and preliminary evaluation data. The preliminary evaluation results form a preliminary evaluation data set to provide input for the subsequent adaptive learning algorithm.

[0159] S52, identifying the evaluation model parameters to be adjusted according to the analysis result of step S51;

[0160] The evaluation model parameters include, but are not limited to, evaluation model weights, threshold settings, or algorithm-specific hyperparameters to optimize the performance of the model and improve its accuracy in predicting pollutant toxicity;

[0161] S53, applying an adaptive learning algorithm to dynamically adjust the identified parameters according to the changes in the preliminary evaluation results obtained in step S4, the adaptive learning algorithm automatically adjusts the learning rate using historical gradient information, adjusts the parameter update step size, and obtains new parameter values; this step is the key to achieving model optimization, ensuring that the model can adapt to new data and environmental changes;

[0162] S54, updating the evaluation model according to the new parameter values ​​provided by the adaptive learning algorithm to form an optimized evaluation model, which can more accurately reflect the toxicity characteristics of the pollutants and provide more accurate toxicity evaluation results;

[0163] S55. Use the updated assessment model to re-evaluate the toxicity of the pollutant characteristic data set to obtain optimized assessment results. These results represent the toxicity level of the pollutants in the form of numerical values, categories or probabilities, providing a scientific basis for environmental management and decision-making, and being able to more accurately predict the toxicity level of pollutants;

[0164] S56. By comparing and analyzing known toxicity data and using cross-validation methods, the optimized evaluation results are verified and tested to ensure the accuracy and reliability of the evaluation model, providing a solid foundation for the toxicity detection and evaluation of environmental pollutants.

[0165] In step S5, the adaptive learning algorithm dynamically adjusts the model parameters according to the difference between the evaluation result and the expected target or historical data to optimize the model performance. The selection of the adaptive learning algorithm includes but is not limited to the gradient descent method, the genetic algorithm and the particle swarm optimization algorithm, etc. The selection of the algorithm is determined according to the complexity of the model and the optimization target.

[0166] The optimized assessment model is used to re-evaluate the toxicity of the pollutant characteristic data set to obtain the optimized assessment results. This step ensures the accuracy and reliability of the assessment results. The optimized assessment results include not only the quantitative description of the toxicity of the pollutants, but also information such as toxicity levels and risk warnings, providing a scientific basis for environmental management and decision-making.

[0167] S6. Using the pre-trained deep learning algorithm, trend analysis and prediction are performed on the optimized evaluation results obtained in step S5 to generate a pollutant toxicity prediction report, which specifically includes the following steps:

[0168] S61, inputting the optimized evaluation result obtained in step S5 into the pre-trained deep learning model;

[0169] The optimized pollutant toxicity assessment results generated by the assessment model include detailed toxicity characteristics of the pollutants and preliminary assessment data, providing rich input information for the deep learning model;

[0170] The pre-trained deep learning model has been trained on a large amount of historical toxicity data and is able to identify patterns and trends in pollutant toxicity, a step that is fundamental to achieving accurate predictions;

[0171] The pre-trained deep learning model is obtained through a series of detailed steps involving data preparation, model selection, training, validation, and optimization. The following are the specific steps of pre-training deep learning models:

[0172] S611. Data collection and preprocessing

[0173] Data Collection: Gather historical monitoring data and known contaminant toxicity information. This data may come from a variety of sources, including laboratory testing, field monitoring, scientific literature, and public databases.

[0174] Data preprocessing: Clean, denoise, and normalize the collected data to ensure data quality. This includes dealing with missing values, outliers, and inconsistent formats.

[0175] S612, Feature Engineering

[0176] Feature extraction: Use big data analysis techniques to extract useful features from pre-processed data. This may include statistical analysis, pattern recognition, and machine learning algorithms to identify potential toxicity patterns of pollutants.

[0177] Feature selection: The most representative feature set is screened out through feature selection algorithms (such as recursive feature elimination or model-based feature selection) to reduce the dimensionality of the feature space and improve the efficiency of model training.

[0178] S613. Model selection and training

[0179] Model selection: Choose an appropriate deep learning architecture, such as a convolutional neural network (CNN), a recurrent neural network (RNN), or a long short-term memory network (LSTM), depending on the characteristics of your data and the requirements of your problem.

