Intelligent intercepting well cloud edge collaborative control method and system

Abstract

This application discloses a cloud-edge collaborative control method and system for intelligent interception wells, belonging to the field of drainage treatment technology. The method includes: first, the perception and execution layer collects various sensor data of the interception well in real time and transmits them to the edge real-time control layer; the edge real-time control layer preprocesses the collected data, generates feature data, and determines the operating state of the interception well. If it is the first operating state, the feature data is encrypted and uploaded to the cloud decision optimization layer; then, the cloud decision optimization layer generates a regional collaborative control target and issues a global optimization collaborative decision instruction to the edge real-time control layer. This layer generates control instructions and issues them to the perception and execution layer to execute control on the interception well. The above method utilizes the powerful real-time data processing capabilities of the edge real-time control layer to quickly analyze and process the data, making real-time control decisions directly on the local machine, reducing data transmission latency, and improving the response speed of the interception well to complex and changing operating conditions.

Description

Technical Field

[0001] This application belongs to the field of drainage treatment technology, and specifically relates to a smart interception well cloud-edge collaborative control method and system. Background Technology

[0002] In existing urban drainage systems, intercepting wells play a crucial role, especially in combined sewer systems. During the dry season, intercepting wells act as loyal guardians, diverting all sewage to wastewater treatment plants. For example, in some older urban areas with a complex network of combined sewer pipes, intercepting wells ensure centralized treatment of domestic sewage, preventing pollution of surrounding rivers and lakes. During the rainy season, the work of intercepting wells becomes more complex and important. In the early stages of rainfall, rainwater washes away the ground, carrying large amounts of silt, oil, garbage, and various chemical pollutants. At this time, intercepting wells quickly intercept this high-concentration initial rainwater and send it to wastewater treatment plants, significantly reducing the pollution load on receiving water bodies. As rainfall continues to increase, when it reaches a certain level, intercepting wells can promptly open overflow channels to allow excess rainwater to be discharged smoothly, preventing urban flooding caused by poor drainage. In cities prone to torrential rain, proper overflow control of interception wells effectively prevents city streets from being flooded by rainwater, ensuring the normal operation of urban traffic and the stability of residents' lives.

[0003] Early intercepting wells were relatively simple in construction and had a single function, often employing a fixed interception ratio. According to the "Code for Design of Outdoor Drainage" (GB50014-2021), the interception ratio was typically between 2 and 5. This method was only suitable for smaller cities with relatively stable sewage volume and water quality. However, with accelerated urbanization, rapid urban population growth, and booming commercial activities, urban drainage systems face increasingly severe challenges.

[0004] To address these challenges, interceptor well technology is constantly being innovated. In terms of control methods, it has gradually shifted from manual operation to automated control. For example, many interceptor wells are equipped with level sensors, rain gauges, and other devices that can automatically control the opening and closing of gates based on changes in water level and rainfall, improving the timeliness and accuracy of interception to a certain extent. In terms of structural design, various new interceptor well structures have emerged, such as composite interceptor wells, which cleverly divide the well into different functional areas, significantly optimizing the separation of rainwater and sewage; and integrated interceptor wells using steel plate shear wall structures, which offer advantages over traditional concrete or brick structures, including high strength, excellent seepage and corrosion resistance, and a short construction period. Material application is also being continuously explored, with more corrosion-resistant, wear-resistant, and long-life materials being used in interceptor well construction, effectively reducing maintenance costs and extending the service life of the interceptor wells. Furthermore, some technical solutions are also attempting to improve management efficiency through a centralized cloud control mode, where all interceptor well sensing data is uploaded to the cloud, where a cloud model uniformly analyzes and issues control commands. For example, some smart drainage cloud platforms aggregate water level and flow data from intercepting wells throughout the city, use cloud-based big data algorithms to generate control strategies, and then distribute these control strategies periodically.

[0005] This type of solution does not introduce edge computing nodes and still relies on the cloud for the entire process of decision-making.

[0006] However, the characteristics and problems of existing technical solutions are as follows: (1) The control model has fundamental defects. Excessive Cloud Dependence: The existing centralized cloud control model concentrates all data processing and decision-making in the cloud. In scenarios with network congestion or large data volumes, this leads to significant response delays. For example, in emergency scenarios such as heavy rain (when the water level rises by 5 cm per second), delayed cloud commands may cause the gates to open only after the water level exceeds the warning threshold, triggering the risk of flooding. If a network outage occurs (such as a base station failure due to extreme weather), the cloud control model will completely fail, and the interception wells will become uncontrollable. Significant Waste of Computing Resources: Decisions for simple operating conditions (such as when the water level is stable during the dry season) do not require complex cloud calculations, but the existing model still uploads data and consumes cloud computing power, resulting in excessive server load and increased operating costs. Approximately 80% of routine simple operating conditions (such as when the water level is stable during the dry season) could be resolved through local logic, but due to reliance on the cloud, more than 30% of computing resources are wasted, significantly increasing operating costs. Significant Data Security Risks: Frequent uploading of sensitive data such as pipeline topology and water quality monitoring poses a risk of leakage, and a single point of failure in the cloud could cause the entire interception well system to become uncontrollable, posing a risk of flooding.

