A water quality detection method and system based on artificial intelligence

Abstract

The present invention discloses a water quality detection method and system based on artificial intelligence, and the method includes data collection, data preprocessing, water quality pre-classification, water quality trend analysis, pollution source tracing prediction and water quality detection. The present invention relates to the field of water quality detection technology, and specifically refers to a water quality detection method and system based on artificial intelligence. The present scheme adopts an artificial intelligence water quality detection method of water quality pre-classification, trend analysis and pollution source tracing; a deep neural network method combining genetic algorithm and random forest model is used for water quality pre-classification; a deep neural network method combining genetic algorithm and random forest model is used for water quality pre-classification and water quality trend analysis; a water quality pre-classification task is used to first pre-classify the data, exclude unpolluted water sources, and then combine the causal-inspired graph neural network, and through the causal-inspired operator and the improved graph neural network, the calculation speed and calculation effect of pollution source tracing are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality detection, and in particular to a water quality detection method and system based on artificial intelligence. Background Art

[0002] Water quality detection methods based on artificial intelligence use machine learning and data analysis technology to quickly and accurately detect and evaluate water samples. These methods collect water quality data through sensors and use artificial intelligence algorithms to process and analyze the data, which can monitor pollutants, bacteria and other harmful substances in water in real time. Its role not only improves the efficiency and accuracy of water quality detection, but also can achieve early warning, help environmental protection and water resources management, and ensure people's drinking water safety and the sustainable development of the ecological environment.

[0003] However, in the existing water quality detection methods, there is a technical problem that traditional intelligent methods usually focus on single-stage water quality detection to analyze specific components, but only detecting the content of the water quality itself cannot meet the deeper needs of industrial production and life; in the existing water quality detection methods, there is a technical problem that multi-stage detection tasks will inevitably reduce the efficiency of calculation, and therefore the calculation speed and real-time performance of the pre-task must be optimized as much as possible; in the existing water quality trend analysis, there is a technical problem that the traditional time series data prediction model is insufficient in understanding the water body composition and water quality changes in the process of water quality trend analysis, which leads to insufficient prediction performance and prediction efficiency; in the existing pollution source tracing prediction methods, there is a technical problem that pollution source tracing involves complex causal inference logic, and in complex water quality data, extracting this part of causal features has great technical difficulty and design difficulty. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a water quality detection method and system based on artificial intelligence. In view of the technical problem that in the existing water quality detection methods, traditional intelligent methods usually focus on single-stage water quality detection to analyze specific components, but only detect the content of the water quality itself and cannot meet the deeper needs of industrial production and life, this solution creatively adopts an artificial intelligence water quality detection method that aggregates three tasks: water quality pre-classification, trend analysis and pollution source tracing. By subdividing the single-stage rough detection into multi-stage segmented detection, it is helpful to implement the single function maintenance of the system of the method, and can also improve the scope of application of the method itself; in view of the technical problem that in the existing water quality detection methods, there is a multi-stage detection task that will inevitably reduce the calculation efficiency, and therefore the calculation speed and real-time performance of the pre-task must be optimized as much as possible, this solution creatively adopts a deep neural network method combining genetic algorithm and random forest model for water quality pre-classification. Through deep neural network, search The combined integration of algorithms and machine learning models can not only improve the accuracy of classification, but also improve the efficiency of the overall task. In the existing water quality trend analysis, there is a technical problem that the traditional time series data prediction model is not able to understand the water body composition and water quality changes in the process of water quality trend analysis, which leads to insufficient prediction performance and prediction efficiency. This solution creatively adopts a long short-term memory neural network combined with cubic chaos weight optimization. The chaotic cubic algorithm can not only optimize the weight update of the model itself, but also update the content of the optimization algorithm used in the model parameter update process, which improves the execution performance of the phased tasks while reducing the implementation cost of the method and system. In the existing pollution source tracing prediction methods, there is a technical problem that pollution source tracing involves complex causal inference logic, and in complex water quality data, extracting this part of the causal features has great technical difficulty and design difficulty. This solution creatively provides a good practical reference and data support for the depth of water quality detection tasks.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides a water quality detection method based on artificial intelligence, which comprises the following steps:

[0006] Step S1: data collection;

[0007] Step S2: data preprocessing;

[0008] Step S3: water quality pre-classification;

[0009] Step S4: water quality trend analysis;

[0010] Step S5: Pollution source prediction;

[0011] Step S6: water quality testing.

[0012] Furthermore, in step S1, the data collection is used for the original data required for water quality detection, specifically, by collecting multi-point data at different frequencies through water quality sensors and environmental sensors to obtain the original data set for water quality detection.

[0013] Further, in step S2, the data preprocessing is used to optimize the original data, specifically, data cleaning, data standardization and feature engineering are performed on the original water quality detection data set to obtain an optimized water quality detection data set;

[0014] The water quality detection optimization data set specifically includes optimizing water quality sensor data, optimizing sampling frequency data, optimizing environmental data set and characteristic data set.

[0015] Further, in step S3, the water quality pre-classification is used to pre-classify the short-term status of water quality, specifically, based on the optimized water quality sensor data and feature data set in the water quality detection optimization data set, a deep neural network method combining a genetic algorithm and a random forest model is used to pre-classify water quality to obtain water quality short-term classification result data;

[0016] The deep neural network method combining genetic algorithm and random forest model specifically includes a deep neural subnet, a genetic optimization operator and a random forest ensemble model;

[0017] The deep neural subnetwork is used to process water quality classification features and perform preliminary classification;

[0018] The genetic optimization operator is used to optimize the model parameters of the deep neural subnet;

[0019] The random forest ensemble model is used to integrate the results of the deep neural subnet to improve the classification accuracy;

[0020] The step of pre-classifying water quality by using a deep neural network method combining a genetic algorithm and a random forest model to obtain water quality short-term classification result data includes:

[0021] Step S31: constructing a deep neural subnet, specifically constructing a standard deep neural subnet including an input layer, a hidden layer and an output layer, and using a loss function to train the subnet;

[0022] The hidden layer specifically includes a first hidden layer, a second hidden layer and a third hidden layer, the number of neurons in the first hidden layer is 128, the number of neurons in the second hidden layer is 64, and the number of neurons in the third hidden layer is 32;

