Water quality health evaluation method and system based on big data

Abstract

The invention relates to the technical field of water quality detection, in particular to a water quality health evaluation method and system based on big data. The method comprises the following steps: acquiring multiple basic water quality data and water body images, preprocessing the basic water quality data and the water body images to obtain preprocessed data and preprocessed images, creating an evaluation model, inputting the preprocessed data and the preprocessed images into the evaluation model to train the evaluation model, and obtaining a trained evaluation model. Then collecting real-time water quality data, inputting the real-time water quality data into the trained evaluation model to obtain a real-time water quality score, finally screening out an observation sample based on the real-time water quality score, predicting a water quality evolution trend of the observation sample in combination with space-time diagram convolution, and screening out a water quality early warning area in combination with the water quality evolution trend. According to the method, the dynamic weight model and the space-time diagram convolution prediction algorithm are combined, multi-dimensional accurate prediction of the water quality of the water body is achieved, unqualified water bodies are screened out in time according to the prediction result, and early warning is given out.

Description

Technical Field

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

[0002] Water quality is the abbreviation of water quality. It indicates the physical (such as color, turbidity, odor, etc.), chemical (inorganic and organic content) and biological (bacteria, microorganisms, plankton, benthic organisms) characteristics of water and its composition. Water quality is used to evaluate the quality of water, and a series of water quality parameters and water quality standards are specified.

[0003] A Chinese patent with publication number CN109858755A discloses a method for evaluating water quality. The method includes performing a distribution test on the single-factor pollution index of each participating indicator to obtain the probability distribution type of each participating indicator; constructing a comprehensive pollution index function, using Monte Carlo sampling simulation to obtain multiple comprehensive pollution index values; constructing a cumulative frequency distribution diagram, and comparing the water quality with the highest probability as the evaluation result. However, the prior art does not predict the future state of the water body, which makes it difficult to detect abnormal water bodies in a timely manner, which is easy to cause damage to the ecological environment. Summary of the invention

[0004] The purpose of the present invention is to address the problems existing in the background technology and to propose a water quality health assessment method and system based on big data.

[0005] The technical solution of the present invention:

[0006] On the one hand, the present application provides a water quality health assessment method based on big data, comprising:

[0007] Collect multiple basic water quality data and water body images, pre-process the basic water quality data and water body images to obtain pre-processed data and pre-processed images;

[0008] Creating an evaluation model, inputting the preprocessed data and the preprocessed image into the evaluation model to train the evaluation model, and obtaining a trained evaluation model;

[0009] Collect real-time water quality data, input the real-time water quality data into the trained evaluation model, and obtain the real-time water quality score;

[0010] Based on the real-time water quality score, the observation samples are screened out, and the water quality evolution trend is predicted by combining the observation samples with the spatiotemporal graph convolution. The water quality warning area is screened based on the water quality evolution trend.

[0011] Preferably, a plurality of basic water quality data and water body images are collected, and the basic water quality data and water body images are preprocessed to obtain preprocessed data and preprocessed images, including:

[0012] Create a water quality data table;

[0013] Set collection parameters, collect multiple basic water quality data based on the collection parameters, and put all the collected basic water quality data into the water quality data table; the basic water quality data includes water body pH value, conductivity, and dissolved oxygen;

[0014] Collect multiple water body images based on the collection parameters, and put all the collected water body images into the water quality data table;

[0015] Normalize each basic water quality data in turn to obtain preprocessed data;

[0016] Filter each water body image in turn to obtain multiple preprocessed images.

[0017] Preferably, create an evaluation model, input the preprocessed data and preprocessed images into the evaluation model to train the evaluation model, and obtain the trained evaluation model, including:

[0018] Create an evaluation model;

[0019] Create a coupling relationship between preprocessed images - preprocessed data according to the corresponding relationship of the collection frequency, and record the preprocessed images, preprocessed data, and the coupling relationship between preprocessed images - preprocessed data as a training sample to obtain multiple training samples;

[0020] Divide all the training samples into a training set and a test set according to a random ratio;

[0021] Input the training set into the evaluation model, so that the evaluation model continuously learns the relationship between water body images and water quality data, and obtain the trained evaluation model;

[0022] Input the test set into the trained evaluation model to verify whether the trained evaluation model is trained successfully.

