Water drainage monitoring and analyzing system based on artificial intelligence

Through the artificial intelligence-based drainage monitoring and analysis system, the water quality analysis model and fragmented processing module are used to analyze and pre-process the drainage monitoring data, which solves the deviation problem of the existing system when processing multi-point and multi-time data, and improves the accuracy of analysis and reliability of early warning.

CN119066434APending Publication Date: 2024-12-03HANGZHOU LINGTU INFORMATION TECH CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411555579.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-04
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

When the existing drainage monitoring system processes and analyzes water quality monitoring data for multiple locations and multiple periods, data processing deviations are prone to occur, resulting in inaccurate analysis and false alarms of water quality data, affecting subsequent water quality monitoring and management.

Method used

Adopt a drainage monitoring and analysis system based on artificial intelligence, train the water quality analysis model by obtaining the national water quality standard index data set, collect and classify monitoring data in real time, conduct hierarchical verification, judge the false alarm status of the early warning module, and use the sharding processing module to match the Data is preprocessed and classified to improve data processing efficiency and analysis accuracy.

Benefits of technology

Achieved a refined analysis of water quality data for multiple locations and multiple periods of drainage is improved, data processing efficiency is avoided, computational load is too large, water quality monitoring is enhanced, false alarms are reduced, and more stable warning information.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066434A_ABST
    Figure CN119066434A_ABST
Patent Text Reader

Abstract

The invention discloses a drainage monitoring and analysis system based on artificial intelligence, and relates to the technical field of data monitoring, the drainage monitoring and analysis system based on artificial intelligence classifies and divides drainage water quality monitoring data collected at multiple sites and multiple periods through a fragmentation processing module, improves data processing efficiency, reduces calculation load, and improves the data processing efficiency. Water quality characteristics can be identified conveniently, and a structured basis is provided for data mapping interaction. Different data sets are subjected to mapping interaction, association of data in different places and time periods is enhanced, multi-dimensional data input is provided, and the recognition capability of a complex model mode is enhanced. The model output is subjected to hierarchical verification, the credibility is improved, the model error is detected through verification, the data quality is ensured, the analysis deviation is avoided, and the model is effectively adjusted. Through multiple times of training, the model generalization is enhanced, the adaptability is improved, overfitting is reduced, a stable and reliable prediction result is provided, and the method is suitable for wide application scenarios.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of data monitoring, and in particular to a drainage monitoring and analysis system based on artificial intelligence. Background Art

[0002] The acceleration of urbanization will inevitably bring greater challenges to urban drainage and sewage treatment. The drainage monitoring and analysis system based on artificial intelligence will improve the efficiency and sustainability of urban drainage and sewage treatment. The traditional drainage monitoring system integrates and centrally supervises the monitoring data of the city's built and scattered sewage treatment plants, pumping stations, overflows, river water quality, etc., but it is often difficult for the system to achieve multi-point, multi-time and multi-indicator drainage water quality monitoring data for simultaneous comparison, analysis and evaluation. It is inevitable that due to the excessive load of processing multiple data at the same time, there will be deviations in data processing, resulting in inaccurate data analysis. At the same time, the system will face false reporting of water quality data, making it difficult to further carry out and supervise subsequent water quality monitoring work;

[0003] The system uses intelligent analysis modules, bandwidth allocation modules, intelligent monitoring modules, security management modules and user terminals to intelligently control wireless network signal transmission and prioritize the connection stability of key equipment. At the same time, the system uses the security risk intelligent monitoring and management function to monitor spectrum data transmission through real-time data, and finally uses user terminals to enable users to independently view network status and adjust device settings, thereby promoting the development of drainage monitoring and analysis systems based on artificial intelligence.

[0004] To sum up, how to prevent the system from processing and analyzing too much data, which may lead to inaccurate analysis data and thus deviation, is an issue that needs to be urgently solved and optimized in drainage monitoring and analysis systems based on artificial intelligence. Summary of the invention

[0005] The present invention provides a drainage monitoring and analysis system based on artificial intelligence, which solves the technical problem in the related art of how to enable the system to avoid processing and analyzing too much data, resulting in inaccurate analysis data and causing deviations.

[0006] In order to solve the above technical problems, the present invention provides a drainage monitoring and analysis system based on artificial intelligence, and the specific technical solution is as follows:

[0007] In a first aspect, a drainage monitoring and analysis system based on artificial intelligence comprises the following steps:

[0008] Obtain a national water quality standard indicator data set, input the national water quality standard indicator data set into a constructed training set to train a water quality analysis model, and output a water quality status result;

[0009] Collect water quality monitoring data from multiple locations and multiple time periods in real time, and classify and divide the water quality monitoring data to obtain a classified data set; the classified data set includes drainage water quality grade data corresponding to multiple locations and multiple time periods;

[0010] Based on the water quality analysis model and the classified data set acquired in real time, an initial analysis data set is obtained; the initial analysis data set includes water quality abnormal state data existing in the classified data set acquired in real time;

[0011] The initial analysis data set is transmitted to the early warning module for receiving. When the initial analysis data set contains abnormal water quality status data, the initial analysis data set is hierarchically verified to obtain a verification data set. The fluctuation trend of multiple abnormal data points is obtained according to the verification data set. The false alarm situation of the early warning module is judged according to the fluctuation trend of multiple abnormal data points. When no false alarm situation occurs, alarm information is obtained and an alarm is issued.

[0012] As a further optimization scheme of the present invention, it also includes:

[0013] Data collection module, slicing processing module, intelligent analysis module and early warning module to intelligently monitor drainage water quality data at multiple locations and over multiple time periods;

[0014] The data acquisition module acquires the water quality monitoring data from each monitoring point in real time to obtain a water quality monitoring data set; based on the water quality monitoring data set, the data acquisition module transmits the water quality monitoring data set in real time to the slicing processing module for classification and division processing to obtain the classified data set; the classified data set includes a first classified data set and a second classified data set; the first classified data set includes drainage water quality monitoring data of multiple different bases; the second classified data set includes monitoring data of multiple different time periods

[0015] The intelligent analysis module outputs the initial analysis data set; the initial analysis data set includes a first analysis data set and a second analysis data set, based on the first analysis data set and the second analysis data set, the first analysis data set includes a plurality of qualified water quality data, and the second analysis data includes abnormal water quality data;

[0016] Based on the first analysis data set and the second analysis data set, hierarchical verification is performed to obtain a water quality analysis verification data set; based on the water quality analysis verification data set, a fluctuation trend of multiple abnormal data points is obtained, and based on the fluctuation trend of the multiple abnormal data points, the false alarm situation of the early warning module is judged.

[0017] As a further optimization solution of the present invention, the shard processing module further includes:

[0018] Acquire the water quality monitoring data set in real time based on the data acquisition module;

[0019] The water quality monitoring data set is preprocessed by the slice processing module to obtain a preprocessed data set; based on the preprocessed data, the preprocessed data set is classified and divided.

