A water quality evaluation method and system based on adaptive fuzzy clustering

Abstract

The application discloses a water quality evaluation method and system based on adaptive fuzzy clustering, and the method comprises the following steps: obtaining water pollution parameters; performing clustering analysis on the water pollution parameters based on an EMT time series clustering algorithm and a fuzzy mathematics algorithm to obtain water pollution parameter indexes; performing water pollution prediction on the water pollution parameter indexes based on a deep learning neural network, and dividing water quality grades according to the water pollution prediction results to obtain a comprehensive water quality evaluation grade; and performing early warning grade division processing according to the comprehensive water quality evaluation grade, and performing water quality risk early warning according to the divided early warning grades. Through the use of the application, the water quality can be objectively evaluated and analyzed from multiple dimensions, so that the accuracy of the water quality prediction result is improved. The application can be widely applied to the field of water quality parameter evaluation as a water quality evaluation method and system based on adaptive fuzzy clustering.

Description

Technical Field

[0001] This invention relates to the field of water quality parameter assessment, and in particular to a water quality assessment method and system based on adaptive fuzzy clustering. Background Technology

[0002] River water quality is affected by numerous water quality indicators, such as pH, dissolved oxygen, turbidity, ammonia nitrogen, permanganate, organic pollutants, and other factors. Common single-factor evaluation methods only select the worst water quality indicator in the target water body as the evaluation parameter, failing to objectively reflect the overall comprehensive state of the water body. However, related technologies utilize machine learning algorithms for river pollutant risk assessment. For example, a water quality assessment method using random forest models to predict the Comprehensive Water Quality Index (WQI) score and optimize PMF (Potentially Flood Factor) source tracing involves: determining parameter weights in the WQI model based on literature experience; assigning values ​​to monitoring data based on the established WQI model; using the random forest model to predict and evaluate performance parameters and determine the optimal WQI model; and further using the performance parameters obtained from the random forest model to optimize PMF. While the source allocation calculation mechanism is available, the weights in the WQI model require human evaluation and determination based on relevant literature or expert opinions regarding the target area, resulting in strong subjectivity. Furthermore, the comprehensive evaluation method, which constructs comprehensive weight coefficients using the standard deviation and interaction coefficients of water quality parameters, involves constructing water quality evaluation weight indicators; screening key indicators based on regression analysis; and using the TOPSIS method to correct the comprehensive water quality evaluation indicators for individual samples. However, this technique only considers the statistical indicators of the water quality indicators themselves when assigning weights to each water quality parameter, without considering the impact of social and ecological factors on water quality evaluation. Therefore, related machine learning techniques only use the worst single water quality indicator to determine the water quality category, failing to objectively assess the overall water quality status. Moreover, the evaluation process involves expert evaluation or reference to experience, resulting in strong subjectivity. Summary of the Invention

[0003] To address the aforementioned technical problems, the present invention aims to provide a water quality assessment method and system based on adaptive fuzzy clustering, which can objectively evaluate and analyze water quality from multiple dimensions, thereby improving the accuracy of water quality prediction results.

[0004] The first technical solution adopted in this invention is: a water quality assessment method based on adaptive fuzzy clustering, comprising the following steps:

[0005] Acquire water pollution parameters, which include environmental monitoring parameters of the target watershed and corresponding eco-social parameters of the environmental monitoring parameters;

[0006] Based on the EMT time series clustering algorithm and fuzzy mathematics algorithm, cluster analysis is performed on water pollution parameters to obtain water pollution parameter indicators;

[0007] Based on deep learning neural networks, water pollution parameters are used to predict water pollution levels, and water quality grades are classified according to the prediction results to obtain a comprehensive water quality evaluation grade.

[0008] The warning levels are classified according to the comprehensive water quality assessment level, and water quality risk warnings are issued based on the classified warning levels.

