Intelligent water supply and drainage control system and method with self-adaptive flow regulation function

Abstract

The invention belongs to the technical field of building water supply and drainage engineering control, and discloses an intelligent water supply and drainage control system and method with a self-adaptive flow regulation function, and the method comprises the steps that a signal collection device collects operation parameters of a water supply and drainage system in real time; the signal processing device adopts adaptive filtering and edge computing technologies to process the originally acquired signals; the intelligent control device realizes working condition discrimination, control strategy optimization and fault tolerance through an intelligent algorithm based on the processed data; the execution adjusting device adjusts the execution mechanism to act according to the control instruction output by the intelligent control device, and control over flow and pressure is achieved. The communication and monitoring device realizes remote transmission, state monitoring and remote control of system data, and supports intelligent management and maintenance. By the adoption of the system and method, dynamic removal of signal noise, real-time optimization of a control strategy, online judgment of working conditions and fault tolerance are achieved, and high-precision flow adjustment, low-energy-consumption operation and intelligent management of a water supply and drainage system are achieved.

Description

Technical Field

[0001] This invention relates to the field of building water supply and drainage engineering control technology, and in particular to an intelligent water supply and drainage control system and method with adaptive flow regulation. Background Technology

[0002] As a core component of building and municipal infrastructure, the water supply and drainage system's operational efficiency, energy conservation level, and intelligent control capabilities directly affect water resource utilization efficiency and user experience. Traditional water supply and drainage systems mostly adopt constant-speed pumps or simple pressure control modes. Constant-speed pump systems suffer from technical problems such as high energy consumption, long response time, and limited adjustment accuracy when multiple operating conditions change. They also struggle to accurately match the dynamic changes in design flow rate during peak and off-peak water usage periods.

[0003] To improve the above situation, variable frequency speed control technology has been widely used in the water supply and drainage field. By dynamically adjusting the pump speed to adapt to water demand, it has achieved a certain degree of energy saving and consumption reduction. However, the control effect of existing variable frequency control systems is highly dependent on the accuracy and real-time performance of data from sensors such as pressure and flow. These sensors are susceptible to external factors such as electromagnetic interference, mechanical vibration, and temperature fluctuations during signal acquisition and transmission, resulting in a large amount of noise mixed in the acquired signal. This interferes with the control logic and causes problems such as unstable pump operation and delayed flow regulation, making it impossible to achieve true adaptive flow regulation.

[0004] Meanwhile, the control strategies of existing control systems are mostly based on fixed preset parameters or simple logic, lacking adaptability to dynamic changes in pipeline operating conditions (such as pipeline aging, seasonal water use fluctuations, leaks and blockages, etc.). This leads to a decrease in control accuracy and a deterioration in stability after long-term operation. Furthermore, fault diagnosis and maintenance rely on manual intervention, resulting in high maintenance costs and untimely responses.

[0005] Therefore, there is an urgent need to develop an adaptive flow regulation system that can intelligently process signal noise, dynamically optimize control strategies, and adapt to complex operating conditions, so as to achieve efficient, stable, and intelligent operation of water supply and drainage systems. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent water supply and drainage control system and method with adaptive flow regulation. By integrating technologies such as adaptive filtering, fuzzy control, neural networks, model predictive control, and cloud-edge collaboration, it achieves dynamic removal of signal noise, real-time optimization of control strategies, online identification of operating conditions, and fault tolerance, ultimately achieving high-precision flow regulation, low-energy operation, intelligent management, and high reliability assurance for the water supply and drainage system.

[0007] To achieve the above objectives, the present invention provides an intelligent water supply and drainage control system with adaptive flow regulation, including a signal acquisition device, a signal processing device, an intelligent control device, an execution regulation device, and a communication and monitoring device. The devices work together to achieve adaptive flow regulation of the water supply and drainage system. Signal acquisition device, used to collect operating parameters of water supply and drainage system in real time; The signal processing device employs adaptive filtering and edge computing techniques to perform noise reduction, fusion, and buffering on the original acquired signals. The intelligent control device, based on the processed data, uses intelligent algorithms to achieve operating condition identification, control strategy optimization, and fault tolerance. The actuator adjusts the action of the actuator according to the control commands output by the intelligent control device, thereby controlling the flow and pressure; Communication and monitoring devices enable remote transmission of system data, status monitoring and remote control, and support intelligent management and maintenance.