[0020] As a further optimization solution of the present invention, the intelligent analysis module further includes:

[0021] Water quality analysis unit and multi-level verification unit;

[0022] Inputting the national water quality standard index data set into the water quality analysis unit, the water quality analysis unit receives the national water quality standard index data set to construct a training set, and then inputs it into the water quality analysis model for model training;

[0023] Inputting the classified data set acquired in real time into the water quality analysis model for analysis to obtain the initial analysis data set, and decomposing the initial analysis data set into the first analysis data set and the second analysis data set;

[0024] Based on the first analysis data set and the second analysis data set, the first analysis data set and the second analysis data set are input into the multi-level verification unit for water quality detection data verification to obtain a verification result.

[0025] As a further optimization scheme of the present invention, the water quality monitoring data set is preprocessed by the slice processing module to obtain a preprocessed data set; and a method for classifying and dividing the preprocessed data set includes:

[0026] Based on the acquired water quality monitoring data set, removing abnormal interference data from the water quality monitoring data set and completing missing values ​​to obtain a preprocessed data set;

[0027] The preprocessed data set is further standardized to obtain a dimensionless preprocessed data set; the dimensionless preprocessed data set includes a plurality of water quality monitoring data with unified standards after multiple preprocessing;

[0028] Based on the dimensionless preprocessed data set, the dimensionless preprocessed data set is set as the water quality monitoring data set P, then P={C i |a1,a2,…,a n}; where a represents the data attributes of water quality monitoring samples at multiple bases, and n represents the number of samples obtained from water quality monitoring bases; C i Indicates the attribute level of sample data, i=1,2,3,4,5;

[0029] according to: ; to obtain the attribute level of water quality detection sample data from multiple bases, where α is the contribution rate of the characteristic attribute, α=0.83; N represents the number of attributes; |D| represents the total number of samples in the water quality monitoring data set P, |D ci | indicates that the category value in the water quality monitoring data set P is c i The number of samples;

[0030] By setting a classification threshold, classification and division are performed based on the attribute levels of the water quality detection sample data of the multiple bases to obtain a first classified data set.

[0031] As a further optimization scheme of the present invention, the method for classifying and dividing the preprocessed data set to obtain a classified molecular data set also includes:

[0032] Based on the dimensionless preprocessed data set, the dimensionless preprocessed data set is set as the drainage water quality monitoring data matrix X, then X={C i |t1, t2, …, t j}; where t represents the data attributes of water quality monitoring samples in multiple time periods, and j represents the number of samples obtained during the water quality monitoring period; C i Indicates the attribute level of sample data, i=1,2,3,4,5;

[0033] according to: ; To obtain the attribute level of water quality test sample data in multiple time periods; where |D ci,tj | indicates that the category in the water quality monitoring dataset P is c i And the multi-period attribute t j The total number of samples of the value, |D ci | indicates that the category value in the water quality monitoring data set P is c i The number of samples, N j Indicates t j The number of possible values;

[0034] By setting a classification threshold, classification and division are performed based on the attribute levels of the water quality detection sample data in the multiple time periods to obtain a second classified data set.

[0035] As a further optimization solution of the present invention, a method for acquiring a classified data set by performing mapping interaction based on the first classified data set and the second classified data set includes:

[0036] Based on the first classified data set and the second classified data set; by: ; Map and interactively integrate the first classification data set and the second classification data set to obtain a classification data set; where x k+1represents the mapping parameter X at the k+1th control node, that is: ; In the formula, k represents the control parameter, that is, k=2, X k ∈(0,1).

[0037] As a further optimization scheme of the present invention, the national water quality standard indicator data set is input into the water quality analysis subunit, and the water quality analysis subunit receives the national water quality standard indicator data set to construct a training set, and then inputs it into the water quality analysis model for model training, comprising:

[0038] The training set is set as an observation vector S, and the observation vector S is encoded into sequence data; the sequence data is input into the water quality analysis model, and the recognition result representing the water quality analysis is output, as follows:

[0039]

[0040] Among them, W u , W r , W c represents the weight parameter, b u , b r , b c represents the bias parameter, represents the dot product, u (t) 、r (t) and c (t) Represent the intermediate states of the first, second, and third hidden layers respectively, S is the observation vector, X (t) represents the tth sequence data item of the sequence data, H (t) and H (t-1) They represent the t-th and t-1-th second hidden state data respectively, n≥t≥1, n represents the total number of input sequence data items, and H (t-1) =X (t) , tanh is the hyperbolic tangent function, σ represents the sigmoid function;

[0041] Based on the trained water quality analysis model, the sequence data of the classified molecular data set obtained in real time is input into the water quality analysis model to output the drainage water quality monitoring data state of the classified molecular data set obtained in real time. The output layer is specifically as follows:

[0042]

[0043] Among them, y t represents the output vector of the t-th intermediate period, where the components of the output vector represent the data items of the input classified molecular data set belonging to the i-th state, and h n Indicates the nth intermediate state represents data, W y is the weight parameter, b yis the bias parameter, and σ represents the sigmoid function.

[0044] As a further optimization scheme of the present invention, based on the first analysis data set and the second analysis data set, the first analysis data set and the second analysis data set are input into the multi-level verification unit for water quality detection data verification to obtain a verification result, comprising:

[0045] The multi-level verification unit establishes a first verification level, wherein the first verification level includes setting a first analysis data set as D={D1, D2, ..., D n}; The first analysis data set is d={d1, d2, ..., d n}; where D n Indicates the nth first analysis data, d n Indicates the nth second analysis data, and sets a verification mark range [L,U], where L indicates the lowest point of data deviation and U indicates the highest point of data deviation;

[0046] Based on the first verification level, the multi-level verification unit establishes a second verification level, wherein the second verification level: ; In the formula, F is the effective value of the data;

[0047] Based on the second verification level, the multi-level verification unit establishes a third verification level, wherein the third verification level: ; to verify the accuracy of the first analysis data set D and the second analysis data set d, where TP represents the true verification result, TN represents the hypothesis verification result, FP represents the actual verification result, and FN represents the actual opposite verification result.

[0048] As a further optimization scheme of the present invention, the fluctuation trend of multiple abnormal data points is obtained according to the verification data set, and the false alarm situation of the early warning module is judged according to the fluctuation trend of multiple abnormal data points. When no false alarm situation occurs, the method of obtaining alarm information and issuing an alarm includes:

[0049] Based on the verification result, the first verification data set and the second verification data set are verified multiple times. When the accuracy parameters of the multiple verification results show a large fluctuation trend, and the fluctuation amplitudes differ greatly each time, the early warning module warns of abnormality. If the early warning module warns of abnormality, a false alarm occurs.

[0050] When the accuracy parameters of multiple verification results show a small fluctuation trend, the warning module warns normally; when the warning module warns normally, the verification data set is input into the warning module; the verification data set triggers the warning threshold of the warning module, and the warning module triggers an alarm.

[0051] The present invention has at least the following beneficial effects: the drainage monitoring and analysis system of the present invention can realize the analysis and monitoring of drainage water quality data at multiple bases and multiple time periods, and can classify and divide the drainage water quality monitoring data collected at multiple bases and multiple time periods through the sharding processing module, thereby improving the efficiency of data processing and avoiding excessive computing load caused by large amounts of data. At the same time, it is easy to identify water quality characteristics under specific conditions, which helps to conduct more refined data analysis, lays the foundation for subsequent data mapping interactions, and organizes the data into a structured form.

