Sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment

Abstract

The invention relates to the technical field of sewage treatment, in particular to a sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment. Comprising the following steps: a real-time monitoring unit; a data processing unit; the early warning analysis unit is used for adopting an improved LSTM-random forest fusion model when the data is sufficient and switching to an improved grey prediction model when the data is insufficient based on a data volume self-adaptive double-model switching strategy; a system cooperation unit; and a terminal interaction unit. According to the invention, the model switching control module determines the data sufficient state according to the sample density, achieves the adaptive switching of the improved LSTM-random forest fusion model and the improved gray prediction model, and carries out the special training of the flow-liquid level cooperative change physical characteristics of the blocking and leakage faults in combination with the fault feature training module. The method can adapt to scenes with different data volumes, and improves the adaptability and accuracy of recognition of two types of faults of blockage and leakage of the sewage pipe network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sewage treatment, in particular to a sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment. BACKGROUND

[0002] The sewage pipe network is the core conveying link of the urban sewage treatment system, and its operation state directly affects the sewage treatment efficiency and water environment safety. In actual operation, the pipe network is prone to blockage due to impurity accumulation and leakage due to pipe aging or external damage, and real-time mastery of key operation parameters such as flow, liquid level and water quality is the premise for accurate identification of faults and timely disposal. At present, sewage pipe network monitoring mainly relies on traditional sensing sampling and data processing methods, but is affected by complex working conditions such as many impurities in the pipe network and large parameter fluctuations, often facing problems such as sampling channel blockage, invalid monitoring data, insufficient fault early warning accuracy and response lag, and it is urgent to build an intelligent monitoring and fault early warning analysis system that adapts to the complex scene of the sewage pipe network.

[0003] For example, Chinese patent CN202510483065.4 discloses a sewage treatment management and control method, system and terminal based on big data intelligent analysis; the method is to obtain the flow difference value of the pipe network inlet and outlet in real time, judge whether the difference value exceeds the set threshold and the duration is qualified, if so, trigger the preliminary leakage early warning; then calculate the Pearson correlation coefficient of the flow sudden increase and pressure sudden drop of the adjacent nodes, if the coefficient is less than the set value, mark the high suspicious area, and finally locate the leakage point according to the pressure wave, which can quickly and accurately locate the leakage point to improve the sewage treatment management and control efficiency and response speed. For example, Chinese patent CN202411167269.9 discloses a smart operation sewage pipe system and its implementation method; through the monitoring equipment, cloud server and other 5G / Internet interactive devices, the operation data is collected by using the Internet of Things, combined with big data, intelligent and image algorithm processing and analysis, the pipe defect prompt, repair and evaluation are realized, the implementation steps include building the system, arranging the equipment, establishing the GIS model, etc., which can effectively dispatch the operation system, allocate operation and maintenance resources, improve the operation efficiency and quality, and solve the problems of external water entering and sewage leakage caused by pipe defects.

[0004] Although the above technical solutions have corresponding design advantages, the above technical solutions still have the following technical defects: firstly, the sampling link lacks anti-pollution design and dynamic adaptation capability, which cannot guarantee the quality of monitoring data: CN202510483065.4 only focuses on the flow and pressure data processing of leakage faults, does not mention the anti-pollution sampling structure for suspended impurities in the pipe network, and does not design a backflushing cleaning mechanism, which is easy to be blocked by impurities in the sampling channel during long-term operation; CN202411167269.9 mentions collecting operation data through the Internet of Things, but does not explicitly design the anti-clogging optimization of the sampling device, and both of them do not use the strategy of "dynamic adjustment of sampling frequency with liquid level fluctuation", which cannot collect data to capture parameter mutation process when the liquid level of the pipe network rises or falls suddenly, resulting in insufficient integrity and effectiveness of the monitoring data; secondly, the fault early warning system is not perfect, and the precision and scene adaptability are lacking: CN202510483065.4 only focuses on leakage fault positioning and does not cover the common blockage fault type of the sewage pipe network, and does not consider the early warning model adaptation scheme when the data is insufficient, which cannot deal with the sample shortage scene caused by the initial monitoring or data transmission interruption; CN202411167269.9 mentions combining intelligent algorithms to process data, but does not explicitly mention the model architecture, and does not fuse the structural characteristics of the pipe network such as pipe diameter and slope with time domain dynamic characteristics such as flow fluctuation amplitude and peak duration, resulting in weak correlation between fault identification and physical properties of the pipe network, which is difficult to meet the high-precision early warning demand in different operation scenarios; thirdly, there is a lack of cooperative closed-loop mechanism between early warning and operation and sewage treatment process, and the fault response efficiency is low: CN202510483065.4 only completes the positioning of leakage points and does not design subsequent disposal links such as intelligent pushing of operation work order and pipe network valve adjustment, which cannot promote the landing of early warning results; the operation scheduling scheme of CN202411167269.9 does not develop differentiated valve adjustment strategies for blockage and leakage faults (such as upstream flow limiting and shunting during blockage, and upstream and downstream valve control range during leakage), and does not link the influent control system of the sewage treatment plant to adapt to the change of pipe network flow, which cannot form a complete closed loop of "early warning-operation disposal-process adaptation", resulting in lagging fault response. In view of this, we propose a sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment. SUMMARY

[0005] The purpose of the present application is to provide a sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment, to solve the problems of lack of anti-pollution design and dynamic adaptation capability in the sampling link, which cannot guarantee the quality of monitoring data, imperfect fault early warning system, lack of precision and scene adaptability, and lack of cooperative closed-loop mechanism between early warning and operation and sewage treatment process, which results in low fault response efficiency.

[0006] To achieve the above technical problems, the purpose of the present application is to provide a sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment, comprising: A real-time monitoring unit is configured to collect flow, liquid level and water quality operation data of the sewage pipe network, adopt a double-stage anti-pollution sampling structure and integrate a backflushing cleaning function, and dynamically adjust the sampling frequency according to the liquid level fluctuation to cope with the pipe network impurity blockage and parameter mutation scenarios. A data processing unit is configured to perform effectiveness screening and fault feature mining based on the monitoring data transmitted by the real-time monitoring unit, pre-judge invalid data in combination with the edge pipe network diameter and slope parameters, and simultaneously perform fusion analysis through the remote time domain features and pipe network structure features to reduce data transmission redundancy and enhance fault correlation. An early warning analysis unit is configured to identify sewage pipe network blockage and leakage faults and output risk early warnings, adopt a double-model switching strategy based on data volume self-adaptation, switch to an improved gray prediction model when the data is insufficient, and simultaneously perform special feature training on the physical characteristics of the “flow-liquid level coordinated change” of the two types of faults to adapt to data fluctuations and improve fault identification accuracy. A system coordination unit is configured to realize the coordinated scheduling of early warning, operation and maintenance, and sewage treatment processes by adopting a fault level driving mechanism, shorten the fault response lag by intelligently pushing operation and maintenance work orders and cooperating with the pipe network valve adjustment strategies for blockage and leakage differentiation. A terminal interaction unit is configured to display real-time monitoring data and early warning results and support basic parameter configuration, and realize multi-terminal data interaction through a Web terminal data display interface and a mobile terminal early warning information push.

[0007] As a further improvement of the technical solution, the real-time monitoring unit comprises a double-stage anti-pollution sampling module and a multi-parameter sensing module, wherein: The double-stage anti-pollution sampling module adopts a series filter structure, the first-stage filter screen is a corrosion-resistant metal filter screen with a large aperture design to intercept large suspended impurities, the second-stage filter screen is a corrosion-resistant high polymer filter screen with a small aperture design to intercept fine suspended solids, the two-stage filter screens are installed in sequence along the water flow direction of the pipe network, and the filter screen edges are attached to the inner wall of the sampling port through a sealing structure to reduce the infiltration of unfiltered sewage. The multi-parameter sensing module integrates flow sensors, liquid level sensors and water quality sensors, and the detection probes of the flow sensors, liquid level sensors and water quality sensors are all arranged on the downstream side of the double-stage anti-pollution sampling module to avoid direct contact with unfiltered sewage.

[0008] As a further improvement of the technical solution, the real-time monitoring unit further comprises a backflushing cleaning module and a sampling frequency adjustment module, wherein: The backflush cleaning module comprises a differential pressure detection submodule, a logic control submodule, a reverse power submodule and an anti-backflow sewage discharge submodule, wherein: The differential pressure detection submodule adopts two industrial-grade pressure sensors, which are respectively fixed on the upstream pipe section of the primary filter screen and the downstream pipe section of the secondary filter screen of the double-stage anti-pollution sampling module. Based on the physical relationship between fluid resistance and pressure difference, the pressure data between the two points are collected and transmitted in real time. The logic control submodule is built-in programmable logic chip, and pre-stores two trigger conditions of "filter screen normal working pressure difference interval" and "longest continuous running time". When the pressure difference transmitted by the differential pressure detection submodule exceeds the normal interval, or the cumulative running time of the backflush cleaning module since the last backflush is reached, the logic control submodule generates a backflush start instruction. The reverse power submodule comprises a micro diaphragm water pump and a water flow switching valve. The water pump inlet is connected to an external cleaning water source, and the outlet is connected to the downstream of the secondary filter screen and the upstream of the primary filter screen through the switching valve. After receiving the start instruction, the water pump first sends reverse water flow to the downstream of the secondary filter screen through the switching valve, and then switches to the upstream of the primary filter screen to send reverse water flow after a preset time. The anti-backflow sewage discharge submodule is composed of a sewage discharge pipe and a one-way check valve. One end of the sewage discharge pipe is connected to the impurity collection area upstream of the primary filter screen, and the other end is connected to the main channel of the sewage pipe network. The sewage containing impurities generated by backflush is discharged through the sewage discharge pipe, and the one-way check valve blocks the reverse flow of sewage from the pipe network into the backflush system, avoiding pollution of the cleaning water source. The sampling frequency adjustment module is data-linked with the liquid level sensor of the multi-parameter sensing module. The sampling frequency is adjusted by real-time identification of the liquid level change trend transmitted by the liquid level sensor: when the liquid level change trend is gentle, the multi-parameter sensing module is controlled to maintain the basic sampling frequency; when the liquid level change trend is significant, the frequency adjustment instruction is issued to the multi-parameter sensing module to encrypt the collection of flow, liquid level and water quality data, so as to ensure the capture of the parameter mutation process of the pipe network, and the frequency adjustment instruction is transmitted in real time through the low-power controller built-in the sampling frequency adjustment module.