[0009] Furthermore, the step of using the EMT-based time series clustering algorithm and fuzzy mathematics algorithm to perform cluster analysis on water pollution parameters and obtain water pollution parameter indicators specifically includes:

[0010] The membership degree of water pollution parameters is obtained by fuzzifying the parameters using a linear function.

[0011] A fitness function is set and a clustering algorithm is introduced to construct a time series clustering algorithm based on EMT. The fitness function includes the maximum silhouette coefficient and the Davidsonburg index.

[0012] The time series clustering algorithm based on EMT is used to perform cluster analysis on the membership degree of water pollution parameters to obtain water pollution parameter indicators.

[0013] Furthermore, the step of fuzzifying the water pollution parameters using a linear function to obtain the membership degree of the water pollution parameters specifically includes:

[0014] Determine the functional relationship between water pollution parameters and water quality.

[0015] If the water pollution parameters are positively correlated with the water quality, then the reciprocal of the water pollution parameters is taken to obtain the water pollution parameters after the first pretreatment.

[0016] If the functional relationship between water pollution parameters and water quality is an inverted V-shape, then the water pollution parameters are subtracted from the optimal value and the absolute value is taken to calculate the water pollution parameters after the second pretreatment.

[0017] The water pollution parameters after the first pretreatment and the water pollution parameters after the second pretreatment are standardized and characterized by membership functions to obtain the membership degree of the water pollution parameters.

[0018] Furthermore, the expression for characterization using the membership function is specifically as follows:

[0019]

[0020]

[0021]

[0022] In the above formula, u ij C represents the membership degree of evaluation factor i of water quality evaluation parameters to water quality standard level j. i S represents the standardized value of evaluation factor i. i,j This represents the j-th level water quality grading standard for evaluation factor i.

[0023] Furthermore, the specific expression for the fitness function is as follows:

[0024]

[0025]

[0026] In the above formula, SI represents the maximum profile coefficient, DBI represents the Davidson-Bolding index, and x i Indicates input data, x represents i The average distance to other samples in the same cluster. x represents i The average distance to other samples in different clusters, R ij Represents two clusters C i and C j The similarity between them, where n represents the number of samples in the dataset and k represents the number of clusters.

[0027] Furthermore, the step of using the EMT-based time series clustering algorithm to perform cluster analysis on the membership degrees of water pollution parameters to obtain water pollution parameter indicators specifically includes:

[0028] The membership degrees of water pollution parameters are subjected to time series clustering to obtain the time series of water pollution parameters;

[0029] The fitness values ​​of the time series of water pollution parameters are obtained through the fitness function;

[0030] New populations of water pollution parameters are generated by cross-pollination, variation, and mutation, and then time series clustering is performed to obtain new time series of water pollution parameters.

[0031] The fitness values ​​of the time series of new water pollution parameters are obtained through the fitness function;

[0032] The process involves cyclic time series clustering, fitness value calculation, and crossover, mutation, and mutation to generate a new population, until the fitness value of the time series of new water pollution parameters reaches the optimal fitness value. The water pollution parameter index corresponding to the optimal fitness value is then output.

[0033] Furthermore, the step of predicting water pollution parameters based on deep learning neural networks and classifying water quality levels according to the prediction results to obtain a comprehensive water quality evaluation level specifically includes:

[0034] Construct a deep learning neural network, wherein the deep learning neural network includes an input layer, a hidden layer and an output layer;

[0035] Water pollution parameters are input into a deep learning neural network;

[0036] Based on the input layer of a deep learning neural network, water quality pollution parameter index data are obtained;

[0037] Based on the hidden layer of a deep learning neural network, the water quality pollution parameter index data are fitted to obtain the fitting results of the water quality pollution parameter index.

[0038] Based on the output layer of the deep learning neural network, the fitting results of the water pollution parameter index are output, and the output water pollution parameter index fitting results are obtained.

[0039] The water quality level is determined by fitting the output water pollution parameters to the evaluation indicators, and a comprehensive water quality evaluation level is obtained.