[0052] The first classified data set and the second classified data set are mapped and interacted to obtain a classified data set; the correlation between multi-period data in different locations is enhanced, multi-dimensional data input is provided for the model, and the model's ability to recognize complex patterns is enhanced.

[0053] The newly acquired initial analysis data set of drainage water quality monitoring data output by the water quality analysis model is subjected to hierarchical verification to obtain a verification data set; the credibility of the analysis model is improved, possible errors of the model are detected through multiple verification levels, data quality is ensured, analysis deviations due to data errors are avoided, and a feedback loop is provided so that the model can be effectively adjusted through verification results.

[0054] The false alarm abnormal status of the early warning module is judged according to the verification data set, thereby enhancing the accuracy of the alarm monitoring data of the early warning module, providing more stable and reliable prediction results, and being suitable for a wider range of practical application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a schematic diagram of a drainage monitoring and analysis system flow based on artificial intelligence provided by an embodiment of the present invention;

[0056] Figure 2 It is a schematic diagram of a method flow of a drainage monitoring and analysis system based on artificial intelligence provided by an embodiment of the present invention;

[0057] Figure 3 It is a schematic diagram of a normal verification band of a verification data set provided by an embodiment of the present invention;

[0058] Figure 4 It is a schematic diagram of abnormal verification bands of a verification data set provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The present application is further described in detail below in conjunction with the accompanying drawings. It is necessary to point out here that the following specific implementation methods are only used to further illustrate the present application and cannot be understood as limiting the scope of protection of the present application. Technical personnel in this field can make some non-essential improvements and adjustments to the present application based on the above application content.

[0060] The present invention provides a drainage monitoring and analysis system based on artificial intelligence, and the specific implementation methods are as follows:

[0061] like Figure 1 As shown, a drainage monitoring and analysis system based on artificial intelligence includes:

[0062] Obtain a national water quality standard indicator data set, input the national water quality standard indicator data set into a constructed training set to train a water quality analysis model, and output a water quality status result;

[0063] Collect water quality monitoring data from multiple locations and multiple time periods in real time, and classify and divide the water quality monitoring data to obtain a classified data set; the classified data set includes drainage water quality grade data corresponding to multiple locations and multiple time periods;

[0064] Based on the water quality analysis model and the classified data set acquired in real time, an initial analysis data set is obtained; the initial analysis data set includes water quality abnormal state data existing in the classified data set acquired in real time;

[0065] The initial analysis data set is transmitted to the early warning module for receiving. When the initial analysis data set contains abnormal water quality status data, the initial analysis data set is hierarchically verified to obtain a verification data set. The fluctuation trend of multiple abnormal data points is obtained according to the verification data set. The false alarm situation of the early warning module is judged according to the fluctuation trend of multiple abnormal data points. When no false alarm situation occurs, alarm information is obtained and an alarm is issued.

[0066] In an embodiment of the present invention, the sharding processing module first collects and classifies the drainage water quality monitoring data of different locations, including the city's built and dispersed sewage treatment plants, pumping stations, overflow outlets, rivers, etc. This means that it preliminarily divides the data according to the geographical location (such as different water stations or drainage outlets), which is called the first classified data set; at the same time, it also divides the monitoring data of different time periods at different locations to form a second classified data set. These time periods can be water quality monitoring data at different time intervals such as hours, days, and weeks. This division helps to improve the efficiency of data processing, because after separating complex data sets by time and place, it can be analyzed more clearly to ensure that data of different dimensions will not be confused, and through the dual division of locations and time periods, the monitored water quality can be flexibly tracked regionally and temporally, which is convenient for discovering regional differences and time trends.

[0067] The mapping interaction based on the first and second classified data sets involves mapping and interacting the above two classified data sets. The so-called mapping interaction is to associate data from different locations and different time periods to form a classified data set containing comprehensive information. This interaction can help combine data from a specific time period and a specific location, thereby providing a more detailed picture of water quality changes. Through mapping interaction, not only can water quality changes be viewed by location, but also water quality development can be monitored by time period within the same location. For water quality managers, this can more intuitively understand the spatial and temporal changes in water quality, so as to take targeted measures.

[0068] The classified data set is transmitted to the intelligent analysis module to build a training set. The classified data set is transmitted to the intelligent analysis module, which is used to build a water quality analysis model. The intelligent analysis module trains a machine learning model based on the transmitted data, enabling it to "learn" the water quality change patterns at different water quality monitoring points and different time periods;

[0069] By building a model, the system can automatically identify potential patterns in water quality changes. For example, water quality may deteriorate in certain areas during certain periods of time, or pollutant emissions from certain drainage outlets may have periodic characteristics. Based on these patterns, more effective water quality monitoring and early warning can be carried out.

[0070] Input the real-time classified molecular data set into the water quality analysis model to obtain the initial analysis data set. After the water quality analysis model training is completed, the system can input the real-time water quality data (i.e., the new classified molecular data set) into the model to conduct a preliminary water quality analysis and form an initial analysis data set. This initial set is the model's first round of judgment on the current water quality. Through this real-time analysis, a rapid response to water quality can be achieved, helping water quality managers to detect abnormal conditions or pollution incidents early during the monitoring process.

[0071] After obtaining the preliminary analysis results, the system will conduct a hierarchical verification of the initial analysis data set to obtain the verification data set. This verification process will review the output of the model according to the established standards and known water quality laws to produce a more reliable verification data set. This verification mechanism can avoid misjudgment of the model, further improve the accuracy and credibility of water quality analysis, and ensure that the final analysis results are closer to the actual water quality situation.

[0072] The validation data set is input into the model again for testing, and the generalization ability of the training model is tested. The validation data set will be input into the water quality analysis model again for testing. Through multiple training and testing, the model will continuously adjust its weights and parameters, gradually improve its ability to analyze different water quality conditions, and enhance the generalization ability of the model, that is, the ability to deal with data from different locations and different time periods.

[0073] Multiple training sessions help the model not only process the currently collected data, but also maintain high-precision analysis when faced with new water quality data in the future. This type of training can reduce the overfitting of the model to special cases and improve its application in a wider range of water quality monitoring environments.

[0074] In a preferred embodiment of the present invention, the above-mentioned drainage monitoring and analysis system based on artificial intelligence also includes:

[0075] Data collection module, slicing processing module, intelligent analysis module and early warning module to intelligently monitor drainage water quality data at multiple locations and over multiple time periods;

[0076] The data acquisition module acquires real-time data of drainage water quality from each monitoring point to obtain a water quality monitoring data set; based on the water quality monitoring data set, the data acquisition module transmits the water quality monitoring data set in real time in batches to the slicing processing module for classification and division processing;

[0077] The intelligent analysis module outputs the initial analysis data set; based on the first analysis data set and the second analysis data set, the first analysis data set includes a plurality of qualified water quality data, and the second analysis data includes abnormal water quality data; hierarchical verification is performed based on the first analysis data set and the second analysis data set to obtain a first verification data set and a second verification data set;

[0078] Based on the first verification data set and the second verification data set, a fusion is performed to obtain a water quality analysis verification data set; the water quality analysis verification data set is input into the water quality analysis model again for testing, and the water quality analysis model is repeatedly tested multiple times to obtain a final version of the water quality analysis model;

[0079] The intelligent analysis module transmits the first verification data set and the second verification data set to the early warning module, sets an early warning threshold, and issues an alarm when the early warning threshold is triggered; based on the abnormal data that issues the alarm, multiple information verifications are performed again through the multi-level verification module to obtain the fluctuation trend of multiple abnormal data points to verify the false alarm situation of the early warning module.