[0009] As a further improvement of the technical solution, the data processing unit comprises an edge data preprocessing module and a remote feature fusion module, wherein: The edge data preprocessing module is used for preliminary processing of the monitoring data transmitted by the real-time monitoring unit, specifically including: Importing the basic parameters of the sewage pipe network and establishing the mapping relationship between the parameters and the corresponding monitoring points; Based on the basic parameters of the sewage pipe network and the pipe flow fluid mechanics law, the reasonable fluctuation interval of the flow and liquid level of each monitoring point is preset. When the original data exceeds the interval, it is marked as invalid data; The invalid data marked is removed, and the missing value of data transmission is supplemented by using an interpolation method based on the trend of adjacent valid data, and complete pre-processing data is output. The remote feature fusion module receives the pre-processing data output by the edge data pre-processing module, and is used for mining fault features, including extracting time domain dynamic features and pipe network structure features of the data, and performing fusion analysis on the time domain dynamic features and the pipe network structure features.

[0010] As a further improvement of the technical solution, the remote feature fusion module comprises a feature extraction submodule and a feature fusion submodule, wherein: The feature extraction submodule is used for extracting two types of key features from the edge pre-processing data, and specifically comprises: The time domain dynamic features include the unit time fluctuation amplitude of the flow and the liquid level, the frequency of continuous abnormal values, and the peak value duration, are obtained by time series slicing analysis of the pre-processing data through a sliding window algorithm, and reflect the time dimension change law of the pipe network operation state; The pipe network structure features are calculated based on the pipe diameter and slope parameters imported by the edge, combined with the aging coefficient of the pipe material, and the actual flow capacity coefficient and resistance loss coefficient of the pipe section, and reflect the basic influence of the physical properties of the pipe network on the operation state; The feature fusion submodule is used for realizing the organic integration of the time domain dynamic features and the pipe network structure features, and specifically comprises: A feature correlation degree evaluation mechanism is established, and based on the corresponding relationship of "time domain dynamic features-structure features-fault type" in historical fault data, dynamic weights are assigned to the time domain dynamic features and the pipe network structure features; A weighted fusion algorithm is used to map the time domain dynamic features and the pipe network structure features to the same feature space, to generate a fusion feature set containing time dynamic changes and physical structure properties, for enhancing the relevance and accuracy of subsequent fault identification.

[0011] As a further improvement of the technical solution, the early warning analysis unit comprises a model switching control module, a double model operation module and a fault feature training module, wherein: The model switching control module collects the fusion feature set output by the data processing unit, defines the effective sample number (total amount of verified effective samples), the data collection time (time span of continuous effective data), the number of monitoring points (total number of pipe network monitoring points in the current analysis area), calculates the sample density , and compares it with the preset density threshold (model accuracy critical value based on historical data) to determine the data sufficiency state; The double model operation module pre-stores an improved LSTM-random forest fusion model and an improved gray prediction model, and calls the corresponding model according to the instruction of the model switching control module; The fault feature training module constructs a feature training set based on the physical characteristics of the blockage fault "flow rate drop-liquid level rise" and the leakage fault "flow rate drop-liquid level drop", which is used for model parameter optimization.

[0012] As a further improvement of the technical solution, the improved gray prediction model in the double model operation module includes a prediction correction submodule and a risk level output submodule, wherein: The prediction correction submodule receives the time series data in the fusion feature set and defines the following parameters: The basic prediction value is the flow rate or liquid level prediction result output by the traditional gray prediction model; The blockage compensation term is set based on the average rise rate of the liquid level in historical blockage faults and a blockage correction coefficient ; wherein, is used to quantify the correction amplitude of ; The leakage compensation term is set based on the average drop rate of the flow rate in historical leakage faults and a leakage correction coefficient ; wherein, is used to quantify the correction amplitude of ; The prediction correction submodule superimposes the blockage scene on the basic prediction value to obtain the corrected prediction value , and superimposes the leakage scene on the basic prediction value to obtain the corrected prediction value ; The risk level output submodule defines the historical normal operation data mean , the normal data standard deviation , and defines the blockage scene deviation and the leakage scene deviation , respectively; according to the size of , , the risk level is output correspondingly, representing low, medium and high risk, respectively.

[0013] As a further improvement of the technical solution, the improved LSTM-random forest fusion model in the double model operation module includes a time series feature enhancement submodule and a classification decision submodule, wherein: ​​The temporal feature enhancement submodule receives the temporal data in the fusion feature set and defines the following parameters: Feature Dimension , is the number of flow, level, and velocity monitoring features; Time series length , is the number of consecutive sampling moments; Attention weight matrix , whose elements For the Class features in The weight of the moment, satisfaction And the sum of the weights of each row is 1; Traffic sudden change threshold and abnormal liquid level fluctuation threshold , are all preset feature thresholds; The temporal feature enhancement submodule satisfies The sudden change characteristics of flow Abnormal liquid level fluctuation characteristics, improve the corresponding position ;in, Indicates the flow difference between adjacent collection times, Indicates the time interval between adjacent acquisition moments set by the sampling frequency adjustment module. Indicates the liquid level difference between adjacent collection moments; The classification decision submodule defines a preset decision threshold , and use the random forest algorithm to classify the enhanced features and output the probability of blockage failure and leakage failure probability , and The value range of is [0,1]; when Reaching the preset decision threshold , it is determined to be a corresponding type of fault.

[0014] As a further improvement of this technical solution, the system collaboration unit includes a fault information analysis module, an operation and maintenance work order scheduling module, a pipe network valve adjustment module and a process adaptation module, wherein: The fault information parsing module receives the blockage and leakage fault type, risk level, and fault location information output by the early warning analysis unit, and simultaneously retrieves the pipe network structure features output by the remote feature fusion module, associates and matches the fault type and risk level with the pipe network structure features, and determines the priority of the coordinated response; The operation and maintenance work order scheduling module generates an operation and maintenance work order containing a fault location, a fault type, a risk level and a corresponding processing suggestion based on the response priority determined by the fault information analysis module; according to a pre-stored mapping relationship of an operation and maintenance team-responsible pipe network area, the work order is automatically pushed to a corresponding operation and maintenance team, and the operation and maintenance scheduling closed loop is formed by receiving the work order receiving state, the on-site disposal progress and the fault repair result fed back by the operation and maintenance team in real time; The pipe network valve adjustment module receives the fault type and location information of the fault information analysis module, combines the real-time flow and liquid level data transmitted by the multi-parameter sensing module, and performs differential adjustment: for the blockage fault, the valve of the associated pipe segment upstream of the fault pipe segment is adjusted to reduce the inflow, and the valve of the standby pipe segment downstream is opened to realize flow distribution; for the leakage fault, the associated control valves upstream and downstream of the fault point are closed to limit the leakage range, and the valve state is fed back to the fault information analysis module during the adjustment process; The process adaptation module establishes data linkage with the pipe network valve adjustment module, receives the actual flow data of the pipe network after valve adjustment, pushes the flow change information to the influent regulation system of the sewage treatment plant, and issues process adjustment instructions to the sewage treatment plant according to the flow change to ensure that the sewage treatment process is adapted to the operation state of the pipe network.

[0015] As a further improvement of the technical solution, the terminal interaction unit includes a data visualization module, an early warning interaction module and a parameter configuration module, wherein: The data visualization module receives the flow and liquid level data of the real-time monitoring unit and the fault early warning result of the early warning analysis unit, builds a pipe network topology visualization interface on the Web end, dynamically identifies and presents the monitoring point position, flow trend and fault state; supports filtering and displaying historical monitoring data according to the monitoring point and time dimension to form a data change curve; The early warning interaction module receives the fault risk level of the early warning analysis unit, triggers a Web end pop-up window reminder and a mobile end instant notification for high-risk faults, and generates a timing summary report for low-risk faults; the user can mark the early warning processing state through the mobile end and feed back the state to the system collaboration unit; The parameter configuration module is used to provide a configuration interface of monitoring frequency, early warning notification mode and data storage period, and the configuration operation is associated with permission management rules (such as administrators can modify all parameters, and operation and maintenance personnel can only adjust the notification receiving mode); after the parameters are modified, the parameters are automatically synchronized to the real-time monitoring unit and the data processing unit to realize multi-unit parameter collaborative update.

