Water quality sampling and monitoring method and system for intelligent water station based on ultraviolet spectrum

Abstract

The invention relates to the technical field of water quality detection, and discloses a water quality sampling monitoring method and system for an intelligent water station based on ultraviolet spectrum, and the method comprises the steps: carrying out the regional division of the intelligent water station, and obtaining a plurality of water areas; performing water quality analysis according to the ultraviolet spectrum data of the water sample at the central sampling point of each water area to obtain a normal water area and an abnormal water area; performing water quality analysis on all sampling point water samples in the abnormal water area to obtain pollution source sampling points; the intelligent water station is subjected to regional treatment, so that the water sample collection quantity is reduced, and the water quality detection process is accelerated; different water quality analysis methods are adopted in different water quality analysis stages, dynamic adjustment of the analysis process is achieved, a sampling point closest to a pollution source is determined, and water station water quality treatment is facilitated; the model parameters of the water quality detection model are optimized by using a multi-objective optimization method, so that local optimum is effectively avoided when a multi-output objective problem is processed, and the accuracy of a water quality detection result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality detection, and in particular to a water quality sampling and monitoring method and system for a smart water station based on ultraviolet spectroscopy. Background Art

[0002] Traditional water quality sampling and testing methods require separate chemical analysis of water samples at each sampling point, resulting in a large amount of analysis work, increasing time and labor costs; traditional testing methods often use fixed analysis processes and single water quality indicators, which cannot be adjusted and optimized according to actual water station conditions, and have poor adaptability to water quality conditions and water station environments.

[0003] For example, a Chinese patent application with publication number CN115060678A discloses a water quality detection system based on liquid core optical fiber and ultraviolet spectrum, including: an ultraviolet LED light source, a liquid core optical fiber, a water sample input head, a water sample output head, a photoelectric sensor, an input multimode optical fiber, an output multimode optical fiber and a single-chip computer; a light beam emitted from the LED light source is transmitted into the liquid core optical fiber through the input optical fiber, and at the same time, the water sample to be tested flows into the liquid core optical fiber through the water sample input head, absorbs the light beam, and transmits the light beam to the photoelectric sensor through the output multimode optical fiber. The sensor receives the light beam, converts the optical signal into an electrical signal and transmits it to the single-chip computer, and processes the result through an artificial neural network algorithm to obtain data on four indicators of total phosphorus, total nitrogen, ammonia nitrogen and COD of the water body to be tested, and sends them to the terminal computer to achieve real-time monitoring.

[0004] The above-mentioned prior art has the problems raised by this background technology: it is necessary to carry out separate chemical analysis of the water sample at each sampling point, which leads to a large amount of analysis work, increasing time and labor costs; it adopts a fixed analysis process and a single water quality indicator, which cannot be adjusted and optimized according to the actual situation of the water station, and has poor adaptability to the water quality conditions and water station environment; in order to solve at least one of the above problems, the present application proposes a water quality sampling and monitoring method and system for a smart water station based on ultraviolet spectroscopy. Summary of the invention

[0005] In view of the shortcomings of the prior art, the main purpose of this application is to provide a water quality sampling and monitoring method and system for a smart water station based on ultraviolet spectroscopy, which can effectively solve the problems in the background technology. The specific technical solutions of this application are as follows: A water quality sampling and monitoring method for a smart water station based on ultraviolet spectroscopy, comprising: The smart water station is divided into regions to obtain multiple water areas; According to the ultraviolet spectrum data of water samples from the central sampling point of each water area, water quality analysis is carried out to obtain normal water areas and abnormal water areas; By analyzing the environmental information of the abnormal water area, the water area boundary is optimized to obtain an updated abnormal water area range, wherein the optimization process includes translation processing and stretching processing; Water quality analysis is performed on water samples from all sampling points within the updated abnormal water area to obtain pollution source sampling points so as to monitor the water quality of the smart water station in real time.

[0006] Specifically, the water quality analysis is performed based on the ultraviolet spectrum data of the water samples at the central sampling point of each water area to obtain normal water areas and abnormal water areas, including: In the preset sliding window, UV spectrum data of water samples at the central sampling point of each water area is collected to obtain multiple sets of UV spectrum data; By analyzing the absorbance values ​​of the multiple groups of ultraviolet spectral data, corresponding feature vectors are constructed, and the water quality of the water samples at the central sampling point of each water area is analyzed to obtain normal water areas and abnormal water areas.

[0007] Specifically, in the preset sliding window, ultraviolet spectrum data is collected for water samples at the central sampling point of each water area, to obtain multiple sets of ultraviolet spectrum data, including: According to the historical UV spectrum data, the sliding window and sliding step size are obtained through the preset window optimization model; According to the sliding step length, ultraviolet spectral data of water samples at each central sampling point of the water area are collected in the sliding window respectively, and the sliding window is filled with ultraviolet spectral data to obtain multiple groups of ultraviolet spectral data.

[0008] Specifically, by analyzing the absorbance values ​​of the multiple sets of ultraviolet spectrum data, corresponding feature vectors are constructed, and the water quality of the water samples at the central sampling point of each water area is analyzed to obtain normal water areas and abnormal water areas, including: Based on historical UV spectrum data and corresponding water quality data, multiple specific wavelengths related to water quality are obtained; At each of the specific wavelengths, the absorbance value of each set of ultraviolet spectrum data is extracted to obtain a plurality of absorbance values; Combine multiple absorbance values ​​of each group to obtain a characteristic vector of each group of ultraviolet spectrum data; The feature vector is input into a pre-trained water quality classification model, and the water quality of the water sample at the central sampling point of each water area is analyzed to obtain normal water areas and abnormal water areas.

[0009] Specifically, the water area boundary is optimized by analyzing the environmental information of the abnormal water area to obtain an updated abnormal water area range, wherein the optimization process includes translation process and stretching process, including: According to the preset pressure sensor, the abnormal water area boundary is translated in combination with the water flow pressure data in the abnormal water area to obtain a moving boundary; According to a preset pollution diffusion model, the moving boundary is stretched to obtain an optimized boundary; According to the optimized boundary, an updated abnormal water area range is obtained.

