CN-121980186-A - Water environment pollution monitoring method, electronic equipment and storage medium

CN121980186ACN 121980186 ACN121980186 ACN 121980186ACN-121980186-A

Abstract

The application provides a method for monitoring water environment pollution, electronic equipment and a storage medium, wherein when target monitoring data of a target water quality index meet an alarm condition, time series data corresponding to a plurality of water environment parameters in a preset time window before the moment that a target monitoring point is met the first alarm condition are obtained, the time series data corresponding to each water environment parameter in the plurality of water environment parameters are subjected to multiple feature extraction in the preset time window to obtain multi-dimensional features of each water environment parameter, the multi-dimensional features of each water environment parameter are utilized to construct a target feature set, the target feature set is input into a pre-trained data confidence prediction model, data confidence used for representing reliability of the target monitoring data is determined, and water environment pollution alarm information is generated based on the data confidence meeting the alarm condition. By the method, the practicability and the alarm accuracy of water environment pollution monitoring are improved.

Inventors

BAO JIE
Request for anonymity
Request for anonymity
Request for anonymity
Request for anonymity

Assignees

芯视界（北京）科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (12)

1. A method for monitoring water environmental pollution, the method comprising: acquiring initial monitoring data of a target water quality index of a target monitoring point position, and correcting the initial monitoring data by utilizing a correction model to obtain target monitoring data of the target water quality index; when the target monitoring data of the target water quality index meets a first alarm condition, acquiring time series data corresponding to various water environment parameters in a preset time window before the moment of meeting the first alarm condition, wherein the various water environment parameters do not comprise the target water quality index; extracting a plurality of characteristics of time series data corresponding to each water environment parameter in the plurality of water environment parameters in the preset time window to obtain multi-dimensional characteristics of each water environment parameter; Constructing a target feature set by utilizing multidimensional features of each water environment parameter, inputting the target feature set into a pre-trained data confidence prediction model, and determining a data confidence used for representing the reliability of the target monitoring data; and generating water environment pollution alarm information of the target monitoring point position when the data confidence degree meets a second alarm condition.
2. The method of monitoring of claim 1, wherein the data confidence prediction model is trained by: Acquiring a sample data set, wherein the sample data set comprises historical time sequence data of the plurality of water environment parameters of the target monitoring point in a historical time period and confidence labels corresponding to the historical time sequence data of the plurality of water environment parameters; Performing characteristic engineering on the historical time sequence data of the multiple water environment parameters to obtain all sample characteristics for representing the characteristics of the historical time sequence data of the multiple water environment parameters; constructing multiple data feature sets by utilizing all sample features, and combining each data feature set with at least one pre-training model to form multiple training combinations; For each training combination, training a pre-training model in the training combination by utilizing a data feature set in the training combination and a confidence coefficient label corresponding to the data feature set to obtain a candidate model corresponding to the training combination; carrying out multi-dimensional comprehensive scoring on the candidate model corresponding to each training combination to obtain a comprehensive score corresponding to each training combination; And taking the training combination with the highest comprehensive score in the training combinations as a target combination, taking a candidate model corresponding to the target combination as the data confidence prediction model, and taking the construction mode of the data feature set corresponding to the target combination as the construction mode of the target feature set.
3. The method of monitoring of claim 2, wherein the plurality of data feature sets includes a first aqueous environment time series feature set, a second aqueous environment time series feature set, and a third aqueous environment time series feature set, the constructing a plurality of data feature sets using the all sample features comprising: Utilizing all the sample characteristics to form the first water environment time sequence characteristic set; Screening all the sample features from the first water environment time sequence feature set to obtain the second water environment time sequence feature set; And determining a characteristic population based on a preset population setting rule and all sample characteristics, and determining the third water environment time sequence characteristic set by utilizing the characteristic population.
4. The method of monitoring according to claim 3, wherein said screening the entire sample features from the first aqueous environment time series feature set to obtain the second aqueous environment time series feature set comprises: Calculating a characteristic contribution degree parameter and a characteristic evaluation parameter corresponding to each sample characteristic in all sample characteristics; And determining a plurality of target features from all the sample features based on the feature contribution degree parameters and the feature evaluation parameters corresponding to each sample feature, and determining the plurality of target features as the second water environment time sequence feature set.
