CN-122020034-A - Multi-index water quality detection system and method based on automation

CN122020034ACN 122020034 ACN122020034 ACN 122020034ACN-122020034-A

Abstract

The invention provides an automatic multi-index water quality detection system and method. Belongs to the technical field of water quality detection. The method comprises the steps of monitoring environmental parameters in real time through environmental perception sensors deployed around a water body, dynamically adjusting a sampling strategy based on a reinforcement learning algorithm, obtaining multi-mode sensing data, and carrying out preliminary preprocessing on the collected multi-mode sensing data to obtain an initial multi-mode data set. Through environmental perception starting and self-adaptive sampling, the water quality change can be responded in time, the multi-mode water quality data can be rapidly obtained, and real-time monitoring and early warning are realized.

Inventors

XU YONG
LIANG GUOBIN
ZHU KAI
XIA LI
PENG TAO
ZHU YONGZHENG
LIN WEI

Assignees

德恩特(江苏)环境科技有限公司
江苏理工学院

Dates

Publication Date: 20260512
Application Date: 20260121

Claims (10)

1. An automatic multi-index water quality detection method, which is characterized by comprising the following steps: S1, monitoring environmental parameters in real time through environmental perception sensors deployed around a water body, dynamically adjusting a sampling strategy based on a reinforcement learning algorithm, acquiring multi-mode sensing data, and performing preliminary preprocessing on the acquired multi-mode sensing data to obtain an initial multi-mode data set; s2, on edge computing equipment, adopting a deep learning model to perform fusion processing on the initial multi-mode data set to generate a comprehensive water quality characteristic vector set; s3, uploading the key abnormal data set generated by the edge end to a cloud server through a wireless communication network, deep mining the uploaded data by the cloud server to obtain a pollution tracing result and a trend prediction report, and optimizing an intelligent analysis model and a sampling strategy of the edge end according to the result obtained by cloud analysis to generate an updating strategy; S4, issuing the update strategy generated by the cloud to the edge computing equipment through a wireless communication network, wherein the edge equipment continues to perform water quality detection according to the updated model and strategy; S5, collecting data related to equipment faults in the running process of the system to obtain a fault associated factor data set, carrying out fault mode probability evaluation according to historical fault case data and the fault associated factor data set to obtain a calibration fault mode probability vector, and carrying out dynamic correction on the calibration fault mode probability vector by combining the real-time fault associated factor data to obtain fault probability distribution data, and sending out a fault early warning signal when the fault probability exceeds a preset threshold value; S6, performing fault probability sorting on the candidate fault mode sorting list according to the fault probability distribution data to obtain a fault mode sorting list, acquiring loop topology data of the water quality detection equipment, performing fault mode loop association analysis according to the fault mode sorting list and the loop topology data to obtain a fault mode loop association table, and performing fault region positioning according to the fault mode loop association table and the fault mode sorting list to obtain fault region positioning result data.
2. The automated multi-index water quality testing method of claim 1, wherein S1 comprises: s11, deploying an environment-aware sensor network, and acquiring surrounding environment parameters of a water body in real time through a wide area network to generate an original environment-aware data stream; s12, comparing the original environment sensing data stream with a preset normal range threshold value based on an environment parameter threshold value judgment model, and triggering the water quality detection system to switch from a standby state to an operating state when any parameter exceeds the threshold value or an emergency abnormal event is detected, so as to generate a system starting instruction; S13, based on a deep reinforcement learning algorithm, using historical sampling data, current environment information and a preset detection target as inputs, constructing a state action rewarding function, dynamically optimizing sampling point coordinates, sampling depth, sampling frequency and single sampling quantity, and generating a self-adaptive sampling strategy; S14, controlling an automatic sampling device to execute sampling operation according to the self-adaptive sampling strategy, and synchronously acquiring multi-mode sensing data through an integrated sensor to generate a multi-mode original data set; S15, preliminary preprocessing is carried out on the multi-mode original data set, and an initial multi-mode data set is generated.
3. The automated multi-index water quality testing method of claim 1, wherein S2 comprises: s21, loading an initial multi-mode data set in an edge computing node, constructing a data dividing rule based on data set feature distribution, and dividing the data dividing rule into a physical index subset and a chemical index subset; s22, constructing a double-branch multi-mode fusion network, and carrying out weighted fusion on the double-branch characteristics through an attention mechanism to generate a comprehensive water quality characteristic vector set; s23, based on a pre-trained intelligent analysis model, classifying and identifying the comprehensive water quality feature vector set, and outputting a water quality normal/abnormal judgment result; S24, triggering a preset response rule according to the type and degree of the abnormal index, and generating a self-cleaning device starting instruction for the abnormal physical index; s25, integrating the abnormal index data, responding to the operation record and the corrected sensing data, and generating a key abnormal data set.
4. The automated multi-index water quality testing method of claim 3, wherein the dual branches comprise physical branches and chemical branches, wherein the physical branches extract space-time features by CNN, and the chemical branches capture related features of contaminant concentration by transducer.