[0040] The second technical solution adopted in this invention is: a water quality assessment system based on adaptive fuzzy clustering, comprising:

[0041] The acquisition module is used to acquire water pollution parameters, which include environmental monitoring parameters of the target watershed and corresponding eco-social parameters.

[0042] The clustering module is used to perform cluster analysis on water pollution parameters based on EMT time series clustering algorithm and fuzzy mathematics algorithm to obtain water pollution parameter indicators;

[0043] The prediction module is used to predict water pollution parameters based on deep learning neural networks, and to classify water quality levels based on the prediction results to obtain a comprehensive water quality evaluation level.

[0044] The early warning module is used to classify early warning levels based on the comprehensive water quality assessment level, and to issue water quality risk warnings based on the classified early warning levels.

[0045] The beneficial effects of the method and system of this invention are as follows: This invention obtains environmental monitoring parameters and corresponding eco-social parameters of the target watershed, and further performs cluster analysis based on the EMT time series clustering algorithm and fuzzy mathematics algorithm. This eliminates the need for subjective methods such as expert evaluation to assign weights to various water quality parameters. The EMT method adaptively selects the number of clusters, further avoiding subjective human influence and making the classification results more objective. The time series clustering method considers the relationship between water quality parameters and time series, identifying possible seasonal or temporal change patterns in the data, improving the accuracy of water quality prediction results. Simultaneously, the data is fuzzified and converted into a membership function to avoid situations where some parameters lack clear classification standards, leading to difficulty in differentiation. Finally, prediction and classification are performed based on neural networks, enabling objective evaluation and analysis of water quality from multiple dimensions, thereby improving the accuracy of water quality prediction results. Attached Figure Description

[0046] Figure 1 This is a flowchart of the steps of a water quality assessment method based on adaptive fuzzy clustering according to the present invention;

[0047] Figure 2 This is a structural block diagram of a water quality assessment system based on adaptive fuzzy clustering according to the present invention;

[0048] Figure 3 This is a flowchart of the steps for water quality assessment and early warning provided in a specific embodiment of the present invention;

[0049] Figure 4 This is a flowchart of the steps for cluster analysis of water pollution parameters provided in a specific embodiment of the present invention;

[0050] Figure 5 This is a schematic diagram of the structure of a deep learning neural network provided in a specific embodiment of the present invention. Detailed Implementation

[0051] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. The step numbers in the following embodiments are only for ease of explanation and do not limit the order of the steps. The execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.

[0052] Before providing a detailed description of the embodiments of this application, some of the nouns and terms involved in the embodiments of this application will be explained first. The nouns and terms involved in the embodiments of this application are subject to the following interpretations.

[0053] 1) Water quality assessment: By measuring and analyzing various physical, chemical and biological parameters in water bodies, and based on specific comprehensive or single-factor evaluation methods, the quality of water bodies is assessed to determine whether they are suitable for specific uses.

[0054] 2) Evolutionary algorithms: A class of computer algorithms that mimic biological evolutionary mechanisms (such as natural selection and heredity) to solve optimization and search problems. They typically simulate the "survival competition" between individuals in a dynamically changing environment to generate the best or feasible solution to the problem.

[0055] 3) Time series clustering: A type of clustering method that divides data into several groups or clusters containing similar groups of each other based on the statistical characteristics, shape, pattern and other measures of time series data.

[0056] 4) Membership degree: A metric in fuzzy mathematics that represents the degree to which an element belongs to a fuzzy set.

[0057] 5) EMT: Multi-task evolutionary algorithm, which assigns different populations to different tasks and allows these groups to communicate appropriately to achieve collaborative operation, in order to achieve an improved class of evolutionary algorithms that solve multiple optimization tasks at the same time.

[0058] Reference Figure 1 This invention provides a water quality assessment method based on adaptive fuzzy clustering, which includes the following steps:

[0059] S100. Obtain water pollution parameters, wherein the water pollution parameters include environmental monitoring parameters of the target watershed and corresponding eco-social parameters of the environmental monitoring parameters.