[0080] In an embodiment of the present invention, the data acquisition module is responsible for acquiring real-time data of drainage water quality from multiple monitoring points, including data aggregation of multiple time periods, to generate a water quality monitoring data set. The present invention relates to, but is not limited to, measuring the acidity and alkalinity of water, i.e., the pH value of water, through a pH sensor; a turbidity sensor measures the number of suspended particles in water, which will affect the transparency of the water, turbidity (the unit may be NTU, i.e., turbidity unit); a dissolved oxygen sensor measures the content of dissolved oxygen in water, which is crucial to assessing the ecological health of the water body, the amount of dissolved oxygen (mg / L or ppm); a conductivity sensor measures the conductivity of water, which indirectly reflects the content of dissolved salts in water, conductivity (usually in μS / cm or mS / cm); a temperature sensor measures the temperature of water, because temperature affects almost all other water quality parameters, temperature (° C or °F); Chemical oxygen demand (COD) sensor measures the chemical oxygen demand of organic matter in water, which is used to assess the degree of pollution in water bodies, COD concentration (mg / L); Ammonia nitrogen sensor measures the ammonia nitrogen content in water, which is an indicator often seen in water quality monitoring, especially in sewage treatment and discharge water, ammonia nitrogen content (mg / L); Total phosphorus and total nitrogen sensors measure the total phosphorus and total nitrogen content in water respectively, these indicators are very important for assessing the degree of eutrophication, total phosphorus and total nitrogen content (mg / L); Heavy metal sensors monitor the content of heavy metals (such as lead, copper, mercury, etc.) in water, which usually come from industrial pollution, the content of specific heavy metals;

[0081] The data acquisition module ensures accurate collection of water quality parameters. This module not only acquires data, but also transmits these data sets in real time to the next module - the sharding processing module, ensuring that the subsequent processing module can receive the latest water quality monitoring information in a timely manner, ensuring the real-time and integrity of the data, and providing a solid foundation for subsequent analysis and processing.

[0082] The fragmentation processing module receives the water quality monitoring data set from the data acquisition module, and classifies and categorizes the data according to preset standards and conditions, dividing it into different segments. The water quality is divided based on the data characteristics, such as by geographical area, time period or water quality type, to form a preliminary classified data set. Through classification and division, the data is made more structured, which is convenient for subsequent intelligent analysis and processing; at the same time, it improves the efficiency of the system in processing large-scale data.

[0083] The intelligent analysis module is one of the core modules of the system. It receives the classified data set from the sharding processing module and outputs the preliminary analysis results. It generates two types of data sets: the first type is data with good or qualified water quality, and the second type is data with poor or questionable water quality. Based on these two types of data, the module performs hierarchical verification to obtain the first verification data set (good water quality) and the second verification data set (problematic water quality). Subsequently, these verification sets are merged to form a more comprehensive water quality analysis verification data set. Through the analysis of intelligent algorithms and repeated tests, it can accurately identify water quality problem areas, reduce misjudgments, and improve the reliability of water quality monitoring results.

[0084] The early warning module further processes the verification data set from the intelligent analysis module. It sets multiple early warning thresholds based on different data attributes (such as water quality parameters, pollutant concentrations, etc.). When the detection data triggers the corresponding threshold, different levels of alarms will be issued. The system also has a supervision mechanism that can monitor the early warning module in real time to see if there are false alarms, and make adjustments and optimizations to ensure the accuracy of the early warning system, so as to achieve early warning of water quality problems, especially for poor or questionable data, so that it can respond quickly to prevent serious water quality problems.

[0085] As a result, the system realizes real-time monitoring and analysis of drainage water quality at multiple locations and in multiple time periods through the close cooperation of the four modules. The data acquisition module ensures the real-time nature of the data, the slicing processing module improves the data processing efficiency, the intelligent analysis module provides accurate water quality analysis results, and the early warning module achieves effective early warning through threshold setting. The system's multiple verifications, integration, and repeated tests have further improved the accuracy and reliability of water quality monitoring, and have strong application value.

[0086] In a preferred embodiment of the present invention, the above-mentioned fragmentation processing module includes:

[0087] Acquire the water quality monitoring data set in real time based on the data acquisition module;

[0088] The water quality monitoring data set is preprocessed by the slice processing module to obtain a preprocessed data set; and the preprocessed data set is classified and divided.

[0089] In this embodiment, the sharding processing module shards the classified data set and transmits it to each water quality analysis subunit. Sharding processing can improve the efficiency of data processing and enable the system to process multiple data blocks in parallel, thereby enabling the system to process a large amount of data in parallel through the sharding processing module, significantly improving the analysis speed.

[0090] Water quality monitoring data may have missing values ​​and duplicate values ​​due to sensor failure or the influence of data collection environment, so data cleaning is required to improve data quality. For null values ​​in the data set, those indicated by "*" or "-" are directly removed. At the same time, water quality indicators such as dissolved oxygen, potassium permanganate index, ammonia nitrogen, total phosphorus, total nitrogen, and pH value are extracted from each sample to ensure the correctness of subsequent water quality classification. Due to the different variable scales of water quality indicators, the order of magnitude of the original indicators is quite different, resulting in indicators with large variable values ​​having a greater impact on the comprehensive evaluation. In order to eliminate this influence, data dimensionless processing is required. Common methods include averaging, normalization, and standardization. Through experiments, it is found that the evaluation results of the data after standardized processing are closer to the actual situation, and almost all current research on PCA adopts this processing method. Therefore, standardized processing is selected to realize dimensionless data.

[0091] In a preferred embodiment of the present invention, the intelligent analysis module includes:

[0092] A water quality analysis unit and a multi-level verification unit; inputting the national water quality standard indicator data set into the water quality analysis subunit, the water quality analysis subunit receiving the national water quality standard indicator data set to construct a training set, which is input into the water quality analysis model for model training;

[0093] Inputting the classified data set acquired in real time into the water quality analysis model for analysis to obtain the initial analysis data set, and decomposing the initial analysis data set into the first analysis data set and the second analysis data set;

[0094] Based on the first analysis data set and the second analysis data set, the first analysis data set and the second analysis data set are input into the multi-level verification unit for water quality detection data verification to obtain a verification result.