[0007] (2) Lack of system coordination and global optimization capabilities The lack of regional coordination mechanisms: Most existing intercepting wells operate independently, resembling isolated islands in different areas, lacking effective information exchange and collaborative control mechanisms. This prevents optimized scheduling from a macro-level perspective of the city's overall drainage system, often resulting in over-interception in some areas, leading to excessive loads on sewage treatment plants, while other areas experience severe overflows, polluting surrounding water bodies. For example, in different city blocks, because intercepting wells operate independently, during heavy rains, some blocks' wells operate at full capacity, causing excessive pressure on the sewage network, while neighboring blocks' wells, failing to adjust in time, experience large overflows of sewage and initial rainwater, causing river pollution. The absence of a target balancing mechanism: It is difficult to coordinate multiple objectives such as "pollution control," "urban flood prevention," and "sewage treatment plant load," often resulting in neglecting one aspect for another (e.g., trying to reduce overflows leading to excessive network pressure and causing urban flooding).

[0008] (3) Poor adaptability to complex working conditions Faced with extreme weather events, such as torrential rains and heavy rainfall from typhoons, the drainage capacity of existing intercepting wells is often insufficient, failing to meet the demand for discharging large amounts of rainwater in a short period, and easily leading to urban flooding. Furthermore, when sewage quality fluctuates significantly, traditional intercepting wells lack the ability to adaptively adjust their interception strategies, failing to respond promptly and effectively to changes in water quality. This negatively impacts the influent quality of subsequent sewage treatment plants, increasing the difficulty and cost of sewage treatment. In addition, in some areas, intercepting wells are frequently clogged with garbage and debris, leading to poor drainage and severely affecting their normal operation; existing technologies are unable to effectively solve this problem.

[0009] (4) Insufficient data processing and decision-making capabilities Although some intercepting wells have achieved automated control, their data processing capabilities are limited. Relying solely on data collected by local sensors for simple decision-making cannot fully tap into the potential value and patterns within vast amounts of historical data. For a complex and dynamic system like urban drainage, the lack of big data analysis and in-depth mining capabilities makes it difficult to accurately predict trends in sewage volume and water quality, and to formulate reasonable interception strategies in advance. Consequently, the scientific accuracy and effectiveness of intercepting well decisions in dealing with complex and ever-changing drainage conditions are significantly compromised. Summary of the Invention

[0010] To address the aforementioned issues, this application provides an intelligent interception well cloud-edge collaborative control method and system, which features fast response and high regional linkage efficiency.

[0011] The purpose of this disclosure is to provide a smart interception well cloud-edge collaborative control method, including, The perception and execution layer collects various sensor data from the interception well in real time and transmits them to the edge real-time control layer. The edge real-time control layer preprocesses the various sensor data collected from the interception well to generate feature data and determines the operating status of the interception well. If the operating status is the first operating status, the feature data is encrypted and uploaded to the cloud decision optimization layer. The cloud-based decision optimization layer performs global optimization based on the received feature data, generates regional collaborative control targets, and sends global optimization collaborative decision instructions to the edge real-time control layer. The edge real-time control layer generates control commands based on the globally optimized collaborative decision-making instructions issued, and then sends them to the perception and execution layer to control the interception well.

[0012] Optionally, various sensor data can be used, including well level, water quality, weather conditions, and valve / equipment status.

[0013] Optionally, it also includes the following: if the working condition is the second working condition, the edge real-time control layer quickly judges the state of the intercepting well based on the rule-based decision model. If the state of the intercepting well exceeds the expectation, it directly generates control commands and sends them to the perception and execution layer to perform control on the intercepting well.

[0014] Optionally, the first working condition and the second working condition are respectively a complex working condition and a simple working condition, wherein, Complex operating conditions include periods of rainfall during the rainy season, sewage from upstream pipe networks flowing in, liquid levels within the overflow risk range, and the stage of multi-equipment collaborative control; Simple operating conditions include the dry season with no rainfall, stable upstream water flow with acceptable water quality, and normal operation of equipment without any abnormalities.

[0015] Optionally, it also includes, The edge real-time control layer also detects the heartbeat and / or confidence level of the cloud decision optimization layer; If the heartbeat interruption time of the cloud decision optimization layer is greater than the preset time or the confidence level is less than the preset value, the edge real-time control layer will perform local autonomous control of the interception well based on the local rule base.

