Search

CN-122020490-A - Water quality anomaly detection and pollution source tracing system based on artificial intelligence

CN122020490ACN 122020490 ACN122020490 ACN 122020490ACN-122020490-A

Abstract

The invention discloses a water quality anomaly detection and pollution source tracing system based on artificial intelligence, and belongs to the technical field of environmental monitoring. The system aims to solve the problems of high false alarm rate and low pollution tracing efficiency in water anomaly detection in the prior art. The technical scheme includes that multidimensional environment data are obtained through a data acquisition module, a space-time diagram neural network anomaly detection module is utilized to extract space-time characteristics based on a water system topological structure to identify anomalies, an intelligent tracing module integrating a knowledge graph is utilized to combine a hydrodynamics reverse model and Bayesian reasoning to locate pollution sources, and finally early warning is conducted through a visualization module. The method is mainly used for monitoring the watershed water environment in real time, can realize high-precision abnormality identification and minute-level automatic tracing, and provides scientific basis for precise pollution control.

Inventors

  • ZHAO TINGTING
  • ZHANG YOU
  • CHENG WEIBO
  • REN XIANGZHENG
  • TAN GUOLIANG
  • WEN JINGJING

Assignees

  • 水发规划设计有限公司

Dates

Publication Date
20260512
Application Date
20260410

Claims (10)

  1. 1. Water quality anomaly detection and pollution source tracing system based on artificial intelligence, which is characterized by comprising: The data acquisition module (100) is used for acquiring water quality parameters, hydrological parameters and meteorological parameters in the monitoring area in real time and preprocessing the parameters; The space-time diagram neural network anomaly detection module (200) is connected with the data acquisition module (100) and is used for constructing a space-time diagram neural network model based on a water system topological structure, extracting and fusing spatial features and time sequence features of the preprocessed data and outputting an anomaly judgment result and an anomaly feature vector; the intelligent tracing module (300) is connected with the space-time diagram neural network anomaly detection module (200) and is used for combining the hydrodynamics reverse model and the pollution source knowledge spectrum after receiving the anomaly judgment result, calculating posterior probability of the potential pollution source through probability reasoning and outputting a tracing result; And the visual early warning module (400) is respectively connected with the space-time diagram neural network anomaly detection module (200) and the intelligent tracing module (300) fused with the knowledge graph and is used for displaying monitoring states, anomaly information and tracing paths and sending early warning information.
  2. 2. The artificial intelligence based water quality anomaly detection and pollution source tracing system of claim 1, wherein said space-time diagram neural network anomaly detection module (200) comprises: The water system topological graph construction unit (210) is used for modeling the monitoring station point as a graph node according to the physical water system structure, constructing a directed edge according to the water flow direction and generating graph structure data containing node characteristics and edge weights; A graph attention convolution network unit (220) for receiving the graph structure data, calculating dynamic attention coefficients between nodes by using a graph attention mechanism, and aggregating neighbor node information to extract spatial features; The multi-scale time sequence coding unit (230) is used for carrying out multi-scale convolution and position coding on the characteristic sequence of the node historical time steps, and extracting time sequence characteristics through a transducer coder; and the space-time feature fusion abnormality detection unit (240) is used for fusing the space features and the time sequence features through a gating mechanism, calculating an abnormality probability score and a reconstruction error based on the fused features, and comprehensively judging whether abnormality occurs.
  3. 3. The artificial intelligence based water quality anomaly detection and pollution source tracing system according to claim 2, wherein in said water system topology construction unit (210), edge weights are used for The calculation formula of (2) is as follows: Wherein, the For a station And (3) with The hydraulic distance between the two hydraulic cylinders is equal to the hydraulic distance between the two hydraulic cylinders, And For the site traffic volume to be the same, For a theoretical lag time for contaminant propagation, In order to be a time decay constant, Is a learnable weight parameter.
  4. 4. The artificial intelligence based water quality anomaly detection and pollution source tracing system of claim 2, wherein said graph annotation force convolution network unit (220) employs a multi-head attention mechanism for any node And its neighbor nodes And introducing edge feature vectors containing hydraulic distance and propagation delay when calculating the dynamic attention coefficient, and extracting spatial features by stacking a multi-layer graph convolution network.
  5. 5. The artificial intelligence based water quality anomaly detection and pollution source tracing system according to claim 2, wherein the multi-scale time sequence encoding unit (230) performs convolution operation by adopting convolution check input feature sequences with different sizes, obtains features of different time receptive fields, and inputs the features into a transform encoder comprising a multi-layer self-attention mechanism after splicing.
  6. 6. The artificial intelligence based water quality anomaly detection and contamination source tracing system of claim 2, wherein the anomaly determination formula of the spatio-temporal feature fusion anomaly detection unit (240) is: Wherein, the In order to classify the anomaly probability score, An error is reconstructed for the data and, In order to average the reconstruction error, And In order for the coefficient of balance to be present, As a result of the abnormality threshold value, As an indication function.
  7. 7. The artificial intelligence based water quality anomaly detection and pollution source tracing system of claim 1, wherein said intelligent tracing module (300) that merges knowledge patterns comprises: A hydrodynamic back propagation engine (310) for back-pushing the emission intensity sequence of each candidate point upstream according to the downstream abnormal concentration sequence based on the inverse solution technique of the one-dimensional convection-diffusion equation; a pollution source knowledge graph library (320) for storing domain knowledge graphs including pollution source entities, pollutant type entities and water system node entities, and providing semantic relationships and vector embedding between the entities; the Bayesian probability reasoning decision network (330) is used for calculating the posterior probability of each candidate pollution source by combining the physical inversion result, the knowledge map priori information and the monitoring data characteristics; And the traceability result interpreter (340) is used for outputting the candidate pollution source with the highest posterior probability and the reasoning interpretation report thereof.
  8. 8. The artificial intelligence based water anomaly detection and pollution source tracing system of claim 7, wherein said bayesian probabilistic inference decision network (330) computes candidate pollution sources The likelihood function on which the probability is based is composed of the product of hydrodynamic evidence, evidence of contaminant feature matching, evidence of time window matching, and evidence of spatial reachability.
  9. 9. The artificial intelligence based water quality anomaly detection and pollution source tracing system of claim 8, wherein said bayesian probabilistic inference decision network (330) further comprises a graph neural network enhancement structure for stitching the embedded vector of the knowledge graph with the anomaly feature vector output by said space-time graph neural network anomaly detection module (200) and outputting a correction factor to adjust the probability distribution.
  10. 10. The artificial intelligence based water quality anomaly detection and pollution source tracing system according to claim 1, wherein the confirmed tracing result output by the intelligent tracing module (300) fused with the knowledge graph is fed back to the space-time diagram neural network anomaly detection module (200) and the pollution source knowledge graph library (320) as a positive sample for incremental update of model parameters.