[0010] Specifically, the water quality analysis is performed on water samples from all sampling points within the updated abnormal water area to obtain the pollution source sampling points so as to monitor the water quality of the smart water station in real time, including: Obtain water samples from all sampling points within the updated abnormal water area; In the preset sliding window, the spectrum data of the water sample at each sampling point is collected respectively to obtain the ultraviolet spectrum data of the water samples at multiple sampling points; Through the preset water quality detection model, the ultraviolet spectrum data of the water sample at each sampling point is analyzed to obtain water quality parameter information; Performing an abnormal analysis on the water quality parameter information to obtain normal sampling points and abnormal sampling points; Among the abnormal sampling points, the sampling points with the highest abnormality in water quality parameters are selected as the pollution source sampling points.

[0011] Specifically, the preset water quality detection model is used to analyze the ultraviolet spectrum data of the water sample at each sampling point to obtain water quality parameter information, including: Based on the historical water quality data of water samples at the sampling points, a pre-trained water quality detection model is obtained in combination with the multi-objective particle swarm optimization method; The ultraviolet spectrum data of the water sample at each sampling point is input into the pre-trained water quality detection model to obtain the water quality parameter information.

[0012] Specifically, the pre-trained water quality detection model is obtained based on the historical water quality data of the water samples at the sampling points and combined with the multi-objective particle swarm optimization method, including: Construct an initial neural network model according to the preset model parameters; Initializing a particle swarm according to the initial neural network model and preset particle swarm algorithm parameters to obtain an initial particle swarm, wherein particles in the initial particle swarm represent weights and bias values ​​of the initial neural network model; According to a preset optimization goal, the initial particle swarm is updated until a preset maximum number of iterations is reached to obtain an optimal solution set; Substituting the optimal solution set into the initial neural network model to obtain a water quality detection model; The water quality detection model is trained using historical ultraviolet spectrum data of water samples at sampling points and corresponding historical water quality data to obtain a pre-trained water quality detection model.

[0013] Specifically, the initial particle swarm is updated according to a preset optimization target until a preset maximum number of iterations is reached to obtain an optimal solution set, including: Analyze the solutions corresponding to the particles according to the preset optimization objectives and obtain multiple non-inferior solutions; storing the plurality of non-inferior releases into a pre-created external archive; Updating the non-inferior solutions in the external file by using the Euclidean distance between the non-inferior solutions and a preset critical distance; The particles in the initial particle swarm are updated until the preset maximum number of iterations is reached and the non-inferior solution meets the preset external file size, and an updated non-inferior solution is obtained; According to the updated non-inferior solutions in the external archive, an optimal solution set is obtained.

[0014] A water quality sampling and monitoring system for a smart water station based on ultraviolet spectroscopy, used to implement a water quality sampling and monitoring method for a smart water station based on ultraviolet spectroscopy, comprising: The water area division module divides the smart water station into regions and obtains multiple water areas; The abnormal water area detection module analyzes water quality based on the ultraviolet spectrum data of water samples at the central sampling point of each water area to obtain normal water areas and abnormal water areas; The pollution source sampling point detection module optimizes the water area boundary by analyzing the environmental information of the abnormal water area to obtain an updated abnormal water area range, wherein the optimization processing includes translation processing and stretching processing. Water quality analysis is performed on water samples from all sampling points within the updated abnormal water area to obtain the pollution source sampling points, so as to monitor the water quality of the smart water station in real time.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This application performs anomaly detection on different areas of the smart water station, first screens out abnormal water areas, and then analyzes the water samples at each sampling point in the abnormal water areas, thereby reducing the amount of data that needs to be analyzed and accelerating the water quality detection process; adopts different water quality analysis methods at different water quality analysis stages to achieve dynamic adjustment of the analysis process to adapt to different water quality conditions, performs water quality parameter analysis on water samples at a single sampling point, and determines the sampling point closest to the pollution source, which is convenient for water station water quality management; based on ultraviolet spectral data, the model parameters of the water quality detection model are optimized using a multi-objective optimization method, which can effectively avoid falling into local optimality when dealing with multi-output target problems, thereby improving the prediction accuracy of the water quality detection model and improving the accuracy of the water quality detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1This is a workflow diagram of a water quality sampling and monitoring method for a smart water station based on ultraviolet spectroscopy in Example 1 of the present invention; Figure 2 A schematic diagram of setting central sampling points of different water areas of the smart water station in Example 1 of the present invention; Figure 3 is a schematic diagram of ultraviolet spectrum data in Example 1 of the present invention; Figure 4 This is a schematic diagram of optimizing the abnormal water area boundary in Example 1 of the present invention; Figure 5 is a workflow diagram of the initial particle swarm update process in Example 1 of the present invention; Figure 6 This is a structural schematic diagram of a water quality sampling and monitoring system for a smart water station based on ultraviolet spectroscopy in Example 2 of the present invention. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0018] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0020] Embodiment 1:

[0021] This embodiment provides a water quality sampling and monitoring method for a smart water station based on ultraviolet spectroscopy. Figure 1 The water quality sampling and monitoring method of a smart water station based on ultraviolet spectroscopy comprises: S101, dividing the smart water station into regions to obtain multiple water areas; S102, performing water quality analysis based on the ultraviolet spectrum data of water samples at the central sampling point of each water area to obtain normal water areas and abnormal water areas; S103. Analyze the water quality of water samples from all sampling points in the abnormal water area to obtain the sampling points of the pollution sources so as to monitor the water quality of the smart water station in real time.

[0022] For smart water stations with normal initial water quality, the water quality in the water station will change over time or in the event of a sudden water pollution incident. Long-term testing of the water quality of the smart water station and identification of the location of the pollution source can quickly manage the water quality at the polluted location, promptly discover potential sources of pollution risks, take preventive measures in advance, reduce the probability of water pollution disasters, and ensure water ecological safety.

[0023] In this embodiment, the smart water station will set up multiple sampling points when monitoring water quality. The traditional water quality detection method directly analyzes the water quality of the water samples collected at each sampling point, which will result in a huge analysis workload. The embodiment of the present application divides the smart water station into water areas to obtain multiple water areas. For each water area, a water sample from the center point of the water area is collected, and water quality detection is performed on the water sample at the center sampling point to preliminarily screen out abnormal water areas with abnormal water quality, reduce the number of water samples collected, and reduce the workload of water sample analysis.