[0023] Preferably, input the training set into the evaluation model, so that the evaluation model continuously learns the relationship between water body images and water quality data, and obtain the trained evaluation model, including:

[0024] Randomly select a preprocessed data from the training set;

[0025] Calculate the initial water quality score corresponding to the preprocessed data based on the preprocessed data through formula 1;

[0026]

[0027] where, F i is the initial water quality score of the i-th preprocessed data, X ij is the i-th preprocessed data matrix, n is the total number of indicators included in the preprocessed data, Wj is the weight of the j-th data type;

[0028] Return a preprocessed data randomly selected from the training set until all the preprocessed data in the training set are selected, and obtain the initial water quality score corresponding to each preprocessed data.

[0029] Preferably, observation samples are screened based on the real-time water quality score, and the water quality evolution trend is predicted for the observation samples by combining spatio-temporal graph convolution. The water quality warning area is screened in combination with the water quality evolution trend, including:

[0030] Calculate the evaluation trend of each real-time water quality score through the TOPSIS algorithm;

[0031] Sort all the real-time water quality scores from small to large according to the evaluation trend, and screen out the real-time water quality scores in the last K; record the real-time water quality data corresponding to the selected K real-time water quality scores as observation samples;

[0032] Predict the water quality evolution trend of the observation samples through spatio-temporal graph convolution to obtain the water quality prediction results of the observation samples;

[0033] Judge whether the water quality prediction results of the observation samples are qualified;

[0034] If the water quality prediction results of the observation samples are unqualified, record the observation samples as unqualified water quality.

[0035] Preferably, calculating the evaluation trend of each real-time water quality score through the TOPSIS algorithm includes:

[0036] Standardize the preprocessing data matrix of the real-time water quality data through formula 2;

[0037]

[0038] where Z ij is the i-th standardized matrix of the real-time water quality data, X ij is the i-th preprocessing data matrix, n is the total number of indicators included in the preprocessing data, and m is the total number of samples in the preprocessing data;

[0039] Perform dimensionless processing on the positive and negative indicators respectively;

[0040] Determine the optimal solution and the worst solution;

[0041] Calculate the positive distance between the standardized matrix and the optimal solution and the negative distance between the standardized matrix and the worst solution respectively.

[0042] Preferably, calculating the evaluation trend of each real-time water quality score through the TOPSIS algorithm also includes:

[0043] Calculate the evaluation trend of the real-time water quality score through Formula 3 by combining the positive distance and the negative distance;

[0044]

[0045] Among them, Si is the evaluation trend of the real-time water quality score, is the positive distance between the standardized matrix and the optimal solution, is the negative distance between the standardized matrix and the worst solution;

[0046] Sort all the evaluation trends of the real-time water quality scores in descending order, and select the last K real-time water quality scores.

[0047] Preferably, predict the water quality evolution trend of the observation sample through spatio-temporal graph convolution to obtain the water quality prediction result of the observation sample, including:

[0048] Divide the water body into multiple structural nodes according to the water body image corresponding to the observation sample, and represent the spatial position relationship between the structural nodes through an adjacency matrix;

[0049] Perform time convolution on the observation samples according to the chronological order of the collection time to obtain the time series dependence relationship;

[0050] Combine the adjacency matrix and the time series dependence relationship for multi-scale spatio-temporal feature joint modeling, and output the water quality prediction image and water quality prediction data of the observation sample at future time steps through the output layer.

[0051] Preferably, if the water quality prediction result of the observation sample is unqualified, then mark the observation sample as unqualified water quality, including:

[0052] Input the water quality prediction image and water quality prediction data into the trained evaluation model, and obtain the predicted water quality score output by the trained evaluation model;

[0053] Judge whether the water quality prediction result is qualified based on the predicted water quality score;

[0054] If the water quality prediction result is unqualified, then issue a warning.