[0016] Compared with the prior art, the present application has the following advantages: 1. The present application can effectively solve the problem of sewage pipe network impurity blockage of the sampling channel, reduce the infiltration of unfiltered sewage, ensure the continuity of monitoring data collection, and avoid key data loss caused by parameter mutation by setting double-stage anti-pollution sampling module and backflush cleaning module in the real-time monitoring unit, adopting series corrosion-resistant filter screen to intercept large and fine suspended impurities, combining differential pressure detection and timing trigger to realize reverse flushing of the filter screen, and cooperating with the strategy of dynamic adjustment of sampling frequency with liquid level fluctuation; 2. The present application can reduce data transmission redundancy, enhance the relevance of monitoring data and pipe network physical properties, and improve the effectiveness of data used for subsequent fault analysis by using edge data preprocessing module, combining basic parameters such as pipe diameter and slope of sewage pipe network with pipe flow hydrodynamic law to predict and eliminate invalid data, using adjacent valid data trend interpolation to supplement missing values, and then extracting time domain dynamic features and pipe network structure features through the remote feature fusion module and performing weighted fusion; 3. The present application can adapt to different data volume scenarios and improve the adaptability and accuracy of sewage pipe network blockage and leakage fault identification by using model switching control module to determine the data sufficiency state according to sample density, realizing adaptive switching of improved LSTM-random forest fusion model and improved gray prediction model, and combining fault feature training module to carry out special training for the physical characteristics of “flow-liquid level coordinated change” of blockage and leakage faults; 4. The system coordination unit of the present application adopts a fault level driven mechanism, determines the response priority based on fault type, risk level and pipe network structure characteristics, generates operation and maintenance work orders and automatically pushes them to the corresponding operation and maintenance team, executes differentiated pipe network valve adjustment strategy for blockage and leakage faults, and links the sewage treatment plant influent regulation system to adapt to flow changes, which can shorten the fault response lag time, realize the coordinated scheduling of early warning, operation and maintenance disposal and sewage treatment process, and form a fault disposal closed loop. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The figure is a system framework diagram of the present application; The meanings of the various numbers in the figure are as follows: 100, real-time monitoring unit; 110, double-stage anti-pollution sampling module; 120, multi-parameter sensing module; 130, backflush cleaning module; 131, differential pressure detection submodule; 132, logic control submodule; 133, reverse power submodule; 134, anti-backflow blowdown submodule; 140, sampling frequency adjustment module; 200, data processing unit; 210, edge data preprocessing module; 220, remote feature fusion module; 221, feature extraction submodule; 222, feature fusion submodule; 300, early warning analysis unit; 310, model switching control module; 320, dual-model operation module; 321, time series feature enhancement submodule; 322, classification decision submodule; 323, prediction correction submodule; 324, risk level output submodule; 330, fault feature training module; 400, system collaboration unit; 410, fault information analysis module; 420, operation and maintenance work order scheduling module; 430, pipeline valve adjustment module; 440, process adaptation module; 500, terminal interaction unit; 510, data visualization module; 520, early warning interaction module; 530, parameter configuration module. DETAILED DESCRIPTION

[0018] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] like Figure 1 As shown, this embodiment provides a sewage pipe network intelligent monitoring and fault warning analysis system for sewage treatment, including: The real-time monitoring unit 100 is used to collect the flow, liquid level, and water quality operation data of the sewage pipe network. It adopts a two-stage anti-fouling sampling structure and integrates a backwash cleaning function. The sampling frequency is dynamically adjusted according to liquid level fluctuations to cope with scenarios such as pipe network impurity blockage and parameter mutations. In this embodiment, the real-time monitoring unit 100 includes a dual-stage anti-pollution sampling module 110 and a multi-parameter sensing module 120, wherein: The two-stage anti-pollution sampling module 110 adopts a serial filtration structure. The first-stage filter is made of corrosion-resistant metal material with a large aperture design to intercept large suspended impurities. The second-stage filter is made of corrosion-resistant polymer material with a small aperture design to intercept fine suspended matter. The two-stage filter is installed in sequence along the water flow direction of the pipe network. The edge of the filter is fitted with the inner wall of the sampling port through a sealing structure to reduce the infiltration of unfiltered sewage. As a further explanation of this embodiment, the first-level filter in this embodiment can be made of 316L stainless steel, which takes advantage of its resistance to chloride ion corrosion to adapt to the sewage environment, and the large-aperture design matches the functional requirement of "intercepting large suspended impurities"; the second-level filter can be made of PVDF (polyvinylidene fluoride), which takes advantage of its acid and alkali resistance and uniform pores to achieve effective interception of fine suspended matter; in addition, the sampling port seal can use a fluororubber sealing ring (resistant to long-term sewage erosion), and the pipeline of the backwash cleaning module 130 is made of UPVC material (taking into account both corrosion resistance and cost, and adapting to reverse water flow pressure).

[0020] The multi-parameter sensing module 120 integrates a flow sensor, a liquid level sensor, and a water quality sensor, and the detection probes of the flow sensor, the liquid level sensor, and the water quality sensor are arranged on the downstream side of the double-stage anti-pollution sampling module 110 to avoid direct contact with unfiltered sewage.

[0021] In the embodiment, the real-time monitoring unit 100 further includes a backflushing cleaning module 130 and a sampling frequency adjusting module 140, wherein: The backflushing cleaning module 130 includes a differential pressure detection sub-module 131, a logic control sub-module 132, a reverse power sub-module 133, and a backflow prevention and sewage discharge sub-module 134, wherein: The differential pressure detection sub-module 131 adopts two industrial-grade pressure sensors, which are respectively fixed on the pipe section upstream of the primary filter screen and the pipe section downstream of the secondary filter screen of the double-stage anti-pollution sampling module 110. Based on the physical relationship between fluid resistance and pressure difference, the differential pressure detection sub-module 131 collects and transmits the pressure data between the two points in real time. The logic control sub-module 132 is built-in with a programmable logic chip and pre-stores two trigger conditions of “normal working pressure difference interval of the filter screen” and “longest continuous running time”. When the pressure difference transmitted by the differential pressure detection sub-module 131 exceeds the normal interval or the cumulative running time of the backflushing cleaning module 130 since the last backflushing is completed reaches the set value, the logic control sub-module 132 generates a backflushing start instruction. The reverse power sub-module 133 includes a micro diaphragm water pump and a water flow switching valve. The water pump inlet is connected to an external cleaning water source, and the outlet is connected to the downstream of the secondary filter screen and the upstream of the primary filter screen through the switching valve. After receiving the start instruction, the water pump first sends reverse water flow to the downstream of the secondary filter screen through the switching valve, and then switches to send reverse water flow to the upstream of the primary filter screen after a preset time. The backflow prevention and sewage discharge sub-module 134 is composed of a sewage discharge pipe and a one-way check valve. One end of the sewage discharge pipe is connected to the impurity collection area upstream of the primary filter screen, and the other end is connected to the main channel of the sewage pipe network. The sewage containing impurities generated by backflushing is discharged through the sewage discharge pipe, and the one-way check valve blocks the reverse flow of sewage in the pipe network into the backflushing system to avoid pollution of the cleaning water source. As a further description of the embodiment, when the backflushing cleaning module 130 is started, the logic control sub-module 132 sends a “backflushing pause instruction” to the sampling frequency adjusting module 140: the multi-parameter sensing module 120 pauses data collection (to avoid interference of the backflushing water flow with the monitoring value); after backflushing (the reverse power sub-module 133 completes water flow switching and a preset time of running), the sampling is automatically restored to ensure data continuity.

[0022] Further, the reverse power submodule 133 in the embodiment is built-in current monitoring circuit, when the water pump working current continuously deviates from the rated range, it is determined that the filter screen is blocked or the water pump is stuck, triggering local sound and light alarm and pushing fault code to the terminal interaction unit; at the same time, the one-way check valve of the anti-backflow sewage submodule 134 integrates a magnetic sensitive state sensor, which synchronously triggers an alarm when abnormally starts and stops. In addition, the multi-parameter sensing module 120 adopts a master-backup dual-probe architecture, automatically switches to the standby probe when the master probe fails, and maintains the basic sampling frequency of the standby probe, ensuring the continuity of core data collection.

[0023] Further, the backflushing cleaning module 130 is configured with a differential pressure state indicator light (green: normal, yellow: warning, red: blockage), combined with the parameter configuration module 530 of the terminal interaction unit 500, to realize remote adjustment of the backflushing period and trigger threshold. The low-power controller of the sampling frequency adjustment module 140 reserves a Micro-USB debugging interface, which not only supports local calibration of liquid level change rate calculation logic, but also can forward upgrade packages through the data processing unit 200 to complete remote program upgrade.

[0024] The sampling frequency adjustment module 140 establishes data linkage with the liquid level sensor of the multi-parameter sensing module 120, and adjusts the sampling frequency by real-time identification of the liquid level change trend transmitted by the liquid level sensor: when the liquid level change trend is gentle, the multi-parameter sensing module 120 maintains the basic sampling frequency; when the liquid level change trend is significant, the frequency adjustment instruction is sent to the multi-parameter sensing module 120 to encrypt the collection of flow, liquid level and water quality data, to ensure the capture of the parameter mutation process of the pipe network, and the frequency adjustment instruction is transmitted in real time through the low-power controller built-in the sampling frequency adjustment module 140.

[0025] As a further illustration of the embodiment, the sampling frequency adjustment module 140 is triggered by "liquid level change rate + duration" double conditions, specifically including: The change rate of the liquid level sensor is calculated in real time for 3 consecutive sampling periods, and the calculation formula is: ; When the change rate exceeds the preset interval (the interval value can be remotely adjusted by the parameter configuration module of the terminal interaction unit 500), and lasts for 2 periods, it is determined that the liquid level change trend is significant, and the frequency is raised.

[0026] It should be noted that the real-time monitoring unit 100 and the data processing unit 200 are connected through the RS485 bus (adapted to the wiring scene of the inspection well), and follow the Modbus-RTU protocol to ensure anti-interference and transmission efficiency; the power supply adopts a compatible scheme of "direct current 24V industrial power supply + lithium battery energy storage module", which preferentially accesses the power of the pipe network inspection well, and the lithium battery supports continuous operation (the time length depends on the device power consumption and monitoring interval configuration) when there is no power supply, and a solar power supplement interface is reserved for outdoor scenes.