[0024] Specifically, ultraviolet spectral data of water samples at the central sampling point of each water area are collected to obtain ultraviolet spectral data of water samples at the central sampling point of each water area. Different water quality components have different absorption characteristics for ultraviolet light, forming specific ultraviolet spectral absorption peaks and absorption valleys. Normal water quality has relatively stable spectral characteristics, while abnormal water quality will cause the spectral characteristics to shift, deform or appear new peaks. The water quality of the corresponding water sample can be detected by analyzing the ultraviolet spectral data. The traditional ultraviolet spectral analysis method uses principal component analysis or statistical methods to extract features and reduce dimensions of spectral data, and compares the spectral features with normal spectral features to obtain abnormal spectra. The traditional analysis method can only capture linear features in spectral data, the analysis is not comprehensive enough, and there are problems of high computational complexity and information loss when processing high-dimensional data. The embodiment of the present application extracts feature vectors from the ultraviolet spectral data of each group of water samples, and based on the feature vectors, uses a pre-trained model to classify water quality, which can mine nonlinear features in spectral data, improve the analysis ability of spectral data, accurately identify abnormal spectra that are significantly different from normal features, and improve the accuracy of abnormal detection.

[0025] When an abnormal water area is identified, the water quality of the abnormal water area needs to be treated. Traditionally, direct treatment of the entire abnormal water area is time-consuming and labor-intensive. This embodiment further analyzes the water quality of all sampling points in the abnormal water area to obtain the pollution source sampling point closest to the pollution source, and treats the water quality from the source, which can reduce costs and improve treatment efficiency.

[0026] Specifically, the water quality of the water samples at each sampling point in the abnormal water area is analyzed separately, and the water quality parameters of the water samples at each sampling point are detected. By determining the water quality parameters, the chemical oxygen demand (COD), heavy metal content and other information in the water samples are obtained, and then the water quality of the water samples and the causes of abnormal water quality are judged. The traditional water quality parameter detection model is prone to fall into the local optimal problem when dealing with multi-output problems, affecting the accuracy of the prediction results. The embodiment of the present application uses a multi-objective optimization algorithm to optimize the model parameters of the water quality detection model to avoid falling into the local optimal problem and improve the accuracy of the predicted water quality parameter results; the prediction results are compared with the water quality standards, the abnormal sampling points are identified, and the abnormal substances and the corresponding causes are determined. The specific sampling points that cause the abnormal water quality can be accurately found, and the sampling points with the highest deviation of water quality parameters from the water quality standards among the abnormal sampling points are used as the pollution source sampling points, which provide accurate basis for subsequent targeted treatment measures, help to timely discover water quality changes, realize real-time and accurate monitoring of water quality in smart water stations, and ensure water resource safety.

[0027] This application performs anomaly detection on different areas of the smart water station, first screens out abnormal water areas, and then analyzes the water samples at each sampling point in the abnormal water areas, thereby reducing the amount of data that needs to be analyzed and accelerating the water quality detection process; adopts different water quality analysis methods at different water quality analysis stages to achieve dynamic adjustment of the analysis process to adapt to different water quality conditions, performs water quality parameter analysis on water samples at a single sampling point, and determines the sampling point closest to the pollution source, which is convenient for water station water quality management; based on ultraviolet spectral data, the model parameters of the water quality detection model are optimized using a multi-objective optimization method, which can effectively avoid falling into local optimality when dealing with multi-output target problems, thereby improving the prediction accuracy of the water quality detection model and improving the accuracy of the water quality detection results.

[0028] Furthermore, the water quality analysis is performed based on the ultraviolet spectrum data of the water samples at the central sampling point of each water area to obtain normal water areas and abnormal water areas, including: S201, within a preset sliding window, collecting ultraviolet spectrum data of water samples at the central sampling point of each water area, to obtain multiple sets of ultraviolet spectrum data; S202. Analyze the water quality of water samples at the central sampling point of each water area according to the multiple sets of ultraviolet spectrum data to obtain normal water areas and abnormal water areas.

[0029] In this embodiment, if Figure 2, obtain water sample data from the central sampling point of each water area, and divide the smart water station into regions according to the scale and monitoring needs of the smart water station. For example, for a medium-sized lake water station, it can be divided into regions according to the lakeside, lake center, water inlet, water outlet and other locations; set the sampling time period according to the change frequency of the historical water quality data of the smart water station and the monitoring needs, collect water samples at each central sampling point according to the time period, and record the sampling point location, sampling time and other information to achieve continuous water quality testing, and be able to capture some periodic or sudden water quality changes in time.

[0030] Specifically, by collecting ultraviolet spectral data, the absorption characteristics of various components in water samples to ultraviolet light can be reflected. Different substances have their specific absorption peaks in the ultraviolet light region. The ultraviolet light region usually refers to the electromagnetic wave region with a wavelength of 200-400nm. The position and intensity of these absorption peaks can be used as characteristic indicators of water quality components and contents, thereby detecting the water quality components in the water sample; the traditional collection method directly collects the ultraviolet spectral data of the water sample, and the obtained data has poor time consistency and comparability; the embodiment of the present application uses a sliding window to collect ultraviolet spectral data of water samples at the central sampling point of each water area, which can provide detailed and continuous spectral information in a unified window, which is helpful to discover tiny changes in spectral characteristics and ensure data consistency and comparability; in a preset sliding window, according to a certain sliding step size, spectral data within the same wavelength range is acquired in a continuous manner to obtain multiple sets of ultraviolet spectral data.

[0031] For each set of ultraviolet spectral data, the characteristics of the ultraviolet spectral data are extracted. The substances and their proportions contained in normal water samples are relatively stable, and the corresponding spectral curves have certain regularity and similarity. Abnormal water samples contain additional pollutants or abnormal changes in substance concentrations, which will cause the spectral curve to shift, new absorption peaks to appear, or abnormal changes in the absorption peak intensity, and the corresponding characteristics will also change abnormally. The characteristics of each set of ultraviolet spectral data are input into a preset water quality classification model, and the water quality of the water samples is classified according to the characteristics. This can accurately and quickly screen out water sample areas with water quality problems. Through the unified analysis of water samples at the central sampling point of each area, the overall monitoring efficiency is improved, and the time and resource consumption of complex analysis of water samples at each sampling point is avoided. Analysis based on spectral data can detect a variety of pollutants, thereby enhancing the comprehensiveness and sensitivity of water quality monitoring.