[0055] On the other hand, the present application also provides a water quality health evaluation system based on big data, including a collection component and a processing component. The water quality data and water body images are collected through the collection component, and the water quality health evaluation method based on big data described in any one of the foregoing is executed through the processing component. The data collected by the collection component is preprocessed through the processing component, and whether the water quality is qualified is judged in combination with the preprocessed data.

[0056] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0057] By collecting a number of basic water quality data and water body images, preprocessing the basic water quality data and water body images to obtain preprocessed data and preprocessed images, then creating an evaluation model, inputting the preprocessed data and preprocessed images into the evaluation model to train the evaluation model, obtaining the trained evaluation model, then collecting real-time water quality data, inputting the real-time water quality data into the trained evaluation model to obtain the real-time water quality score, and finally screening out observation samples based on the real-time water quality score, and predicting the water quality evolution trend for the observation samples by combining spatio-temporal graph convolution, screening out water quality warning areas by combining the water quality evolution trend. This application combines a dynamic weight model with a spatio-temporal graph convolution prediction algorithm to achieve accurate multi-dimensional prediction of water body water quality, and timely screens out unqualified water bodies based on the prediction results and issues warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a schematic flowchart of a water quality health evaluation method based on big data proposed by the present invention;

[0059] Figure 2 It is a schematic structural diagram of a water quality health evaluation system based on big data proposed by the present invention;

[0060] Reference numerals: 100, acquisition component; 200, processing component. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] Example 1, as Figure 1 shown, a water quality health evaluation method based on big data proposed by the present invention includes:

[0062] S100, collecting a number of basic water quality data and water body images, preprocessing the basic water quality data and water body images to obtain preprocessed data and preprocessed images;

[0063] S200, creating an evaluation model, inputting the preprocessed data and preprocessed images into the evaluation model to train the evaluation model, obtaining the trained evaluation model;

[0064] S300, collecting real-time water quality data, inputting the real-time water quality data into the trained evaluation model to obtain the real-time water quality score;

[0065] S400, screening out observation samples based on the real-time water quality score, and predicting the water quality evolution trend for the observation samples by combining spatio-temporal graph convolution, screening out water quality warning areas by combining the water quality evolution trend;

[0066] Specifically, according to the water environment function and protection objectives of surface water, it is divided into five categories in descending order of function, namely Class I to Class V, where Class I is the source water and Class V is agricultural water.

[0067] In the present invention, by collecting a plurality of basic water quality data and water body images, preprocessing the basic water quality data and water body images to obtain preprocessed data and preprocessed images, then creating an evaluation model, inputting the preprocessed data and preprocessed images into the evaluation model to train the evaluation model to obtain a trained evaluation model, then collecting real-time water quality data, inputting the real-time water quality data into the trained evaluation model to obtain a real-time water quality score, and finally screening out observation samples based on the real-time water quality score, and predicting the water quality evolution trend for the observation samples in combination with a spatio-temporal graph convolution, and screening out water quality early warning areas in combination with the water quality evolution trend. The present application realizes precise prediction of the multi-dimensions of water body water quality by combining a dynamic weight model and a spatio-temporal graph convolution prediction algorithm, and timely screens out unqualified water bodies based on the prediction results and issues early warnings.

[0068] In an alternative embodiment, the 100 includes:

[0069] S110, create a water quality data table;

[0070] S120, set collection parameters, collect a plurality of basic water quality data based on the collection parameters, and put all the collected basic water quality data into the water quality data table; the basic water quality data includes the water body pH value, conductivity, and dissolved oxygen;

[0071] S130, collect a plurality of water body images based on the collection parameters, and put all the collected water body images into the water quality data table;

[0072] S140, perform normalization processing on each basic water quality data in turn to obtain preprocessed data;

[0073] S150, perform filtering processing on each water body image in turn to obtain a plurality of preprocessed images;

[0074] Specifically, the water body image can be preprocessed by Gaussian filtering to smooth the water body image.