[0180] Model training: Use the data after feature selection to train the deep learning model. This process involves setting the loss function (such as cross entropy loss), choosing an optimizer (such as Adam or SGD), and adjusting hyperparameters (such as learning rate, batch size, and number of iterations).

[0181] S614, Model Validation and Optimization

[0182] Model Validation: Validate the performance of deep learning models through cross-validation and comparison with other models. Evaluate the performance metrics of the model such as accuracy, recall, and F1 score.

[0183] Model optimization: Based on the validation results, further adjust and optimize the model parameters and structure to improve the model's prediction accuracy and generalization ability.

[0184] S615. Preservation and deployment of pre-trained models

[0185] Model saving: Save the trained and optimized deep learning model into a deployable format for use in real-world applications.

[0186] Model deployment: Deploy the pre-trained deep learning model to the central data processing unit, enabling it to receive real-time pollutant characteristic data sets and perform toxicity assessment and trend prediction.

[0187] S616, Continuous Learning and Model Update

[0188] Continuous learning: As new data is continuously collected, the pre-trained deep learning model is regularly fine-tuned with new data to adapt to environmental changes and new toxicity patterns.

[0189] Model update: Update the deployed model based on the results of continuous learning to ensure that the model always maintains the best predictive performance.

[0190] Through the above steps, the pre-trained deep learning model in the present invention can accurately predict and analyze the toxicity trends of environmental pollutants, providing strong technical support for environmental management and decision-making.

[0191] S62, executing the forward propagation process of the deep learning model, and calculating the toxicity assessment result of the deep learning model for each sample;

[0192] Perform the forward propagation process of the deep learning model, where the model extracts trends and patterns in the time series data through its multi-layer structure. This process involves the accumulation of activation functions and weights of all layers in the model.

[0193] S63. In the output layer of the deep learning model, a softmax function is applied to convert the output of the deep learning model into a probability distribution, representing the probabilities of different toxicity categories;

[0194] At the output layer of the model, a softmax function is applied to convert the output of the model into a probability distribution, representing the probability of different toxicity categories. This step converts the output of the model into an interpretable toxicity prediction category.

[0195] S64. Determine the predicted category of pollutant toxicity, i.e., the most likely toxicity level, based on the probability distribution output by the deep learning model;

[0196] This step is key to achieving pollutant toxicity prediction and provides clear guidance for environmental management.

[0197] S65. Analyze and predict the toxicity trend of pollutants using the prediction results of the deep learning model in combination with time series analysis technology, and generate the predicted change trend of pollutant toxicity in the future time period;

[0198] This step takes advantage of deep learning's ability to process complex data structures and is able to capture subtle trends in changes in pollutant toxicity.

[0199] S66. Deep learning model generates prediction reports on pollutant toxicity;

[0200] The forecast report includes the predicted toxicity level, changing trend and expected impact;

[0201] The report describes in detail the predicted trends, potential risks and recommended countermeasures of pollutant toxicity, providing a scientific basis for environmental management and decision-making. The generation of the report is the ultimate goal of this invention, which converts the prediction results of the deep learning model into decision support in practical applications.

[0202] S67. Compare and verify the forecast report with historical data and real-time monitoring data to evaluate the accuracy and reliability of the model forecast. This verification step is an important part of model optimization, which ensures the reliability of the model prediction results and provides feedback for further optimization of the model;

[0203] S68. According to the verification results, the pre-trained deep learning model is adjusted and retrained to obtain an optimized deep learning model.

[0204] This step ensures that the model can adapt to new data and environmental changes, improving the model's performance in future trend analysis and prediction.

[0205] The present invention realizes accurate assessment and effective management of the toxicity of environmental pollutants through real-time monitoring, in-depth data analysis, customized assessment models and adaptive learning algorithms; it can ensure the accuracy and reliability of environmental pollutant toxicity detection and assessment methods, and provide a scientific basis for environmental management and decision-making; through the application of adaptive learning algorithms, it realizes continuous optimization and improvement of assessment models, and improves the accuracy and real-time nature of assessment; through the application of deep learning algorithms, it realizes accurate prediction of pollutant toxicity trends, and provides strong technical support for environmental management and decision-making.

[0206] The applicant declares that the above is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Those skilled in the art should understand that any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention shall fall within the protection scope and disclosure scope of the present invention.