[0032] Furthermore, within the preset sliding window, ultraviolet spectrum data is collected for water samples at the central sampling point of each water area to obtain multiple sets of ultraviolet spectrum data, including: S301, according to the historical ultraviolet spectrum data, a sliding window and a sliding step length are obtained through a preset window optimization model; S302. Collect ultraviolet spectral data of water samples at each central sampling point of the water area in the sliding window according to the sliding step length, fill the sliding window with ultraviolet spectral data, and obtain multiple groups of ultraviolet spectral data.

[0033] In this embodiment, the historical ultraviolet spectral data includes various information from previous spectral analysis of water samples, reflecting the spectral characteristics at different times and under different water quality conditions. The preset window optimization model is used to analyze the statistical characteristics and internal laws of the historical data, so as to determine the sliding window and sliding step that are most suitable for the current monitoring situation, thereby avoiding information loss or redundancy due to the window being too large or too small. For example, for the historical ultraviolet spectral data of a lake water station, which includes the spectral information of water samples in different seasons and at different times of the past year, PCA is used as the window optimization model. After the data is input into PCA, it is found that the principal component contribution rate of the wavelength in the range of 230-280nm is the largest, indicating that this range contains important information about the main substances in the water sample. Therefore, the sliding window is set to 230-280nm. At the same time, analysis finds that in the range of 230-250nm, the discreteness of the data is large, while in the range of 250-280nm, the discreteness of the data is small. Therefore, for the range of 230-250nm, the sliding step is set to 1nm, and for the range of 250-280nm, the sliding step is set to 5nm. According to the historical ultraviolet spectrum data and the window optimization model, the optimal sliding window and sliding step can be determined, avoiding the subjectivity and blindness of manually setting the window and step, and improving the efficiency and quality of data collection.

[0034] Specifically, the measurement parameters of the ultraviolet spectrometer are set according to the determined sliding window and sliding step size, for example, the sliding window is 230-250nm, the step size is 1 nm, the measurement wavelength range and scanning interval of the ultraviolet spectrometer are set to corresponding values, each group of water samples is placed in the sample pool of the ultraviolet spectrometer in turn, and the ultraviolet spectrometer is started for measurement. For each group of water samples, starting from the starting wavelength of the sliding window, the absorbance of each wavelength point is gradually measured according to the sliding step size until the end wavelength of the sliding window is reached, and the absorbance data of each group of water samples in the sliding window is recorded to form a complete set of ultraviolet spectrum data, ensuring that the data at each wavelength point is accurately measured and recorded, and obtaining continuous spectral information, which comprehensively and accurately reflects the absorption characteristics of water samples within a specific wavelength range, thereby reflecting the various components of water quality and their concentration changes.

[0035] Furthermore, the water quality of the water samples at the central sampling point of each water area is analyzed according to the multiple sets of ultraviolet spectrum data to obtain normal water areas and abnormal water areas, including: S401, obtaining a plurality of specific wavelengths related to water quality according to historical ultraviolet spectrum data and corresponding water quality data; S402, extracting the absorbance value of each group of ultraviolet spectrum data at each specific wavelength to obtain multiple absorbance values; S403, combining multiple absorbance values ​​of each group to obtain a feature vector of each group of ultraviolet spectrum data; S404, inputting the feature vector into a pre-trained water quality classification model, analyzing the water quality of water samples at the central sampling point of each water area, and obtaining normal water areas and abnormal water areas.

[0036] In this embodiment, if Figure 3 , curves of different colors represent the changes in the absorbance values ​​of different water quality components with wavelength. The historical ultraviolet spectrum data and the corresponding water quality data include detailed information on the previous monitoring of water samples. Different water quality components have their own unique absorption characteristics on the ultraviolet spectrum, and these characteristics will be manifested at specific wavelengths. By analyzing the historical data, it can be determined which wavelengths are closely related to water quality indicators. These specific wavelength positions are highly sensitive to changes in water quality, and specific wavelengths are used as key feature points for water quality analysis; for example, in the process of historical data analysis, it was found that when the COD concentration in the water sample increased, the absorbance at 220nm always changed significantly, and the correlation coefficient reached 0.85; when the ammonia nitrogen content changed, the absorbance at 254nm showed a significant positive correlation with it, and the correlation coefficient was 0.78; for the presence of certain heavy metal ions, the absorbance change at 280nm is more sensitive, with a correlation coefficient of 0.70; 220nm, 254nm and 280nm can be determined as specific wavelengths related to water quality.

[0037] Specifically, after determining the specific wavelength, the absorbance values ​​at these specific wavelengths are extracted from each set of ultraviolet spectral data. The absorbance values ​​at these specific wavelengths can more accurately reflect the important information of specific components or water quality indicators in the water sample. By extracting these values, complex spectral information can be simplified into several key data points, while retaining information closely related to water quality, reducing interference from irrelevant information, and improving the validity of the data; the absorbance values ​​extracted at multiple specific wavelengths are combined into a feature vector, and the information is integrated to obtain a comprehensive feature representation of each group of water samples at a specific wavelength, including the absorption characteristics of water samples at different water quality sensitive wavelengths, which is convenient for quantification and comparison of multiple groups of water samples.

[0038] Specifically, a large amount of historical data is used to train the water quality classification model to obtain a pre-trained water quality classification model, and the feature vector patterns of normal water quality and abnormal water quality are learned. When a new feature vector is input, the model can judge whether the water sample represented by the feature vector belongs to the normal group or the abnormal group according to its internal learning rules and training results. For example, using a trained SVM model, the feature vector [0.2, 0.25, 0.3] is input into the model. The model judges that the feature vector belongs to the normal group according to its internal decision function and support vector, and the output is 1, then the group of water samples is judged as a normal water area with normal water quality; for the feature vector [0.5, 0.6, 0.7] of another group of water samples, the model output is 0, and it is judged as an abnormal water area with abnormal water quality. The pre-trained model can quickly and accurately classify a large number of water sample groups, improve the automation and accuracy of water quality monitoring, and is suitable for processing a large amount of water sample data and complex water quality changes.