[0075] It should be noted that when setting the collection parameters, it is necessary to set the collection frequency and collection period, so as to collect a plurality of basic water quality data according to the collection frequency within the collection period, and when collecting water body images, the same collection parameters also need to be maintained to ensure that the water body images and the basic water quality data can correspond one by one.

[0076] In an alternative embodiment, the S200 includes:

[0077] S210, create an evaluation model;

[0078] S220. Create a coupling relationship between the preprocessed image and the preprocessed data according to the corresponding relationship of the acquisition frequency, and record the preprocessed image, the preprocessed data, and the coupling relationship between the preprocessed image and the preprocessed data as a training sample to obtain multiple training samples;

[0079] S230. Divide all the training samples into a training set and a test set according to a random ratio;

[0080] S240. Input the training set into the evaluation model, enabling the evaluation model to continuously learn the relationship between the water body image and the water quality data to obtain a trained evaluation model;

[0081] S250. Input the test set into the trained evaluation model to verify whether the trained evaluation model is trained successfully.

[0082] It should be noted that by sequentially inputting multiple training samples into the evaluation model, a trained evaluation model is obtained after multiple trainings. Subsequently, it is necessary to verify whether the trained evaluation model is trained successfully. During the verification, the accuracy rate can be used as the judgment criterion. When the accuracy rate of the water quality score output by the trained evaluation model reaches the standard, it can be determined that the trained evaluation model has been trained successfully. The trained evaluation model has the ability to automatically output the initial water quality score corresponding to the water body based on the input water body image and water quality data.

[0083] In an optional embodiment, the S240 includes:

[0084] S241. Randomly select a preprocessed data from the training set;

[0085] S242. Calculate the initial water quality score corresponding to the preprocessed data based on the preprocessed data through Formula 1;

[0086]

[0087] where, F i is the initial water quality score of the i-th preprocessed data, X ij is the i-th preprocessed data matrix, n is the total number of indicators included in the preprocessed data, and W j is the weight of the j-th data type;

[0088] S243. Return to randomly select a preprocessed data from the training set until all the preprocessed data in the training set are selected, and obtain the initial water quality score corresponding to each preprocessed data.

[0089] It should be noted that since the evaluation model calculates the initial water quality score based on each indicator in the preprocessed data and the weight corresponding to each indicator, the obtained initial water quality score is a result obtained by integrating multi-source data and is more reliable.

[0090] In the actual calculation process, the weight corresponding to each index can be adjusted in real time. For example, the seasonal change coefficient is adjusted according to different seasons, so as to realize the dynamic correction of the weight factor. However, when calculating different water body targets, the consistency of the weights needs to be maintained to ensure that the initial water quality scores for different water bodies are obtained under the same evaluation criteria.

[0091] In an alternative embodiment, the S400 includes:

[0092] S410, calculating the evaluation trend of each real-time water quality score through the TOPSIS algorithm;

[0093] S420, sorting all the real-time water quality scores from small to large according to the evaluation trend, and screening out the real-time water quality scores in the last K; recording the real-time water quality data corresponding to the selected K real-time water quality scores as the observation samples;

[0094] S430, predicting the water quality evolution trend of the observation samples through spatio-temporal graph convolutional network to obtain the water quality prediction results of the observation samples;

[0095] S440, judging whether the water quality prediction results of the observation samples are qualified;

[0096] S450, if the water quality prediction results of the observation samples are unqualified, then record the observation samples as unqualified water quality.

[0097] It should be noted that in this application, the TOPSIS algorithm and spatio-temporal graph convolution are combined to predict the water quality trend of the observation samples, so as to timely detect unqualified water bodies in the observation samples according to the water quality trend, so as to issue a warning for early treatment, thereby reducing the ecological impact.