[0095] In this embodiment, the water quality analysis unit is the core part of the system and is responsible for preliminary analysis of water quality. It contains multiple water quality analysis subunits, which have specific functions and can analyze different types of data. The following are the main working steps of this unit:

[0096] Each subunit is responsible for receiving the classified data set after sharding. The water quality analysis subunit will use the national water quality standard indicator data set to build a training set, and then input it into the water quality analysis model for model training; the water quality analysis model is trained with the received training data set to improve the model's ability to recognize and analyze new data. This stage is very important, as it determines the accuracy and effectiveness of the model;

[0097] When there are classified data sets acquired in real time, they will be input into the trained water quality analysis model for analysis to generate an initial analysis data set. After the initial data set is analyzed, the system will decompose it into a first analysis data set and a second analysis data set. This data classification can further optimize the subsequent data verification steps, reduce data redundancy, and improve the accuracy of the analysis.

[0098] The multi-level verification unit is a module in the system responsible for verifying the results of water quality analysis. Its role is to ensure the accuracy of the analysis results through different levels of verification methods. Verification process: The system inputs the first and second analysis data sets into the multi-level verification unit. The unit verifies the reliability and accuracy of the water quality analysis data through a multi-level verification mechanism. This may include methods such as statistical verification, model verification, or cross-validation. Finally, the verification unit will output the verification results and feed them back to the analysis module for further decision-making or adjustment.

[0099] Construction of training sets based on classified data helps improve the accuracy of water quality analysis models and ensures that the models maintain high predictive capabilities when faced with new data; multi-level verification improves accuracy: multi-level verification units cross-validate analysis results through verification methods at different levels to ensure high data reliability and reduce false positives and negatives; through the configuration of multiple sub-units, the system can be easily expanded to meet different types of water quality monitoring needs.

[0100] In a preferred embodiment of the present invention, the above-mentioned fragmentation processing module further includes:

[0101] Based on the acquired water quality monitoring data set, removing abnormal interference data from the water quality monitoring data set and completing missing values ​​to obtain a preprocessed data set;

[0102] The preprocessed data set is further standardized to obtain a dimensionless preprocessed data set; the dimensionless preprocessed data set includes a plurality of water quality monitoring data with unified standards after multiple preprocessing;

[0103] Based on the dimensionless preprocessed data set, the dimensionless preprocessed data set is set as the water quality monitoring data set P, then P={C i |a1,a2,…,a n}; where a represents the data attributes of water quality monitoring samples at multiple bases, and n represents the number of samples obtained from water quality monitoring bases; C i Indicates the attribute level of sample data, i=1,2,3,4,5;

[0104] according to: ; to obtain the attribute level of water quality detection sample data from multiple bases, where α is the contribution rate of the characteristic attribute, α=0.83; N represents the number of attributes; |D| represents the total number of samples in the water quality monitoring data set P, |D ci | indicates that the category value in the water quality monitoring data set P is c i The number of samples;

[0105] By setting a classification threshold, classification and division are performed based on the attribute levels of the water quality detection sample data of the multiple bases to obtain a first classified data set.

[0106] In this embodiment, the water quality monitoring data set is obtained by removing abnormal interference data and filling missing values: First, the system obtains the monitoring data of drainage water quality. These data come from multiple sensors at different monitoring points and cover multiple water quality attributes. Since water quality data may be affected by environmental factors or equipment errors, there will be some abnormal data. The goal of this step is to identify and remove these abnormal data through a certain data detection method (such as Z value test or IQR method) to ensure the validity of the data. During the monitoring process, some data may be missing due to sensor failure or other reasons. By using statistical methods (such as mean filling, interpolation or KNN algorithm) to fill missing values, incomplete data can be avoided to affect subsequent analysis. The preprocessed data set may be inconsistent due to different attribute units or ranges. Therefore, it is necessary to standardize the data to make it dimensionless, that is, to scale the data of different attributes to the same dimension according to certain rules (for example, normalized to [0,1]). This makes subsequent analysis and calculations more reasonable and avoids the impact of certain data on the results due to excessive or small dimensions. After standardization, the dimensionless data set obtained can eliminate the differences between the attribute values ​​of different monitoring points, making the attribute dimensions of the data set unified and laying the foundation for subsequent model processing.

[0107] Set the water quality monitoring data set P. The system sets the dimensionless preprocessed data set as the water quality monitoring data set P, thereby obtaining the level assigned to the attributes of each water quality sample (such as level 1, 2, 3, 4, 5, etc.), and calculates the corresponding attribute levels of the water quality monitoring sample data of multiple monitoring points. These levels reflect the performance of different monitoring points in various water quality parameters, which is convenient for subsequent classification analysis;

[0108] The system sets classification thresholds based on different water quality parameter levels. The classification thresholds determine how different water quality attributes are classified, such as "good", "qualified", "abnormal" and other categories. Based on these set classification thresholds, the system classifies and divides water quality test samples from multiple monitoring points according to the standard indicator range, including: Physical indicators: pH value: generally 6.5-8.5

[0109] Turbidity: Usually less than 1NTU (drinking water)

[0110] Chemical indicators: Dissolved oxygen (DO): greater than 5 mg / L (surface water)

[0111] Potassium permanganate index: less than 6 mg / L (surface water)

[0112] Ammonia nitrogen (NH4⁺-N): less than 0.5 mg / L (surface water)

[0113] Total phosphorus (TP): less than 0.1 mg / L (surface water)

[0114] Total nitrogen (TN): less than 1 mg / L (partial surface water)

[0115] Biochemical oxygen demand (BOD): less than 3 mg / L (surface water)

[0116] Chemical oxygen demand (COD): less than 20 mg / L (surface water)

[0117] Heavy metals: Lead: less than 0.01 mg / L (drinking water)

[0118] Mercury: less than 0.001 mg / L (drinking water)

[0119] Cadmium: less than 0.005 mg / L (drinking water)

[0120] Biological indicators: Coliform bacteria: non-drinking water should be less than 1000 CFU / 100mL, and drinking water should not be detected. These samples are divided into different categories. These categories correspond to different water quality states. After the classification is completed, the system generates the first classification data set. Each data point in this data set has been classified into a specific water quality level category for further analysis and processing.

[0121] Removing abnormal data and filling missing values ​​ensures the integrity and consistency of the data, reducing the impact of incomplete or erroneous data on the analysis results. Dimensionless standardization eliminates the impact of data dimensions, allowing data with different attributes to be compared and analyzed on the same scale, ensuring the rationality of the analysis results.

[0122] By calculating the contribution rate and setting the classification threshold, the system can perform fine classification analysis on water quality samples from different monitoring points. This can more accurately reflect the water quality status of each monitoring point, facilitate supervision and governance, and through classification, the system can convert complex water quality monitoring data into classified molecular data sets that are easy to understand and process, providing a basis for further model training and analysis.

[0123] In another preferred embodiment of the present invention, the above-mentioned fragmentation processing module further includes:

[0124] Based on the dimensionless preprocessed data set, the dimensionless preprocessed data set is set as the drainage water quality monitoring data matrix X, then X={C i |t1, t2, …, t j}; where t represents the data attributes of water quality monitoring samples in multiple time periods, and j represents the number of samples obtained during the water quality monitoring period; C i Indicates the attribute level of sample data, i=1,2,3,4,5;

[0125] according to: ; To obtain the attribute level of water quality test sample data in multiple time periods; where |D ci,tj | indicates that the category in the water quality monitoring dataset P is c i And the multi-period attribute t j The total number of samples of the value, |D ci | indicates that the category value in the water quality monitoring data set P is c i The number of samples, N j Indicates t j The number of possible values;

[0126] By setting a classification threshold, classification and division are performed based on the attribute levels of the water quality detection sample data in the multiple time periods to obtain a second classified data set.