[0023] The loss function specifically adopts the minimum mean square error loss function;

[0024] Step S32: constructing a genetic optimization operator, specifically, performing iterative model parameter combination calculation through population initialization and fitness evaluation to obtain the optimal deep neural subnet model parameter combination;

[0025] Step S33: constructing a random forest ensemble model, specifically, combining a voting classification mechanism to perform random forest decision tree ensemble learning, and performing fine classification based on the preliminary classification results of the deep neural subnet using the optimal deep neural subnet model parameter combination to obtain a fine classification output of the water quality classification results;

[0026] Step S34: water quality pre-classification model training, specifically, training the water quality pre-classification model through the deep neural subnet, the genetic optimization operator and the random forest ensemble model to obtain the water quality pre-classification model Model WC ;

[0027] Step S35: water quality pre-classification, specifically using the water quality pre-classification model Model WC , performing water quality pre-classification according to the optimized water quality sensor data and feature data set in the water quality detection optimization data set to obtain water quality short-term classification result data;

[0028] The water quality short-term classification result data specifically refers to the water quality pre-classification level prediction data, including excellent water quality, good water quality, slightly polluted water quality, moderately polluted water quality and severely polluted water quality.

[0029] Further, in step S4, the water quality trend analysis is used to classify the trend of water quality changes from the long-term trend, specifically based on the optimized water quality sensor data, optimized sampling frequency data and feature data set in the water quality detection optimization data set, combined with the water quality short-term classification result data, using the long short-term memory neural network combined with cubic chaos weight optimization to perform water quality trend analysis, and obtain water quality long-term trend analysis result data;

[0030] The long short-term memory neural network combined with cubic chaos weight optimization specifically includes a standard long short-term memory neural network and a cubic chaos weight optimization search operator;

[0031] The standard long short-term memory neural network is used to process time series data and features in water quality trend analysis;

[0032] The cubic chaos weight optimization search operator is used to improve the performance of the long short-term memory neural network for trend analysis through the cubic chaos weight optimization search algorithm;

[0033] The step of using a long short-term memory neural network combined with cubic chaos weight optimization to perform water quality trend analysis and obtain water quality long-term trend analysis result data includes:

[0034] Step S41: constructing a standard long short-term memory neural network, specifically constructing a standard long short-term memory neural network including an input gate, a forget gate and an output gate as a basic prediction network of the model;

[0035] Step S42: constructing a cubic chaos random optimization operator, specifically constructing a random chaos dynamic system, and generating a cubic chaos random state vector through the random chaos dynamic system, wherein the cubic chaos random state vector is used to replace the weight of the standard long short-term memory neural network, and optimize the search operator to obtain a cubic chaos optimized long short-term neural network;

[0036] Step S43: constructing a search operator, specifically constructing a bone crest chicken search algorithm, calculating the optimal solution of the model parameters of the standard long short-term memory neural network, and obtaining an initial search operator parameter combination output;

[0037] Step S44: cubic chaos random optimization search algorithm optimization, specifically executing the random chaos dynamic system and performing cubic chaos initialization, regenerating the first random weight coefficient and the second random weight coefficient in the search operator by obtaining the cubic chaos random state vector, and obtaining the cubic chaos random optimization parameter combination of the cubic chaos optimization long-term and short-term neural network through iterative training;

[0038] Step S45: training the water quality trend analysis model, specifically, optimizing the standard long short-term memory neural network through the cubic chaos random optimization search algorithm to obtain the cubic chaos optimized long short-term neural network, and training the water quality trend analysis model according to the cubic chaos random optimization parameter combination to obtain the water quality trend analysis model Model FR ;

[0039] Step S46: water quality trend analysis, specifically using the water quality trend analysis model Model FR , based on the optimized water quality sensor data, optimized sampling frequency data and feature data set in the water quality detection optimization data set, and combined with the data of the water quality short-term classification result data, the pre-classification results of which are lightly polluted water quality, moderately polluted water quality and heavily polluted water quality, a water quality trend analysis is performed to obtain the water quality long-term trend analysis result data;

[0040] The water quality long-term trend analysis result data specifically includes water quality pollution degree change trend prediction data.

[0041] Further, in step S5, the pollution source prediction is used to predict potential pollution sources with short-term and long-term impacts on water quality in combination with environmental data, specifically, based on the optimized water quality sensor data and feature data set in the water quality detection optimization data set, combined with the water quality short-term classification result data, using a graph neural network combined with causal inspiration to perform pollution source prediction, and obtain pollution source prediction result data;

[0042] The graph neural network combined with causal inspiration specifically includes a graph variational autoencoder, a causal inspiration operator, a linear decoder and a basic classifier;

[0043] The graph variational autoencoder is used to construct an improved graph neural network and perform feature encoding;

[0044] The causal heuristic operator is used to set the causal effect estimator to enhance the potential feature mining capability of pollution source tracing;

[0045] The linear decoder is used to perform linear decoding;

[0046] The basic classifier is used to perform prediction and classification according to the decoding result to obtain the pollution college prediction result;

[0047] The step of using a graph neural network combined with causal inspiration to perform pollution source tracing prediction and obtain pollution source tracing prediction result data includes:

[0048] Step S51: constructing a graph variational autoencoder, specifically, reconstructing the graph adjacency matrix and node features of the standard graph convolutional neural network through a multilayer perceptron and a linear inner product decoder, and improving the graph convolutional neural network by constructing a graph variational autoencoder objective function to obtain the graph variational autoencoder;

[0049] Step S52: constructing a causal heuristic operator, specifically, by constructing a causal effect estimator objective function, calculating the characteristic relationship between the pollution source causal characteristics and the pollution source non-causal characteristics, and by constructing the causal effect estimator objective function, and by integrating the variational auto-encoding objective function and the causal effect estimator objective function into a loss function, the potential characteristics of pollution source tracing are enhanced to obtain causal heuristic feature data;

[0050] Step S53: constructing a linear decoder, specifically constructing a linear inner product decoder, and linearly decoding the causal heuristic graph convolution output features to obtain decoded feature data;

[0051] Step S54: constructing a basic classifier, specifically by constructing a standard random forest model as a basic classifier, and obtaining a pollution source tracing classification prediction result according to the decoded feature data;

[0052] Step S55: Training the pollution source tracing prediction model, specifically, training the pollution source tracing prediction model through the graph variational autoencoder, the causal heuristic operator, the linear decoder and the basic classifier, combined with the model loss function, to obtain the pollution source tracing prediction model Model WP ;

[0053] Step S56: Pollution source prediction, specifically using the pollution source prediction model Model WP , based on the optimized water quality sensor data and feature data set in the water quality detection optimization data set, and combined with the water quality short-term classification result data, the pre-classification results are lightly polluted water quality, moderately polluted water quality and heavily polluted water quality, to perform pollution source tracing prediction and obtain pollution source tracing prediction result data;

[0054] The pollution source tracing prediction result data specifically includes pollution source classification results, pollution level classification results and pollution time factor classification results.

