Dynamic detection method for water quality of natural mineral water

Abstract

The invention provides a dynamic detection method for water quality of natural mineral water, which comprises the following steps: marking one or more detection points in a detection area according to longitude and latitude, and arranging a plurality of sensor devices and a water sample collection device for data collection; constructing a water quality dynamic prediction model, mapping the change of the environmental data and the change of the water quality data, and when the environmental data is changed, collecting a water sample and detecting the water quality data; after the water quality dynamic prediction model is constructed, water quality data detection is periodically carried out on the detection points, and water quality data at corresponding moments in the water quality dynamic prediction model are corrected; and classifying the predicted water quality data through a random forest algorithm, and detecting the water quality data again when the classification result is that the water quality does not reach the standard. The water quality data of the next moment are predicted by constructing the water quality dynamic prediction model, and the water quality dynamic prediction model is optimized through data correction, so that the prediction accuracy is improved.

Description

Technical Field

[0001] The invention relates to the technical field of water quality safety detection, and in particular to a dynamic detection method for natural mineral water quality. Background Art

[0002] Natural mineral water contains many chemical substances and trace elements that are beneficial to human health.

[0003] Water quality monitoring is the process of monitoring and measuring the types of pollutants in water bodies, the concentrations of various pollutants and their changing trends, and evaluating the water quality, including unpolluted and polluted natural water and various industrial drainage. The main monitoring can be divided into two categories: one is the comprehensive indicators reflecting the water quality, such as temperature, color, turbidity, pH value, conductivity, suspended solids, dissolved oxygen, chemical oxygen demand and biochemical oxygen demand; the other is some toxic substances, such as phenol, cyanide, arsenic, lead, chromium, cadmium, mercury and organic pesticides.

[0004] Currently, existing detection methods mainly rely on laboratories to detect harmful substances or use water quality analyzers to measure some specific physical parameters. However, neither laboratory testing nor water quality analyzer measurements can dynamically detect changes in water quality, let alone predict changes in water quality in complex environments. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a method for dynamic detection of natural mineral water quality, which realizes dynamic detection of water quality by collecting environmental data and water quality data and entering them into a pre-built water quality detection model.

[0006] To achieve the above-mentioned object of the invention, the present invention provides a method for dynamic detection of water quality of natural mineral water, the method comprising:

[0007] One or more detection points are marked in the detection area according to longitude and latitude, and multiple sensor devices and water sample collection devices are set to collect data; the multiple sensor devices record the environmental data in the detection area in real time, and send the environmental data to the data center; the data center controls the water sample collection device to collect water samples at the detection points and detect water quality data through the Internet of Things technology;

[0008] Construct a dynamic water quality prediction model, map the changes in environmental data with the changes in water quality data, collect water samples and test the water quality data when the environmental data changes; after the dynamic water quality prediction model is constructed, periodically test the water quality data at the test points and calibrate the water quality data at the corresponding time in the dynamic water quality prediction model;

[0009] The predicted water quality data is classified by the random forest algorithm. When the classification result is that the water quality does not meet the standard, the water quality data is tested again.

[0010] Furthermore, the environmental data include but are not limited to: time, longitude and latitude, water temperature, rainfall, water level, water flow direction, water flow speed, and pH value;

[0011] The water quality data include but are not limited to: chromaticity, turbidity, taste, odor, foreign matter, element content, organic matter content, inorganic matter content, and microbial content.

[0012] Further, the change of environmental data is mapped with the change of water quality data, including: at time T1, the environmental data is set H1, and the water quality data is set S1; at time T2, the environmental data changes from set H1 to set H2, and the water quality data set S2 at this time is recorded; the environmental data and water quality data at times T1, T2...TK are collected to complete the data mapping of the first adjacent time;

[0013] Based on the data mapping of the first adjacent moments, the water quality data at the next moment is predicted.