[0039] Furthermore, the water quality analysis of water samples from all sampling points in the abnormal water area is performed to obtain the sampling points of the pollution source so as to monitor the water quality of the smart water station in real time, including: S501, optimizing the water area boundary according to the environmental information of the abnormal water area to obtain an updated abnormal water area range; S502, obtaining water samples from all sampling points within the updated abnormal water area; S503, within a preset sliding window, respectively collect spectral data of the water sample at each sampling point to obtain ultraviolet spectral data of the water samples at multiple sampling points; S504, analyzing the ultraviolet spectrum data of the water sample at each sampling point through a preset water quality detection model to obtain water quality parameter information; S505, performing abnormal analysis on the water quality parameter information to obtain normal sampling points and abnormal sampling points; S506. Select the sampling point with the highest degree of water quality parameter abnormality from the abnormal sampling points as the pollution source sampling point.

[0040] Furthermore, the current ultraviolet spectrum water quality detection algorithm uses ultraviolet light to detect water quality in the test area, but it cannot quickly and accurately find the source of pollution. The present application first introduces the abnormal area of ​​pollutants and calculates the ideal abnormal area of ​​pollutants. Then, it combines the water flow model and considers the impact of water flow on the diffusion of pollutants, thereby updating the actual abnormal area based on the water flow model. Since the actual abnormal area fully considers the impact of water flow, it can quickly and accurately locate the actual impact of the pollution source, and then find the specific location of the pollution source through ultraviolet light. The present application can quickly detect the specific location of the pollution source in the early stage of the pollution source entering, thereby quickly responding to and processing the pollution source, ensuring the health of water quality.

[0041] In this embodiment, the boundary of the abnormal water area is optimized according to the environmental information of the abnormal water area, the actual boundary range of the abnormal water area is determined more accurately, and the influence of different environmental factors on the diffusion range of pollutants in the water area is eliminated. For example, the direction and speed of the water flow will cause the pollutants to diffuse downstream, the wind direction will affect the distribution of pollutants on the water surface, and the topography such as the bend of the river channel and the valley topography will cause pollutants to gather in certain areas or change the diffusion path. At this time, the initial abnormal water area division range cannot accurately reflect the scope of polluted water. Within the updated abnormal water area, water samples from all sampling points within the abnormal water area are obtained, the water quality in the abnormal water area is further analyzed, and the water quality information in the abnormal water area is fully obtained, providing a sufficient data basis for accurately analyzing the water quality status and determining the pollution source.

[0042] Specifically, within a preset sliding window, spectral data of water samples at each sampling point in the abnormal water area are collected to obtain ultraviolet spectral data of the water samples at each sampling point; the ultraviolet spectral data of the water samples at each sampling point are respectively input into a preset water quality detection model, and the model extracts features related to water quality parameters from the ultraviolet spectral data, and maps these features to specific water quality parameter values. There is a certain quantitative relationship between the spectral absorption characteristics of different substances and water quality parameters. The ultraviolet spectral data is converted into actual quantifiable water quality parameters, such as chemical oxygen demand (COD), ammonia nitrogen content, total phosphorus content and other parameter data through the model, to obtain specific water quality parameter information.

[0043] Specifically, the water quality parameters are compared and analyzed according to the normal water quality parameter range determined by national water quality standards or expert experience, and the water quality parameter information of the water sample at each sampling point is compared with the normal water quality parameter range to determine whether it exceeds the normal range, thereby determining whether the sampling point is normal or abnormal. For example, the normal range of COD is set to 0-20 mg / L, ammonia nitrogen is 0-1 mg / L, and total phosphorus is 0-0.2 mg / L. The COD of sampling point A is 30 mg / L, then sampling point A is determined to be an abnormal sampling point, and the corresponding water quality abnormality is determined according to its COD abnormality. The reasons are identified and corresponding countermeasures are taken; if all water quality parameters of sampling point B are within the normal range, it is determined to be a normal sampling point; through the analysis of water quality parameters at the sampling points, the abnormal sampling points can be accurately located, and the sampling points with the highest degree of abnormal water quality parameters can be screened out among the abnormal sampling points. As the pollution source sampling points, it is helpful to quickly discover the source of water pollution, provide clear information for subsequent targeted governance and adjustments, realize real-time and accurate monitoring of the water quality of smart water stations, ensure the safety and stability of the water quality of water stations, take timely measures to solve water quality problems, and improve the efficiency and effectiveness of water resources management.

[0044] Furthermore, the water area boundary is optimized according to the environmental information of the abnormal water area to obtain an updated abnormal water area range, including: S601, according to a preset pressure sensor, in the abnormal water area, combined with water flow pressure data, the boundary of the abnormal water area is translated to obtain a moving boundary; S602, stretching the moving boundary according to a preset pollution diffusion model to obtain an optimized boundary; S603: Obtain an updated abnormal water area based on the optimized boundary.

[0045] In this embodiment, as the water flow diffuses, the pollutants also diffuse, and the diffusion direction and speed are closely related to factors such as the speed and direction of the water flow. In abnormal waters, the water flow speed and direction are analyzed according to the water pressure, and the water flow pressure data is obtained through a preset pressure sensor. The flow direction of the water flow is determined according to the water flow pressure data. For example, if the water flow pressure data in a certain area shows that the water flow has a tendency to flow or gather in a specific direction, then the boundary of the abnormal water area will move in that direction. Correspondingly, the original abnormal water area boundary is translated to obtain a moving boundary. Through boundary movement, the changes in the abnormal water area boundary caused by the water flow can be dynamically tracked, so that the boundary of the abnormal water area is more in line with the actual situation, avoiding the inaccurate boundary caused by the movement of the water flow, and providing a more accurate range for determining the location of the pollution source.