[0098] In an alternative embodiment, the S410 includes:

[0099] S411, standardizing the preprocessed data matrix of the real-time water quality data through formula 2;

[0100]

[0101] where, Z ij is the i-th standardized matrix of the real-time water quality data, X ij is the i-th preprocessed data matrix, n is the total number of indicators included in the preprocessed data, and m is the total number of samples in the preprocessed data;

[0102] S412, dimensionless processing of positive and negative indicators respectively;

[0103] S413, determining the optimal solution and the worst solution;

[0104] Specifically, the optimal solution is composed of the maximum values of each column of all the standardized matrices, that is, Z+ = (maxZ i1 , maxZ i2 ,..., maxZ in ), and the worst solution is composed of the minimum values of each column of all the standardized matrices, that is, Z- = (minZ i1 , minZ i2 ,..., minZ in );

[0105] S414. Calculate the positive distance between the standardized matrix and the optimal solution and the negative distance between the standardized matrix and the worst solution respectively;

[0106] Specifically, the positive distance between the standardized matrix and the optimal solution is calculated by formula 4, and the negative distance between the standardized matrix and the worst solution is calculated by formula 5;

[0107]

[0108] Among them, is the positive distance between the standardized matrix and the optimal solution, and Z ij is the i-th standardized matrix of the real-time water quality data;

[0109]

[0110] Among them, is the negative distance between the standardized matrix and the optimal solution, and Z ij is the i-th standardized matrix of the real-time water quality data.

[0111] It should be noted that all the observed samples are sorted by the TOPSIS algorithm. Thus, based on the sorted results, the observed samples are obtained from multiple water bodies. Based on the normalized original data matrix, the optimal solution and the worst solution in the finite solutions are found by the cosine method. Then, the distances between each evaluation object and the optimal solution and the worst solution are calculated respectively, and the relative closeness degree of each evaluation object to the optimal solution is obtained, which is used as the basis for evaluating the quality of the observed samples.

[0112] In an alternative embodiment, the S410 includes:

[0113] S415. Calculate the evaluation trend of the real-time water quality score by formula 3 by combining the positive distance and the negative distance;

[0114]

[0115] Among them, Si is the evaluation trend of the real-time water quality score, is the positive distance between the standardized matrix and the optimal solution, is the negative distance between the standardized matrix and the worst solution;

[0116] S416. Sort all the evaluation trends of the real-time water quality scores in descending order, and screen out the last K real-time water quality scores.

[0117] It should be noted that in Formula 3, since the denominator is greater than the numerator, the evaluation trend of the real-time water quality score is always less than 1. And as the evaluation trend of the real-time water quality score gets closer to 1, it means that the negative distance between the standardized matrix and the worst solution is smaller, which means that the water quality of the water body corresponding to the standardized matrix is better. Therefore, for the observation samples with a smaller evaluation trend of the real-time water quality score, it reflects that the water quality score of the corresponding water body may be poor. So these water bodies need to be taken as key observation objects to timely detect water quality problems and issue warnings.

[0118] In an alternative embodiment, S430 includes:

[0119] S431. Divide the water body into multiple structural nodes according to the water body image corresponding to the observation sample, and represent the spatial position relationship between the structural nodes through an adjacency matrix;

[0120] S432. Perform temporal convolution on the observation samples according to the chronological order of the acquisition time to obtain the temporal dependence relationship;

[0121] S433. Combine the adjacency matrix and the temporal dependence relationship for multi-scale spatio-temporal feature joint modeling, and output the water quality prediction image and water quality prediction data of the observation sample at future time steps through the output layer.

[0122] It should be noted that spatio-temporal graph convolution combines a graph convolutional network and a temporal convolutional network to capture the spatial and temporal features of the water body, so as to predict the water quality prediction image and water quality prediction data of the observation sample at future time steps, and make a further judgment on the observation sample according to the prediction results of spatio-temporal graph convolution to improve the reliability of the judgment of water body quality. When processing sequence data (such as time series, text, audio, etc.), a time step refers to the smallest unit into which the sequence is divided.