[0055] Further, in step S6, the water quality detection is used to perform comprehensive water quality detection in combination with short-term water quality classification, long-term trend analysis and pollution source tracing prediction, specifically, by combining the water quality short-term classification result data, the water quality long-term trend analysis result data and the pollution source tracing prediction result data, a comprehensive water quality detection is performed to obtain a comprehensive water quality detection reference report;

[0056] The comprehensive water quality testing reference report specifically includes water quality grade assessment classification data, key water quality parameter reference data, water quality historical trend comparison data, water quality expected trend data and pollution source analysis data.

[0057] The present invention provides an artificial intelligence-based water quality detection system, which includes a data collection and processing module, a water quality pre-classification module, a water quality trend analysis module, a pollution source tracing prediction module and a water quality detection module;

[0058] The data collection and processing module is used for data collection and data preprocessing. Through data collection and data preprocessing, a water quality detection optimization data set is obtained, and the water quality detection optimization data set is sent to the water quality pre-classification module, the water quality trend analysis module and the pollution source tracing prediction module;

[0059] The water quality pre-classification module is used for water quality pre-classification, obtains water quality short-term classification result data through water quality pre-classification, and sends the water quality short-term classification result data to the water quality trend analysis module, the pollution source tracing prediction module and the water quality detection module;

[0060] The water quality trend analysis module is used for water quality trend analysis, obtains water quality long-term trend analysis result data through water quality trend analysis, and sends the water quality long-term trend analysis result data to the water quality detection module;

[0061] The pollution source tracing prediction module is used for pollution source tracing prediction, obtains pollution source tracing prediction result data through pollution source tracing prediction, and sends the pollution source tracing prediction result data to the water quality detection module;

[0062] The water quality detection module is used for water quality detection, and a comprehensive water quality detection reference report is obtained through water quality detection.

[0063] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0064] (1) In view of the technical problem that traditional intelligent methods usually focus on single-stage water quality detection to analyze specific components, but only detect the content of water quality itself and cannot meet the deeper needs of industrial production and life, this solution creatively adopts an artificial intelligence water quality detection method that aggregates three tasks: water quality pre-classification, trend analysis, and pollution source tracing. By subdividing the single-stage rough detection into multi-stage segmented detection, it is helpful to maintain the single function of the system implementing the method and improve the scope of application of the method itself.

[0065] (2) In the existing water quality detection methods, there are multi-stage detection tasks that will inevitably reduce the efficiency of calculation. Therefore, the calculation speed and real-time performance of the pre-task must be optimized as much as possible. This solution creatively uses a deep neural network method that combines genetic algorithm and random forest model to pre-classify water quality. Through the combined integration of deep neural network, search algorithm and machine learning model, it can not only improve the classification accuracy, but also improve the efficiency of the overall task.

[0066] (3) In the existing water quality trend analysis, there is a technical problem that the traditional time series data prediction model is not able to understand the water body composition and water quality changes in the process of water quality trend analysis, which leads to insufficient prediction performance and prediction efficiency. This solution creatively adopts a long short-term memory neural network combined with cubic chaos weight optimization. The chaotic cubic algorithm can not only optimize the weight update of the model itself, but also update the content of the optimization algorithm used in the model parameter update process, thereby improving the execution performance of the phased task and reducing the implementation cost of the method and system.

[0067] (4) In view of the technical problem that pollution source tracing involves complex causal inference logic in existing pollution source tracing prediction methods, and extracting this part of causal features from complex water quality data has great technical and design difficulties, this solution creatively adopts the water quality pre-classification task to first pre-classify the data to exclude unpolluted water sources, and then combines it with the causal-inspired graph neural network. Through the causal-inspired operator and the improved graph neural network, the calculation speed and effect of pollution source tracing are improved, and the depth of water quality detection tasks is provided with better practical reference and data support. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A schematic diagram of a water quality detection method based on artificial intelligence provided by the present invention;

[0069] Figure 2 A schematic diagram of a water quality detection system based on artificial intelligence provided by the present invention;

[0070] Figure 3 This is a schematic diagram of the process of data preprocessing in step S2;

[0071] Figure 4 This is a schematic diagram of the process of water quality pre-classification in step S3;

[0072] Figure 5 This is a schematic diagram of the process of water quality trend analysis in step S4;

[0073] Figure 6 This is a flow chart of the pollution source prediction process in step S5.

[0074] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0075] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0076] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0077] Example 1, see Figure 1 The present invention provides a water quality detection method based on artificial intelligence, which comprises the following steps:

[0078] Step S1: data collection;

[0079] Step S2: data preprocessing;

[0080] Step S3: water quality pre-classification;

[0081] Step S4: water quality trend analysis;

[0082] Step S5: Pollution source prediction;

[0083] Step S6: water quality testing.

[0084] By performing the above operations, in view of the technical problem that in the existing water quality detection methods, traditional intelligent methods usually focus on single-stage water quality detection to analyze specific components, but only detecting the content of the water quality itself cannot meet the deeper needs of industrial production and life, this solution creatively adopts an artificial intelligence water quality detection method that aggregates three tasks: water quality pre-classification, trend analysis, and pollution source tracing. By subdividing the single-stage rough detection into multi-stage segmented detection, it not only helps to implement the single function maintenance of the system of the method, but also improves the scope of application of the method itself.