Claims

1. A method for detecting and evaluating the toxicity of environmental pollutants, characterized in that: The following steps are involved: S1. Deploy a multi-parameter sensor array in the target monitoring environment to obtain initial monitoring data; S2. Use big data analysis and cloud computing technology to perform in-depth mining of the initial monitoring data obtained in step S1 and extract toxicity characteristics of pollutants to generate a pollutant characteristic data set; S3. According to the monitoring requirements of the target environment, preset the configuration of monitoring parameters and evaluation models to form a customized evaluation model framework; S4, inputting the pollutant characteristic data set generated in step S2 into a preset customized assessment model to conduct a toxicity assessment of environmental pollutants and obtain a preliminary assessment result; S5. Adopt an adaptive learning algorithm, based on the preliminary evaluation results obtained in step S4, automatically adjust the parameters of the customized evaluation model to obtain an optimized evaluation model, use the optimized evaluation model to perform toxicity evaluation on the pollutant characteristic data set again, and obtain an optimized evaluation result.

2. The method for detecting and evaluating the toxicity of environmental pollutants according to claim 1, characterized in that: The step S1 deploys a multi-parameter sensor array in the target monitoring environment to obtain initial monitoring data, specifically including the following steps: S11. Install and calibrate the multi-parameter sensor array; The multi-parameter sensor array includes chemical sensors, biological sensors and physical sensors; S12, collecting pollutant data in the target monitoring environment in real time through a multi-parameter sensor array; S13, preprocessing the collected pollutant data, wherein the preprocessing includes data cleaning, denoising and normalization; S14. Integrate the preprocessed data into an initial monitoring data set.

3. The method for detecting and evaluating the toxicity of environmental pollutants according to claim 1, characterized in that: The step S2 uses big data analysis and cloud computing technology to perform deep mining on the initial monitoring data set obtained in step S1 and extract the toxicity characteristics of pollutants to generate a pollutant characteristic data set, which specifically includes the following steps: S21, performing data cleaning on the initial monitoring data set obtained in step S1 to form a cleaned monitoring data set; The data cleaning includes removing outliers, filling missing data and data standardization; S22. Apply big data analysis techniques to conduct statistical analysis and pattern recognition on the cleaned monitoring data sets to identify potential toxicity patterns of pollutants; S23. Use cloud computing resources to conduct in-depth analysis of the identified potential toxicity patterns and extract the toxicity characteristics of pollutants; Said in-depth analysis includes machine learning classification and regression analysis; S24, screening the extracted pollutant toxicity characteristics through a feature selection algorithm to determine the most representative pollutant toxicity characteristic set; S25. The most representative pollutant toxicity characteristics are collected after screening to generate a pollutant characteristic data set.

4. The method for detecting and evaluating the toxicity of environmental pollutants according to claim 1, characterized in that: The step S3 presets the configuration of monitoring parameters and evaluation model according to the target environment monitoring requirements to form a customized evaluation model framework, which specifically includes the following steps: S31. Analyze the specific needs of the target monitoring environment and determine the preliminary settings of monitoring parameters; The specific requirements of the target monitoring environment include pollutant types, monitoring frequency, data accuracy and environmental conditions; S32, selecting parameters of the multi-parameter sensor array according to the specific requirements analysis results of the target detection environment; S33. Design the architecture of the evaluation model based on the specific requirements analysis results of the target detection environment; The framework of the evaluation model includes determining the input parameters, output results and processing algorithms of the model; S34. Use historical monitoring data and known pollutant toxicity information to conduct preliminary training and parameter adjustment of the assessment model to form a basic assessment model; S35, applying the basic evaluation model to the actual monitoring data, performing model verification, and obtaining performance indicators of the evaluation model; The performance indicators of the evaluation model include precision, recall and F1 score; S36, further adjusting and optimizing the parameters of the evaluation model according to the performance indicators of the evaluation model to obtain an optimized evaluation model; S37. Integrate the optimized assessment model with monitoring parameters and data processing procedures to form a customized assessment model framework.