[0127] In this embodiment, the dimensionless data is constructed as a matrix X={C i |t1, t2, …, t j},

[0128] i=1,2,3,4,5; this matrix representation facilitates efficient storage and calculation of multi-dimensional time series data, and is also convenient for application in subsequent analysis models. By calculating the classification set of water quality monitoring data in multiple time periods, the sample distribution of each category in each time period can be determined. The core of this process is to extract and analyze the attribute distribution of water quality samples in each time period, which can help reveal the trend of water quality changes in different time periods and category distribution characteristics, laying the foundation for subsequent classification.

[0129] Based on the national water quality standard range data, the classification threshold is set with different time periods of the day as the measurement unit, and the water quality monitoring situation at different time periods of the day is summarized into weekly, monthly and annual reports. The daily water quality monitoring situation of different cycles is summarized to check the changes in drainage water quality at different time periods, and the attribute levels of water quality monitoring samples are classified into different water quality categories (such as "good", "medium", "poor", etc.). The complex continuous data is divided into discrete categories through the setting of classification thresholds, which is convenient for further data analysis and processing. Through classification and division, the water quality status can be understood more intuitively and provide a basis for water quality management.

[0130] Based on the above classification, a new data set is finally obtained, namely the "second classification data set". This data set contains the classified water quality sample data, which can be used for further analysis, model building or prediction. After reasonable classification, the second classification data set can more clearly present the overall trend and problem points of water quality, providing valuable information for decision support.

[0131] Furthermore, in a preferred embodiment of the present invention, the above-mentioned fragmentation processing module further includes:

[0132] Based on the first classified data set and the second classified data set; by: ; Map and interactively integrate the first classification data set and the second classification data set to obtain a classification data set; where x k+1 represents the mapping parameter X at the k+1th control node, that is: ; In the formula, k represents the control parameter, that is, k=2, X k ∈(0,1).

[0133] In this embodiment, two different molecular data sets (first classification and second classification) are mapped and interacted, which means that the data from the two sets are paired and fused in some way, which may involve data on molecular structure, chemical properties or biological activity, etc. The mapping algorithm is used to effectively integrate the information in the two data sets.

[0134] During the mapping process, X k+1 Indicates the mapping parameter X at the k+1th control node. k is used as the control parameter (here k=2), indicating that the mapping process or iterative process is controlled by the progressive rule. The mapping parameter X k Changing within the interval (0,1) means that when processing the mapping of two sets of data in the mapping function, the data is ensured to be processed within a specified range to avoid excessive deviation during mapping, which makes the attributes of the data points in the two data sets mapped to each other differ too much, resulting in mismatch.

[0135] In a preferred embodiment of the present invention, the water quality analysis model includes:

[0136] The training set is set as an observation vector S, and the observation vector S is encoded into sequence data; the sequence data is input into the water quality analysis model, and the recognition result representing the water quality analysis is output, as follows:

[0137]

[0138] Among them, W u , W r , W crepresents the weight parameter, b u , b r , b c represents the bias parameter, represents the dot product, u (t) 、r (t) and c (t) Represent the intermediate states of the first, second, and third hidden layers respectively, S is the observation vector, X (t) represents the tth sequence data item of the sequence data, H (t) and H (t-1) They represent the t-th and t-1-th second hidden state data respectively, n≥t≥1, n represents the total number of input sequence data items, and H (t-1) =X (t) , tanh is the hyperbolic tangent function, σ represents the sigmoid function;

[0139] Based on the trained water quality analysis model, the sequence data of the classified molecular data set obtained in real time is input into the water quality analysis model to output the drainage water quality monitoring data state of the classified molecular data set obtained in real time. The output layer is specifically as follows:

[0140]

[0141] Among them, y t represents the output vector of the t-th intermediate period, where the components of the output vector represent the data items of the input classified molecular data set belonging to the i-th state, and h n Indicates the nth intermediate state represents data, W y is the weight parameter, b y is the bias parameter, and σ represents the sigmoid function.

[0142] In this embodiment, the acquired national water quality standard indicator data set is used to construct a water quality analysis model. The model constructed by the national water quality standard provides an accurate evaluation benchmark, ensures the accuracy of the test results, improves the ability to identify abnormal water quality, reduces the probability of false alarms and missed reports, and processes monitoring data in real time to make water quality status identification faster, detect water quality changes in a timely manner, help identify long-term trends and potential risks, provide a basis for policy making, improve resource utilization efficiency, reduce monitoring costs, and release human resources for other environmental protection activities.

[0143] Through trend analysis, we can predict future water quality problems and plan preventive measures in advance to reduce potential environmental

[0144] Ensure the safety of environmental water bodies, protect public health, and enhance public confidence in environmental quality through reliable results. Overall, the water quality analysis model has greatly improved the scientificity and effectiveness of water environment management by efficiently and accurately identifying water quality status.

[0145] In a preferred embodiment of the present invention, the hierarchical verification unit comprises:

[0146] The multi-level verification unit establishes a first verification level, wherein the first verification level includes setting a first analysis data set as D={D1, D2, ..., D n}; The first analysis data set is d={d1, d2, ..., d n}; where D n Indicates the nth first analysis data, d n Indicates the nth second analysis data, and sets a verification mark range [L,U], where L indicates the lowest point of data deviation and U indicates the highest point of data deviation;

[0147] Based on the first verification level, the multi-level verification unit establishes a second verification level, wherein the second verification level: ; In the formula, F is the effective value of the data;

[0148] Based on the second verification level, the multi-level verification unit establishes a third verification level, wherein the third verification level: ; to verify the accuracy of the first analysis data set D and the second analysis data set d, where TP represents the true verification result, TN represents the hypothesis verification result, FP represents the actual verification result, and FN represents the actual opposite verification result.

[0149] In this embodiment, in the first verification level, all data points of the first analysis data set need to be evaluated within the range of [L, U] to determine whether their deviations are within the acceptable range. The data comparison between D and d can be used to preliminarily detect data consistency or deviation. Beneficial effects: The first verification level provides a basic data screening process. By setting the range [L, U], abnormal data points that may have deviations can be quickly identified. This helps to improve the efficiency of subsequent verification levels and reduce unnecessary complex calculations.

[0150] Based on the results of the first verification level, the second verification level further sets the validity judgment of the data, that is, F is used to represent the valid value of the data. In this verification level, data that does not meet the conditions can be obtained by judging the validity of the data. Further analysis, such as statistical analysis or model prediction, is performed on the data screened by the first level to determine whether the data meets the preset validity criteria, thereby improving the accuracy of the verification results. By screening the valid value F of the data, the risk of error propagation can be reduced.

[0151] In the third validation level, more refined validation criteria are set to evaluate the accuracy of the first analysis data set D and the second analysis data set d. In this level, four indicators are used for classification validation:

[0152] TP (True Positive): represents the number of true verification results, that is, the actual data points where verification was successful.