[0123] In an alternative embodiment, S440 includes:

[0124] S441. Input the water quality prediction image and water quality prediction data into the trained evaluation model, and obtain the predicted water quality score output by the trained evaluation model;

[0125] S442. Judge whether the water quality prediction result is qualified based on the predicted water quality score;

[0126] S443. If the water quality prediction result is unqualified, a warning is issued.

[0127] It should be noted that after obtaining the water quality prediction image and water quality prediction data of the observation sample in the future time step through spatio-temporal graph convolution, the water quality prediction image and water quality prediction data need to be input into the trained evaluation model to obtain the predicted water quality score of the water body in the future, so as to judge whether the water body is qualified based on the predicted water quality score, and a warning is issued in time when an unqualified water body is found.

[0128] As Figure 2 shown, the present application also provides a water quality health evaluation system based on big data, including a collection component and a processing component. The water quality data and water body images are collected through the collection component, and the water quality health evaluation method based on big data described in any one of Embodiment 1 is executed through the processing component. The data collected by the collection component is preprocessed through the processing component, and whether the water quality is qualified is judged in combination with the preprocessed data.

[0129] It should be noted that the collection component and the processing component are communicatively connected to facilitate the transmission of the data collected by the collection component into the processing component. After receiving the data, the processing component preprocesses all the data and judges whether the water body is qualified in combination with the preprocessed data.

[0130] The embodiments of the present invention have been described in detail above with reference to the drawings. However, the present invention is not limited thereto. Various changes can be made without departing from the spirit of the present invention within the knowledge scope of those skilled in the art.

Claims

1. A water quality health assessment method based on big data, characterized in that: include: Collect multiple basic water quality data and water body images, pre-process the basic water quality data and water body images to obtain pre-processed data and pre-processed images; Creating an evaluation model, inputting the preprocessed data and the preprocessed image into the evaluation model to train the evaluation model, and obtaining a trained evaluation model; Collect real-time water quality data, input the real-time water quality data into the trained evaluation model, and obtain the real-time water quality score; Based on the real-time water quality score, the observation samples are screened out, and the water quality evolution trend is predicted by combining the observation samples with the spatiotemporal graph convolution. The water quality warning area is screened based on the water quality evolution trend.

2. According to the big data-based water quality health assessment method of claim 1, it is characterized in that: Collect multiple basic water quality data and water body images, pre-process the basic water quality data and water body images, and obtain pre-processed data and pre-processed images, including: Create water quality data tables; Set collection parameters, collect multiple basic water quality data based on the collection parameters, and put all the collected basic water quality data into a water quality data table; the basic water quality data includes water body pH value, conductivity and dissolved oxygen; Collect multiple water body images based on the acquisition parameters, and put all the collected water body images into a water quality data table; Each basic water quality data is normalized in turn to obtain preprocessed data; Each water body image is filtered in turn to obtain multiple pre-processed images.

3. A water quality health assessment method based on big data according to claim 2, characterized in that: Create an evaluation model, input the preprocessed data and the preprocessed image into the evaluation model to train the evaluation model, and obtain the trained evaluation model, including: Create evaluation models; A coupling relationship between a preprocessed image and preprocessed data is created according to the corresponding relationship of the acquisition frequency, and the preprocessed image, the preprocessed data, and the coupling relationship between the preprocessed image and the preprocessed data are recorded as a training sample to obtain multiple training samples; Divide all training samples into training sets and test sets according to random proportions; The training set is input into the evaluation model, so that the evaluation model continuously learns the relationship between the water body image and the water quality data, and obtains the trained evaluation model; Input the test set into the trained evaluation model to verify whether the trained evaluation model is trained.