5. The method according to claim 4, wherein the determining a plurality of target features from the all sample features based on the feature contribution parameter and the feature evaluation parameter corresponding to each sample feature includes: Aiming at each sample feature, obtaining a first joint screening index of the sample feature based on the feature contribution parameter preset weight, the feature evaluation parameter preset weight, the feature contribution parameter corresponding to the sample feature and the feature evaluation parameter; and determining a plurality of target features from all the sample features based on the first joint screening index corresponding to each sample feature.
6. The method according to claim 4, wherein the determining a plurality of target features from the all sample features based on the feature contribution parameter and the feature evaluation parameter corresponding to each sample feature includes: For each training round of each pre-training model, determining a plurality of first features from all sample features based on feature contribution parameters corresponding to each sample feature, and determining a plurality of second features from all sample features based on feature evaluation parameters corresponding to each sample feature; Determining a first model performance evaluation index based on the pre-training models of the plurality of first features and the current training round, and determining a second model performance evaluation index based on the pre-training models of the plurality of second features and the current training round; Determining a characteristic contribution parameter target weight and a characteristic evaluation parameter target weight of the current training round based on the first model performance evaluation index, the second model performance evaluation index and the characteristic contribution parameter weight of the last training round; determining a feature union between the first plurality of features and the second plurality of features; Aiming at each third feature in the feature union, obtaining a second joint screening index of the third feature in the current training round based on the feature contribution parameter target weight, the feature evaluation parameter target weight, the feature contribution parameter corresponding to the third feature and the feature evaluation parameter; And determining a plurality of target features from all the third features based on the second joint screening index corresponding to each third feature.
7. The method of monitoring of claim 3, wherein said determining said third water environment time series feature set using said feature population comprises: Performing selection, crossing and mutation operations on the characteristic population to generate a new generation characteristic population; calculating the fitness of the new generation characteristic population, and judging whether a preset termination condition is met or not based on the fitness; If yes, taking the characteristic of the new generation characteristic population, the adaptability of which reaches a preset threshold value, as the time sequence characteristic set of the third water environment; If not, taking the new generation of characteristic population as the characteristic population, and returning to execute the step of executing the selection, crossing and mutation operation on the characteristic population until the adaptability meets the preset termination condition.
8. The method of monitoring according to claim 2, wherein the step of performing multi-dimensional composite scoring on the candidate model corresponding to each training combination to obtain a composite score corresponding to each training combination includes: Calculating scores of the candidate models under a plurality of preset dimensions aiming at the candidate models corresponding to each training combination, wherein the plurality of preset dimensions comprise a prediction performance dimension, a robustness dimension, a complexity dimension and a target consistency dimension; and carrying out weighted summation on the scores to obtain a comprehensive score corresponding to the candidate model, and taking the comprehensive score corresponding to the candidate model as the comprehensive score corresponding to the training combination.
9. The method of monitoring according to claim 1, wherein determining whether the target monitoring data of the target water quality indicator meets a first alarm condition is performed by: comparing the target monitoring data of the target water quality index with an alarm threshold, and judging that a first alarm condition is met when the target monitoring data exceeds the alarm threshold; Or alternatively And determining an output result of the pollution result prediction model based on the target monitoring data and a pre-trained pollution result prediction model, and judging that a first alarm condition is met when the output result is pollution.
10. The method of monitoring according to claim 1, wherein determining whether the data confidence level meets a second alarm condition is performed by: comparing the data confidence level with a first confidence level threshold; when the data confidence coefficient exceeds the first confidence coefficient threshold value, judging that a second alarm condition is met; when the data confidence coefficient is the same as the first confidence coefficient threshold value, determining an alarm confidence coefficient used for representing the reliability of an alarm result based on the data confidence coefficient and a pre-trained alarm confidence coefficient prediction model; and when the alarm confidence coefficient exceeds a second confidence coefficient threshold value, judging that a second alarm condition is met.
11. An electronic device comprising a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory in communication via the bus when the electronic device is in operation, the machine-readable instructions being executable by the processor to perform the steps of the method for monitoring water environmental pollution of any one of claims 1 to 10.
12. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, performs the steps of the method for monitoring water environmental pollution according to any of claims 1 to 10.