[0060] Specifically, environmental monitoring parameters (pH, dissolved oxygen, ammonia nitrogen, permanganate index, organic pollutant concentration, etc.) related to the target watershed are selected from the hydrological information dataset. At the same time, corresponding ecological or social parameters (such as the toxic effects of organic matter) and corresponding classification standards are determined based on the selected parameters.

[0061] S200, based on EMT time series clustering algorithm and fuzzy mathematics algorithm, performs cluster analysis on water pollution parameters to obtain water pollution parameter indicators;

[0062] S210. The water pollution parameters are fuzzified using a linear function to obtain the membership degree of the water pollution parameters;

[0063] Specifically, the functional relationship between water pollution parameters and water quality is determined. If the functional relationship between water pollution parameters and water quality is positively correlated, the reciprocal of the water pollution parameters is taken to obtain the first pre-treated water pollution parameters. If the functional relationship between water pollution parameters and water quality is inverted V-shaped, the difference between the water pollution parameters and the optimal value is calculated and the absolute value is taken to obtain the second pre-treated water pollution parameters. The first and second pre-treated water pollution parameters are standardized and characterized by a membership function to obtain the membership degree of the water pollution parameters.

[0064] In this embodiment, using the data obtained in step S100, for parameters where larger values ​​are considered better (such as dissolved oxygen), their reciprocals are taken to convert them into parameters where smaller values ​​are considered better; for parameters where the intermediate value is optimal (such as pH at 7 being optimal), the absolute value of the difference between the intermediate and optimal values ​​is taken; subsequently, the data is standardized. The membership function of the standardized parameters is constructed. Since after preprocessing, all parameters are considered better the smaller they are, a linear function is chosen to represent their membership function, and its expression is:

[0065]

[0066]

[0067]

[0068] In the above formula, u ij C represents the membership degree of evaluation factor i of water quality evaluation parameters to water quality standard level j. i S represents the standardized value of evaluation factor i. i,j This represents the j-th level water quality grading standard for evaluation factor i.

[0069] It should be noted that j≠1 and j≠5 in formula (2). If there is no grading standard for the parameters or the grading standard is less than 5 categories, the first 15%, 15%-35%, 35%-60%, 60%-85%, and 85%-100% of the parameters arranged from smallest to largest are defined as categories I to V, respectively.

[0070] S220. Set the fitness function and introduce a clustering algorithm to construct a time series clustering algorithm based on EMT. The fitness function includes the maximum silhouette coefficient and the Davidsonburg index.

[0071] Specifically, the EMT framework is built using the DEAP library on the Python platform. The fitness functions are set as the maximum silhouette coefficient (SI, which judges the intra-cluster clustering effect, with a value between 0 and 1, the closer to 1 the better the clustering effect) and the Davidson-Bolding index (DBI, which judges the inter-cluster clustering effect, the closer to 0 the better the clustering effect). The calculation methods for the two types of parameters are as follows:

[0072]

[0073]

[0074] In the above formula, SI represents the maximum profile coefficient, DBI represents the Davidson-Bolding index, and x i Indicates input data, x represents i The average distance to other samples in the same cluster. x represents i The average distance to other samples in different clusters, R ij Represents two clusters C i and C j The similarity between them, where n represents the number of samples in the dataset and k represents the number of clusters.

[0075] S230. Based on the EMT time series clustering algorithm, cluster analysis is performed on the membership degree of water pollution parameters to obtain water pollution parameter indicators.