[0027] The data processing unit 200 performs validity screening and fault feature mining based on the monitoring data transmitted by the real-time monitoring unit 100, predicts invalid data in combination with edge pipe network diameter and slope parameters, synchronously performs fusion analysis through remote time domain features and pipe network structure features, reduces data transmission redundancy, and enhances fault correlation; In the embodiment, the data processing unit 200 includes an edge data preprocessing module 210 and a remote feature fusion module 220, wherein: The edge data preprocessing module 210 is configured to perform preliminary processing on the monitoring data transmitted by the real-time monitoring unit 100, and specifically includes: Importing the basic parameters of the sewage pipe network and establishing a mapping relationship between the parameters and the corresponding monitoring points; Based on the basic parameters of the sewage pipe network and the pipe flow fluid mechanics law, the reasonable fluctuation range of the flow and liquid level of each monitoring point is preset, and the original data is marked as invalid data when it is out of the range; The marked invalid data is removed, and the interpolation method based on the trend of adjacent valid data is used to supplement the missing values of data transmission, and the complete preprocessed data is output; As a further description of the embodiment, the core function of the edge data preprocessing module 210 is to preliminarily purify and complete the original monitoring data transmitted by the real-time monitoring unit 100. First, the edge data preprocessing module 210 obtains the basic parameters (diameter, slope, pipe material, etc.) of the sewage pipe network through the pipe network GIS system. If the GIS system data is missing, it is supplemented by manual checking and recording, and then a one-to-one mapping relationship between the monitoring point ID and the basic parameters (for example, the monitoring point number 101 corresponds to the pipe diameter, slope 0.3‰, and ductile iron material parameters) is established. All parameters are stored in the EEPROM chip built-in the edge data preprocessing module 210, which supports manual update through local USB interface or remote rewriting through terminal interaction unit 500.

[0028] Further, in order to accurately determine invalid data, the edge data preprocessing module 210 presets the reasonable fluctuation range of the flow and liquid level of each monitoring point based on the pipe flow fluid mechanics law. Among them, the theoretical range of flow is calculated by using the Manning formula, and the theoretical value formula of flow under full flow state is . In the formula, is the pipe roughness (such as 0.013 for ductile iron material and 0.009 for PVC material, which can be assigned according to the actual pipe material table), is the pipe cross-sectional area (the calculation formula is , is the pipe diameter), is the hydraulic radius (under full flow state ), Pipe network slope (thousandth percentage value directly substituted, such as 0.3 ‰ is substituted 0.0003). Based on the theoretical value, the edge end data preprocessing module 210 presets the flow fluctuation interval as , reserving 20% of the working condition fluctuation margin; The liquid level theoretical range is based on the full-flow liquid level of the pipeline ( The inner diameter of the pipeline), combined with the correlation between flow and liquid level, and preset as , avoiding extreme invalid values in the empty pipe or full overflow state. When the original data exceeds the above interval, if it exceeds for 3 consecutive sampling periods, it is determined as persistent invalid data and is directly excluded, and if only a single period is abnormal, it is marked as suspicious data for temporary storage for verification. For missing values occurring in the data transmission process, the edge end data preprocessing module 210 uses a linear trend interpolation method based on the trend of adjacent valid data to supplement. Let the time stamps and monitoring values of the valid data before and after the missing point be , wherein is the time stamp, is the flow or liquid level monitoring value, then the interpolation of the missing point is calculated according to the formula , the trend value of the missing point is fitted through the time and value relationship of adjacent valid data, and the integrity of the output preprocessed data is ensured.

[0029] The remote feature fusion module 220 receives the preprocessed data output by the edge end data preprocessing module 210, and is used to mine fault features, including extracting time domain dynamic features and pipe network structure features of the data, and fusing and analyzing the time domain dynamic features and the pipe network structure features.

[0030] In this embodiment, the remote feature fusion module 220 includes a feature extraction submodule 221 and a feature fusion submodule 222, wherein: The feature extraction submodule 221 is used to extract two types of key features from the edge end preprocessed data, specifically including: Time domain dynamic features: including the unit time fluctuation amplitude of flow and liquid level, the frequency of continuous abnormal values, and the peak duration, which are obtained by time series slicing analysis of preprocessed data through a sliding window algorithm, reflecting the time dimension change law of the pipe network operation state; Pipe network structure features: based on the pipe diameter and slope parameters imported by the edge end, combined with the aging coefficient of the pipe material, the actual flow capacity coefficient and resistance loss coefficient of the pipe section are calculated, reflecting the basic influence of the physical properties of the pipe network on the operation state; As a further illustration of this embodiment, the feature extraction submodule 221 extracts two types of key features, time domain dynamic features and pipe network structure features, from the preprocessed data, wherein: The time-domain dynamic characteristics are obtained by a sliding window algorithm, and the window length W is set to 10 sampling periods by default (supporting remote adjustment by the terminal interaction unit 500); after time series slicing of the preprocessed data, the fluctuation amplitude of the flow and the liquid level per unit time is calculated the frequency of occurrence of continuous abnormal values the time span of continuous peak values (the peak value is defined as a value greater than the mean value of the monitoring value + 1.5 times the standard deviation), so as to reflect the time dimension variation law of the pipe network operation state; The pipe network structure characteristics are calculated based on the basic parameters introduced by the edge: first, the actual flow capacity coefficient is calculated, and the formula is In the formula, is the aging correction roughness, wherein is the pipe aging coefficient, which is assigned according to the service life of the pipe and the corrosion detection result, and the new pipe , the cast iron pipe served for 10 years .2, is the hydraulic radius, is the pipe network slope (the percentage value is directly substituted). Then the resistance loss coefficient is calculated (K is a proportional constant, and the experience value is 1), so as to reflect the basic influence of the physical properties of the pipe network on the operation state.

[0031] The feature fusion sub-module 222 is configured to realize the organic integration of the time-domain dynamic characteristics and the pipe network structure characteristics, and specifically includes: establishing a feature correlation degree evaluation mechanism, assigning dynamic weights to the time-domain dynamic characteristics and the pipe network structure characteristics based on the corresponding relationship of “time-domain dynamic characteristics-structure characteristics-fault type” in the historical fault data; using a weighted fusion algorithm to map the time-domain dynamic characteristics and the pipe network structure characteristics to the same feature space to generate a fusion feature set containing time dynamic changes and physical structure properties, so as to enhance the relevance and accuracy of subsequent fault identification.

[0032] As a further description of the embodiment, the feature fusion sub-module 222 realizes the organic integration of the two types of characteristics by establishing a feature correlation degree evaluation mechanism. The feature correlation degree evaluation mechanism calculates the Pearson correlation coefficient of the time-domain dynamic characteristics , the pipe network structure characteristics and the fault type based on the corresponding relationship of “time-domain dynamic characteristics-structure characteristics-fault type” in the historical fault data, and assigns dynamic weights according to the correlation coefficients. The weight calculation formula is: ​​​ ; ; wherein is a time domain dynamic feature weight, is a pipe network structure feature weight, and satisfies ; Subsequently, the feature fusion sub-module 222 adopts a weighted fusion algorithm to map the two types of features to the same feature space, and fuses the features The calculation formula of the fused features is Finally, the fused feature set containing the time dynamic change and the physical structure attribute is generated, providing high correlation data support for subsequent fault identification.

[0033] It can be understood that, in order to ensure the stable operation and efficient cooperation of the data processing unit 200, in addition to the internal interaction of the edge data preprocessing module 210 and the remote feature fusion module 220, cross-unit linkage of the data processing unit 200 and the real-time monitoring unit 100, the system cooperation unit 400, and the terminal interaction unit 500 is also involved, and the specific mechanism is as follows: On the one hand, the edge data preprocessing module 210 adopts the LZ4 lossless compression algorithm to perform lightweight processing on the output preprocessed data, and transmits the preprocessed data to the remote end through the RS485 bus (adapted to the check well wiring scene) or the LoRa module (adapted to the non-wiring dispersed deployment scene). At the same time, the module is built-in with a cache chip, which can temporarily store preprocessed data for nearly 24 hours when the network is interrupted, and automatically continue transmission after the network is restored, avoiding data loss.

[0034] On the other hand, if the invalid data accounts for more than 30% in the edge end for 10 consecutive sampling periods, it is determined that the monitoring point data is abnormal. At this time, the edge data preprocessing module 210 sends a "data anomaly warning" instruction to the real-time monitoring unit 100, triggers the sampling frequency adjustment module 140 to increase the sampling frequency to encrypt the collected data to verify the validity, and at the same time pushes the "suggestion for maintenance of monitoring equipment" prompt to the terminal interaction unit 500, and generates a maintenance task through the operation and maintenance work order scheduling module 420 to ensure that the monitoring equipment fault can be timely investigated and analyzed.

[0035] It needs to be supplemented that, on the one hand, in terms of operation and configuration design, the edge data preprocessing module 210 reserves a USB configuration interface, supports local import of pipe network parameters (such as pipe diameter change, slope adjustment, etc.) by operation and maintenance personnel, and can also update the fluctuation interval threshold of flow and liquid level, the parameters of the linear interpolation algorithm (such as the time interval determination standard of adjacent valid data) and the sliding window length ; On the other hand, the remote feature fusion module 220 supports offline import of pipe network historical fault data (such as the blockage and leakage fault records stored in the operation and maintenance database), optimizes the feature correlation degree evaluation mechanism by updating the correspondence relationship of "time domain dynamic features-structural features-fault types", and automatically adjusts the dynamic weight distribution logic; when a new fault type occurs in the pipe network, the labeled fault data can be uploaded through the terminal interaction unit 500 to expand the feature adaptation range of the feature fusion sub-module 222, and ensure that the fault feature mining capability of the data processing unit 200 matches the actual operation demand of the pipe network.