[0016] Optionally, the cloud-based decision optimization layer performs global optimization based on the received feature data to generate regional collaborative control objectives, including: The objective functions are defined as minimizing the total amount of pollutant overflow and minimizing the energy consumption of the pipeline network. ； in, For the total amount of pollutant overflow target item, For pipeline network energy consumption target items, and These are the weighting coefficients. It is a control increment penalty item. This is the penalty coefficient; Based on a defined objective function, a digital twin model of the urban drainage network is constructed using a graph neural network, with real-time data from each intercepting well as input. The global optimal solution is obtained by using the particle swarm optimization algorithm to generate a regional cooperative control target. The regional collaborative control objectives are converted into control commands for the corresponding interception wells and sent to the edge real-time control layer.

[0017] Optionally, it also includes, The global model is trained using the desensitized data from each edge node uploaded by the edge real-time control layer through a federated learning mechanism. Based on the analysis of historical hydrological data and meteorological radar echoes using a trained global model, the dry season or rainy season pattern is identified, and optimized edge control strategies are periodically generated and distributed to the edge real-time control layer after being encrypted using an algorithm.

[0018] Another objective of this disclosure is to provide an intelligent interception well cloud-edge collaborative control system, comprising a perception and execution layer, an edge real-time control layer, and a cloud-based decision optimization layer, wherein... The perception and execution layer is used to collect various sensor data from the interception well in real time and transmit them to the edge real-time control layer. The edge real-time control layer is used to preprocess various sensor data collected from the interception well, generate feature data, and determine the operating status of the interception well. If the operating status is the first operating status, the feature data is encrypted and uploaded to the cloud decision optimization layer. The cloud-based decision optimization layer is used to perform global optimization based on the received feature data, generate regional collaborative control targets, and issue global optimization collaborative decision instructions to the edge real-time control layer. The edge real-time control layer is also used to generate control commands based on the issued global optimization and collaborative decision-making instructions, and send them to the perception and execution layer to execute and control the interception well.

[0019] Optionally, the perception and execution layer includes, The data acquisition module is used to collect multi-sensor data in real time and transmit it to the edge real-time control layer.

[0020] The execution and feedback module is used to convert control commands issued by the edge real-time control layer into valve and motor actions, and collect execution results to feed back to the edge real-time control layer.

[0021] Optionally, the edge real-time control layer includes, The data preprocessing module is used to preprocess the received multi-sensor data and upload it to the cloud-based decision optimization layer. The edge real-time decision-making module is used to determine the status of the interception well based on the second working condition and the rule-based decision-making model. If the status of the interception well exceeds the expectations, control commands are generated directly. The edge autonomy module is used to automatically switch to local autonomy mode when the cloud heartbeat interruption is detected to be greater than a preset time or the confidence level is less than a preset value, and to perform autonomous control of the interception well based on the local rule base.

[0022] The encrypted transmission module is used to desensitize and encrypt multi-sensor data.

[0023] Optionally, the cloud-based decision optimization layer includes, The global optimization module is used to determine the objective functions of minimizing the total amount of pollutant overflow and minimizing the energy consumption of the pipe network. Based on the determined objective functions, a digital twin model of the urban drainage pipe network is built using a graph neural network, with real-time data from each intercepting well as input; and the global optimal solution is obtained through a particle swarm optimization algorithm to generate regional collaborative control objectives. The collaborative decision-making module is used to convert regional collaborative control objectives into control commands for corresponding interception wells and send them to the edge real-time control layer. It also tracks the command execution results and sends the feedback data back to the global optimization module and the model dynamic update module. The model dynamic update module is used to train a global model using desensitized data from each edge node uploaded by the edge real-time control layer through a federated learning mechanism; and to analyze historical hydrological data and meteorological radar echoes based on the trained global model to identify dry season or rainy season patterns, periodically generate optimized edge control strategies, and send them to the edge real-time control layer after encryption algorithms.

[0024] Optionally, the cloud-based decision optimization layer also includes, The data storage and analysis platform is used to store historical data using a time-series database, and to generate water quality trend and energy consumption analysis reports.

[0025] The secure communication module is used for direct data upload and download between the edge real-time control layer and the cloud decision optimization layer, as well as for data transmission via encrypted data.

[0026] Compared with the prior art, this application has the following advantages: The aforementioned method partially offloads data processing and decision-making processes traditionally performed in the cloud or on a central server to the edge real-time control layer local to the interceptor well. Leveraging the powerful real-time data processing capabilities of the edge real-time control layer, it rapidly analyzes and processes the large amounts of sensing data collected on-site, making real-time control decisions directly on-site. This reduces data transmission latency and improves the interceptor well's response speed to complex and changing operating conditions. Furthermore, collaborative control with the cloud-based decision optimization layer enables millisecond-level response, edge autonomy, regional linkage, and dynamic optimization of the interceptor well control.