[0076] Specifically, the membership degrees of water pollution parameters are subjected to time-series clustering to obtain time series of water pollution parameters; the fitness values ​​of the time series of water pollution parameters are obtained through a fitness function; crossover, mutation, and mutation of water pollution parameters are performed to generate new populations, and time-series clustering is performed again to obtain new time series of water pollution parameters; the fitness values ​​of the new time series of water pollution parameters are obtained through a fitness function; the time-series clustering process, fitness value calculation process, and crossover, mutation, and mutation to generate new populations are repeated until the fitness value of the new time series of water pollution parameters reaches the optimal fitness value, and the water pollution parameter index corresponding to the optimal fitness value is output.

[0077] In this embodiment, time series clustering provided by the tslearn library in the Python platform is used within the established EMT framework, with the classic k-means clustering method selected. Individuals with optimal fitness are obtained through the EMT framework and used as the optimal cluster size for cluster analysis, ultimately yielding the water quality classification results. Each original result is then assigned a category label. The overall process is as follows: Figure 4As shown, time series clustering is performed by inputting the membership degrees of the parameters obtained after processing in step S210. The SI and DBI indicators in step S220 are selected, and a new population is generated through crossover, mutation, and mutation to re-cluster. The above process is repeated until the fitness is the highest. The output individuals at this time are the optimal number of clusters.

[0078] S300. Based on deep learning neural networks, water pollution parameters are predicted, and water quality grades are classified according to the prediction results to obtain a comprehensive water quality evaluation grade.

[0079] Specifically, a deep learning neural network is constructed, comprising an input layer, a hidden layer, and an output layer. Water pollution parameters are input into the deep learning neural network. Based on the input layer, water pollution parameter data is obtained. Based on the hidden layer, the water pollution parameter data is fitted to obtain a fitting result. Based on the output layer, the fitting result is output to obtain the final water pollution parameter fitting result. Finally, the final water pollution parameter fitting result is used to classify the water quality level using evaluation indicators to obtain a comprehensive water quality evaluation level.

[0080] In this embodiment, the classified water quality parameters obtained in step S200 are used as the training set and input into the deep learning model to discriminate and predict the categories of subsequently newly input water quality data. A schematic diagram of the deep learning model structure is shown below. Figure 5 As shown. The model consists of three layers (input layer, hidden layer, and output layer), and the algorithm can be described as follows:

[0081] L(W,B∣J)

[0082] In the above formula, W is the weight matrix and set {W i} 1:N-1 For N networks, there are i connected layers; J represents the training samples; B is the bias vector of the first layer. The activation function can be a rectified linear function, as shown in the following formula:

[0083] f(α) = max(0, α)

[0084] In the above formula, α is the weighting factor α=∑iw i x i +b;x i With w i is the neuron input value; b is the weighting constant.

[0085] The deep learning model is implemented using the TensorFlow package on the Python platform, with three hidden layers. The first fully connected layer has 128 neurons, and the second fully connected layer has 64 neurons. The training criterion is defined as minimizing the loss, and the cross-entropy function is used as the loss function, as shown in the formula below, to ensure a high degree of fit in the training results.

[0086]

[0087] In the above formula, t and O are the predicted value and the actual input value, respectively; y is the output value.

[0088] After the target parameters are input into the output layer, they undergo hidden layer calculations and fitting to finally obtain the output result. Five parameters can be used to evaluate the model's fit: Accuracy (ACC), which evaluates the model's accuracy; Precision, which evaluates the model's precision; Recall, which evaluates the model's sensitivity; F1 score, which assesses the overall performance of the model and balances precision and recall; and Kappa, which measures the consistency between the classifier's predictions and random predictions. Their respective calculation formulas are as follows:

[0089]

[0090]

[0091]

[0092]

[0093]

[0094] In the above formula, TP, TN, FP, and FN represent true positive, true negative, false positive, and false negative, respectively; a c The number of measured values ​​in each category; b c The number of simulated values ​​in each category; n is the total number of samples.

[0095] S400. Based on the comprehensive water quality assessment level, the early warning level is divided and processed, and water quality risk early warning is issued based on the divided early warning level.