[0014] Furthermore, when environmental data changes, water samples are collected and water quality data is tested, including:

[0015] The water temperature value in the detection area is recorded in real time, and the upper and lower water temperature thresholds and the water temperature change thresholds are set. When the water temperature value at a certain moment exceeds the upper and lower water temperature thresholds, or the water temperature change rate before and after exceeds the water temperature change threshold, the data center controls the water sample collection device to collect water samples at the detection point;

[0016] The rainfall and water level in the detection area are recorded in real time. When rainfall occurs or the water level rises, the data center controls the water sample collection device to collect water samples from the detection point after the rainfall stops or the water level rises.

[0017] The water flow direction and water flow velocity in the detection area are recorded in real time, and a flow velocity change threshold is set. When the water flow direction at a certain moment changes, or the change rate of the water flow velocity at the previous and next moments exceeds the flow velocity change threshold, the data center controls the water sample collection device to collect water samples at the detection point;

[0018] The pH value in the detection area is recorded in real time, and the acid-base pH threshold is set. When the pH value at a certain moment exceeds the acid-base pH threshold, the data center controls the water sample collection device to collect water samples at the detection point.

[0019] Furthermore, when the environmental data changes, the first error threshold is set to detect the accuracy of the water quality prediction model by comparing the water quality data at the next moment predicted by the water quality dynamic prediction model and the water quality data at the next moment detected by the water sample collection device;

[0020] If the error between the detected water quality data and the predicted water quality data is less than a first error threshold, the detected water quality data is used as the water quality data at the corresponding moment;

[0021] If the error between the detected water quality data and the predicted water quality data is greater than the first error threshold, data inversion is performed to eliminate invalid data. At the same time, the next moment is taken as moment T11, and environmental data and water quality data at moments T11, T12…T1 K are collected to complete the data mapping of the second adjacent moments.

[0022] Furthermore, after the water quality dynamic prediction model is constructed, a detection cycle is set, and the water sample collection device is controlled to periodically detect water quality data at the detection point, and the water quality data detected by the water sample collection device is compared with the water quality data predicted by the water quality dynamic prediction model, and a second error threshold is set. If the error between the detected water quality data and the predicted water quality data is less than the second error threshold, the detected water quality data is used as the water quality data at the current moment; if the error between the detected water quality data and the predicted water quality data is greater than the second error threshold, data inversion is performed to eliminate invalid data. At the same time, the next moment is taken as moment T21, and environmental data and water quality data at moments T21, T22...T2K are collected to complete the data mapping of the third adjacent moments.

[0023] Furthermore, the water quality data were divided into those that met the water quality standards and those that did not meet the water quality standards through the random forest algorithm;

[0024] The water quality data contains N detection features, and several decision trees are constructed. Each decision tree contains any S detection features, and N>S. Each detection feature is a node of the decision tree, and the node contains two branches: up to standard and not up to standard. The current node is connected to the up-to-standard branch of the previous node; if the result of the current node is not up to standard, the output result water quality is not up to standard; if the result of the current node is up to standard, enter the next node until all nodes meet the standard, and the output result water quality meets the standard.

[0025] Furthermore, the output results of all decision trees are counted, and the result with the most occurrences is used as the classification result of the water quality data.

[0026] Furthermore, when the water quality data at the current moment predicted by the water quality dynamic prediction model is output as water quality not meeting the standard, water samples are collected at the detection point and the water quality data is detected.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] 1. By establishing a mapping between environmental data and water quality data, water quality data can be predicted based on changes in environmental data;

[0029] 2. By collecting data mapping at three different adjacent moments, the data mapping can be supplemented and improved in different seasons or environments;

[0030] 3. By setting the error threshold, when the error exceeds the threshold, the model is optimized to improve the model accuracy;

[0031] 4. The technical solution of this application adopts a random forest algorithm. Since there may be invalid detection features in the predicted water quality data, the random forest algorithm can maintain the accuracy of the classification results even if a large part of the features are lost. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0033] Figure 1 This is a flow chart of a dynamic detection method for natural mineral water quality provided by an embodiment of the present invention;

[0034] Figure 2 It is a model diagram of a decision tree provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0036] Embodiment 1

[0037] This embodiment provides a method for dynamic detection of natural mineral water quality, which constructs a dynamic water quality prediction model to dynamically predict the water quality data at the next moment, and classifies the predicted water quality data through a random forest algorithm to improve the accuracy of the prediction.