[0085] Example 2, see Figure 1 and Figure 2 In step S1, the data collection is used to collect the original data required for water quality detection, specifically by collecting multi-point data at different frequencies through water quality sensors and environmental sensors to obtain the original data set for water quality detection;

[0086] The water quality sensor specifically includes a pH sensor, a turbidity sensor, a dissolved oxygen sensor, a heavy metal component sensor, a conductivity sensor, a chemical oxygen demand sensor and a water temperature sensor;

[0087] The environmental sensors specifically include temperature sensors, humidity sensors, wind speed sensors, time sensors and rainfall sensors;

[0088] In the said frequency-divided multi-point data collection, frequency-divided specifically refers to minute frequency, hour frequency and daily frequency data collection, and multi-point specifically refers to water surface, middle layer and bottom layer sampling;

[0089] The water quality detection original data set specifically includes water quality original data, sampling frequency original data and environment original data set;

[0090] The raw water quality data specifically include physical data, chemical data and biological data;

[0091] The physical data specifically include turbidity data, conductivity data and temperature data;

[0092] The turbidity data is used to indicate the concentration of suspended particles in water;

[0093] The conductivity data is used to indicate the total dissolved solids concentration in the water body;

[0094] The temperature data is used to indicate the temperature of the water body;

[0095] The chemical data specifically includes pH data, dissolved oxygen data, heavy metal concentration data and chemical oxygen demand data;

[0096] The pH data is used to indicate the acidity or alkalinity of water;

[0097] The dissolved oxygen data is used to indicate the dissolved oxygen content in water;

[0098] The heavy metal concentration data is used to indicate the degree of heavy metal pollution in water;

[0099] The chemical oxygen demand data is used to indicate the content of organic pollutants in water bodies;

[0100] The biological data refers specifically to algae concentration data, which is used as a reference indicator of biological activity in water;

[0101] The sampling frequency raw data specifically includes minute-level sampling data, hour-level sampling data, and day-level sampling data;

[0102] The environmental raw data specifically includes temperature data, humidity data, wind speed data, rainfall data and timestamp data.

[0103] Example 3, see Figure 1 , Figure 2 and Figure 3 This embodiment is based on the above embodiment. In step S2, the data preprocessing is used to optimize the original data, specifically to perform data cleaning, data standardization and feature engineering on the original water quality detection data set to obtain an optimized water quality detection data set, which specifically includes the following steps:

[0104] Step S21: Data cleaning, specifically performing missing value processing, outlier detection and denoising on the original water quality detection data set in sequence to obtain a cleaned and optimized data set, specifically including the following steps:

[0105] Step S211: missing value processing, specifically using interpolation and average value filling methods to optimize the actual sensor data;

[0106] Step S212: outlier detection, specifically, outlier detection and processing based on statistical Z-score combined with time series analysis method;

[0107] Step S213: De-noising, specifically, using a wavelet transform method to de-noise the noise signal in the sensor data;

[0108] Step S22: data standardization, specifically, using a maximum-minimum standard method to perform standardization processing on the cleaned and optimized data set to obtain a standardized processed data set;

[0109] Step S23: feature engineering, specifically, using artificial feature engineering design and statistical feature extraction methods to extract features from the standardized processed data set to obtain a feature data set;

[0110] The feature data set specifically includes statistical features, temporal features and environmental interaction features;

[0111] The statistical features specifically include mean features, standard deviation features, extreme value features and slope features;

[0112] The time series characteristics specifically include autocorrelation time series characteristics and trend time series characteristics;

[0113] The environmental interaction characteristics specifically include temperature-dissolved oxygen interaction characteristics and precipitation-turbidity interaction characteristics data;

[0114] Step S24: data optimization, specifically, optimizing the original data through the data cleaning, data standardization and feature engineering to obtain a water quality detection optimized data set;

[0115] The water quality detection optimization data set specifically includes optimizing water quality sensor data, optimizing sampling frequency data, optimizing environmental data set and characteristic data set.

[0116] Example 4, see Figure 1 , Figure 2 and Figure 4This embodiment is based on the above embodiment. In step S3, the water quality pre-classification is used to pre-classify the short-term status of water quality. Specifically, based on the optimized water quality sensor data and feature data set in the water quality detection optimization data set, a deep neural network method combining a genetic algorithm and a random forest model is used to pre-classify water quality to obtain water quality short-term classification result data;

[0117] The deep neural network method combining genetic algorithm and random forest model specifically includes a deep neural subnet, a genetic optimization operator and a random forest ensemble model;

[0118] The deep neural subnetwork is used to process water quality classification features and perform preliminary classification;

[0119] The genetic optimization operator is used to optimize the model parameters of the deep neural subnet;

[0120] The random forest ensemble model is used to integrate the results of the deep neural subnet to improve the classification accuracy;

[0121] The step of pre-classifying water quality by using a deep neural network method combining a genetic algorithm and a random forest model to obtain water quality short-term classification result data includes:

[0122] Step S31: constructing a deep neural subnet, specifically constructing a standard deep neural subnet including an input layer, a hidden layer and an output layer, and using a loss function to train the subnet;

[0123] The hidden layer specifically includes a first hidden layer, a second hidden layer and a third hidden layer, the number of neurons in the first hidden layer is 128, the number of neurons in the second hidden layer is 64, and the number of neurons in the third hidden layer is 32;

[0124] The loss function specifically adopts the minimum mean square error loss function;

[0125] Step S32: construct a genetic optimization operator, specifically, perform iterative model parameter combination calculation through population initialization and fitness evaluation to obtain the optimal deep neural subnet model parameter combination. The calculation formula of the fitness evaluation is:

[0126] ;

[0127] Where F is the fitness function, N is the total number of samples, and n is the sample index. is the predicted classification result obtained by the initial classification of the deep neural subnet, is the true classification result of the sample;

[0128] Step S33: constructing a random forest ensemble model, specifically, combining a voting classification mechanism to perform random forest decision tree ensemble learning, and performing fine classification based on the preliminary classification results of the deep neural subnet using the optimal deep neural subnet model parameter combination to obtain a fine classification output of the water quality classification results;

[0129] The calculation formula for random forest decision tree ensemble learning combined with the voting classification mechanism is:

[0130] ;

[0131] In the formula, is the fine classification output of water quality classification results, M V (·) is the voting classification mechanism operation function, T1(X) is the prediction result of the first decision tree, X is the random forest input data, the preliminary classification result of the deep neural subnet using the optimal deep neural subnet model parameter combination, and m is the decision tree index;

[0132] Step S34: water quality pre-classification model training, specifically, training the water quality pre-classification model through the deep neural subnet, the genetic optimization operator and the random forest ensemble model to obtain the water quality pre-classification model Model WC ;

[0133] Step S35: water quality pre-classification, specifically using the water quality pre-classification model Model WC , performing water quality pre-classification according to the optimized water quality sensor data and feature data set in the water quality detection optimization data set to obtain water quality short-term classification result data;

[0134] The water quality short-term classification result data specifically refers to the water quality pre-classification level prediction data, including excellent water quality, good water quality, slightly polluted water quality, moderately polluted water quality and severely polluted water quality.