Description

Water environment pollution monitoring method, electronic equipment and storage medium Technical Field The application relates to the technical field of water environment monitoring, in particular to a monitoring method for water environment pollution, electronic equipment and a storage medium. Background The on-line monitoring of water quality is an important means for environmental protection, and by arranging monitoring equipment at the positions of surface water, underground water, a pipe well and the like, key water quality parameters such as Chemical Oxygen Demand (COD), turbidity, pH value and the like in the water body are monitored in real time, so that the pollution of the water body is discovered in time, and an alarm is sent to an environmental protection department, thereby having important significance for guaranteeing the safety of the water environment. However, the on-line monitoring device faces a plurality of interference factors in practical application, which results in reduced accuracy of the monitored data and increased false alarm rate. Among them, turbidity interference is one of the most common and most significant problems. When the turbidity of the water body rises due to factors such as rainfall, hydrologic variation and the like, the COD value measured by the on-line monitoring instrument based on the optical principle often has obvious deviation, so that false alarm is generated. Aiming at the problem of turbidity interference, the prior art mainly adopts two quality control methods, namely a sensor self-calibration method, which reduces measurement deviation through regular calibration of standard liquid, and a mathematical model correction method, which carries out data correction by establishing a correlation between turbidity and COD measurement deviation. Although the data quality is improved to a certain extent by the method, the model often presents the problem of insufficient adaptability in complex and changeable natural water environments, particularly when sudden turbidity changes or the coaction of various interference factors are encountered, the correction effect is obviously reduced, a large amount of false alarms are still generated, and the accuracy and the efficiency of environmental protection supervision are influenced. Disclosure of Invention In view of the above, the application aims to provide a monitoring method, electronic equipment and storage medium for water environment pollution, which can effectively identify and shield false alarms generated by unreliable data caused by environmental interference, remarkably improve the accuracy and the intelligent level of alarm decision, effectively reduce the false alarm rate and improve the practicability and the alarm accuracy of water environment pollution monitoring. In a first aspect, an embodiment of the present application provides a method for monitoring water environmental pollution, where the method includes: acquiring initial monitoring data of a target water quality index of a target monitoring point position, and correcting the initial monitoring data by utilizing a correction model to obtain target monitoring data of the target water quality index; when the target monitoring data of the target water quality index meet a first alarm condition, acquiring time sequence data corresponding to various water environment parameters in a preset time window before the moment of meeting the first alarm condition; extracting a plurality of characteristics of time series data corresponding to each water environment parameter in the plurality of water environment parameters in the preset time window to obtain multi-dimensional characteristics of each water environment parameter; Constructing a target feature set by utilizing multidimensional features of each water environment parameter, inputting the target feature set into a pre-trained data confidence prediction model, and determining a data confidence used for representing the reliability of the target monitoring data; and generating water environment pollution alarm information of the target monitoring point position when the data confidence degree meets a second alarm condition. Further, the data confidence prediction model is trained by: Acquiring a sample data set, wherein the sample data set comprises historical time sequence data of the plurality of water environment parameters of the target monitoring point in a historical time period and confidence labels corresponding to the historical time sequence data of the plurality of water environment parameters; Performing characteristic engineering on the historical time sequence data of the multiple water environment parameters to obtain all sample characteristics for representing the characteristics of the historical time sequence data of the multiple water environment parameters; constructing multiple data feature sets by utilizing all sample features, and combining each data feature set with at least o