5. The automated multi-index water quality testing method of claim 1, wherein S3 comprises: s31, packaging and encrypting a key abnormal data set, real-time environment parameters, a self-adaptive sampling strategy and an equipment operation log generated by an edge end through a 5G slicing technology, and uploading the key abnormal data set, the real-time environment parameters, the self-adaptive sampling strategy and the equipment operation log to a cloud server to generate a cloud input data packet; s32, the cloud server unpacks and secondarily pre-processes the input data packet to construct a standardized analysis data set; S33, constructing a pollution tracing model, fusing river channel topology data of a geographic information system, water flow velocity field data of a hydrological model and a standardized analysis data set, simulating a pollutant diffusion path through a reverse track tracking algorithm, positioning a pollution source, and generating a pollution tracing result; S34, adopting an LSTM time sequence prediction model, taking water quality index data of about 72 hours as input, predicting the index change trend of 48 hours in the future, and generating a trend prediction report by combining a pollution tracing result; s35, based on the trend prediction report and the historical detection precision data, adjusting weight parameters of the intelligent analysis model at the edge end and optimization factors of the self-adaptive sampling strategy through a reinforcement learning algorithm to generate model updating parameters and a sampling strategy adjustment scheme; S36, integrating the model updating parameters and the sampling strategy adjustment scheme to generate a structured updating strategy.
6. The automated multi-index water quality testing method of claim 1, wherein S4 comprises: s41, the cloud server transmits an update strategy to the edge computing equipment through an encryption communication protocol, and a strategy transmitting log is generated; S42, the edge computing equipment receives an update strategy, ensures data integrity through CRC, and analyzes to obtain model update parameters and a sampling strategy adjustment scheme; s43, based on the model updating parameters, performing incremental updating on the local intelligent analysis model, and reconfiguring the operation parameters of the automatic sampling device according to the sampling strategy adjustment scheme; and S44, after updating, the edge equipment re-collects and analyzes the water quality data based on the new model and the strategy to form a continuous evolution closed loop.
7. The automated multi-index water quality testing method of claim 6, wherein S42 comprises: The communication module of the edge computing equipment monitors a designated port, receives an encryption strategy data packet issued by a cloud, temporarily stores the encryption strategy data packet in a local cache area, and generates a data packet receiving record; carrying out integrity check on the encryption strategy data packet in the buffer area through a CRC32 check module, if the encryption strategy data packet passes the check, executing the next step, and if the encryption strategy data packet fails the check, sending a retransmission request to the cloud end to generate a check result report; Decrypting the encrypted policy data packet by using a locally stored equipment exclusive key to obtain a policy structured document, comparing the UUID with the local equipment ID to verify the attribution of the data packet, and generating a decryption verification result; analyzing the strategy structured document, extracting model updating parameters and sampling strategy adjustment schemes, respectively storing the model updating parameters and the sampling strategy adjustment schemes into a parameter configuration area and a strategy configuration area, and generating an analysis result list.
8. The automated multi-index water quality testing method of claim 1, wherein S5 comprises: S51, collecting equipment fault associated data in real time through a state monitoring module and a communication module which are built in a sensor, and generating a fault original data set; s52, based on a historical fault case library, constructing a fault mode probability evaluation model by adopting a Bayesian probability statistical method, inputting a fault original data set, calculating initial occurrence probability of each mode, and generating an initial fault mode probability vector; s53, combining the real-time fault associated data, and dynamically correcting the probability vector of the initial fault mode through a Kalman filtering algorithm to obtain real-time fault probability distribution data; S54, comparing the fault probability distribution data with a preset threshold value, and triggering a fault early warning signal if the fault probability distribution data exceeds the threshold value; s55, making a hierarchical maintenance decision according to the fault mode types and the probability sequence, and generating a maintenance decision list.
9. The automated multi-index water quality testing method of claim 1, wherein S6 comprises: S61, descending order sorting is carried out on the candidate fault mode sorting list according to the fault probability distribution data, so that a fault mode sorting list arranged according to the occurrence probability is obtained; s62, loop topology data of the water quality detection equipment are called from a system configuration library, association degrees of all modes and loop nodes in a fault mode ordering list are analyzed through a graph theory algorithm, and a fault mode loop association table is generated; S63, positioning loop areas in the fault set by adopting cluster analysis based on a fault mode loop association table to obtain fault area positioning result data; S64, calling a fault mode feature description library, and deeply analyzing fault reasons by combining fault region positioning result data to generate a detailed water quality detection system fault reason analysis report; And S65, if the system has no fault, integrating a water quality detection result, a pollution tracing result, a trend prediction report and an equipment running state to generate a water quality detection comprehensive report.
10. An automated multi-index water quality detection system comprising: one or more processors; A memory for storing one or more programs, Wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1 to 9.