[0027] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A schematic flowchart of a smart interception well cloud-edge collaborative control method according to an embodiment of this disclosure is shown; Figure 2 A decision-making flowchart of an edge real-time control layer in an embodiment of this disclosure is shown; Figure 3 A flowchart illustrating the dynamic update of a cloud-based decision optimization layer model in an embodiment of this disclosure is shown. Figure 4 A structural diagram of an intelligent interception well cloud-edge collaborative control system according to an embodiment of this disclosure is shown; Figure 5 This invention discloses a structural diagram of another intelligent interception well cloud-edge collaborative control system in an embodiment of the present disclosure; Figure 6 A flowchart illustrating a cross-regional interception well collaborative control method is shown in an embodiment of this disclosure. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] like Figure 1As shown in the embodiments of this disclosure, a cloud-edge collaborative control method for intelligent interception wells is introduced. The method includes: first, the perception and execution layer collects various sensor data of the interception well in real time and transmits them to the edge real-time control layer; second, the edge real-time control layer preprocesses the collected various sensor data of the interception well to generate feature data and determines the operating status of the interception well. If the operating status is the first operating status, the feature data is encrypted and uploaded to the cloud decision optimization layer; then, the cloud decision optimization layer performs global optimization based on the received feature data, generates regional collaborative control targets, and issues global optimization collaborative decision instructions to the edge real-time control layer; finally, the edge real-time control layer generates control instructions based on the issued global optimization collaborative decision instructions and issues them to the perception and execution layer to execute control of the interception well. The aforementioned method partially offloads data processing and decision-making processes traditionally performed in the cloud or on a central server to the edge real-time control layer local to the interceptor well. Leveraging the powerful real-time data processing capabilities of the edge real-time control layer, it rapidly analyzes and processes the large amounts of sensing data collected on-site, making real-time control decisions directly on-site. This reduces data transmission latency and improves the interceptor well's response speed to complex and changing operating conditions. Furthermore, collaborative control with the cloud-based decision optimization layer enables millisecond-level response, edge autonomy, regional linkage, and dynamic optimization of the interceptor well control.

[0032] Specifically, various sensor data include the diversion well level, water quality, weather conditions, and valve status. The data acquisition module obtains the corresponding sensor data through deployed sensors of multiple types (level, water quality, rainfall, gate valve status, etc.). The sensing and execution layer is the data acquisition source, providing raw input for control decisions and executing control commands. Furthermore, various sensor data are collected and transmitted in real-time to the edge real-time control layer via the industrial Ethernet protocol. The types of sensors are not limited to those mentioned above; other sensors used to measure relevant parameters of the diversion well are also applicable to this disclosure.

[0033] In this embodiment of the disclosure, such as Figure 2 As shown, the edge real-time control layer preprocesses various sensor data collected from the interception well to generate feature data. This includes data cleaning, feature extraction, and compression of various sensor data transmitted from the perception layer. Preferably, only key feature data is uploaded to the cloud decision optimization layer, reducing the transmission volume by 70% and solving the bandwidth waste problem caused by uploading all data in existing technologies. Key feature data consists of key values ​​and event labels for judging trends and risks, including liquid level statistics (e.g., average liquid level over the past 5 minutes, peak value) and event indicators (e.g., warning liquid level reached, overflow weir).

[0034] Furthermore, the operating status of the interception well includes a first operating status and a second operating status, which are respectively a complex operating status and a simple operating status. The simple operating status includes the normal operation stage during the dry season when there is no rainfall, the upstream water inflow is stable and the water quality meets the standards, and the equipment is operating normally without abnormalities. The stability of the water inflow and the compliance of the water quality can be determined by setting the range of flow rate and the range of water quality compliance, as water quality varies from place to place; these values ​​are usually reserved in the system to adapt to the system's operation under actual conditions. For example, if the water level is stable during the dry season, the edge real-time control layer quickly determines the status of the interception well based on a rule-based decision model. If the status of the interception well exceeds expectations, a control command is directly generated and sent to the perception and execution layer to execute control of the interception well, solving the problem of second-level latency in traditional cloud control. The rule-based decision-making model includes rules based on preset rules, such as the dry season wastewater control rules: Rule 1: Gate control, if "dry season mode is in effect", then "keep the intercepting gate open and the confluence gate closed to ensure all incoming water enters the wastewater treatment plant"; Rule 2: Liquid level control, if "well liquid level > wastewater pump start-up level", then "start the wastewater pump to discharge wastewater to the wastewater treatment plant"; if "well liquid level < wastewater pump stop-up level", then "stop the wastewater pump". The status of the intercepting well includes the water level. When the water level is greater than the preset water level value, a control command is generated. For example, if the preset water level value is 80% (percentage) of the water level threshold, then the generated control command is to open the overflow gate valve by 30%. The preset water level value and the opening state of the overflow gate valve are not limited to these; adaptive modifications based on the application environment are applicable to this disclosure.