[0038] It should be noted that according to the provisions of the "Geological Exploration Specifications for Natural Mineral Water Resources", drinking natural mineral water needs to be evaluated based on sensory requirements, boundary indicators, limit indicators, pollutant indicators and microbial indicators; therapeutic natural mineral water is evaluated based on the exploration specifications for total dissolved solids, carbon dioxide, total hydrogen sulfide, orthosilicic acid, metaboric acid, bromine, iodine, total iron, arsenic, radon, water temperature and other indicators.

[0039] Figure 1This is a flow chart of a method for dynamic detection of natural mineral water quality provided by Embodiment 1 of the present invention, such as Figure 1 As shown, the text-image pedestrian re-identification method for noise association provided in this embodiment is applicable to the case of dynamic water quality detection in a complex environment, and specifically includes the following steps:

[0040] S110: Mark one or more detection points in the detection area according to longitude and latitude, and set up multiple sensor devices and water sample collection devices to collect data; the multiple sensor devices record the environmental data in the detection area in real time, and send the environmental data to the data center; the data center controls the water sample collection device through the Internet of Things technology to collect water samples at the detection points and detect water quality data.

[0041] It should be noted that the sensor device can be equipped with GPS or Beidou positioning, and can be equipped with temperature sensors, rainfall sensors, water level sensors, speed sensors, pH sensors, etc., and can also be equipped with ammonia nitrogen sensors, nitrite sensors, turbidity sensors, salinity sensors, dissolved oxygen sensors, etc. The sensor device can be fixed at a certain point in the detection water area, or can be moved on the water surface with the water flow, or the sensor device can be remotely controlled to move on the water surface, and the collected data can be uploaded to the data center through wireless communication.

[0042] The function of the water sample collection device is to receive instructions from the data center, collect water samples from one or more detection points, and mark different water samples by time and longitude and latitude. The water sample collection device can quickly determine the content values ​​of COD, ammonia nitrogen, total phosphorus, total nitrogen, total chromium, hexavalent chromium, copper, nickel, zinc, iron, phosphate, nitrate, nitrite, chlorine dioxide, residual chlorine, total chlorine, sulfate, sulfide, color, turbidity, suspended matter, silicon dioxide, cyanide, fluoride and other parameters in water. You can also choose to only mark and store water samples and conduct centralized laboratory testing.

[0043] Both the sensor device and the water sample collection device can choose to use the Internet of Things technology. The sensor device and the water sample collection device can be controlled by the Internet of Things to move to the longitude and latitude of the detection point to achieve remote control, saving manpower and material resources, or the sensor device and the water sample collection device can be fixed at the detection point for detection.

[0044] The environmental data in the technical solution of this application include but are not limited to: time, longitude and latitude, water temperature, rainfall, water level, water flow direction, water flow speed, pH value, etc.

[0045] The water quality data in the technical solution of the present application include but are not limited to: chromaticity, turbidity, taste, odor, foreign matter, element content, organic content, inorganic content, microbial content, etc.

[0046] Since water quality data is affected by multiple factors, preferably, multiple factors can be added to the environmental data and principal component analysis can be performed to obtain the main factors that affect water quality changes in the environmental data, or, when one or more items in the environmental data change, one or more items in the corresponding water quality data change, and it is considered that one or more changes in the environmental data cause one or more items in the water quality data to change, so as to establish a corresponding mapping relationship.