[0135] By executing the above operations, in view of the technical problem that in the existing water quality detection methods, there are multi-stage detection tasks that will inevitably reduce the calculation efficiency, and therefore the calculation speed and real-time performance of the pre-task must be optimized as much as possible, this solution creatively adopts a deep neural network method combining genetic algorithm and random forest model to pre-classify water quality. Through the combined integration of deep neural network, search algorithm and machine learning model, it can not only improve the classification accuracy, but also improve the efficiency of the overall task.

[0136] Example 5, see Figure 1 , Figure 2 and Figure 5This embodiment is based on the above embodiment. In step S4, the water quality trend analysis is used to classify the trend of water quality changes from the long-term trend. Specifically, based on the optimized water quality sensor data, optimized sampling frequency data and feature data set in the water quality detection optimization data set, combined with the water quality short-term classification result data, a long short-term memory neural network combined with cubic chaos weight optimization is used to perform water quality trend analysis to obtain water quality long-term trend analysis result data;

[0137] The long short-term memory neural network combined with cubic chaos weight optimization specifically includes a standard long short-term memory neural network and a cubic chaos weight optimization search operator;

[0138] The standard long short-term memory neural network is used to process time series data and features in water quality trend analysis;

[0139] The cubic chaos weight optimization search operator is used to improve the performance of the long short-term memory neural network for trend analysis through the cubic chaos weight optimization search algorithm;

[0140] The step of using a long short-term memory neural network combined with cubic chaos weight optimization to perform water quality trend analysis and obtain water quality long-term trend analysis result data includes:

[0141] Step S41: constructing a standard long short-term memory neural network, specifically constructing a standard long short-term memory neural network including an input gate, a forget gate and an output gate as a basic prediction network of the model;

[0142] Step S42: constructing a cubic chaos random optimization operator, specifically constructing a random chaotic dynamic system, and generating a cubic chaos random state vector through the random chaotic dynamic system. The cubic chaos random state vector is used to replace the weight of the standard long short-term memory neural network, and optimize the search operator to obtain a cubic chaos optimized long short-term neural network. The calculation formula for constructing the random chaotic dynamic system is:

[0143] ;

[0144] In the formula, x t+1 is a cubic chaotic random state vector used to generate weights to replace the standard long short-term memory neural network and to generate search path optimization search operators, f 3 (·) is the chaos equation function used to represent the three chaotic mapping operations, x t is the chaotic equation input, t is the time index;

[0145] Step S43: constructing a search operator, specifically constructing a bone crest chicken search algorithm, calculating the optimal solution of the model parameters of the standard long short-term memory neural network, and obtaining the initial search operator parameter combination output. The calculation formula is:

[0146] ;

[0147] Where, X new is the initial search operator parameter combination output, X current is the current parameter combination, r1 is the first random weight coefficient, X best is the optimal solution of the current parameter combination, r2 is the second random weight coefficient, X random is a random parameter combination;

[0148] Step S44: cubic chaos random optimization search algorithm optimization, specifically executing the random chaos dynamic system and performing cubic chaos initialization, regenerating the first random weight coefficient and the second random weight coefficient in the search operator by obtaining the cubic chaos random state vector, and obtaining the cubic chaos random optimization parameter combination of the cubic chaos optimization long-term and short-term neural network through iterative training;

[0149] Step S45: training the water quality trend analysis model, specifically, optimizing the standard long short-term memory neural network through the cubic chaos random optimization search algorithm to obtain the cubic chaos optimized long short-term neural network, and training the water quality trend analysis model according to the cubic chaos random optimization parameter combination to obtain the water quality trend analysis model Model FR ;

[0150] Step S46: water quality trend analysis, specifically using the water quality trend analysis model Model FR , based on the optimized water quality sensor data, optimized sampling frequency data and feature data set in the water quality detection optimization data set, and combined with the data of the water quality short-term classification result data, the pre-classification results of which are lightly polluted water quality, moderately polluted water quality and heavily polluted water quality, a water quality trend analysis is performed to obtain the water quality long-term trend analysis result data;

[0151] The water quality long-term trend analysis result data specifically includes water quality pollution degree change trend prediction data.

[0152] By executing the above operations, in order to address the technical problem that in the existing water quality trend analysis, the traditional time series data prediction model has insufficient understanding of water body composition and water quality changes in the process of water quality trend analysis, which leads to insufficient prediction performance and prediction efficiency, this solution creatively adopts a long short-term memory neural network combined with cubic chaos weight optimization. The chaotic cubic algorithm can not only optimize the weight update of the model itself, but also update the content of the optimization algorithm used in the model parameter update process, which improves the execution performance of the phased tasks while reducing the implementation cost of the method and system.