Description

Water quality anomaly detection and pollution source tracing system based on artificial intelligence Technical Field The invention relates to the technical field of environmental monitoring, in particular to a water quality anomaly detection and pollution source tracing system based on artificial intelligence. Background Along with the acceleration of the industrialization process, the water environment safety problem is increasingly prominent, and the establishment of an efficient water quality monitoring and management system is important. Although the existing water quality monitoring system is widely deployed with a sensor network, the existing water quality monitoring system still has significant defects in the aspects of intelligent treatment of anomaly detection and pollution tracing, and is mainly characterized in the following two aspects: Firstly, the existing water quality abnormality detection technology mostly adopts a single-point single-parameter threshold judgment method, and is difficult to effectively identify multi-parameter coupling type abnormality and cross-regional progressive pollution. Moreover, the existing machine learning method often does not fully consider the constraint effect of a water system topological structure on pollutant transmission, lacks the fine modeling of space-time dependency relationship of different monitoring stations, causes hysteresis in detection of upstream and downstream association pollution events, and is easy to generate a large number of false alarms due to normal fluctuation of environmental background values. Secondly, the existing pollution source tracing technology mainly relies on manual field investigation or simple upstream and downstream concentration comparison, and is low in efficiency and limited in coverage range. Although a part of researches introduce a modeling traceability method, the physical propagation rule of hydrodynamics and priori knowledge (such as enterprise pollution discharge characteristics, position relations and the like) of regional pollution sources are difficult to effectively fuse, a probability comprehensive evaluation mechanism for a plurality of potential pollution sources is lacked, the interpretability and the confidence of traceability results are insufficient, and the requirement of rapid and accurate emergency response of a supervision department is difficult to meet. Disclosure of Invention The invention mainly aims to provide a water quality anomaly detection and pollution source tracing system based on artificial intelligence so as to solve the problems in the related art. In order to achieve the above objective, according to one aspect of the present invention, there is provided an artificial intelligence-based water quality anomaly detection and pollution source tracing system, which includes a data acquisition module for acquiring water quality parameters, hydrological parameters and meteorological parameters in a monitoring area in real time and performing pretreatment; The space-time diagram neural network anomaly detection module is connected with the data acquisition module and is used for constructing a space-time diagram neural network model based on a water system topological structure, extracting and fusing spatial features and time sequence features of the preprocessed data and outputting an anomaly judgment result and an anomaly feature vector; The intelligent tracing module is connected with the space-time diagram neural network anomaly detection module and is used for combining the hydrodynamics reverse model and the pollution source knowledge spectrum after receiving the anomaly judgment result, calculating the posterior probability of the potential pollution source through probabilistic reasoning and outputting a tracing result; And the visual early warning module is respectively connected with the space-time diagram neural network anomaly detection module and the intelligent tracing module fused with the knowledge graph and is used for displaying monitoring states, anomaly information and tracing paths and sending early warning information. Further, the space-time diagram neural network anomaly detection module includes: The water system topological graph construction unit is used for modeling the monitoring station point as a graph node according to the physical water system structure, constructing a directed edge according to the water flow direction and generating graph structure data containing node characteristics and edge weights; the graph attention convolution network unit is used for receiving the graph structure data, calculating dynamic attention coefficients among nodes by utilizing a graph attention mechanism, and aggregating neighbor node information to extract spatial features; the multi-scale time sequence coding unit is used for carrying out multi-scale convolution and position coding on the characteristic sequence of the node historical time step, and extracting time seq