[0047] It should be noted that chromaticity, turbidity, taste, odor, and foreign matter refer to whether the water is colorless, odorless and free of foreign matter; element content refers to the content of lithium, strontium, zinc, selenium, arsenic, antimony, lead, cadmium, mercury, tin, copper, barium, chromium, manganese, nickel, silver and other content as well as the content of some radioactive elements; organic matter content refers to the content of volatile phenols, mineral oil and other organic matter; inorganic matter content refers to the content of inorganic matter such as total dissolved solids, free carbon dioxide, orthosilicic acid, hydrogen sulfide, orthosilicic acid, metaboric acid, fluoride, cyanide and other inorganic matter; microbial content refers to the content of microorganisms such as Escherichia coli, fecal Streptococcus, Pseudomonas aeruginosa, and Clostridium perfringens.

[0048] S120: Construct a dynamic water quality prediction model, map the changes in environmental data with the changes in water quality data, collect water samples when the environmental data changes, and test the water quality data; after the dynamic water quality prediction model is constructed, periodically test the water quality data at the test points and correct the water quality data at the corresponding time in the dynamic water quality prediction model.

[0049] It should be noted that since water quality data is affected by the seasons, the rainfall and evaporation in the detection area are different in different seasons, resulting in little reference between the water quality data collected in different seasons. That is, only collecting environmental data and water quality data within a certain period is not enough to predict water quality data in a complex environment. For this reason, the technical solution of the present application sets up data mapping at three adjacent moments to supplement the reference data.

[0050] Specifically, the data mapping of the first adjacent moments is collected at random moments, the data mapping of the second adjacent moments is performed when there is a large error between the predicted water quality data and the actual water quality data when the environment changes, and the data mapping of the third adjacent moments occurs during the periodic detection process, when there is a large error between the water quality data predicted by the periodic detection and the actual water quality data.

[0051] Mapping the changes in environmental data with the changes in water quality data includes: at time T1, the environmental data is set H1, and the water quality data is set S1; at time T2, the environmental data changes from set H1 to set H2, and the water quality data set S2 at this time is recorded; the environmental data and water quality data at times T1, T2...TK are collected to complete the data mapping of the first adjacent time;

[0052] Based on the data mapping of the first adjacent moments, the water quality data at the next moment is predicted.

[0053] For ease of understanding, for example: the interval between adjacent moments is one day, that is, T2 and T1 are 24 hours apart. At T1, the environmental data is set H1, and the water quality data is set S1. At T2, the environmental data is set H2, and the water quality data is set S2. By collecting environmental data and water quality data at T1, T2...TK, dynamic predictions are made for environmental data and water quality data at future moments. For example: at TK, its environmental data and water quality data are similar to those at T3, and the environmental data at the next moment is similar to those at T4, then the water quality data at T4 is used as the water quality data at the next moment of TK.

[0054] There is an error between the water quality data predicted by the first data mapping at adjacent moments and the actual water quality data. Therefore, the second and third data mapping at adjacent moments are required to correct the data.

[0055] When environmental data changes, water samples are collected and water quality data is tested, including:

[0056] The water temperature value in the detection area is recorded in real time, and the upper and lower water temperature thresholds and the water temperature change thresholds are set. When the water temperature value at a certain moment exceeds the upper and lower water temperature thresholds, or the water temperature change rate before and after exceeds the water temperature change threshold, the data center controls the water sample collection device to collect water samples at the detection point;

[0057] The rainfall and water level in the detection area are recorded in real time. When rainfall occurs or the water level rises, the data center controls the water sample collection device to collect water samples from the detection point after the rainfall stops or the water level rises.

[0058] The water flow direction and water flow velocity in the detection area are recorded in real time, and a flow velocity change threshold is set. When the water flow direction at a certain moment changes, or the change rate of the water flow velocity at the previous and next moments exceeds the flow velocity change threshold, the data center controls the water sample collection device to collect water samples at the detection point;

[0059] The pH value in the detection area is recorded in real time, and the acid-base pH threshold is set. When the pH value at a certain moment exceeds the acid-base pH threshold, the data center controls the water sample collection device to collect water samples at the detection point.