[0153] Example 6, see Figure 1 , Figure 2 and Figure 6 This embodiment is based on the above embodiment. In step S5, the pollution source prediction is used to predict potential pollution sources with short-term and long-term impacts on water quality in combination with environmental data. Specifically, based on the optimized water quality sensor data and feature data set in the water quality detection optimization data set, combined with the water quality short-term classification result data, a graph neural network combined with causal inspiration is used to perform pollution source prediction to obtain pollution source prediction result data;

[0154] The graph neural network combined with causal inspiration specifically includes a graph variational autoencoder, a causal inspiration operator, a linear decoder and a basic classifier;

[0155] The graph variational autoencoder is used to construct an improved graph neural network and perform feature encoding;

[0156] The causal heuristic operator is used to set the causal effect estimator to enhance the potential feature mining capability of pollution source tracing;

[0157] The linear decoder is used to perform linear decoding;

[0158] The basic classifier is used to perform prediction and classification according to the decoding result to obtain the pollution college prediction result;

[0159] The step of using a graph neural network combined with causal inspiration to perform pollution source tracing prediction and obtain pollution source tracing prediction result data includes:

[0160] Step S51: constructing a graph variational autoencoder, specifically, reconstructing the graph adjacency matrix and node features of the standard graph convolutional neural network through a multilayer perceptron and a linear inner product decoder, and improving the graph convolutional neural network by constructing a graph variational autoencoder objective function to obtain the graph variational autoencoder;

[0161] The calculation formula for reconstructing the graph adjacency matrix of the standard graph convolutional neural network through the multi-layer perceptron and the linear inner product decoder is:

[0162] ;

[0163] In the formula, Z k is the output of the k-th layer of the graph convolutional neural network, Z is the total output of the graph convolutional neural network, sig(·) is the S-type activation function, is the normalized graph adjacency matrix, Z k-1 is the output of the k-1th layer of graph convolutional neural network, W k-1 is the output of the k-1th layer of graph convolutional neural network;

[0164] A c is the graph adjacency matrix reconstructed by linear inner product decoding, Z T is the transposed matrix of the total output of the graph convolutional neural network, X c It is the node feature after being processed by the multi-layer perceptron;

[0165] The calculation formula of the variational auto-encoding objective function is:

[0166] GraV=E[||XX c || F ]+E[||AA c || F ]-E[D KL [q(Z|A,X)||p(Z)]];

[0167] Where GraV is the variational auto-encoding objective function, which is used to obtain the maximum optimization model, E[·] is the expected function, X is the original node feature of the graph convolutional neural network, and X c is the node feature after being processed by the multi-layer perceptron, ||·|| F is the Frobenius norm operator, A is the original graph adjacency matrix of the graph convolutional neural network, A c is the graph adjacency matrix reconstructed by linear inner product decoding, D KL [·||·] is the KL divergence calculation function, which is used to measure the similarity of two probability distributions. q(Z|A,X) is the posterior distribution, which is used to represent the probability distribution of the total output Z of the graph convolutional neural network with respect to the original graph adjacency matrix A and the original node feature X. p(Z) is the prior distribution, which is used to represent the Gaussian distribution of the total output Z of the graph convolutional neural network.

[0168] Step S52: constructing a causal heuristic operator, specifically, by constructing a causal effect estimator objective function, calculating the characteristic relationship between the pollution source causal characteristics and the pollution source non-causal characteristics, and by constructing the causal effect estimator objective function, and by integrating the variational auto-encoding objective function and the causal effect estimator objective function into a loss function, the potential characteristics of pollution source tracing are enhanced to obtain causal heuristic feature data;

[0169] The calculation formula of the objective function of the causal effect estimator is:

[0170] ;

[0171] Wherein, ReaV is the objective function of the causal effect estimation, max is the maximum value function, Lo(a,Y) is the causal impact measurement function, where a is the pollution source causal feature, Y is the pollution source feature set, which is used to represent the feature data set in the water quality detection optimization data set, I(a;b) is the mutual information calculation function, b is the pollution source non-causal feature, and the pollution source causal feature a and the pollution source non-causal feature b are specifically obtained through manual labeling;

[0172] The calculation formula for integrating the variational autoencoding objective function and the causal effect estimator objective function into the loss function is:

[0173] ;

[0174] Where L is the model loss function, GraV is the variational autoencoding objective function, and ReaV is the causal effect estimator objective function. is a trainable hyperparameter;

[0175] Step S53: construct a linear decoder, specifically, construct a linear inner product decoder, and linearly decode the causal heuristic graph convolution output feature to obtain decoded feature data, and the calculation formula is:

[0176] Y decode =sig(Z GR Z GR T );

[0177] Where Y decode is the decoded feature data, Z GR is the causal heuristic feature data, Z GR T is the transpose of the causal-inspired feature data;

[0178] Step S54: constructing a basic classifier, specifically by constructing a standard random forest model as a basic classifier, and obtaining a pollution source tracing classification prediction result according to the decoded feature data;

[0179] Step S55: Training the pollution source tracing prediction model, specifically, training the pollution source tracing prediction model through the graph variational autoencoder, the causal heuristic operator, the linear decoder and the basic classifier, combined with the model loss function, to obtain the pollution source tracing prediction model Model WP ;

[0180] Step S56: Pollution source prediction, specifically using the pollution source prediction model Model WP, based on the optimized water quality sensor data and feature data set in the water quality detection optimization data set, and combined with the water quality short-term classification result data, the pre-classification results are lightly polluted water quality, moderately polluted water quality and heavily polluted water quality, to perform pollution source tracing prediction and obtain pollution source tracing prediction result data;

[0181] The pollution source tracing prediction result data specifically includes pollution source classification results, pollution level classification results and pollution time factor classification results.

[0182] In step S6, the water quality detection is used to perform comprehensive water quality detection in combination with short-term water quality classification, long-term trend analysis and pollution source tracing prediction, specifically, by combining the water quality short-term classification result data, the water quality long-term trend analysis result data and the pollution source tracing prediction result data, a comprehensive water quality detection reference report is obtained;

[0183] The comprehensive water quality testing reference report specifically includes water quality grade assessment classification data, key water quality parameter reference data, water quality historical trend comparison data, water quality expected trend data and pollution source analysis data.

[0184] By executing the above operations, in view of the technical problem that in the existing pollution source tracing prediction methods, pollution source tracing involves complex causal inference logic, and in complex water quality data, extracting this part of causal features has great technical and design difficulties, this solution creatively adopts the water quality pre-classification task to first pre-classify the data to exclude unpolluted water sources, and then combines the causal-inspired graph neural network. Through the causal-inspired operator and the improved graph neural network, the calculation speed and effect of pollution source tracing are improved, and it also provides a better practical reference and data support for the depth of water quality detection tasks.