[0036] The early warning analysis unit 300 is used for identifying sewage pipe network blockage and leakage faults and outputting risk early warning. Based on the data adaptive double model switching strategy, the improved LSTM-random forest fusion model is used when the data is sufficient, and the improved gray prediction model is switched to when the data is insufficient. Meanwhile, special feature training is carried out for the "flow-liquid level collaborative change" physical characteristics of the two types of faults, so as to adapt to data fluctuations and improve fault recognition accuracy. In the embodiment, the early warning analysis unit 300 includes a model switching control module 310, a double model operation module 320 and a fault feature training module 330, wherein: The model switching control module 310 collects the fusion feature set output by the data processing unit 200, defines the effective sample number (total amount of verified effective samples), data collection time length (time span of continuous effective data), number of monitoring points (total number of pipe network monitoring points in the current analysis area), calculates the sample density , and compares it with the preset density threshold (model accuracy critical value based on historical data) to determine the data sufficient state; As a further description of the embodiment, the sample density calculation and model switching logic in the model switching control module 310 of the embodiment specifically include: based on the fusion feature set output by the data processing unit 200, the data sufficient state is determined by defining parameters, calculating sample density and comparing threshold, and then the corresponding model is called by the double model operation module 320; Further, the sample density determination and model switching of the model switching control module 310 of the embodiment include the following steps: First, the fusion feature set output by the data processing unit 200 is collected, the effective sample number (total amount of verified effective samples), data collection time length (time span of continuous effective data), number of monitoring points (total number of pipe network monitoring points in the current analysis area) is defined; Subsequently, the sample density is calculated according to the formula Calculate sample density ; Then, the calculated sample density is compared with a preset density threshold ; Next, if , it is determined as a "data sufficient scenario", triggering the dual model operation module 320 to call the improved LSTM-random forest fusion model; If , it is determined as a "data insufficient scenario" (such as the initial monitoring period or network interruption recovery period), triggering the call of the improved gray prediction model; If is in the preset critical interval ( is the lower limit of the data volume threshold, which is configured by the terminal), the current model is maintained, and a "data volume critical prompt" is pushed to the terminal interaction unit 500, prompting the operation and maintenance personnel to check the data acquisition state.

[0037] Finally, the last three groups of output results of the previous model are retained during model switching, and the consistency of the trend is verified with the first three groups of output results of the new model (such as the change amplitude of the fault risk level ≤2 levels, then directly switch; otherwise, use "weighted transition": the weight of the previous model result is reduced by 50% gradually, and the weight of the new model result is increased by 50% gradually), to avoid sudden changes in warning results.

[0038] The dual model operation module 320 pre-stores the improved LSTM-random forest fusion model and the improved gray prediction model, and calls the corresponding model according to the instruction of the model switching control module 310; The fault feature training module 330 constructs a feature training set based on the physical characteristics of the blockage fault "flow decrease-liquid level rise" and the leakage fault "flow decrease-liquid level decrease", which is used for model parameter optimization.

[0039] As a further description of this embodiment, in the model iteration of the fault feature training module 330 of this embodiment, backtesting specifically includes: when the cumulative number of newly added training samples reaches 50 (configurable) or an iteration instruction is received, the model parameters are updated by incremental training, and the consistency of historical sample backtesting is verified for performance, with the formula being: ; wherein represents the consistency percentage of historical sample backtesting (reflecting the matching degree of model prediction and actual fault); If , the model parameters are updated, otherwise the original parameters are retained and an iteration failure prompt is pushed.

[0040] In this embodiment, the improved grey prediction model in the dual-model operation module 320 includes a prediction correction submodule 323 and a risk level output submodule 324, wherein: The prediction and correction submodule 323 receives the time series data in the fusion feature set and defines the following parameters: Basic forecast value , is the flow or liquid level prediction result output by the traditional grey prediction model; Congestion compensation item , based on the average rate of rise of the liquid level in historical blockage faults and blockage correction factor Setting; among them, For quantification right the extent of the correction; Leakage compensation item , based on the average rate of decline of flow in historical leakage failures and leakage correction factor Setting; among them, For quantification right the extent of the correction; The prediction and correction submodule 323 superimposes the congestion scene Get the revised forecast value , superimpose the leakage scene Get the revised forecast value ; The risk level output submodule 324 defines the mean of historical normal operation data , normal data standard deviation , respectively define the congestion scene deviation , Leakage scenario deviation ;according to 、 The size of the corresponding output risk level , representing low, medium, and high risks respectively.

[0041] In this embodiment, the improved LSTM-random forest fusion model in the dual-model operation module 320 includes a temporal feature enhancement submodule 321 and a classification decision submodule 322, wherein: The temporal feature enhancement submodule 321 receives the temporal data in the fusion feature set and defines the following parameters: Feature Dimension , is the number of flow, level, and velocity monitoring features; Time series length , is the number of consecutive sampling moments; Attention weight matrix , whose elements For the first class feature in the first moment, the weight satisfies and the weight sum of each row is 1; the flow sudden change threshold and the liquid level abnormal fluctuation threshold are both preset feature thresholds; The time sequence feature enhancement submodule 321 enhances the corresponding position of the flow sudden change feature satisfying and the liquid level abnormal fluctuation feature satisfying ; wherein, represents the flow difference value of adjacent collection moments, represents the time interval of adjacent collection moments set by the sampling frequency adjustment module 140, represents the liquid level difference value of adjacent collection moments; The classification decision submodule 322 defines a preset decision threshold , and uses a random forest algorithm to classify the enhanced features, and outputs the blockage fault probability and the leakage fault probability , and both have a value range of [0, 1]; when reaches the preset decision threshold , it is determined as a fault of the corresponding type.

[0042] As a further description of the present embodiment, the improved LSTM-random forest fusion model (data sufficient scenario) in the double model operation module 320 of the present embodiment specifically includes: The time sequence feature enhancement submodule 321 receives the fusion feature set output by the remote feature fusion module 220, and enhances the change trend of the flow and liquid level related time domain features through a sliding time window + feature difference, and corrects the pipe network structure features in combination with the pipe material aging coefficient ; the classification decision submodule 322 calls an LSTM submodel with a 3-layer hidden layer structure to predict the flow and liquid level values in the next 15 minutes, constructs a random forest submodel with 100 decision trees, outputs the fault type probability based on the pipe network structure features and the LSTM prediction residual, and completes the fusion decision through the double model weight allocation based on the prediction residual size; the prediction correction submodule 323 calls the historical fault feature library of the fault feature training module 330, calculates the cosine similarity between the current features and the historical features through the formula , and corrects the fault probability in combination with the pipe segment position information; wherein, represents the current feature vector (multi-dimensional features of fusion flow, liquid level, and pipe network structure), represents the historical feature vector (typical feature vector of historical fault scene).

[0043] Furthermore, in the prediction correction submodule of the dual model operation module 320 of this embodiment, the improvement of the grey prediction model specifically includes: introducing a trend correction item for the scenario of insufficient data. (The average change rate of the first three samples, the upward trend is positive and the downward trend is negative) Optimize the traditional grey prediction model, and the improved formula is: , to enhance the model's adaptability to the short-term trend of the data; among them, Indicates the first Step prediction value (used to restore the change trend of flow / liquid level); Indicates the first sample value (initial baseline value) of the original monitoring data; Indicates the development coefficient (reflects the strength of the data change trend, The smaller the trend, the more stable it is); Indicates the gray action (reflects the degree of influence of external interference on the data, The larger the value, the more significant the interference); Represents the trend correction term (the average rate of change of the first three samples, which is positive if it rises and negative if it falls, to compensate for short-term trend deviations).

[0044] System collaboration unit 400 uses a fault-level driven mechanism to achieve coordinated scheduling of early warning, operation and maintenance, and sewage treatment processes. It reduces fault response lags by intelligently pushing operation and maintenance work orders and adopting differentiated pipe network valve adjustment strategies for blockages and leaks. In this embodiment, the system collaboration unit 400 includes a fault information analysis module 410, an operation and maintenance work order scheduling module 420, a pipe network valve adjustment module 430, and a process adaptation module 440, wherein: The fault information parsing module 410 receives the blockage and leakage fault type, risk level, and fault location information output by the early warning analysis unit 300, and simultaneously retrieves the pipe network structure features output by the remote feature fusion module 220, correlates and matches the fault type and risk level with the pipe network structure features, and determines the priority of the coordinated response; As a further illustration of this embodiment, the specific dimensions and priority rules of the association matching of the fault information analysis module 410 are as follows: Pipeline function attributes: If the faulty pipe section is a trunk pipe (carrying more than 50% of the regional sewage transport volume), regardless of the risk level, the priority is increased by one level by default (e.g., medium risk - high priority); if it is a branch pipe (serving only a single community or road section), the priority is implemented according to the original risk level; Service scope: Count the number of users covered by the faulty pipe section. If the coverage exceeds 500 households, the priority will be increased by one level based on the original risk level (low risk - medium priority, medium risk - high priority); Key node attribute: if the fault pipe section contains key nodes such as pump stations and valve wells, the priority is automatically set to the highest to avoid affecting the overall operation of the subsequent pipe network.

[0045] As a further illustration of the present embodiment, the present embodiment finally divides the priority into three levels, specifically including: emergency response (high priority) (main pipe failure, high-risk failure, covering more than 500 households), regular response (medium priority) (medium-risk failure, branch pipe and covering 200-500 households), delayed response (low priority) (low-risk failure, branch pipe and covering less than 200 households), and the priority result is synchronously pushed to the operation and maintenance work order scheduling module 420 and the pipe network valve adjustment module 430 through the internal data bus as the core basis for subsequent actions.

[0046] Further, the fault information analysis module 410 has an information verification mechanism: if the received fault information is missing (such as no fault location or risk level is not marked), the "information completion request" is automatically sent to the early warning analysis unit 300, and the received information is temporarily stored, and after completion, the correlation matching is performed again to avoid scheduling deviation due to incomplete information.