[0060] When the environmental data changes, the first error threshold is set to detect the accuracy of the dynamic water quality prediction model by comparing the water quality data at the next moment predicted by the dynamic water quality prediction model and the water quality data at the next moment detected by the water sample collection device;

[0061] If the error between the detected water quality data and the predicted water quality data is less than a first error threshold, the detected water quality data is used as the water quality data at the corresponding moment;

[0062] If the error between the detected water quality data and the predicted water quality data is greater than the first error threshold, data inversion is performed to eliminate invalid data. At the same time, the next moment is taken as moment T11, and environmental data and water quality data at moments T11, T12…T1 K are collected to complete the data mapping of the second adjacent moments.

[0063] The process of data inversion is to infer the water quality data of the previous moment through the environmental data and water quality data of the current moment, and according to the environmental data of the previous moment, which is exactly the opposite of the prediction process of water quality data. Take the time from T3 to T5 as an example, for example: the environmental data and water quality data of the time T3 have been detected as H3 and S3, the environmental data of the time T4 is H4, and the water quality data of the time T4 is predicted to be S4, the environmental data of the time T5 is H5, and the water quality data of the time T5 is predicted to be S5. After the water quality test at the time T5, it is found that there is a large error between the real water quality data L5 at the time T5 and the predicted water quality data S5. Then, through H5 and L5 at the time T5, combined with H4 at the time T4, another predicted water quality data L4 at the time T4 is inverted. Similarly, the predicted water quality data L3 at the time T3 is obtained. If the error between L3 and S3 is small, S3, L4, and L5 are added to the data mapping of adjacent moments; if the error between L3 and S3 is large, the water quality data at the time T4 is eliminated as invalid data.

[0064] That is, there are real water quality data in the data mapping, and there are also predicted water quality data whose errors meet the requirements.

[0065] It should be noted that when the error between the detected water quality data and the predicted water quality data is greater than the first error threshold, it means that the reference data of T1, T2...TK are no longer applicable to the water quality data prediction under the current complex environment. Therefore, the second data mapping of adjacent moments is started to adapt to or supplement the water quality data prediction under the current complex environment.

[0066] After the water quality dynamic prediction model is constructed, a detection cycle is set, and the water sample collection device is controlled to periodically detect water quality data at the detection point. The water quality data detected by the water sample collection device is compared with the water quality data predicted by the water quality dynamic prediction model, and a second error threshold is set. If the error between the detected water quality data and the predicted water quality data is less than the second error threshold, the detected water quality data is used as the water quality data at the current moment; if the error between the detected water quality data and the predicted water quality data is greater than the second error threshold, data inversion is performed to eliminate invalid data. At the same time, the next moment is taken as moment T21, and environmental data and water quality data at moments T21, T22...T2K are collected to complete the data mapping of the third adjacent moments.

[0067] The data mapping of the third adjacent time is similar to the data mapping of the second adjacent time, but the second is based on the change of environmental data, and the third is a periodic detection. The purpose is to supplement the water quality data prediction in the current complex environment and reduce the contingency.

[0068] S130: Classify the predicted water quality data using a random forest algorithm. When the classification result indicates that the water quality does not meet the standard, perform water quality data detection again.

[0069] Random forest is an integrated algorithm that combines multiple weak classifiers. The final result is voted or averaged, so that the overall model has high accuracy and generalization performance. It can achieve good results, mainly due to "random" and "forest", one makes it resistant to overfitting, and the other makes it more accurate. It can produce very high-dimensional (many features) data without dimensionality reduction or feature selection.

[0070] Random forest is chosen not only because it can determine the importance of features, determine the mutual influence between different features, has a relatively fast training speed, is easy to make into a parallel method, and is relatively simple to implement, but also for unbalanced data sets, it can balance errors and maintain accuracy if a large part of the features are missing.

[0071] Furthermore, the output results of all decision trees are counted, and the result with the most occurrences is used as the classification result of the water quality data.

[0072] Random forest belongs to the bootstrap aggregation method, which is an ensemble technique that trains classifiers by reselecting k new data sets from the original data set through sampling with replacement. It uses a collection of trained classifiers to classify new samples, and then uses majority voting or the method of averaging the output to count the classification results of all classifiers. For classification problems, simple voting is usually used, and the category or one of the categories that receives the most votes is the final model output.