[0046] Specifically, the diffusion of pollutants in water bodies has certain regularity. According to the preset pollution diffusion model, the pollution range is further analyzed. The pollution diffusion model is selected according to the water body type and water flow velocity of the abnormal water area, and the two-dimensional convection-diffusion equation is used to describe the spatiotemporal distribution of pollutants in the water area. The model equation is as follows: ; Where C is the pollutant concentration of the water sample at the central sampling point, , are the speeds of the water flow in the x and y directions, and is the turbulent diffusion coefficient, which can be set based on the actual water characteristics. S is the pollution source term. When there is no new pollution source, S is 0. t is the time. is the change of pollutant concentration over time, is the change in pollutant concentration in the x direction, is the change of pollutant concentration in the y direction, is the changing speed of pollutant concentration in the x direction, is the rate of change of pollutant concentration in the y direction. According to the pollution diffusion model, the maximum diffusion distance of pollutants is determined, the boundary of the polluted area is stretched, and the diffusion trend of pollutants is comprehensively considered to avoid determining the boundary based only on the currently detected pollution range, so as to obtain a more comprehensive polluted area, which is conducive to more thorough water quality testing and pollution source investigation, and reduce the omission of polluted areas.

[0047] Specifically, according to the direction and speed of the water flow and combined with the two-dimensional convection-diffusion equation, the diffusion range of pollutants within a preset time is solved, such as the maximum diffusion range of pollutants within an hour. According to the maximum diffusion range, the boundary of the abnormal water area is stretched and optimized to obtain an updated abnormal water area range, which can better reflect the actual range of polluted water areas.

[0048] In this embodiment, by comprehensively considering the boundary movement caused by water flow pressure and the scope of pollution diffusion, a more accurate and comprehensive abnormal water area can be determined based on the optimized boundary, which can more truly reflect the area actually affected by pollution. Figure 4 , it is detected that the direction of the water flow is to the right, the initial square abnormal water area boundary is translated to the right, and the pollutant diffusion range is solved according to the pollution diffusion model. The initial square abnormal water area boundary is stretched into an elliptical water area boundary to obtain the optimized boundary. Within the optimized boundary, the water quality of the abnormal water area is further analyzed.

[0049] Furthermore, the preset water quality detection model is used to analyze the ultraviolet spectrum data of the water sample at each sampling point to obtain water quality parameter information, including: S701, according to the historical water quality data of the water samples at the sampling points, combined with the multi-objective particle swarm optimization method, a pre-trained water quality detection model is obtained; S702: Input the ultraviolet spectrum data of the water sample at each sampling point into the pre-trained water quality detection model to obtain water quality parameter information.

[0050] In this embodiment, a neural network model is used as a water quality detection model to map the input ultraviolet spectrum data to information of multiple water quality parameters. When dealing with multi-objective output problems, traditional neural network models are prone to fall into local optimality, resulting in low accuracy of the output solution. The embodiment of the present application uses a multi-objective particle swarm optimization method to optimize the parameters of the neural network model, improve the performance of the model, avoid falling into local optimality, and improve the accuracy of the model prediction results; use historical water quality data of water samples from a large number of sampling points to train the optimized neural network model to obtain a pre-trained water quality detection model.

[0051] Specifically, the pre-trained water quality detection model learns the mapping relationship between spectral features and water quality parameters through historical data and optimization algorithms, and inputs the ultraviolet spectrum data of the water sample at each sampling point into the model respectively. The model can convert it into corresponding water quality parameter information according to the learned mapping relationship. For example, for a water sample currently collected at a sampling point, its ultraviolet spectrum data is in the range of 200-300nm. It is input into the pre-trained model, and the model outputs chemical oxygen demand of 25mg / L, ammonia nitrogen of 1.2mg / L, and total phosphorus of 0.3mg / L. Through the water quality detection model, the spectral data can be quickly and accurately converted into actual water quality parameters, avoiding the tediousness and time-consuming of traditional chemical analysis methods, and providing an efficient analysis method for water quality monitoring.

[0052] Furthermore, the pre-trained water quality detection model is obtained based on the historical water quality data of the water samples at the sampling points and combined with the multi-objective particle swarm optimization method, including: S801, constructing an initial neural network model according to preset model parameters; S802, initializing a particle swarm according to the initial neural network model and preset particle swarm algorithm parameters to obtain an initial particle swarm, wherein particles in the initial particle swarm represent weights and bias values ​​of the initial neural network model; S803, according to the preset optimization target, the initial particle swarm is updated until the preset maximum number of iterations is reached to obtain the optimal solution set; S804, substituting the optimal solution set into the initial neural network model to obtain a water quality detection model; S805. Use the historical ultraviolet spectrum data of the water samples at the sampling points and the corresponding historical water quality data to train the water quality detection model to obtain a pre-trained water quality detection model.

[0053] In this embodiment, the input layer, hidden layer and output layer of the neural network are designed according to the dimension of the input ultraviolet spectrum data (assuming n wavelength points) and the number of water quality parameters expected to be output (assuming m parameters). For example, the input layer has n nodes, corresponding to the ultraviolet spectrum data of n wavelength points; one or more hidden layers can be set, and the number of nodes in each hidden layer is determined based on experience or experiments, and a hidden layer is set with k nodes; the output layer has m nodes, corresponding to m water quality parameters, and the weights and biases of the neural network are randomly initialized, and the weights and biases are initialized to smaller random numbers to ensure that the network has a certain degree of randomness at the beginning of training, so as to better learn different data patterns and obtain an initial neural network model.

[0054] Specifically, according to preset particle swarm algorithm parameters, including the number of particles, initial particle positions, etc., an initial particle swarm is obtained, wherein each particle in the particle swarm represents a set of weights and bias values ​​of the neural network, and by continuously updating the positions of the particles, a parameter combination that optimizes the performance of the neural network is found. Through parallel search of a large number of particles, a wide range of explorations can be made in the parameter space of the neural network to avoid falling into a local optimal solution, and the global optimal solution can be quickly found, thereby improving the optimization efficiency and effect. According to a preset optimization goal, the positions of the particles are continuously updated, and the speed and position of the particles are adjusted according to the performance of the current position and the global optimal position, so as to guide the particles to move to a better area, and the parameters of the neural network are continuously adjusted until the preset maximum number of iterations is reached, so as to find a set of weights and biases that minimize the prediction error, thereby improving the prediction performance of the model.