4. A water quality health assessment method based on big data according to claim 3, characterized in that: The training set is input into the evaluation model so that the evaluation model continuously learns the relationship between water body images and water quality data, and the trained evaluation model is obtained, including: Randomly select a preprocessed data from the training set; Based on the pre-processed data, the initial water quality score corresponding to the pre-processed data is calculated by formula 1; Among them, F i is the initial water quality score of the i-th preprocessed data, X ij is the i-th preprocessed data matrix, n is the total number of indicators contained in the preprocessed data, W j is the weight of the jth data type; Return to randomly select a preprocessed data from the training set until all preprocessed data in the training set are selected, and obtain the initial water quality score corresponding to each preprocessed data.

5. A water quality health assessment method based on big data according to claim 4, characterized in that: Based on the real-time water quality score, the observation samples are screened out, and the water quality evolution trend is predicted by combining the observation samples with the spatiotemporal graph convolution. The water quality early warning areas are screened based on the water quality evolution trend, including: The evaluation trend of each real-time water quality score is calculated using the TOPSIS algorithm; All real-time water quality scores are sorted from small to large according to the evaluation trend, and the real-time water quality scores at the bottom K are screened out; the real-time water quality data corresponding to the selected K real-time water quality scores are recorded as observation samples; The water quality prediction results of the observed samples are obtained by predicting the water quality evolution trend of the observed samples through spatiotemporal graph convolution. Determine whether the water quality prediction results of the observed samples are qualified; If the water quality prediction result of the observed sample is unqualified, the observed sample will be recorded as unqualified water quality.

6. A water quality health assessment method based on big data according to claim 5, characterized in that: The evaluation trend of each real-time water quality score is calculated by the TOPSIS algorithm, including: The preprocessed data matrix of real-time water quality data is standardized by formula 2; Among them, Z ij is the i-th standardized matrix of real-time water quality data, X ij is the i-th preprocessed data matrix, n is the total number of indicators contained in the preprocessed data, and m is the total number of samples in the preprocessed data; The positive and negative indicators are dimensionless respectively; Determine the best and worst solutions; The positive distance between the normalized matrix and the optimal solution and the negative distance between the normalized matrix and the worst solution are calculated respectively.

7. A water quality health assessment method based on big data according to claim 6, characterized in that: The evaluation trend of each real-time water quality score is calculated by the TOPSIS algorithm, including: The evaluation trend of the real-time water quality score is calculated by combining the positive distance and the negative distance through Formula 3; Among them, Si is the evaluation trend of real-time water quality score, is the positive distance between the normalized matrix and the optimal solution, is the negative distance between the normalized matrix and the worst solution; The evaluation trends of all real-time water quality scores are sorted in order from large to small, and the real-time water quality scores at the bottom K are screened out.

8. A water quality health assessment method based on big data according to claim 7, characterized in that: The water quality evolution trend of the observed samples is predicted by spatiotemporal graph convolution, and the water quality prediction results of the observed samples are obtained, including: The water body is divided into multiple structural nodes according to the water body image corresponding to the observed sample, and the spatial position relationship between the structural nodes is represented by the adjacency matrix; Perform time convolution on the observed samples according to the order of acquisition time to obtain the temporal dependency; The adjacency matrix and temporal dependency are combined to jointly model multi-scale spatiotemporal features, and the water quality prediction image and water quality prediction data of the observed samples in the future time step are output through the output layer.

9. A water quality health assessment method based on big data according to claim 8, characterized in that: If the water quality prediction result of the observed sample is unqualified, the observed sample will be recorded as unqualified water quality, including: Inputting the water quality prediction image and the water quality prediction data into the trained evaluation model to obtain the predicted water quality score output by the trained evaluation model; Determine whether the water quality prediction result is qualified based on the predicted water quality score; If the water quality forecast result is unsatisfactory, an early warning will be issued.

10. A water quality health assessment system based on big data according to claim 9, characterized in that: include: A collection component, through which water quality data and water body images are collected; A processing component, wherein the water quality health assessment method based on big data described in any one of claims 1 to 9 is executed by the processing component, the data collected by the collection component is preprocessed by the processing component, and whether the water quality is qualified is determined based on the preprocessed data.

Images