[0185] Embodiment 7, see Figure 1 and Figure 2 , this embodiment is based on the above embodiment, and the present invention provides an artificial intelligence-based water quality detection system, including a data collection and processing module, a water quality pre-classification module, a water quality trend analysis module, a pollution source tracing prediction module and a water quality detection module;

[0186] The data collection and processing module is used for data collection and data preprocessing. Through data collection and data preprocessing, a water quality detection optimization data set is obtained, and the water quality detection optimization data set is sent to the water quality pre-classification module, the water quality trend analysis module and the pollution source tracing prediction module;

[0187] The water quality pre-classification module is used for water quality pre-classification, obtains water quality short-term classification result data through water quality pre-classification, and sends the water quality short-term classification result data to the water quality trend analysis module, the pollution source tracing prediction module and the water quality detection module;

[0188] The water quality trend analysis module is used for water quality trend analysis, obtains water quality long-term trend analysis result data through water quality trend analysis, and sends the water quality long-term trend analysis result data to the water quality detection module;

[0189] The pollution source tracing prediction module is used for pollution source tracing prediction, obtains pollution source tracing prediction result data through pollution source tracing prediction, and sends the pollution source tracing prediction result data to the water quality detection module;

[0190] The water quality detection module is used for water quality detection, and a comprehensive water quality detection reference report is obtained through water quality detection.

[0191] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0192] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0193] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A water quality detection method based on artificial intelligence, characterized in that: The method comprises the following steps: Step S1: Data collection, obtaining the original data set of water quality detection; Step S2: data preprocessing to obtain optimized data set for water quality detection; Step S3: pre-classify water quality by using a deep neural network method combining a genetic algorithm and a random forest model to pre-classify water quality and obtain short-term classification result data of water quality; Step S4: water quality trend analysis, using a long short-term memory neural network combined with cubic chaos weight optimization to perform water quality trend analysis and obtain water quality long-term trend analysis result data; The step of using a long short-term memory neural network combined with cubic chaos weight optimization to perform water quality trend analysis and obtain water quality long-term trend analysis result data includes: Step S41: constructing a standard long short-term memory neural network, specifically constructing a standard long short-term memory neural network including an input gate, a forget gate and an output gate as a basic prediction network of the model; Step S42: constructing a cubic chaos random optimization operator, specifically constructing a random chaos dynamic system, and generating a cubic chaos random state vector through the random chaos dynamic system, wherein the cubic chaos random state vector is used to replace the weight of the standard long short-term memory neural network, and optimize the search operator to obtain a cubic chaos optimized long short-term neural network; Step S43: constructing a search operator, specifically constructing a bone crest chicken search algorithm, calculating the optimal solution of the model parameters of the standard long short-term memory neural network, and obtaining an initial search operator parameter combination output; Step S44: cubic chaos random optimization search algorithm optimization, specifically executing the random chaos dynamic system and performing cubic chaos initialization, regenerating the first random weight coefficient and the second random weight coefficient in the search operator by obtaining the cubic chaos random state vector, and obtaining the cubic chaos random optimization parameter combination of the cubic chaos optimization long-term and short-term neural network through iterative training; Step S45: training the water quality trend analysis model, specifically, optimizing the standard long short-term memory neural network through the cubic chaos random optimization search algorithm to obtain the cubic chaos optimized long short-term neural network, and training the water quality trend analysis model according to the cubic chaos random optimization parameter combination to obtain the water quality trend analysis model Model FR ; Step S46: water quality trend analysis, specifically using the water quality trend analysis model Model FR , based on the optimized water quality sensor data, optimized sampling frequency data and feature data set in the water quality detection optimization data set, and combined with the data of the water quality short-term classification result data, the pre-classification results of which are lightly polluted water quality, moderately polluted water quality and heavily polluted water quality, a water quality trend analysis is performed to obtain the water quality long-term trend analysis result data; The water quality long-term trend analysis result data, specifically the water quality pollution degree change trend forecast data; Step S5: Pollution source prediction, using a graph neural network combined with causal inspiration to perform pollution source prediction and obtain pollution source prediction result data; The step of using a graph neural network combined with causal inspiration to perform pollution source tracing prediction and obtain pollution source tracing prediction result data includes: Step S51: constructing a graph variational autoencoder, specifically, reconstructing the graph adjacency matrix and node features of the standard graph convolutional neural network through a multilayer perceptron and a linear inner product decoder, and improving the graph convolutional neural network by constructing a graph variational autoencoder objective function to obtain the graph variational autoencoder; Step S52: constructing a causal heuristic operator, specifically, by constructing a causal effect estimator objective function, calculating the characteristic relationship between the pollution source causal characteristics and the pollution source non-causal characteristics, and by constructing the causal effect estimator objective function, and by integrating the variational auto-encoding objective function and the causal effect estimator objective function into a loss function, the potential characteristics of pollution source tracing are enhanced to obtain causal heuristic feature data; Step S53: constructing a linear decoder, specifically constructing a linear inner product decoder, and linearly decoding the causal heuristic graph convolution output features to obtain decoded feature data; Step S54: constructing a basic classifier, specifically by constructing a standard random forest model as a basic classifier, and obtaining a pollution source tracing classification prediction result according to the decoded feature data; Step S55: Training the pollution source tracing prediction model, specifically, training the pollution source tracing prediction model through the graph variational autoencoder, the causal heuristic operator, the linear decoder and the basic classifier, combined with the model loss function, to obtain the pollution source tracing prediction model Model WP ; Step S56: Pollution source prediction, specifically using the pollution source prediction model Model WP , based on the optimized water quality sensor data and feature data set in the water quality detection optimization data set, and combined with the water quality short-term classification result data, the pre-classification results are lightly polluted water quality, moderately polluted water quality and heavily polluted water quality, to perform pollution source tracing prediction and obtain pollution source tracing prediction result data; The pollution source tracing prediction result data specifically includes the pollution source classification results, pollution level classification results and pollution time factor classification results; Step S6: water quality testing, by combining the water quality short-term classification result data, the water quality long-term trend analysis result data and the pollution source tracing prediction result data, a comprehensive water quality testing is performed to obtain a comprehensive water quality testing reference report.