[0035] Complex operating conditions include periods of heavy rainfall during the rainy season, the influx of sewage from upstream pipe networks, liquid levels within the overflow risk range, and the stage of multi-device collaborative control. The liquid level within the overflow risk range generally means the liquid level is close to overflow, i.e., within the preset overflow value range. For complex operating conditions, the preset value range needs to be set based on the interception well design parameters and actual site conditions. For example, in cases of sudden rainstorms and water quality changes, the corresponding feature values ​​are encrypted and uploaded to the cloud-based decision optimization layer, awaiting global optimization and collaborative decision-making instructions, i.e., cloud-based collaborative processing. The edge real-time control layer detects the heartbeat and / or confidence level of the cloud-based decision optimization layer. Specifically, the cloud-based decision optimization layer sends "heartbeat packets" and "confidence levels" to the edge layer at a fixed frequency (e.g., once per second), and the edge real-time control layer actively and in real-time detects this. If the heartbeat interruption time of the cloud-based decision optimization layer is greater than a preset time or the confidence level is less than a preset value, the edge real-time control layer performs autonomous control of the interception well based on its local rule base. Local autonomy includes a built-in "dual-active degradation" rule base in the local rule base, storing 32 expert rules and the best historical strategies from the past three months. For example, when a cloud heartbeat interruption is detected for >5 seconds or a confidence level <80%, the system automatically switches to local autonomy mode, maintaining basic control functions through an offline model to ensure that the pollution interception efficiency is not less than 80% during network outages, overcoming the shortcomings of existing technologies where network interruptions result in loss of control. The preset values ​​for time and confidence level are not limited to these; other preset values ​​also apply to this disclosure. It should be noted that the "dual-active degradation" rule base is a hierarchical, degradable set of local control strategies built into the edge real-time control layer. Its core objective is to achieve "dual-active" capabilities: "collaboration when the cloud is online, and seamless degradation autonomy when the cloud fails." The 32 expert rules are scenario-based, deterministic emergency control rules extracted from the experience of experts in the field of interception well operation and maintenance. The 32 rules are not mandatory; they represent a roughly feasible number. The "best historical strategies from the past three months" is a scenario-based control case library stored locally at the edge layer. Local autonomy enables local control when the edge real-time control layer fails in the cloud. It optimizes the global collaborative model of the cloud-based decision-making optimization layer, rather than a model that directly generates control commands locally. The local control model is relatively fixed, requiring only optimization of threshold parameters to improve control performance.

[0036] In this embodiment of the disclosure, the cloud-based decision optimization layer performs global optimization based on the received feature data to generate regional collaborative control objectives, including... The objective functions are determined to be minimizing the total amount of pollutant overflow and minimizing the energy consumption of the pipeline network; the objective function needs to be constructed based on the regional collaborative control scenario, and the objective function satisfies: ； in, For the total amount of pollutant overflow target item, For pipeline network energy consumption target items, and Weighting coefficient (during heavy rain) Take 0.8-0.9, during the dry season. (The value can be 0.6-0.7, but is not limited to this; other values ​​also apply to this disclosure). It controls incremental penalties to prevent frequent start-ups and shutdowns of valves and pump stations. This is the penalty coefficient.

[0037] Based on a defined objective function, a digital twin model of the urban drainage network is constructed using a graph neural network. Real-time data from each intercepting well is input, including liquid level, water quality, rainfall, status of electric gate valve equipment, and status of water pump equipment. The global optimal solution is obtained through particle swarm optimization, and a regional cooperative control objective is generated. F(X) = ω1 f1(x)+ω2 f2(x)+ω3 f3(x)+ω4 f4(x) Where ω1, ω2, ω3, and ω4 are all target weights, determined by engineering requirements, and ω1 + ω2 + ω3 + ω4 = 1, f i (x) is the normalized sub-objective function; The goal of regional collaborative control is to coordinate the operational status of all interception wells within the region and achieve comprehensive benefits of "maximizing pollution interception, minimizing overflow risk, optimizing system energy consumption, and maximizing facility lifespan." Therefore, f1(x) to f4(x) represent the sub-objective functions for maximizing pollution interception, minimizing overflow risk, optimizing system energy consumption, and maximizing facility lifespan, respectively.

[0038] The regional collaborative control objectives are transformed into corresponding control commands for interception wells and sent to the edge real-time control layer. The regional collaborative control objectives (such as prioritizing flood prevention while also considering sewage interception) are first decomposed into specific objectives and assigned to individual well objectives, such as the liquid level objective of well #1 in the region (e.g., storing more water and preventing overflow). After receiving the objective, the individual well analyzes it according to its local strategy and then generates specific commands to adjust the opening of pumps and valves, ultimately achieving the overall objective for the region.

[0039] In this embodiment of the disclosure, such as Figure 3As shown, the intelligent interception well cloud-edge collaborative control method further includes a cloud-based decision optimization layer that trains a global model using desensitized data from each edge node uploaded by the edge real-time control layer through a federated learning mechanism. Based on the trained global model, historical hydrological data and meteorological radar echoes are analyzed to identify dry or rainy season patterns. Optimized edge control strategies are periodically generated and distributed to the edge real-time control layer after encryption. Identifying dry or rainy season patterns includes predicting rainfall levels in the next few hours using meteorological radar echoes and combining this with historical hydrological data from the same period to identify upcoming rainy season rainfall patterns or confirm the current stable dry season without rain. The edge control strategies include various decision models for both simple and complex operating conditions.