[0153] TN (True Negative): represents the number of hypothesis verification results, that is, data points that failed verification but met the hypothesis conditions.

[0154] FP (False Positive): Indicates the number of actual verification failures, that is, data points that are mistakenly considered to be successful verification.

[0155] FN (False Negative): Indicates the number of actual opposite verification results, that is, data points that are mistakenly considered to have failed verification.

[0156] These indicators can be used to calculate evaluation indicators such as accuracy, precision, and recall, and comprehensively measure the verification effect of the data set. The third verification level uses more comprehensive verification standards to evaluate the classification accuracy of the data, making the verification results more reliable and scientific. This can not only improve the accuracy of data verification, but also provide more powerful data support for the decision-making process.

[0157] according to Figure 2 As shown, in a preferred embodiment of the present invention, the early warning module includes:

[0158] Based on the verification result, the first verification data set and the second verification data set are input into the early warning module; when the early warning threshold of the early warning module is triggered, the early warning module triggers an alarm;

[0159] Feedback the abnormal data that triggers the alarm to the hierarchical verification module for multiple verifications. When the accuracy parameters of the multiple verification results show a small fluctuation trend, the early warning module warns that it is normal;

[0160] When the accuracy parameters of multiple verification results show a large fluctuation trend, and the amplitude of each fluctuation is greatly different, the early warning module warns of abnormality and a false alarm occurs.

[0161] according to Figure 3 and Figure 4 In this embodiment, the early warning module receives verification results from the first and second verification data sets. When these results trigger the set early warning threshold:

[0162] Threshold setting based on the degree of deviation: Usually based on the standard deviation or coefficient of variation of the data. A common approach is to set the threshold to a certain percentage of the mean (such as ±2% or ±5%) or several standard deviations (such as ±2σ or ±3σ, where σ represents the standard deviation). If the mean of the data is 100 and the standard deviation is 10, then the ±2σ threshold will be set between 80 and 120.

[0163] Threshold setting for abnormal data frequency: It is set based on the statistical analysis of abnormal frequency in historical data. For example, it can be set to the maximum abnormal event ratio allowed before triggering an alarm, such as 5 abnormal values ​​in every 1,000 data points. If there are usually 20 abnormal values ​​in the past 10,000 data points, then the abnormal frequency threshold can be set to 0.2%.

[0164] Setting warning thresholds for specific patterns: This involves identifying non-random patterns in the data, such as periodic fluctuations, persistent trends, or recurring data combinations. The threshold can be set to the number or intensity of occurrences of a specific pattern. When a pattern is detected to occur more than 5 times in a row, an alarm is triggered.

[0165] The early warning module will trigger an alarm. This threshold may be based on a variety of factors, such as the degree of data deviation, the frequency of abnormal data, or the appearance of a specific pattern. By automatically triggering an alarm, the system can respond to potential data problems in a timely manner to prevent erroneous data from affecting subsequent decisions or operations. This timely response mechanism can significantly reduce the risk of wrong decisions and improve overall operational efficiency and safety.

[0166] When the early warning module triggers an alarm, the relevant abnormal data will be fed back to the hierarchical verification module. Here, the data will be verified multiple times to evaluate its accuracy. The verification process can show the fluctuation trend of monitoring accuracy parameters (such as error rate, consistency, etc.). Multiple verifications can provide a more comprehensive data evaluation, ensuring that each data point is fully checked, thereby enhancing the reliability of the final data quality. This method can effectively screen out real data anomalies and avoid making wrong judgments due to single occasional data misreading.

[0167] Normal fluctuation: Multiple verification results show that the fluctuation trend of the accuracy parameter is small, indicating that the data is basically stable, and the early warning module will determine that the early warning is normal. This means that the initial triggering of the early warning may be just an accidental or non-critical problem.

[0168] Abnormal fluctuation: If the accuracy parameter fluctuates greatly and the amplitude of each fluctuation is significantly different, the warning module will determine it as an abnormal warning. This indicates that there may be major data problems or systematic errors, which require further inspection and processing.

[0169] Handling of normal fluctuations: Avoids over-response due to occasional small errors, reduces unnecessary operational intervention, and optimizes resource allocation.

[0170] Handling of abnormal fluctuations: Ability to promptly detect and respond to major data problems or system errors to ensure the stability of system operation and the accuracy of data.

[0171] The purpose of the present invention can also be achieved by running a program or a group of programs on any computing device. The computing device can be a well-known general-purpose device. Therefore, the purpose of the present invention can also be achieved by simply providing a program product containing a program code that implements the method or device. That is to say, such a program product also constitutes the present invention, and the storage medium storing such a program product also constitutes the present invention. Obviously, the storage medium can be any well-known storage medium or any storage medium developed in the future. It should also be pointed out that in the device and method of the present invention, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. In addition, the steps of performing the above-mentioned series of processing can naturally be performed in chronological order according to the order of description, but it is not necessary to perform them in chronological order. Some steps can be performed in parallel or independently of each other.

[0172] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

[0173] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A drainage monitoring and analysis system based on artificial intelligence, characterized in that: include: Obtain a water quality standard indicator data set, input the water quality standard indicator data set into a constructed training set to train a water quality analysis model, and output a water quality status result; Collect water quality monitoring data from multiple locations and multiple time periods in real time, and classify and divide the water quality monitoring data to obtain a classified data set; the classified data set includes drainage water quality grade data corresponding to multiple locations and multiple time periods; Based on the water quality analysis model and the classified data set acquired in real time, an initial analysis data set is obtained; the initial analysis data set includes water quality abnormal state data existing in the classified data set acquired in real time; The initial analysis data set is transmitted to the early warning module for receiving, and when the initial analysis data set contains abnormal water quality state data, the initial analysis data set is subjected to hierarchical verification to obtain a verification data set; The fluctuation trend of multiple abnormal data points is obtained according to the verification data set, and the false alarm situation of the early warning module is judged according to the fluctuation trend of multiple abnormal data points. When no false alarm situation occurs, the alarm information is obtained and the alarm is issued.

2. The drainage monitoring and analysis system based on artificial intelligence according to claim 1 is characterized in that: Also includes: Data collection module, sharding processing module, intelligent analysis module and early warning module; The data acquisition module acquires the water quality monitoring data from each monitoring point in real time to obtain a water quality monitoring data set; The slicing processing module classifies and divides the water quality monitoring data set to obtain the classified data set; The intelligent analysis module is provided with a water quality analysis model, and the classified data set is input into the water quality analysis model to output the initial analysis data set; Performing hierarchical verification on the initial analysis data set to obtain a water quality analysis verification data set; The false alarm situation of the early warning module is determined based on the water quality analysis verification data set.

3. The drainage monitoring and analysis system based on artificial intelligence according to claim 2 is characterized in that: The fragment processing module further includes: Acquire the water quality monitoring data set in real time based on the data acquisition module; Preprocessing the water quality monitoring data set through the slicing processing module to obtain a preprocessed data set; classifying and dividing the preprocessed data set based on the preprocessed data to obtain the classified data set; The classified data set includes a first classified data set and a second classified data set; the first classified data set includes drainage water quality monitoring data of multiple different locations; and the second classified data set includes monitoring data of multiple different time periods.