[0047] The operation and maintenance work order scheduling module 420 generates an operation and maintenance work order containing the fault location, fault type, risk level and corresponding processing suggestion based on the response priority determined by the fault information analysis module 410; according to the pre-stored "operation and maintenance team-responsible pipe network area" mapping relationship, the work order is automatically pushed to the corresponding operation and maintenance team, and the work order receiving state, on-site disposal progress and fault repair result fed back by the operation and maintenance team are received in real time, forming an operation and maintenance scheduling closed loop; As a further illustration of the present embodiment, the operation and maintenance work order of the operation and maintenance work order scheduling module 420 contains the following core fields: Fault identification information: fault unique code (composed of "year + month + random 6 digits", such as 202408123456), fault occurrence time (accurate to minutes), fault pipe section GIS coordinates (supporting mobile APP navigation); Disposal guide information: fault type (blockage / leakage), risk level, recommended disposal time limit (emergency response within 4 hours, regular response within 8 hours, delayed response within 24 hours), preliminary disposal suggestion (blockage fault suggests carrying out dredging equipment, leakage fault suggests carrying out leak detector and repair materials); Associated data attachments: flow / liquid level change curve 1 hour before the fault, pipe network structure diagram of the fault pipe section (annotating upstream and downstream associated valve positions).

[0048] Further, the operation and maintenance work order scheduling module 420 has a built-in "operation and maintenance team-responsible area" mapping database, which stores content including operation and maintenance team name, contact information of the person in charge, mobile APP account, and responsible pipe network area (divided by pipe section number), and supports remote updating (such as adding operation and maintenance teams and adjusting the boundaries of the responsible area) by the terminal interaction unit 500. At the same time, its push logic is: according to the fault pipe section number, the corresponding responsible operation and maintenance team is matched, and the work order is pushed to the mobile APP of the team leader through the 4G / 5G network, and a short message is sent for reminding (including a work order link, which can be clicked to jump to the APP to view details). If no "received" feedback is received within 15 minutes after the work order is pushed, the operation and maintenance work order scheduling module 420 automatically forwards the work order to the standby operation and maintenance team (2-3 groups of operation and maintenance teams are preset in the same area), and pushes a "main operation and maintenance team does not respond" prompt to the terminal interaction unit 500, so as to avoid work order backlog.

[0049] In addition, the operation and maintenance team feeds back the disposal progress through the mobile APP, and the state is divided into "received", "in disposal", "to be accepted", and "completed": "In disposal" needs to synchronously upload on-site photos (such as comparison before and after the unblocking of the blockage point, and repair of the leakage point); "When to be accepted", the operation and maintenance work order scheduling module 420 automatically retrieves the real-time monitoring data (whether the fault pipe section flow / liquid level is restored to normal) of the early warning analysis unit 300, and generates an acceptance form if the data is normal; "After completed", the operation and maintenance work order scheduling module 420 stores the disposal record (including disposal time, used equipment, and repair result) into the historical database, which supports subsequent traceability query.

[0050] The pipe network valve adjustment module 430 receives the fault type and location information of the fault information analysis module 410, combines the real-time flow and liquid level data transmitted by the multi-parameter sensing module 120, and performs differentiated adjustment: for blockage fault, adjust the valve of the associated pipe section upstream of the fault pipe section to reduce the inflow, and open the valve of the standby pipe section downstream to realize flow distribution; for leakage fault, close the associated control valve upstream and downstream of the fault point to limit the leakage range, and synchronously feed back the valve state to the fault information analysis module 410 during the adjustment process; As a further description of the present embodiment, for blockage fault (determined by the early warning analysis unit 300 as "flow reduction-liquid level rise"), the adjustment target is "reduce the inflow of the fault pipe section + enable standby pipe section flow distribution", which specifically includes: Upstream associated pipe section valve adjustment: Obtain real-time flow data of three key valves upstream of the fault pipe section. If the liquid level of the fault pipe section is higher than 10% of the normal interval, adjust the valve opening according to the "near first, far second" principle - the upstream valve closest to the fault point, dynamically adjust according to the liquid level change (every 5% higher than the normal interval, close 1 / 4 of the valve opening), to avoid excessive adjustment range causing abnormal liquid level in the upstream pipe section; Downstream standby pipe section activation: When the fault pipe section flow exceeds 80% of the design flow (the design flow is retrieved from the remote feature fusion module 220), open the valve of the downstream standby pipe section. The opening process is completed in three times (1 / 3 opening each time, interval 5 minutes), while monitoring the standby pipe section flow to ensure that the fault pipe section flow is reduced to below 60% of the design flow after shunting; Adjustment result feedback: After valve adjustment, monitor the liquid level change of the fault pipe section through the liquid level sensor. If the liquid level does not decrease to the normal interval within 30 minutes, readjust the upstream valve opening (increase the closing range) and push the "adjustment to be optimized" prompt to the fault information analysis module 410.

[0051] Further, for leakage faults (determined by the early warning analysis unit 300 "flow drop-liquid level drop"), the adjustment target is "limit leakage range + avoid pipe section negative pressure", the specific steps are: Valve closing sequence: First, close one valve downstream of the fault point (to prevent the spread of sewage to the downstream pipe section), and then close one valve upstream (to prevent the formation of negative pressure in the pipe section, which may cause air to enter and affect subsequent repair); Leakage range isolation: If the length of the fault pipe section exceeds 50 meters, close two valves upstream and downstream (20 meters apart) to form an "isolated section", and open the sewage valve of the isolated section (if any) to drain the remaining sewage into the emergency collection tank; State monitoring and abnormal handling: After the valve is closed, monitor the liquid level change of the isolated section through the multi-parameter sensing module 120. If the liquid level continues to drop (more than 20% of the normal interval), it is determined that there is another leakage point, and the "leakage point not completely isolated" warning is pushed to the fault information analysis module 410, triggering a secondary fault positioning.

[0052] In addition, the pipe network valve adjustment module 430 also has a "adjustment protection threshold": if the torque sensor detects that the torque exceeds the rated value (such as the rated torque of an electric valve 10 N·m) during valve adjustment, the adjustment action is immediately stopped, and the "valve stuck" fault prompt is pushed to the terminal interaction unit 500 to avoid valve damage; At the same time, record the opening at the time of interruption for reference in subsequent operation and maintenance.

[0053] The process adaptation module 440 establishes data linkage with the pipe network valve adjustment module 430, receives the actual pipe network flow data after valve adjustment, pushes the flow change information to the sewage treatment plant inflow control system, and according to the flow change condition, issues process adjustment instructions to the sewage treatment plant to ensure that the sewage treatment process is adapted to the pipe network operation state.

[0054] As a further illustration of the present embodiment, the process adaptation module 440 obtains the following data from the pipe network valve adjustment module 430 through the internal data bus: the fault pipe segment flow before and after valve adjustment, the current pipe network total inflow flow (the sum of all monitoring point flows), and the flow change trend (rise / fall / stable). The data is preprocessed: abnormal values with flow fluctuations exceeding 30% of the normal range are removed (judged as sensor error), and a sliding window algorithm (window length 10 minutes) is used to calculate the average flow as the basis for process adjustment.

[0055] As a further illustration of the present embodiment, the present embodiment generates differentiated process adjustment instructions according to the change ratio of average flow and historical same period normal flow, and the instruction content focuses on the core links of sewage treatment (aeration, chemical dosing, sludge return), specifically including: Flow increase scenario: if the current flow is increased by 10%-20% compared with the historical same period normal flow, issue the instruction of "aeration intensity increased by 5%-10%, chemical dosing increased by 8%-15%" to the sewage treatment plant (the flow increase amplitude is positively correlated with the process adjustment amplitude); if the flow increase exceeds 20%, additionally issue the instruction of "enable standby aeration tank" to avoid insufficient aeration leading to COD removal rate decrease; Flow reduction scenario: if the current flow is reduced by 10%-20% compared with the historical same period normal flow, issue the instruction of "aeration intensity reduced by 5%-10%, chemical dosing reduced by 5%-12%"; if the flow reduction exceeds 20%, issue the instruction of "reduce sludge return ratio (from 50% to 30%)" to avoid high sludge concentration leading to increased load of sedimentation tank; Flow stable scenario: maintain the current process parameters, and send "parameter retention suggestion" to the sewage treatment plant every 30 minutes to ensure process stability.

[0056] Further, the process adaptation module 440 in the present embodiment has a two-way communication channel with the sewage treatment plant inflow control system: If no "instruction received" feedback is received from the sewage treatment plant within 10 minutes, automatically resend the instruction (maximum 3 times), and if the resend still fails, push the "process linkage interruption" warning to the terminal interaction unit 500, prompting the operation and maintenance personnel to manually coordinate; If the sewage treatment plant feedbacks that "process adjustment exceeds the equipment capacity" (such as the aeration system has reached the maximum load), the process adaptation module 440 adjusts the instruction content (such as when the flow rate increases by more than 20%, it is recommended to "extend the hydraulic retention time" instead of activating the standby aeration tank), to ensure that the instruction can be executed.

[0057] It should be noted that, as a further illustration of the present embodiment, the system coordination unit 400 ensures the stability of inter-module coordination and the integrity of fault response through the following mechanisms: Data linkage mode: Each module realizes real-time data transmission through the internal CAN bus. The "priority result" output by the fault information analysis module 410 serves as the core trigger signal, synchronously driving the operation and maintenance work order scheduling, valve regulation, and process adaptation, to avoid timing errors in actions; Emergency bottom-up strategy: If any module (such as the pipe network valve regulation module 430) fails, the fault information analysis module 410 automatically raises the priority by one level and pushes a "module failure requires manual intervention" prompt to the terminal interaction unit 500, guiding the operation and maintenance personnel to complete valve regulation or process coordination through manual operation; Log recording and tracing: All coordinated actions (work order pushing, valve regulation, process instruction issuance) are recorded in the log system built into each module. The log includes "action execution time, execution subject, action result, and associated fault identifier", supporting the terminal interaction unit 500 for querying and exporting, facilitating subsequent operation and maintenance review and responsibility tracing.

[0058] The terminal interaction unit 500 is used to display real-time monitoring data, warning results, and support basic parameter configuration. Through the Web data display interface and mobile terminal warning information push, multi-terminal data interaction is realized.