[0040] It should be noted that the global model is a multi-dimensional fusion prediction and decision-making model deployed in the cloud and serving the entire regional cutoff well system. It is not a single model, but a composite model system that integrates operating condition identification, parameter prediction, and control optimization. Its core function is to coordinate the collaborative operation of all cutoff wells within the region.

[0041] like Figure 4 As shown in the embodiments of this disclosure, an intelligent interception well cloud-edge collaborative control system capable of executing the above-described method is also introduced. The system includes a perception and execution layer, an edge real-time control layer, and a cloud decision optimization layer. The perception and execution layer is used to collect various sensor data of the interception well in real time and transmit them to the edge real-time control layer. The edge real-time control layer is used to preprocess the collected various sensor data of the interception well, generate feature data, and determine the operating status of the interception well. If the operating status is the first operating status, the feature data is encrypted and uploaded to the cloud decision optimization layer. The cloud decision optimization layer is used to perform global optimization based on the received feature data, generate regional collaborative control targets, and issue global optimization collaborative decision instructions to the edge real-time control layer. The edge real-time control layer is also used to generate control instructions based on the issued global optimization collaborative decision instructions and issue them to the perception and execution layer to perform execution control on the interception well. The aforementioned system constructs a distributed cloud-edge collaborative architecture of "edge computing + cloud decision-making" to resolve the contradiction between response speed and global optimization. By clarifying the functional boundaries between the edge layer and the cloud layer, it achieves an organic combination of "rapid edge response + global cloud optimization". The architecture is divided into three layers: "end-edge-cloud", and each layer interacts through standardized interfaces.

[0042] The perception and execution layer serves as the data acquisition source, providing raw input for control decisions and executing control commands. This layer deploys various types of sensors (such as liquid level, water quality, rainfall, and gate valve status), data acquisition modules, electric gate valves, water pumps (variable frequency control), and execution and feedback modules. The data acquisition modules collect multi-sensor data in real time and transmit it to the edge real-time control layer. The execution and feedback modules translate control commands from the edge real-time control layer into valve and motor actions to control the electric gate valves, and collect execution results to feed back to the edge real-time control layer.

[0043] The edge real-time control layer is the central hub for real-time decision-making. Each interception well is equipped with one edge controller (running a Linux system, equipped with two Ethernet ports, four RS485 ports, and supporting 5G wireless communication), constituting one edge node. This setup differs from existing technologies that use controllers without local computing capabilities; the controllers in this layer possess edge computing and independent decision-making capabilities. Furthermore, the edge real-time control layer includes a data preprocessing module, an edge real-time decision-making module, an edge autonomy module, and an encrypted transmission module. The data preprocessing module preprocesses the received multi-sensor data and uploads it to the cloud-based decision optimization layer. Specifically, it cleans, extracts features, and compresses various sensor data transmitted from the perception and execution layers to generate feature data. This feature data includes, but is not limited to, water level change rate, water quality exceedance flags (water quality status indicators), and rainfall intensity levels. The edge real-time control layer only uploads key feature data to the cloud-based decision optimization layer, reducing transmission volume by 70% and resolving the bandwidth waste problem caused by uploading all data in existing technologies.

[0044] The edge real-time decision module is equipped with a lightweight real-time decision algorithm to determine the control strategy based on the operating conditions. The operating conditions include a first operating condition and a second operating condition, which are respectively a complex operating condition and a simple operating condition. If the operating condition is the second operating condition, the state of the intercepting well is judged based on the second operating condition and a rule-based decision model. If the state of the intercepting well exceeds the expectations, a control command is directly generated. This method solves the problem of second-level latency in traditional cloud control. The edge autonomy module automatically switches to local autonomy mode when a cloud-based decision optimization layer heartbeat interruption exceeds a preset time or the confidence level falls below a preset value. This mode uses a local rule base to autonomously control the interception well. The edge autonomy module includes a built-in "dual-active degradation" rule base, storing 32 expert rules and the best historical strategies from the past three months. In local autonomy mode, basic control functions are maintained through an offline model, ensuring that the interception efficiency is no less than 80% during network outages, overcoming the limitation of existing technologies where network interruptions lead to loss of control.

[0045] The encrypted transmission module is used for de-identification and encryption of multi-sensor data. A lightweight encrypted transmission and data compression mechanism is designed to resolve the conflict between security and efficiency in data transmission, ensuring reliable system operation in complex network environments. De-identification methods include data generalization, converting real-time data into interval representations; and timestamp offsetting, among others.