4. The drainage monitoring and analysis system based on artificial intelligence according to claim 2 is characterized in that: The intelligent analysis module further includes: Water quality analysis unit and multi-level verification unit; Inputting the water quality standard index data set into the water quality analysis unit, the water quality analysis unit receives the water quality standard index data set to construct a training set, and then inputs it into the water quality analysis model for model training; Inputting the classified data set acquired in real time into the water quality analysis model for analysis to obtain the initial analysis data set; Decomposing the initial analysis data set into the first analysis data set and the second analysis data set; the first analysis data set includes a plurality of qualified water quality data, and the second analysis data includes abnormal water quality data; Based on the first analysis data set and the second analysis data set, the first analysis data set and the second analysis data set are input into the multi-level verification unit for water quality detection data verification to obtain a verification result.

5. The drainage monitoring and analysis system based on artificial intelligence according to claim 3 is characterized in that: Preprocessing the water quality monitoring data set by the slice processing module to obtain a preprocessed data set; The method for classifying and dividing the preprocessed data set includes: Based on the acquired water quality monitoring data set, removing abnormal interference data from the water quality monitoring data set and completing missing values ​​to obtain a preprocessed data set; The preprocessed data set is further standardized to obtain a dimensionless preprocessed data set; the dimensionless preprocessed data set includes a plurality of water quality monitoring data with unified standards after multiple preprocessing; Based on the dimensionless preprocessed data set, the dimensionless preprocessed data set is set as the water quality monitoring data set P, then P={C i |a1,a2,…,a n }; where a represents the data attributes of water quality monitoring samples at multiple bases, and n represents the number of samples obtained from water quality monitoring bases; C i Indicates the attribute level of sample data, i=1,2,3,4,5; according to: ; to obtain the attribute level of water quality detection sample data from multiple bases, where α is the contribution rate of the characteristic attribute, α=0.83; N represents the number of attributes; |D| represents the total number of samples in the water quality monitoring data set P, |D ci | indicates that the category value in the water quality monitoring data set P is c i The number of samples; By setting a classification threshold, classification and division are performed based on the attribute levels of the water quality detection sample data of the multiple bases to obtain a first classified data set.

6. The drainage monitoring and analysis system based on artificial intelligence according to claim 5 is characterized in that: The method of classifying and dividing the preprocessed data set to obtain a classified molecular data set also includes: Based on the dimensionless preprocessed data set, the dimensionless preprocessed data set is set as the drainage water quality monitoring data matrix X, then X={C i |t1,t2,…,t j }; where t represents the data attributes of water quality monitoring samples in multiple time periods, and j represents the number of samples obtained during the water quality monitoring period; C i Indicates the attribute level of sample data, i=1,2,3,4,5; according to: ; To obtain the attribute level of water quality test sample data in multiple time periods; where |D ci,tj | indicates that the category in the water quality monitoring dataset P is c i And the multi-period attribute t j The total number of samples of the value, |D ci | indicates that the category value in the water quality monitoring data set P is c i The number of samples, N j Indicates the property j The number of possible values; By setting a classification threshold, classification and division are performed based on the attribute levels of the water quality detection sample data in the multiple time periods to obtain a second classified data set.

7. The drainage monitoring and analysis system based on artificial intelligence according to claim 6 is characterized in that: The method for performing mapping interaction based on the first classified data set and the second classified data set to obtain a classified data set includes: Based on the first classified data set and the second classified data set; by: ; Map and interactively integrate the first classification data set and the second classification data set to obtain a classification data set; where x k+1 represents the mapping parameter X at the k+1th control node, that is: ; In the formula, k represents the control parameter, that is, k=2, X k ∈(0,1).

8. The drainage monitoring and analysis system based on artificial intelligence according to claim 7 is characterized in that: The method of inputting the water quality standard index data set into the water quality analysis subunit, wherein the water quality analysis subunit receives the water quality standard index data set to construct a training set, and then inputs the training set into the water quality analysis model for model training, comprises: The training set is set as an observation vector S, and the observation vector S is encoded into sequence data; the sequence data is input into the water quality analysis model, and the recognition result representing the water quality analysis is output, as follows: ; Among them, W u , W r , W c represents the weight parameter, b u , b r , b c represents the bias parameter, represents the dot product, u (t) 、r (t) and c (t) Represent the intermediate states of the first, second, and third hidden layers respectively, S is the observation vector, X (t) represents the tth sequence data item of the sequence data, H (t) and H (t-1) They represent the t-th and t-1-th second hidden state data respectively, n≥t≥1, n represents the total number of input sequence data items, and H (t-1) =X (t) , tanh is the hyperbolic tangent function, and σ represents the sigmoid function.

9. The drainage monitoring and analysis system based on artificial intelligence according to claim 8 is characterized in that: Based on the first analysis data set and the second analysis data set, the method of inputting the first analysis data set and the second analysis data set into the multi-level verification unit for water quality detection data verification to obtain a verification result includes: The multi-level verification unit establishes a first verification level, wherein the first verification level includes setting a first analysis data set as D={D1, D2, ..., D n }; The first analysis data set is d={d1, d2, ..., d n }; where D n Indicates the nth first analysis data, d n Indicates the nth second analysis data, and sets a verification mark range [L,U], where L indicates the lowest point of data deviation and U indicates the highest point of data deviation; Based on the first verification level, the multi-level verification unit establishes a second verification level, wherein the second verification level: ; In the formula, F is the effective value of the data; Based on the second verification level, the multi-level verification unit establishes a third verification level, wherein the third verification level: ; to verify the accuracy of the first analysis data set D and the second analysis data set d, where TP represents the true verification result, TN represents the hypothesis verification result, FP represents the actual verification result, and FN represents the actual opposite verification result.

10. The drainage monitoring and analysis system based on artificial intelligence according to claim 9, characterized in that: The method of obtaining the fluctuation trend of multiple abnormal data points according to the verification data set, judging the false alarm situation of the early warning module according to the fluctuation trend of the multiple abnormal data points, and obtaining the alarm information and issuing the alarm when no false alarm situation occurs, comprises: Based on the verification result, the first verification data set and the second verification data set are verified multiple times. When the accuracy parameters of the multiple verification results show a large fluctuation trend, and the fluctuation amplitudes differ greatly each time, the early warning module warns of abnormality. If the early warning module warns of abnormality, a false alarm occurs. When the accuracy parameters of multiple verification results show a small fluctuation trend, the warning module warns normally; when the warning module warns normally, the verification data set is input into the warning module; the verification data set triggers the warning threshold of the warning module, and the warning module triggers an alarm.

Citation Information

Patent Citations

  • Verification system and method for low-voltage ride-through simulation model of wind power unit

    CN105356498A

  • Microflora-based sewage treatment aeration system control method

    CN114573096A

  • Transformer anomaly detection model training and deployment method and device

    CN114999527A

  • Water quality prediction method based on multi-task learning and deep learning model

    CN115983493A

  • Surface water quality index prediction system

    CN118035726A