[0059] In the present embodiment, the terminal interaction unit 500 includes a data visualization module 510, a warning interaction module 520, and a parameter configuration module 530, wherein: The data visualization module 510 receives the flow rate and liquid level data of the real-time monitoring unit 100 and the fault warning results of the warning analysis unit 300, and builds a pipe network topology visualization interface on the Web to dynamically identify and present the monitoring point location, flow rate trend, and fault status. It supports filtering and displaying historical monitoring data by monitoring point and time dimension, forming data change curves. As a further illustration of the present embodiment, the data visualization module 510 of the present embodiment builds a Web interactive interface based on the B / S architecture, and receives the flow rate and liquid level data of the real-time monitoring unit 100 (multi-parameter sensing module 120) and the fault warning results of the warning analysis unit 300 in real time through the WebSocket protocol. It builds a hierarchical visualization interface in combination with a 1:1000 scale pipe network GIS map; Meanwhile, the monitoring points in the interface are distinguished by green (normal), yellow (early warning), and red (fault) circular icons, and the icon diameters are classified as 8 mm for the main pipe and 5 mm for the branch pipe. Clicking the icon can view real-time data and a thumbnail of the last one hour. The fault pipe segment is highlighted in red dotted line and labeled with the fault type and risk level. The "flow-liquid level trend panel" on the right side displays a 24-hour line chart by default, and supports single monitoring point selection and 1-hour to 7-day period adjustment. Meanwhile, the data visualization module 510 also supports filtering historical data by "region-pipe segment-monitoring point" tree structure and "today / yesterday / near 7 days / custom" time dimension. The filtering results can be exported as Excel / CSV files containing fields such as "monitoring point number, sampling time, data status", etc. After the export task is completed, a Web notification will prompt the download. If the data loading times out, a yellow prompt bar will be displayed on the interface and a retry mechanism will be triggered every 30 seconds for a total of 5 times. If the retry fails, a network check will be prompted and the last cached data will be retained (with an update time label).

[0060] The early warning interaction module 520 receives the fault risk level from the early warning analysis unit 300, triggers a Web pop-up window reminder and a mobile terminal instant notification for high-risk faults, and generates a timed summary report for low-risk faults. Users can mark the early warning processing status through the mobile terminal and feed back the status to the system collaboration unit 400. As a further description of the present embodiment, the early warning interaction module 520 of the present embodiment differentiates the push of early warning information based on risk level, specifically including: High-risk faults trigger a Web central red pop-up window (containing fault details and "view details / mark processing" buttons), mobile terminal APP push, and operation and maintenance responsible person SMS notification. If no feedback is received within 15 minutes, the information will be pushed to the backup personnel again. Medium-risk faults only trigger a Web top yellow prompt and APP push, and a summary report is generated every 2 hours. Low-risk faults push a daily summary report at 20:00 every day. Meanwhile, the early warning interaction module 520 supports users to mark the "to be processed / processing / finished / false alarm" status through Web / APP. "Processing" requires filling in the processor and estimated time and synchronizing to the system collaboration unit 400. "Finished" requires uploading the processing description and 1-3 on-site photos and triggering the early warning analysis unit 300 for review. "False alarm" requires selecting the reason and storing the record for model optimization. In addition, the early warning interaction module 520 also has a built-in "early warning notification record" query function, which supports filtering by time, risk level, and processing status. The record contains the push channel, recipient, and feedback information.

[0061] The parameter configuration module 530 is configured to provide a configuration interface of a monitoring frequency, a pre-warning notification mode, and a data storage period, and configure an operation association permission management rule (for example, an administrator can modify all parameters, and an operation and maintenance personnel can only adjust the notification receiving mode); after the parameters are modified, the parameters are automatically synchronized to the real-time monitoring unit 100 and the data processing unit 200, and the multi-unit parameter collaborative update is realized.

[0062] As a further description of the present embodiment, the parameter configuration module 530 of the present embodiment provides a three-tab visual interface of "monitoring configuration", "pre-warning configuration", and "data storage configuration", wherein: The "monitoring configuration" supports 1-30 minute sampling frequency selection and data validity threshold switch configuration; The "pre-warning configuration" can select multiple notification modes (Web / APP / short message / email) and set the middle and low risk report push time; The "data storage configuration" can select a real-time data storage period of 7 days to 1 year, a historical fault permanent storage, and a 7-day retention period of an exported file; Meanwhile, based on the RBAC permission rule distribution operation range, the administrator has all configuration permissions, the operation and maintenance personnel can only modify the pre-warning notification mode, and the viewing personnel have no configuration permission. When there is no permission operation, a red prompt is popped up and a log is recorded; Meanwhile, after the parameters are modified, the rationality is verified first (for example, the sampling frequency is greater than or equal to 1 minute), and if the verification is passed, the parameters are synchronized to the associated units such as the real-time monitoring unit 100 (the sampling frequency adjustment module 140) and the data processing unit 200 (the edge data preprocessing module 210) through the MQTT protocol. If the confirmation is not received within 5 minutes, a synchronization failure prompt is pushed to the administrator and the original parameters are retained, and a retry push is triggered once every 10 minutes for a total of three times.

[0063] It should be noted that the multi-terminal collaboration and compatibility design in the terminal interaction unit 500 of the present embodiment specifically includes that the Web terminal supports Chrome, Firefox, and Edge browsers of nearly three versions, and uses a responsive design to adapt to computers and tablets. The tablet terminal can call out a hidden monitoring point tree list through a sliding gesture; the mobile terminal APP supports Android >= 8.0 and iOS >= 12.0 systems, and synchronizes the pre-warning state and configuration parameters with the Web terminal in real time; the account security supports password and mobile phone verification code login, the administrator account is forced to use two-factor authentication, the account is automatically locked after 30 minutes of idling, and a login log query function including login time, device, and IP is provided to facilitate abnormal login tracing.

[0064] Those skilled in the art can understand that the process of implementing all or part of the steps of the above-mentioned embodiments can be completed by hardware, or by a program to instruct related hardware.

[0065] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Various changes and improvements can be made to the present application without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. The sewage pipe network intelligent monitoring and fault warning analysis system for sewage treatment is characterized by: include: A real-time monitoring unit (100) is used to collect flow, liquid level, and water quality operation data of the sewage pipe network, adopts a two-stage anti-fouling sampling structure and integrates a backwash cleaning function, and dynamically adjusts the sampling frequency according to liquid level fluctuations to cope with pipe network impurity blockage and parameter mutation scenarios; A data processing unit (200), wherein the data processing unit (200) performs validity screening and fault feature mining based on the monitoring data transmitted by the real-time monitoring unit (100), combines the pipe diameter and slope parameters of the edge end pipe network to predict invalid data, and simultaneously analyzes the remote time domain characteristics and the pipe network structure characteristics through fusion, thereby reducing data transmission redundancy and enhancing fault correlation; An early warning analysis unit (300) is used to identify blockage and leakage faults in the sewage pipe network and output risk warnings. Based on a dual-model switching strategy that is adaptive to data volume, an improved LSTM-random forest fusion model is used when data is sufficient, and an improved grey prediction model is switched when data is insufficient. At the same time, special feature training is carried out for the physical characteristics of "flow-level coordinated change" of the two types of faults to adapt to data fluctuations and improve fault identification accuracy. A system coordination unit (400) adopts a fault level driving mechanism to realize coordinated scheduling of early warning, operation and maintenance, and sewage treatment processes, and shortens fault response lag by intelligently pushing operation and maintenance work orders and coordinating pipe network valve adjustment strategies differentiated by blockage and leakage; The terminal interaction unit (500) is used to display real-time monitoring data and warning results and support basic parameter configuration, and realize multi-terminal data interaction through a Web-side data display interface and mobile-side warning information push.

2. The sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment according to claim 1 is characterized in that: The real-time monitoring unit (100) comprises a two-stage anti-pollution sampling module (110) and a multi-parameter sensing module (120), wherein: The two-stage anti-pollution sampling module (110) adopts a serial filtering structure, the first-stage filter is a corrosion-resistant metal filter, and adopts a large aperture design to intercept large suspended impurities, and the second-stage filter is a corrosion-resistant polymer filter, and adopts a small aperture design to intercept fine suspended matter. The two-stage filter is installed in sequence along the water flow direction of the pipe network, and the edge of the filter is fitted with the inner wall of the sampling port through a sealing structure to reduce the infiltration of unfiltered sewage; The multi-parameter sensing module (120) integrates a flow sensor, a liquid level sensor, and a water quality sensor, and the detection probes of the flow sensor, the liquid level sensor, and the water quality sensor are all arranged on the downstream side of the two-stage anti-pollution sampling module (110) to avoid direct contact with unfiltered sewage.