[0073] The water quality data is divided into those that meet the water quality standards and those that do not meet the water quality standards through the random forest algorithm.

[0074] The water quality data contains N detection features, and several decision trees are constructed. Each decision tree contains any S detection features, and N>S. Each detection feature is a node of the decision tree, and the node contains two branches: up to standard and not up to standard. The current node is connected to the up-to-standard branch of the previous node; if the result of the current node is not up to standard, the output result water quality is not up to standard; if the result of the current node is up to standard, enter the next node until all nodes meet the standard, and the output result water quality meets the standard.

[0075] Figure 2 A model diagram of a decision tree provided in the first embodiment of the present invention is shown as follows: Figure 2 As shown, taking two features as an example, feature 1 and feature 2 are nodes of the decision tree. Feature 1 has two branches: meeting the standard and not meeting the standard. The meeting branch is connected to node feature 2, and the not meeting the standard branch is connected to the output result that the water quality does not meet the standard.

[0076] If feature 1 corresponding to the water quality data meets the water quality index, feature 2 will be used to continue the judgment. If feature 1 does not meet the water quality index, the direct output result is that the water quality does not meet the standard; if both feature 1 and feature 2 meet the standard, the output result is that the water quality meets the standard.

[0077] When the water quality data at the current moment predicted by the water quality dynamic prediction model is output as water quality not meeting the standard, water samples are collected at the detection point and the water quality data is detected.

[0078] The technical solution of this application aims to dynamically predict whether the water quality is polluted and whether it meets the drinking standards, rather than to measure the water quality data in real time and accurately. Therefore, a slight deviation between the predicted data and the actual data is allowed without affecting the judgment of whether the water quality meets the standards. The dynamic water quality prediction model provides rich experience for water quality prediction by storing a large amount of historical test data and historical prediction data with errors that meet the requirements.

[0079] The dynamic water quality prediction model not only dynamically predicts water quality data, but also continuously updates itself over time to cope with data prediction in complex environments.

[0080] This embodiment provides a method for dynamic detection of natural mineral water quality, which collects environmental data and water quality data through a sensor device and a water sample collection device, and constructs a water quality dynamic prediction model to predict water quality data using the real data that has been collected and the environmental data at each moment. In the process of water quality prediction, the water quality dynamic prediction model is updated and optimized through environmental changes or periodic detection, and finally the predicted water quality data is classified through a random forest algorithm. Even if there are some invalid features in the predicted data, the accuracy of the prediction can still be maintained.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A dynamic detection method for natural mineral water quality, characterized in that: The method comprises: One or more detection points are marked in the detection area according to longitude and latitude, and multiple sensor devices and water sample collection devices are set to collect data; the multiple sensor devices record the environmental data in the detection area in real time, and send the environmental data to the data center; the data center controls the water sample collection device to collect water samples at the detection points and detect water quality data through the Internet of Things technology; Construct a dynamic water quality prediction model, map the changes in environmental data with the changes in water quality data, collect water samples and test the water quality data when the environmental data changes; after the dynamic water quality prediction model is constructed, periodically test the water quality data at the test points and calibrate the water quality data at the corresponding time in the dynamic water quality prediction model; The predicted water quality data is classified by the random forest algorithm. When the classification result is that the water quality does not meet the standard, the water quality data is tested again.

2. A dynamic detection method for natural mineral water quality according to claim 1, characterized in that: The environmental data include but are not limited to: time, longitude and latitude, water temperature, rainfall, water level, water flow direction, water flow speed, and pH value; The water quality data include but are not limited to: chromaticity, turbidity, taste, odor, foreign matter, element content, organic matter content, inorganic matter content, and microbial content.