[0046] The cloud-based decision optimization layer includes a global optimization module, a collaborative decision-making module, and a model dynamic update module. The global optimization module is used to determine the objective functions of minimizing the total amount of pollutant overflow and minimizing the energy consumption of the pipe network. Based on the determined objective functions, a digital twin model of the urban drainage pipe network is built using a graph neural network, with real-time data from each intercepting well as input. The module also uses a particle swarm optimization algorithm to solve the global optimal solution and generate regional collaborative control objectives. This addresses the regional imbalance problem caused by the "information silos" of existing technologies. Unlike traditional cloud-based systems that only perform data storage and basic functional analysis, this module has proactive optimization and cross-regional scheduling capabilities.

[0047] The collaborative decision-making module is used to translate regional collaborative control objectives into control commands for corresponding interception wells, which are then sent to the edge real-time control layer. It also tracks the command execution results and transmits feedback data back to the global optimization module and the model dynamic update module. The control commands for the corresponding interception wells include, for example, uniformly reducing the interception ratio of all interception wells in a certain area to 3.

[0048] The model dynamic update module trains a global model using desensitized data from each edge node uploaded by the edge real-time control layer via a federated learning mechanism. Based on the trained global model, it analyzes historical hydrological data and meteorological radar echoes to identify dry or rainy season patterns, periodically generating optimized edge control strategies, which are then encrypted and sent to the edge real-time control layer. The edge control strategies include a rainy season-specific fuzzy control rule table. The desensitized data includes water quality-water level correlation features.

[0049] The cloud-based decision optimization layer also includes a data storage and analysis platform and a secure communication module. The data storage and analysis platform uses a time-series database to store historical data and generate reports such as water quality trends and energy consumption analysis. The secure communication module is used for direct data upload and download between the edge real-time control layer and the cloud-based decision optimization layer, as well as for encrypted data transmission.

[0050] The interactions between the modules in the above layers are as follows: Figure 5 As shown. Further, as... Figure 6As shown, the interception well in region A includes edge nodes A1 and A2, and the interception well in region B includes edge nodes B1 and B2. These interact with the global optimization module and collaborative decision-making module in the cloud-based decision optimization process, respectively. The cloud-edge collaboration mechanism employs a federated learning framework: edge nodes train the water level prediction model locally, while the cloud aggregates parameters to generate a global model and distributes it. A bidirectional interaction protocol is designed, with the edge layer uploading compressed key data (such as feature values ​​and control effects) every 5 minutes, and the cloud distributing the optimized model every 24 hours. The interaction utilizes 5G slicing technology and a lightweight encrypted transmission mechanism (data digest + digital signature) to ensure security, resolving the inherent trade-off between data transmission efficiency and security in existing technologies. Furthermore, each edge node is a node in the edge real-time control layer, and the water level prediction model can employ different deep learning or machine learning models.

[0051] A collaborative control server is deployed in the cloud layer, and information synchronization among multiple edge nodes is achieved through a distributed consensus algorithm. When the load of a certain area's interception well exceeds the limit, the cloud sends collaborative instructions to the edge nodes of adjacent areas to adjust their interception ratio (e.g., from 5 to 3), thereby achieving regional load balancing and solving the local overload problem caused by the independent operation of existing technologies.

[0052] This disclosure proposes a collaborative control architecture, under which different types of models or algorithms can be used to achieve different task objectives, thereby improving the control efficiency and reliability of the interception well.

[0053] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for intelligent interception well cloud-edge collaborative control, characterized in that, include, The perception and execution layer collects various sensor data from the interception well in real time and transmits them to the edge real-time control layer. The edge real-time control layer preprocesses the various sensor data collected from the interception well to generate feature data and determines the operating status of the interception well. If the operating status is the first operating status, the feature data is encrypted and uploaded to the cloud decision optimization layer. The cloud-based decision optimization layer performs global optimization based on the received feature data, generates regional collaborative control targets, and sends global optimization collaborative decision instructions to the edge real-time control layer. The edge real-time control layer generates control commands based on the globally optimized collaborative decision-making instructions issued, and then sends them to the perception and execution layer to control the interception well.

2. The intelligent interception well cloud-edge collaborative control method according to claim 1, characterized in that, Multiple sensor data include interception well liquid level, water quality, weather, and valve equipment status.

3. The intelligent interception well cloud-edge collaborative control method according to claim 2, characterized in that, It also includes the following: if the working condition is the second working condition, the edge real-time control layer quickly judges the state of the interception well based on the rule-based decision model. If the state of the interception well exceeds the expectations, it directly generates control commands and sends them to the perception and execution layer to perform control on the interception well.

4. The intelligent interception well cloud-edge collaborative control method according to claim 3, characterized in that, The first and second operating conditions are complex and simple, respectively. Complex operating conditions include periods of rainfall during the rainy season, sewage from upstream pipe networks flowing in, liquid levels within the overflow risk range, and the stage of multi-equipment collaborative control; Simple operating conditions include the dry season with no rainfall, stable upstream water flow with acceptable water quality, and normal operation of equipment without any abnormalities.