[0055] Specifically, the optimal set of weights and biases in the optimal solution set are substituted into the initial neural network model to form an optimized water quality detection model. The model is further trained using the historical ultraviolet spectral data of water samples at the sampling points and the corresponding historical water quality data to enable the model to better fit the data, improve the generalization ability of the model, better process the input ultraviolet spectral data, and improve the accuracy of the model output results.

[0056] Further, such as Figure 5 , the initial particle swarm is updated according to the preset optimization target until the preset maximum number of iterations is reached to obtain the optimal solution set, including: S901, analyzing the solutions corresponding to the particles according to the preset optimization objectives to obtain multiple non-inferior solutions; S902, storing the plurality of non-inferior releases into a pre-created external archive; S903, updating the non-inferior solutions in the external file according to the Euclidean distance between the non-inferior solutions and the preset critical distance; S904, updating the particles in the initial particle swarm until a preset maximum number of iterations is reached and the non-inferior solution satisfies a preset external file size, thereby obtaining an updated non-inferior solution; S905. Obtain an optimal solution set according to the updated non-inferior solutions in the external archive.

[0057] In this embodiment, the optimization target is constructed according to minimizing the prediction error of the neural network model, the prediction error of each water quality parameter is minimized, and the optimization target is obtained. The solution corresponding to the particle is evaluated using the optimization target, and non-inferior solutions are screened out. Non-inferior solutions refer to solutions that do not have other solutions that are superior to them in all objectives at the same time under multiple objectives. By screening non-inferior solutions, it is possible to avoid focusing on the optimality of a single objective and ignoring other important objectives in multi-objective optimization, and multiple excellent options can be provided for the final optimization results. The pre-constructed external archive stores non-inferior solutions, which are used to save the excellent solutions found in each iteration process, and the newly found non-inferior solutions are stored in the external archive, and the different excellent solutions generated in the entire optimization process are retained to avoid the loss of excellent solutions, and provide a reference for the final selection of the optimal solution set.

[0058] Specifically, the Euclidean distance between the non-inferior solutions in the external archive is calculated and compared with the preset critical distance. If the Euclidean distance between two non-inferior solutions is less than the preset critical distance, it means that they are too close in the solution space and are considered to be similar solutions. At this time, the better solution is retained to update the external archive to ensure that the solutions in the archive have sufficient diversity, avoid the non-inferior solutions in the external archive from being too dense, and improve the quality of the final solution.

[0059] Specifically, in the particle swarm algorithm, the position and velocity of the particles are continuously updated until the maximum number of iterations is reached, and a better solution is searched in the solution space. Combined with the update of the external archive, it is ensured that when the maximum number of iterations is reached, the number of non-inferior solutions in the external archive reaches the preset external archive size, making the optimization process more targeted and efficient, and ensuring that enough excellent solutions are finally obtained. For example, the maximum number of iterations is initially set to 1000 times, and the preset size of the external archive is 50 non-inferior solutions. At the 100th iteration, the particles are updated, new non-inferior solutions are calculated and the external archive is updated. This process continues until 1000 iterations are reached and the number of non-inferior solutions in the external archive reaches 50, and an updated non-inferior solution is obtained. The optimal solution set with the best comprehensive performance on all optimization objectives is selected from the updated non-inferior solutions. Combining the search of the particle swarm and the update of the external archive makes the optimization process more systematic and controllable, ensuring that the final non-inferior solution has a certain number and diversity, and avoiding premature convergence and falling into the local optimum.

[0060] Embodiment 2:

[0061] In this embodiment, if Figure 6 , a water quality sampling and monitoring system for a smart water station based on ultraviolet spectroscopy is provided, which is used to implement the water quality sampling and monitoring method for a smart water station based on ultraviolet spectroscopy, comprising: The water area division module divides the smart water station into regions and obtains multiple water areas; The abnormal water area detection module analyzes water quality based on the ultraviolet spectrum data of water samples at the central sampling point of each water area to obtain normal water areas and abnormal water areas; The pollution source sampling point detection module analyzes the water quality of water samples from all sampling points in abnormal water areas and obtains the pollution source sampling points to monitor the water quality of the smart water station in real time.

[0062] In this embodiment, the water area division module includes a geographic information processing unit and a regional division algorithm unit, which divides the area covered by the smart water station into multiple different water areas, providing a basic framework for subsequent water quality monitoring in different areas, making the monitoring work more targeted and systematic; the abnormal water area detection module includes a water sample collection unit, an ultraviolet spectrum collection unit, and a water quality analysis algorithm unit. By performing water quality analysis on the ultraviolet spectrum data of water samples at the central sampling point of each water area, all water areas are divided into normal water areas and abnormal water areas, and areas with water quality problems are quickly located, providing guidance for subsequent key monitoring of abnormal water areas, effectively narrowing the monitoring scope, and improving monitoring efficiency; the pollution source sampling point detection module includes a water sample comprehensive collection unit, an ultraviolet spectrum secondary collection unit, a deep water quality analysis unit, and an abnormal analysis and screening unit. Within the determined abnormal water area, all sampling point water samples are deeply analyzed to accurately find the pollution source sampling point, realize real-time monitoring of the water quality of the smart water station, and provide key support for timely discovery and treatment of pollution sources and ensuring the water quality safety of the water station.

[0063] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A water quality sampling and monitoring method for a smart water station based on ultraviolet spectroscopy, characterized in that: include: The smart water station is divided into regions to obtain multiple water areas; According to the ultraviolet spectrum data of water samples from the central sampling point of each water area, water quality analysis is carried out to obtain normal water areas and abnormal water areas; By analyzing the environmental information of the abnormal water area, the water area boundary is optimized to obtain an updated abnormal water area range, wherein the optimization process includes translation processing and stretching processing; Water quality analysis is performed on water samples from all sampling points within the updated abnormal water area to obtain pollution source sampling points so as to monitor the water quality of the smart water station in real time.

2. The water quality sampling and monitoring method of a smart water station based on ultraviolet spectroscopy according to claim 1 is characterized in that: The water quality analysis is performed based on the ultraviolet spectrum data of the water samples at the central sampling point of each water area to obtain normal water areas and abnormal water areas, including: In the preset sliding window, UV spectrum data of water samples at the central sampling point of each water area is collected to obtain multiple sets of UV spectrum data; By analyzing the absorbance values ​​of the multiple groups of ultraviolet spectral data, corresponding feature vectors are constructed, and the water quality of the water samples at the central sampling point of each water area is analyzed to obtain normal water areas and abnormal water areas.