2. The water quality detection method based on artificial intelligence according to claim 1 is characterized in that: In step S1, the data collection is used for the original data required for water quality detection, specifically, by collecting multi-point data at different frequencies through water quality sensors and environmental sensors to obtain the original data set for water quality detection; In step S2, the data preprocessing is used to optimize the original data, specifically, data cleaning, data standardization and feature engineering are performed on the original water quality detection data set to obtain an optimized water quality detection data set; The water quality detection optimization data set specifically includes optimizing water quality sensor data, optimizing sampling frequency data, optimizing environmental data set and characteristic data set.

3. The water quality detection method based on artificial intelligence according to claim 2 is characterized in that: In step S3, the water quality pre-classification is used to pre-classify the short-term status of water quality, specifically, based on the optimized water quality sensor data and feature data set in the water quality detection optimization data set, a deep neural network method combining a genetic algorithm and a random forest model is used to pre-classify water quality to obtain water quality short-term classification result data; The deep neural network method combining genetic algorithm and random forest model specifically includes a deep neural subnet, a genetic optimization operator and a random forest ensemble model; The deep neural subnetwork is used to process water quality classification features and perform preliminary classification; The genetic optimization operator is used to optimize the model parameters of the deep neural subnet; The random forest ensemble model is used to integrate the results of the deep neural subnet to improve the classification accuracy; The step of pre-classifying water quality by using a deep neural network method combining a genetic algorithm and a random forest model to obtain water quality short-term classification result data includes: Step S31: constructing a deep neural subnet, specifically constructing a standard deep neural subnet including an input layer, a hidden layer and an output layer, and using a loss function to train the subnet; The loss function specifically adopts the minimum mean square error loss function; Step S32: constructing a genetic optimization operator, specifically, performing iterative model parameter combination calculation through population initialization and fitness evaluation to obtain the optimal deep neural subnet model parameter combination; Step S33: constructing a random forest ensemble model, specifically, combining a voting classification mechanism to perform random forest decision tree ensemble learning, and performing fine classification based on the preliminary classification results of the deep neural subnet using the optimal deep neural subnet model parameter combination to obtain a fine classification output of the water quality classification results; Step S34: water quality pre-classification model training, specifically, training the water quality pre-classification model through the deep neural subnet, the genetic optimization operator and the random forest ensemble model to obtain the water quality pre-classification model Model WC ; Step S35: water quality pre-classification, specifically using the water quality pre-classification model Model WC , performing water quality pre-classification according to the optimized water quality sensor data and feature data set in the water quality detection optimization data set to obtain water quality short-term classification result data; The water quality short-term classification result data specifically refers to the water quality pre-classification level prediction data, including excellent water quality, good water quality, slightly polluted water quality, moderately polluted water quality and severely polluted water quality.

4. The water quality detection method based on artificial intelligence according to claim 3 is characterized in that: In step S4, the water quality trend analysis is used to classify the trend of water quality changes from the long-term trend, specifically based on the optimized water quality sensor data, optimized sampling frequency data and feature data set in the water quality detection optimization data set, combined with the water quality short-term classification result data, using the long short-term memory neural network combined with cubic chaos weight optimization to perform water quality trend analysis to obtain water quality long-term trend analysis result data; The long short-term memory neural network combined with cubic chaos weight optimization specifically includes a standard long short-term memory neural network and a cubic chaos weight optimization search operator; The standard long short-term memory neural network is used to process time series data and features in water quality trend analysis; The cubic chaos weight optimization search operator is used to improve the performance of the long short-term memory neural network for trend analysis through the cubic chaos weight optimization search algorithm.

5. The water quality detection method based on artificial intelligence according to claim 4 is characterized in that: In step S5, the pollution source prediction is used to predict potential pollution sources with short-term and long-term impacts on water quality in combination with environmental data. Specifically, based on the optimized water quality sensor data and feature data set in the water quality detection optimization data set, combined with the water quality short-term classification result data, a graph neural network combined with causal inspiration is used to perform pollution source prediction to obtain pollution source prediction result data; The graph neural network combined with causal inspiration specifically includes a graph variational autoencoder, a causal inspiration operator, a linear decoder and a basic classifier; The graph variational autoencoder is used to construct an improved graph neural network and perform feature encoding; The causal heuristic operator is used to set the causal effect estimator to enhance the potential feature mining capability of pollution source tracing; The linear decoder is used to perform linear decoding; The basic classifier is used to perform prediction and classification according to the decoding results to obtain the pollution academy prediction results.

6. The water quality detection method based on artificial intelligence according to claim 5 is characterized in that: In step S6, the water quality detection is used to perform comprehensive water quality detection in combination with short-term water quality classification, long-term trend analysis and pollution source tracing prediction, specifically, by combining the water quality short-term classification result data, the water quality long-term trend analysis result data and the pollution source tracing prediction result data, a comprehensive water quality detection reference report is obtained; The comprehensive water quality testing reference report specifically includes water quality grade assessment classification data, key water quality parameter reference data, water quality historical trend comparison data, water quality expected trend data and pollution source analysis data.

7. An artificial intelligence-based water quality detection system, used to implement an artificial intelligence-based water quality detection method as claimed in any one of claims 1 to 6, characterized in that: It includes data collection and processing module, water quality pre-classification module, water quality trend analysis module, pollution source tracing and prediction module and water quality detection module.

8. The water quality detection system based on artificial intelligence according to claim 7 is characterized in that: The data collection and processing module is used for data collection and data preprocessing. Through data collection and data preprocessing, a water quality detection optimization data set is obtained, and the water quality detection optimization data set is sent to the water quality pre-classification module, the water quality trend analysis module and the pollution source tracing prediction module; The water quality pre-classification module is used for water quality pre-classification, obtains water quality short-term classification result data through water quality pre-classification, and sends the water quality short-term classification result data to the water quality trend analysis module, the pollution source tracing prediction module and the water quality detection module; The water quality trend analysis module is used for water quality trend analysis, obtains water quality long-term trend analysis result data through water quality trend analysis, and sends the water quality long-term trend analysis result data to the water quality detection module; The pollution source tracing prediction module is used for pollution source tracing prediction, obtains pollution source tracing prediction result data through pollution source tracing prediction, and sends the pollution source tracing prediction result data to the water quality detection module; The water quality detection module is used for water quality detection, and a comprehensive water quality detection reference report is obtained through water quality detection.

Images