Description

Multi-index water quality detection system and method based on automation Technical Field The invention provides an automatic multi-index water quality detection system and method, and belongs to the technical field of water quality detection. Background The traditional water quality detection method generally relies on manual sampling and laboratory analysis, and has the problems of long detection period, poor real-time performance, incapability of timely coping with sudden water quality pollution events and the like. Meanwhile, single index detection is difficult to comprehensively reflect the water quality condition, and is lack of self-adaptive maintenance and intelligent optimization mechanisms for detection equipment, so that the high requirements of modern water quality monitoring are difficult to meet. Disclosure of Invention The invention provides an automatic multi-index water quality detection system and method, which are used for solving the problems mentioned in the background art: The invention provides an automatic multi-index water quality detection method, which comprises the following steps: S1, monitoring environmental parameters in real time through environmental perception sensors deployed around a water body, dynamically adjusting a sampling strategy based on a reinforcement learning algorithm, acquiring multi-mode sensing data, and performing preliminary preprocessing on the acquired multi-mode sensing data to obtain an initial multi-mode data set; S2, on the edge computing equipment, adopting a deep learning model to conduct fusion processing on the initial multi-mode data set, generating a comprehensive water quality characteristic vector set, analyzing the comprehensive water quality characteristic vector set by utilizing a pre-trained intelligent analysis model, judging whether the current water quality is abnormal, and obtaining a key abnormal data set. S3, uploading a key abnormal data set generated by the edge end to a cloud server through a wireless communication network, deeply mining the uploaded data by the cloud server to obtain a pollution tracing result and a trend prediction report, optimizing an intelligent analysis model and a sampling strategy of the edge end according to a result obtained by cloud analysis, and generating an updating strategy; S4, issuing the update strategy generated by the cloud to the edge computing equipment through a wireless communication network, wherein the edge equipment continues to perform water quality detection according to the updated model and strategy; S5, collecting data related to equipment faults in the running process of the system to obtain a fault associated factor data set, carrying out fault mode probability evaluation according to historical fault case data and the fault associated factor data set to obtain a calibration fault mode probability vector, and carrying out dynamic correction on the calibration fault mode probability vector by combining the real-time fault associated factor data to obtain fault probability distribution data, and sending out a fault early warning signal when the fault probability exceeds a preset threshold value; S6, performing fault probability sorting on the candidate fault mode sorting list according to the fault probability distribution data to obtain a fault mode sorting list, acquiring loop topology data of the water quality detection equipment, performing fault mode loop association analysis according to the fault mode sorting list and the loop topology data to obtain a fault mode loop association table, and performing fault region positioning according to the fault mode loop association table and the fault mode sorting list to obtain fault region positioning result data. The invention provides an automatic multi-index water quality detection system, which comprises: one or more processors; A memory for storing one or more programs, Wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of the preceding claims. The intelligent water quality monitoring system has the advantages that the intelligent water quality monitoring system can respond to water quality changes in time, multi-mode water quality data can be obtained quickly, real-time monitoring and early warning are achieved, the accuracy and pertinence of water quality detection are improved by adopting a self-adaptive sampling and edge-cloud collaborative intelligent analysis method driven by reinforcement learning, multi-index water quality conditions such as river water and the like can be detected accurately, a continuous evolution closed loop of perception-decision-execution-learning is formed, the system can continuously optimize models and sampling strategies according to actual running conditions, adapt to different water quality environments and detection requirements, the reliability and stability of the system are enhanced through the exec