[0096] In this embodiment, the comprehensive water quality evaluation level of the new input data is obtained after deep learning output. It is possible to set a specific water quality category (e.g., category IV or V) for which an early warning will be issued, and subsequent response measures will be taken to handle the alarm. For example, if adaptive classification yields six water quality categories, the early warning levels can be divided according to Table 1.

[0097] Table 1 Correspondence between Water Quality Categories and Warning Levels

[0098]

[0099]

[0100] In summary, the embodiments of the present invention do not require subjective methods such as expert evaluation to assign weights to various water quality parameters. Instead, they use historical data of the study area itself for clustering as classification. At the same time, the EMT method is used to adaptively select the number of clusters, which further avoids the influence of human subjectivity, making the classification results more objective and ensuring the excellent predictive and early warning effect.

[0101] The time series clustering method used in this embodiment of the invention takes into account the relationship between water quality parameters and time series, identifies possible seasonal or temporal change patterns in the data, is closer to the actual situation, and the prediction and early warning results are more realistic.

[0102] This invention also incorporates ecological and socially relevant parameters (such as toxicity parameters of toxic substances) in addition to conventional water quality parameters, making the water quality assessment system more comprehensive and complete. Simultaneously, the data is processed by fuzzification and converted into membership functions to avoid situations where some parameters lack clear classification standards, leading to difficulty in differentiation.

[0103] Reference Figure 2 A water quality assessment system based on adaptive fuzzy clustering includes:

[0104] The acquisition module 201 is used to acquire water pollution parameters, which include environmental monitoring parameters of the target watershed and corresponding eco-social parameters.

[0105] Clustering module 202 is used to perform cluster analysis on water pollution parameters based on EMT time series clustering algorithm and fuzzy mathematics algorithm to obtain water pollution parameter indicators;

[0106] The prediction module 203 is used to predict water pollution parameters based on deep learning neural networks, and to classify water quality levels based on the water pollution prediction results to obtain a comprehensive water quality evaluation level.

[0107] The early warning module 204 is used to classify early warning levels based on the comprehensive water quality assessment level, and to issue water quality risk warnings based on the classified early warning levels.

[0108] The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0109] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A water quality assessment method based on adaptive fuzzy clustering, characterized in that, Includes the following steps: Acquire water pollution parameters, which include environmental monitoring parameters of the target watershed and corresponding eco-social parameters of the environmental monitoring parameters; Determine the functional relationship between water pollution parameters and water quality. If the water pollution parameters are positively correlated with the water quality, then the reciprocal of the water pollution parameters is taken to obtain the water pollution parameters after the first pretreatment. If the functional relationship between water pollution parameters and water quality is an inverted V-shaped relationship, then the water pollution parameters are subtracted from the optimal value and the absolute value is taken to calculate the water pollution parameters after the second pretreatment. The water pollution parameters after the first pretreatment and the water pollution parameters after the second pretreatment are standardized and characterized by membership functions to obtain the membership degree of the water pollution parameters. The expression for characterization using the membership function is specifically shown below: ； ； ； In the above formula, Evaluation factors representing water quality evaluation parameters right Membership degree of the Class I water quality standard Indicates evaluation factors The standardized value, Indicates evaluation factors The Grade I water quality grading standards; A fitness function is set and a clustering algorithm is introduced to construct a time series clustering algorithm based on multi-task evolutionary EMT. The fitness function includes the maximum silhouette coefficient and the Davidsonburg index. Based on the time series clustering algorithm of multi-task evolutionary EMT, cluster analysis is performed on the membership degree of water pollution parameters to obtain water pollution parameter indicators. Based on deep learning neural networks, water pollution parameters are used to predict water pollution levels, and water quality grades are classified according to the prediction results to obtain a comprehensive water quality evaluation grade. The warning levels are classified according to the comprehensive water quality assessment level, and water quality risk warnings are issued based on the classified warning levels.