3. The sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment according to claim 2 is characterized in that: The real-time monitoring unit (100) further includes a backwash cleaning module (130) and a sampling frequency adjustment module (140), wherein: The backflushing cleaning module (130) comprises a pressure difference detection submodule (131), a logic control submodule (132), a reverse power submodule (133), and a backflow prevention and sewage discharge submodule (134), wherein: The pressure difference detection submodule (131) uses two industrial-grade pressure sensors, which are respectively fixed to the upstream pipe section of the first filter and the downstream pipe section of the second filter of the two-stage anti-pollution sampling module (110), and based on the physical relationship between fluid resistance and pressure difference, the pressure data between the two points is collected and transmitted in real time; The logic control submodule (132) has a built-in programmable logic chip and pre-stores two trigger conditions: "normal working pressure difference range of the filter" and "maximum continuous operation time". When the pressure difference transmitted by the pressure difference detection submodule (131) exceeds the normal range, or the cumulative operation time of the backwash cleaning module (130) since the last backwash is completed reaches a set value, the logic control submodule (132) generates a backwash start instruction; The reverse power submodule (133) comprises a micro-diaphragm water pump and a water flow switching valve, wherein the water inlet of the water pump is connected to an external clean water source, and the water outlet is connected to the downstream of the secondary filter and the upstream of the primary filter respectively through the switching valve; after receiving a start command, the water pump first transmits reverse water flow to the downstream of the secondary filter through the switching valve, and after a preset time, the switching valve switches to the upstream of the primary filter to transmit reverse water flow; The anti-backflow sewage submodule (134) is composed of a sewage pipe and a one-way check valve. One end of the sewage pipe is connected to the impurity collection area upstream of the first-level filter, and the other end is connected to the main channel of the sewage pipe network. The impurity-containing sewage generated by backwashing is discharged through the sewage pipe, and the one-way check valve blocks the sewage in the pipe network from flowing back into the backwash system to avoid contamination of clean water sources. The sampling frequency adjustment module (140) establishes data linkage with the liquid level sensor of the multi-parameter sensing module (120), and adjusts the sampling frequency by real-time identification of the liquid level change trend transmitted by the liquid level sensor: when the liquid level change trend is gentle, the multi-parameter sensing module (120) is controlled to maintain the basic sampling frequency; when the liquid level change trend is significant, a frequency increase instruction is sent to the multi-parameter sensing module (120) to encrypt the collected flow, liquid level, and water quality data, ensuring that the sudden change process of the pipe network parameters is captured, and the frequency adjustment instruction is transmitted in real time through the low-power controller built into the sampling frequency adjustment module (140).

4. The sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment according to claim 3 is characterized in that: The data processing unit (200) comprises an edge data pre-processing module (210) and a remote feature fusion module (220), wherein: The edge data pre-processing module (210) is used to perform preliminary processing on the monitoring data transmitted by the real-time monitoring unit (100), specifically including: Import basic parameters of the sewage network and establish a mapping relationship between the parameters and the corresponding monitoring points; Based on the basic parameters of the sewage pipe network and the fluid mechanics of pipe flow, a reasonable fluctuation range of flow and liquid level at each monitoring point is preset. When the original data exceeds this range, it is marked as invalid data. Eliminate the marked invalid data, and use the interpolation method based on the trend of adjacent valid data to supplement the missing values ​​of data transmission, and output complete preprocessed data; The remote feature fusion module (220) receives the pre-processed data output by the edge data pre-processing module (210) and is used to mine fault features, including extracting the time domain dynamic features and pipe network structure features of the data, and performing fusion analysis on the time domain dynamic features and pipe network structure features.

5. The sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment according to claim 4 is characterized in that: The remote feature fusion module (220) includes a feature extraction submodule (221) and a feature fusion submodule (222), wherein: The feature extraction submodule (221) is used to extract two types of key features from the edge pre-processed data, specifically including: Time domain dynamic characteristics: including the unit time fluctuation amplitude of flow and liquid level, the frequency of continuous abnormal values, and the duration of peak values. These characteristics are obtained by performing time series slicing analysis on pre-processed data using a sliding window algorithm, reflecting the temporal variation pattern of the pipeline network operation status. Pipeline network structural characteristics: Based on the pipe diameter and slope parameters imported from the edge end and the aging coefficient of the pipe material, the actual flow capacity coefficient and resistance loss coefficient of the pipe section are calculated, reflecting the fundamental impact of the physical properties of the pipeline network on the operating status; The feature fusion submodule (222) is used to realize the organic integration of time domain dynamic features and pipe network structure features, specifically including: Establish a feature correlation evaluation mechanism to assign dynamic weights to time-domain dynamic features and pipe network structural features based on the correspondence between "time-domain dynamic features, structural features, and fault types" in historical fault data. A weighted fusion algorithm is used to map the time domain dynamic features and the pipe network structure features into the same feature space, generating a fusion feature set that includes time dynamic changes and physical structure properties, which is used to enhance the relevance and accuracy of subsequent fault identification.

6. The sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment according to claim 5 is characterized in that: The early warning analysis unit (300) comprises a model switching control module (310), a dual-model operation module (320) and a fault feature training module (330), wherein: The model switching control module (310) collects the fusion feature set output by the data processing unit (200) and defines the effective sample number , data collection duration , number of monitoring points , calculate the sample density and with a preset density threshold Compare and determine the data sufficiency status; The dual-model operation module (320) pre-stores an improved LSTM-random forest fusion model and an improved grey prediction model, and calls the corresponding model according to the instruction of the model switching control module (310); The fault feature training module (330) constructs a feature training set based on the physical characteristics of a blockage fault "flow rate drop-liquid level rise" and a leakage fault "flow rate drop-liquid level drop" for model parameter optimization.

7. The sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment according to claim 6 is characterized in that: The improved grey prediction model in the dual-model operation module (320) includes a prediction correction submodule (323) and a risk level output submodule (324), wherein: The prediction correction submodule (323) receives the time series data in the fusion feature set and defines the following parameters: Basic forecast value , is the flow or liquid level prediction result output by the traditional grey prediction model; Congestion compensation item , based on the average rate of rise of the liquid level in historical blockage faults and blockage correction factor Setting; among them, For quantification right the extent of the correction; Leakage compensation item , based on the average rate of decline of flow in historical leakage failures and leakage correction factor Setting; among them, For quantification right the extent of the correction; The prediction correction submodule (323) superimposes the congestion scene Get the revised forecast value , superimpose the leakage scenario Get the revised forecast value ; The risk level output submodule (324) defines the mean value of historical normal operation data , normal data standard deviation , respectively define the congestion scene deviation , Leakage scenario deviation ;according to 、 The size of the corresponding output risk level , representing low, medium, and high risks respectively.

8. The sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment according to claim 7 is characterized in that: The improved LSTM-random forest fusion model in the dual-model operation module (320) includes a temporal feature reinforcement submodule (321) and a classification decision submodule (322), wherein: The temporal feature enhancement submodule (321) receives the temporal data in the fusion feature set and defines the following parameters: Feature Dimension , is the number of flow, level, and velocity monitoring features; Time series length , is the number of consecutive sampling moments; Attention weight matrix , whose elements For the Class features in The weight of the moment, satisfaction And the sum of the weights of each row is 1; Traffic sudden change threshold and abnormal liquid level fluctuation threshold , are all preset feature thresholds; The temporal feature reinforcement submodule (321) satisfies The sudden change characteristics of flow Abnormal liquid level fluctuation characteristics, improve the corresponding position ;in, Indicates the flow difference between adjacent collection times, represents the time interval between adjacent acquisition moments set by the sampling frequency adjustment module (140), Indicates the liquid level difference between adjacent collection moments; The classification decision submodule (322) defines a preset decision threshold , and use the random forest algorithm to classify the enhanced features and output the probability of blockage failure and leakage failure probability , and The value range of is [0,1]; when Reaching the preset decision threshold , it is determined to be a corresponding type of fault.

9. The sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment according to claim 8 is characterized in that: The system collaboration unit (400) includes a fault information analysis module (410), an operation and maintenance work order scheduling module (420), a pipe network valve adjustment module (430) and a process adaptation module (440), wherein: The fault information analysis module (410) receives the blockage and leakage fault types, risk levels, and fault location information output by the early warning analysis unit (300), and simultaneously retrieves the pipe network structure features output by the remote feature fusion module (220), associates and matches the fault types, risk levels, and pipe network structure features, and determines the priority of the coordinated response; The operation and maintenance work order scheduling module (420) generates an operation and maintenance work order including the fault location, fault type, risk level and corresponding handling suggestions based on the response priority determined by the fault information analysis module (410); automatically pushes the work order to the corresponding operation and maintenance team according to the pre-stored "operation and maintenance team-responsible pipe network area" mapping relationship, and receives the work order receipt status, on-site handling progress and fault repair results fed back by the operation and maintenance team in real time, thus forming an operation and maintenance scheduling closed loop; The pipe network valve adjustment module (430) receives the fault type and location information from the fault information analysis module (410), and performs differential adjustment in combination with the real-time flow and liquid level data transmitted by the multi-parameter sensing module (120): for a blockage fault, the valve of the upstream associated pipe section of the faulty pipe section is adjusted to reduce the incoming flow, while the valve of the downstream spare pipe section is opened to achieve diversion; for a leakage fault, the control valves associated upstream and downstream of the fault point are closed to limit the leakage range, and the valve status is synchronously fed back to the fault information analysis module (410) during the adjustment process; The process adaptation module (440) establishes data linkage with the pipe network valve adjustment module (430), receives actual pipe network flow data after valve adjustment, and pushes flow change information to the sewage treatment plant inlet control system; based on the flow change, a process adjustment instruction is issued to the sewage treatment plant to ensure that the sewage treatment process is compatible with the pipe network operation status.

10. The sewage pipe network intelligent monitoring and fault early warning analysis system for sewage treatment according to claim 9 is characterized in that: The terminal interaction unit (500) includes a data visualization module (510), an early warning interaction module (520) and a parameter configuration module (530), wherein: The data visualization module (510) receives the flow and liquid level data of the real-time monitoring unit (100) and the fault warning results of the warning analysis unit (300), and constructs a pipe network topology visualization interface on the Web side, presenting the monitoring point location, flow trend and fault status with dynamic identification; supports filtering and displaying historical monitoring data according to monitoring point and time dimensions, and forms a data change curve; The early warning interaction module (520) receives the fault risk level of the early warning analysis unit (300), triggers a pop-up window reminder on the web end and an instant notification on the mobile end for high-risk faults, and generates a scheduled summary report push for medium and low-risk faults; supports users to mark the early warning processing status through the mobile end, and feeds back the status to the system collaboration unit (400); The parameter configuration module (530) is used to provide a configuration interface for monitoring frequency, early warning notification mode, and data storage period, and configure operation-related authority management rules; after the parameters are modified, they are automatically synchronized to the real-time monitoring unit (100) and the data processing unit (200), realizing multi-unit parameter collaborative updating.

Images