3. The water quality sampling and monitoring method of a smart water station based on ultraviolet spectroscopy according to claim 2 is characterized in that: In the preset sliding window, ultraviolet spectrum data is collected for water samples at the central sampling point of each water area, and multiple sets of ultraviolet spectrum data are obtained, including: According to the historical UV spectrum data, the sliding window and sliding step size are obtained through the preset window optimization model; According to the sliding step length, ultraviolet spectral data of water samples at each central sampling point of the water area are collected in the sliding window respectively, and the sliding window is filled with ultraviolet spectral data to obtain multiple groups of ultraviolet spectral data.

4. The water quality sampling and monitoring method of a smart water station based on ultraviolet spectroscopy according to claim 2 is characterized in that: By analyzing the absorbance values ​​of the multiple sets of ultraviolet spectrum data, corresponding feature vectors are constructed, and the water quality of the water samples at the central sampling point of each water area is analyzed to obtain normal water areas and abnormal water areas, including: Based on historical UV spectrum data and corresponding water quality data, multiple specific wavelengths related to water quality are obtained; At each of the specific wavelengths, the absorbance value of each set of ultraviolet spectrum data is extracted to obtain a plurality of absorbance values; Combine multiple absorbance values ​​of each group to obtain a characteristic vector of each group of ultraviolet spectrum data; The feature vector is input into a pre-trained water quality classification model, and the water quality of the water sample at the central sampling point of each water area is analyzed to obtain normal water areas and abnormal water areas.

5. The water quality sampling and monitoring method of a smart water station based on ultraviolet spectroscopy according to claim 1 is characterized in that: The water area boundary is optimized by analyzing the environmental information of the abnormal water area to obtain an updated abnormal water area range, wherein the optimization process includes translation process and stretching process, including: According to the preset pressure sensor, the abnormal water area boundary is translated in combination with the water flow pressure data in the abnormal water area to obtain a moving boundary; According to a preset pollution diffusion model, the moving boundary is stretched to obtain an optimized boundary; According to the optimized boundary, an updated abnormal water area range is obtained.

6. The water quality sampling and monitoring method of a smart water station based on ultraviolet spectroscopy according to claim 1 is characterized in that: The water quality analysis of all the water samples at the sampling points within the updated abnormal water area is performed to obtain the pollution source sampling points so as to monitor the water quality of the smart water station in real time, including: Obtain water samples from all sampling points within the updated abnormal water area; In the preset sliding window, the spectrum data of the water sample at each sampling point is collected respectively to obtain the ultraviolet spectrum data of the water samples at multiple sampling points; Through the preset water quality detection model, the ultraviolet spectrum data of the water sample at each sampling point is analyzed to obtain water quality parameter information; Performing an abnormal analysis on the water quality parameter information to obtain normal sampling points and abnormal sampling points; Among the abnormal sampling points, the sampling points with the highest abnormality in water quality parameters are selected as the pollution source sampling points.

7. The water quality sampling and monitoring method of a smart water station based on ultraviolet spectroscopy according to claim 6 is characterized in that: The preset water quality detection model is used to analyze the ultraviolet spectrum data of the water sample at each sampling point to obtain water quality parameter information, including: Based on the historical water quality data of water samples at the sampling points, a pre-trained water quality detection model is obtained in combination with the multi-objective particle swarm optimization method; The ultraviolet spectrum data of the water sample at each sampling point is input into the pre-trained water quality detection model to obtain the water quality parameter information.

8. The water quality sampling and monitoring method of a smart water station based on ultraviolet spectroscopy according to claim 7 is characterized in that: The pre-trained water quality detection model is obtained based on the historical water quality data of the water samples at the sampling points and combined with the multi-objective particle swarm optimization method, including: Construct an initial neural network model according to the preset model parameters; Initializing a particle swarm according to the initial neural network model and preset particle swarm algorithm parameters to obtain an initial particle swarm, wherein particles in the initial particle swarm represent weights and bias values ​​of the initial neural network model; According to a preset optimization goal, the initial particle swarm is updated until a preset maximum number of iterations is reached to obtain an optimal solution set; Substituting the optimal solution set into the initial neural network model to obtain a water quality detection model; The water quality detection model is trained using historical ultraviolet spectrum data of water samples at sampling points and corresponding historical water quality data to obtain a pre-trained water quality detection model.

9. The water quality sampling and monitoring method of a smart water station based on ultraviolet spectroscopy according to claim 8 is characterized in that: The initial particle swarm is updated according to a preset optimization target until a preset maximum number of iterations is reached to obtain an optimal solution set, including: Analyze the solutions corresponding to the particles according to the preset optimization objectives and obtain multiple non-inferior solutions; storing the plurality of non-inferior releases into a pre-created external archive; Updating the non-inferior solutions in the external file by using the Euclidean distance between the non-inferior solutions and a preset critical distance; The particles in the initial particle swarm are updated until the preset maximum number of iterations is reached and the non-inferior solution meets the preset external file size, and an updated non-inferior solution is obtained; According to the updated non-inferior solutions in the external archive, an optimal solution set is obtained.

10. A water quality sampling and monitoring system for a smart water station based on ultraviolet spectroscopy, characterized in that: A method for sampling and monitoring water quality of a smart water station based on ultraviolet spectroscopy according to any one of claims 1 to 9, comprising: The water area division module divides the smart water station into regions and obtains multiple water areas; The abnormal water area detection module analyzes water quality based on the ultraviolet spectrum data of water samples at the central sampling point of each water area to obtain normal water areas and abnormal water areas; The pollution source sampling point detection module optimizes the water area boundary by analyzing the environmental information of the abnormal water area to obtain an updated abnormal water area range, wherein the optimization processing includes translation processing and stretching processing, and performs water quality analysis on water samples of all sampling points within the updated abnormal water area to obtain the pollution source sampling points, so as to monitor the water quality of the smart water station in real time.

Images