3. A dynamic detection method for natural mineral water quality according to claim 2, characterized in that: Mapping the changes in environmental data with the changes in water quality data includes: at time T1, the environmental data is set H1, and the water quality data is set S1; at time T2, the environmental data changes from set H1 to set H2, and the water quality data set S2 at this time is recorded; the environmental data and water quality data at times T1, T2...TK are collected to complete the data mapping of the first adjacent time; Based on the data mapping of the first adjacent moments, the water quality data at the next moment is predicted.

4. A dynamic detection method for natural mineral water quality according to claim 3, characterized in that: When environmental data changes, water samples are collected and water quality data is tested, including: The water temperature value in the detection area is recorded in real time, and the upper and lower water temperature thresholds and the water temperature change thresholds are set. When the water temperature value at a certain moment exceeds the upper and lower water temperature thresholds, or the water temperature change rate before and after exceeds the water temperature change threshold, the data center controls the water sample collection device to collect water samples at the detection point; The rainfall and water level in the detection area are recorded in real time. When rainfall occurs or the water level rises, the data center controls the water sample collection device to collect water samples from the detection point after the rainfall stops or the water level rises. The water flow direction and water flow velocity in the detection area are recorded in real time, and a flow velocity change threshold is set. When the water flow direction at a certain moment changes, or the change rate of the water flow velocity at the previous and next moments exceeds the flow velocity change threshold, the data center controls the water sample collection device to collect water samples at the detection point; The pH value in the detection area is recorded in real time, and the acid-base pH threshold is set. When the pH value at a certain moment exceeds the acid-base pH threshold, the data center controls the water sample collection device to collect water samples at the detection point.

5. A dynamic detection method for natural mineral water quality according to claim 4, characterized in that: When the environmental data changes, the first error threshold is set to detect the accuracy of the dynamic water quality prediction model by comparing the water quality data at the next moment predicted by the dynamic water quality prediction model and the water quality data at the next moment detected by the water sample collection device; If the error between the detected water quality data and the predicted water quality data is less than a first error threshold, the detected water quality data is used as the water quality data at the corresponding moment; If the error between the detected water quality data and the predicted water quality data is greater than the first error threshold, data inversion is performed to eliminate invalid data. At the same time, the next moment is taken as moment T11, and environmental data and water quality data at moments T11, T12…T1 K are collected to complete the data mapping of the second adjacent moments.

6. A method for dynamic detection of natural mineral water quality according to claim 5, characterized in that: After the water quality dynamic prediction model is constructed, a detection cycle is set, and the water sample collection device is controlled to periodically detect water quality data at the detection point. The water quality data detected by the water sample collection device is compared with the water quality data predicted by the water quality dynamic prediction model, and a second error threshold is set. If the error between the detected water quality data and the predicted water quality data is less than the second error threshold, the detected water quality data is used as the water quality data at the current moment; if the error between the detected water quality data and the predicted water quality data is greater than the second error threshold, data inversion is performed to eliminate invalid data. At the same time, the next moment is taken as moment T21, and environmental data and water quality data at moments T21, T22...T2K are collected to complete the data mapping of the third adjacent moments.

7. A method for dynamic detection of natural mineral water quality according to claim 6, characterized in that: The water quality data is divided into those that meet the water quality standards and those that do not meet the water quality standards through the random forest algorithm. The water quality data contains N detection features, and several decision trees are constructed. Each decision tree contains any S detection features, and N>S. Each detection feature is a node of the decision tree. The node contains two branches: reaching the standard and not reaching the standard. The current node is connected to the reaching branch of the previous node. If the result of the current node is not up to standard, the output result water quality is not up to standard; if the result of the current node is up to standard, it will enter the next node until all nodes are up to standard, and the output result water quality is up to standard.

8. A method for dynamic detection of natural mineral water quality according to claim 7, characterized in that: The output results of all decision trees are counted, and the result with the most occurrences is taken as the classification result of water quality data.

9. A method for dynamic detection of natural mineral water quality according to claim 8, characterized in that: When the water quality data at the current moment predicted by the water quality dynamic prediction model is output as water quality not meeting the standard, water samples are collected at the detection point and the water quality data is detected.

Images