5. The method for detecting and evaluating the toxicity of environmental pollutants according to claim 1, characterized in that: The step S4 inputs the generated pollutant characteristic data set into a preset customized assessment model to perform toxicity assessment of environmental pollutants and obtain preliminary assessment results, which specifically includes the following steps: S41, inputting the pollutant characteristic data set generated in step S2 into the optimized evaluation model obtained in step S3 according to the preset format and requirements; S42, in the evaluation model obtained in step S3, executing a toxicity evaluation algorithm according to the input pollutant characteristic data set, the toxicity evaluation algorithm performs calculations according to the monitoring parameters and evaluation model configuration preset in step S37 to evaluate the toxicity of the pollutant; S43. The assessment model processes the input pollutant characteristic data set and generates preliminary toxicity assessment results based on built-in logic and parameters.

6. The method for detecting and evaluating the toxicity of environmental pollutants according to claim 5, characterized in that: The toxicity assessment algorithm specifically includes the following steps: S421. Define a loss function for evaluating the model; The loss function is used to measure the difference between the predicted value of the evaluation model and the actual pollutant toxicity value; S422, based on the defined loss function, calculate the initial loss value of the evaluation model; by inputting the pollutant characteristic data set obtained in step S2 into the evaluation model, and comparing the predicted output of the model with the actual toxicity value, the formula is expressed as: Where: L represents the overall loss, N is the number of samples, l is the loss function, y i is the actual toxicity value of the i-th sample, is the toxicity value predicted by the model; S423, using a gradient descent algorithm, calculating the gradient of the evaluation model parameters according to the initial loss value; The gradient calculation formula is as follows: Where: θ represents the model parameters, represents the loss function, L represents the gradient with respect to the model parameter θ; S424. Update the parameters of the evaluation model according to the calculated gradient. The parameter update rule is determined according to the adaptive learning algorithm. The update formula is: Among them, θ t represents the model parameters at time step t, η t represents the learning rate, represents the gradient at time step t; S425. Repeat steps S422 to S424 until the loss value of the evaluation model converges or reaches a predetermined number of iterations.

7. The method for detecting and evaluating the toxicity of environmental pollutants according to claim 1, characterized in that: The step S5 adopts an adaptive learning algorithm to automatically adjust the parameters of the customized evaluation model based on the obtained preliminary evaluation results to obtain an optimized evaluation model, and uses the optimized evaluation model to perform toxicity evaluation on the pollutant characteristic data set again to obtain an optimized evaluation result, which specifically includes the following steps: S51, analyzing the accuracy and reliability of the preliminary evaluation results obtained in step S4 to determine whether there are systematic deviations or parameters to be optimized; The system includes calculating the accuracy, recall and F1 score of the predictions of the evaluation model; S52, identifying the evaluation model parameters to be adjusted according to the analysis result of step S51; The evaluation model parameters include evaluation model weights, threshold settings, or algorithm-specific hyperparameters; S53, applying an adaptive learning algorithm to dynamically adjust the identified parameters according to changes in the preliminary evaluation results obtained in step S4, wherein the adaptive learning algorithm uses historical gradient information to automatically adjust the learning rate and the parameter update step size to obtain new parameter values; S54, updating the evaluation model according to the new parameter value provided by the adaptive learning algorithm; S55. Use the updated assessment model to conduct toxicity assessment on the pollutant characteristic data set again to obtain optimized assessment results.

8. The method for detecting and evaluating the toxicity of environmental pollutants according to claim 1, characterized in that: The step S6 uses a pre-trained deep learning algorithm to perform trend analysis and prediction on the optimized evaluation results obtained in step S5 to generate a prediction report on the toxicity of pollutants, which specifically includes the following steps: S61, inputting the optimized evaluation result obtained in step S5 into the pre-trained deep learning model; S62, executing the forward propagation process of the deep learning model, and calculating the toxicity assessment result of the deep learning model for each sample; S63, applying a softmax function at the output layer of the deep learning model to convert the output of the deep learning model into a probability distribution, representing the probabilities of different toxicity categories; S64. Determine the predicted category of pollutant toxicity based on the probability distribution output by the deep learning model; S65. Analyze and predict the toxicity trend of pollutants using the prediction results of the deep learning model in combination with time series analysis technology, and generate the predicted change trend of pollutant toxicity in the future time period; S66. Deep learning model generates prediction reports on pollutant toxicity; The forecast report includes predicted toxicity levels, changing trends and expected impacts.

Images