5. The intelligent interception well cloud-edge collaborative control method according to any one of claims 1-3, characterized in that, It also includes, The edge real-time control layer also detects the heartbeat and / or confidence level of the cloud decision optimization layer; If the heartbeat interruption time of the cloud decision optimization layer is greater than the preset time or the confidence level is less than the preset value, the edge real-time control layer will perform local autonomous control of the interception well based on the local rule base.

6. The intelligent interception well cloud-edge collaborative control method according to claim 5, characterized in that, The cloud-based decision optimization layer performs global optimization based on the received feature data to generate regional collaborative control objectives, including... The objective functions are defined as minimizing the total amount of pollutant overflow and minimizing the energy consumption of the pipeline network. ； in, For the total amount of pollutant overflow target item, For pipeline network energy consumption target items, and These are the weighting coefficients. It is a control increment penalty item. This is the penalty coefficient; Based on a defined objective function, a digital twin model of the urban drainage network is constructed using a graph neural network, with real-time data from each intercepting well as input. The global optimal solution is obtained by using the particle swarm optimization algorithm to generate a regional cooperative control target. The regional collaborative control objectives are converted into control commands for the corresponding interception wells and sent to the edge real-time control layer.

7. The intelligent interception well cloud-edge collaborative control method according to claim 6, characterized in that, It also includes, The global model is trained using the desensitized data from each edge node uploaded by the edge real-time control layer through a federated learning mechanism. Based on the analysis of historical hydrological data and meteorological radar echoes using a trained global model, the dry season or rainy season pattern is identified, and optimized edge control strategies are periodically generated and distributed to the edge real-time control layer after being encrypted using an algorithm.

8. A smart interception well cloud-edge collaborative control system, characterized in that, It includes a perception and execution layer, an edge real-time control layer, and a cloud-based decision optimization layer, among which, The perception and execution layer is used to collect various sensor data from the interception well in real time and transmit them to the edge real-time control layer. The edge real-time control layer is used to preprocess various sensor data collected from the interception well, generate feature data, and determine the operating status of the interception well. If the operating status is the first operating status, the feature data is encrypted and uploaded to the cloud decision optimization layer. The cloud-based decision optimization layer is used to perform global optimization based on the received feature data, generate regional collaborative control targets, and issue global optimization collaborative decision instructions to the edge real-time control layer. The edge real-time control layer is also used to generate control commands based on the issued global optimization and collaborative decision-making instructions, and send them to the perception and execution layer to execute and control the interception well.

9. The intelligent interception well cloud-edge collaborative control system according to claim 8, characterized in that, The perception and execution layer includes, The data acquisition module is used to collect multi-sensor data in real time and transmit it to the edge real-time control layer; The execution and feedback module is used to convert control commands issued by the edge real-time control layer into valve and motor actions, and collect execution results to feed back to the edge real-time control layer.

10. The intelligent interception well cloud-edge collaborative control system according to claim 9, characterized in that, The edge real-time control layer includes, The data preprocessing module is used to preprocess the received multi-sensor data and upload it to the cloud-based decision optimization layer. The edge real-time decision-making module is used to determine the status of the interception well based on the second working condition and the rule-based decision-making model. If the status of the interception well exceeds the expectations, control commands are generated directly. The edge autonomy module is used to automatically switch to local autonomy mode when the cloud heartbeat interruption is detected to be greater than a preset time or the confidence level is less than a preset value, and to perform autonomous control of the interception well based on the local rule base; The encrypted transmission module is used to desensitize and encrypt multi-sensor data.

11. The intelligent interception well cloud-edge collaborative control system according to claim 10, characterized in that, The cloud-based decision optimization layer includes, The global optimization module is used to determine the objective functions of minimizing the total amount of pollutant overflow and minimizing the energy consumption of the pipe network. Based on the determined objective functions, a digital twin model of the urban drainage pipe network is built using a graph neural network, with real-time data from each intercepting well as input; and the global optimal solution is obtained through a particle swarm optimization algorithm to generate regional collaborative control objectives. The collaborative decision-making module is used to convert regional collaborative control objectives into control commands for corresponding interception wells and send them to the edge real-time control layer. It also tracks the command execution results and sends the feedback data back to the global optimization module and the model dynamic update module. The model dynamic update module is used to train a global model using desensitized data from each edge node uploaded by the edge real-time control layer through a federated learning mechanism; and to analyze historical hydrological data and meteorological radar echoes based on the trained global model to identify dry season or rainy season patterns, periodically generate optimized edge control strategies, and send them to the edge real-time control layer after encryption algorithms.

12. The intelligent interception well cloud-edge collaborative control system according to claim 11, characterized in that, The cloud-based decision optimization layer also includes, The data storage and analysis platform is used to store historical data using a time-series database, and to generate water quality trend and energy consumption analysis reports. The secure communication module is used for direct data upload and download between the edge real-time control layer and the cloud decision optimization layer, as well as for data transmission via encrypted data.

Images