2. The water quality assessment method based on adaptive fuzzy clustering according to claim 1, characterized in that, The specific expression for the fitness function is as follows: ； ； In the above formula, Indicates the maximum profile coefficient. This refers to the Davidson-Borgin index. Indicates input data, express The average distance to other samples in the same cluster. express The average distance to other samples in different clusters Indicates two clusters and The similarity between them Indicates the number of samples in the dataset. Indicates the number of clusters.

3. The water quality assessment method based on adaptive fuzzy clustering according to claim 2, characterized in that, The time series clustering algorithm based on multi-task evolutionary EMT, which performs cluster analysis on the membership degrees of water pollution parameters to obtain water pollution parameter indicators, specifically includes the following steps: The membership degrees of water pollution parameters are subjected to time series clustering to obtain the time series of water pollution parameters; The fitness values ​​of the time series of water pollution parameters are obtained through the fitness function; New populations of water pollution parameters are generated by cross-pollination, variation, and mutation, and then time series clustering is performed to obtain new time series of water pollution parameters. The fitness values ​​of the time series of new water pollution parameters are obtained through the fitness function; The process involves cyclic time series clustering, fitness value calculation, and crossover, mutation, and mutation to generate a new population, until the fitness value of the time series of new water pollution parameters reaches the optimal fitness value. The water pollution parameter index corresponding to the optimal fitness value is then output.

4. The water quality assessment method based on adaptive fuzzy clustering according to claim 3, characterized in that, The step of predicting water pollution parameters based on deep learning neural networks and classifying water quality levels according to the prediction results to obtain a comprehensive water quality evaluation level specifically includes: Construct a deep learning neural network, wherein the deep learning neural network includes an input layer, a hidden layer and an output layer; Water pollution parameters are input into a deep learning neural network; Based on the input layer of a deep learning neural network, water quality pollution parameter index data are obtained; Based on the hidden layer of a deep learning neural network, the water quality pollution parameter index data are fitted to obtain the fitting results of the water quality pollution parameter index. Based on the output layer of the deep learning neural network, the fitting results of the water pollution parameter index are output, and the output water pollution parameter index fitting results are obtained. The water quality level is determined by fitting the output water pollution parameters to the evaluation indicators, and a comprehensive water quality evaluation level is obtained.

5. A water quality assessment system based on adaptive fuzzy clustering, characterized in that, Includes the following modules: The acquisition module is used to acquire water pollution parameters, which include environmental monitoring parameters of the target watershed and corresponding eco-social parameters. The clustering module is used to determine the functional relationship between water pollution parameters and water quality. If the water pollution parameters are positively correlated with the water quality, then the reciprocal of the water pollution parameters is taken to obtain the water pollution parameters after the first pretreatment. If the functional relationship between water pollution parameters and water quality is an inverted V-shaped relationship, then the water pollution parameters are subtracted from the optimal value and the absolute value is taken to calculate the water pollution parameters after the second pretreatment. The water pollution parameters after the first pretreatment and the water pollution parameters after the second pretreatment are standardized and characterized by membership functions to obtain the membership degree of the water pollution parameters. The expression for characterization using the membership function is specifically shown below: ； ； ； In the above formula, Evaluation factors representing water quality evaluation parameters right Membership degree of the Class I water quality standard Indicates evaluation factors The standardized value, Indicates evaluation factors The Grade I water quality grading standards; A fitness function is set and a clustering algorithm is introduced to construct a time series clustering algorithm based on multi-task evolutionary EMT. The fitness function includes the maximum silhouette coefficient and the Davidsonburg index. Based on the time series clustering algorithm of multi-task evolutionary EMT, cluster analysis is performed on the membership degree of water pollution parameters to obtain water pollution parameter indicators. The prediction module is used to predict water pollution parameters based on deep learning neural networks, and to classify water quality levels based on the prediction results to obtain a comprehensive water quality evaluation level. The early warning module is used to classify early warning levels based on the comprehensive water quality assessment level, and to issue water quality risk warnings based on the classified early warning levels.

Images