[0008] Preferably, the signal acquisition device includes: a flow rate acquisition element, a water level acquisition element, and an auxiliary acquisition element; Flow acquisition elements: Flow sensors are used and deployed at the inlet and outlet of the storage tank and at pipeline nodes for online calculation of pump efficiency and total flow measurement. Water level acquisition element: An anti-interference water level sensor is used, deployed in the regulating reservoir and pump station forebay, to monitor water level data in real time and trigger water level alarms; the conditions for triggering water level alarms are as follows: (1) Triggering conditions for high water level alarm: When the real-time water level is greater than or equal to 90% of the pool depth, an audible and visual alarm will be triggered immediately and pushed remotely to prevent the storage tank from overflowing; (2) Low water level alarm triggering conditions: When the real-time water level is less than or equal to 10%, an alarm is triggered and the pump set is controlled to reduce the load to prevent the water pump from running dry and being damaged. Meanwhile, the system will record the duration of the water level exceeding the limit: if the exceeding state lasts for more than 30 seconds, the emergency plan will be automatically activated, the backup storage tank will be switched, and the operation and maintenance personnel will be notified. Auxiliary acquisition components include pressure sensors and water flow acoustic signal spectrum characteristic sensors, which respectively collect pipeline pressure data and water flow acoustic signals to provide multi-dimensional data support for operating condition judgment.

[0009] Preferably, the signal processing device includes: an edge computing node element, a signal preprocessing element, and a data buffer and breakpoint resume element; Edge computing node components: Based on industrial-grade ARM chips, lightweight data acquisition programs, preprocessing algorithm kernels and local control logic are deployed in the chip, integrating data acquisition, preprocessing and local control functions to achieve offline autonomy and multi-node collaboration, ensuring that the system can perform normal traffic regulation when the network is disconnected; Signal preprocessing components: Relying on the computing power of the ARM chip in the edge computing node, an adaptive Kalman filter algorithm is adopted, combined with multi-sensor data fusion technology, to dynamically adjust the filter coefficients to remove noise and transform the noisy raw signal into effective data; at the same time, through data consistency verification, conflicting data is eliminated to ensure data credibility. Data caching and breakpoint resume components: A circular caching mechanism is adopted to cache compressed data for 72 hours; data is continuously recorded when the network is interrupted, and the data is resumed according to the data sequence number after the network is restored, ensuring that the integrity rate of cloud model data is not less than 99.9%; at the same time, it supports seamless connection of control logic to avoid traffic overshoot caused by restarting after communication interruption.

[0010] Preferably, the intelligent control device includes: an online condition discrimination element, an adaptive model update element, a model predictive control element, a cloud-edge collaborative element, and a fault-tolerant element; Online operating condition discrimination element: Construct a three-dimensional feature database of flow-pressure-water flow acoustic signal spectrum features, adopt a support vector machine algorithm, compare the collected data with the feature database data in real time, extract the fault frequency domain feature vector, and output the operating condition discrimination result and confidence level within 50ms, triggering the control strategy switching; When using the Support Vector Machine (SVM) algorithm, the radial basis function (RBF) kernel function should be selected as the kernel type. As shown below: ; in, The feature vector of the working condition to be determined; The sample vectors in the feature library; Indicates the index of the sample; The square of the Euclidean distance; The kernel function bandwidth parameter, with a value between 0.01 and 0.1, is used to adapt to the nonlinear distribution of the acoustic spectrum characteristics of flow-pressure-water flow, thereby improving the accuracy of fault condition discrimination. This represents an exponential function with base e; Adaptive model update component: The built-in miniature long short-term memory network LSTM model retrains the model weights based on the working condition discrimination results and real-time flow error, realizing the dynamic update of the flow prediction model and adapting to changes in pipeline characteristics. The built-in miniature Long Short-Term Memory (LSTM) network model employs a two-layer hidden layer structure, as detailed below: The first hidden layer has 64 neurons, and the activation function used is ReLU. The second hidden layer has 32 neurons, and the activation function used is tanh. The output layer is a fully connected layer with 1 neuron, used to output the predicted flow rate. Model predictive control element: Based on the updated LSTM model, the flow rate change trend is predicted. Combined with system constraints, the model predictive control algorithm is used to calculate the optimal control parameters. Cloud-edge collaborative components: Based on the model parameter synchronization mechanism between edge nodes and cloud servers, edge nodes upload 24-hour compressed data to the cloud, and the cloud retrains the LSTM model with a large-scale dataset and sends the optimized model parameters back to the edge nodes, realizing local response and cloud closed-loop control; Fault-tolerant components: A three-modal redundancy voting mechanism is adopted to perform redundancy verification on sensor data and actuator feedback data; when a sensor or actuator failure is detected, the faulty component is automatically isolated within 10ms, switched to the backup channel, and the downgraded control table is invoked.

[0011] Preferably, the regulating device includes: an electric regulating valve, a variable frequency centrifugal pump, a storage tank gate valve, and electrical safety components; Electric regulating valve: It adopts an electric actuator to adjust the opening degree in real time according to control commands; Variable frequency centrifugal pump: adopts PLC + fuzzy control architecture, and the pump group adopts a multi-pump parallel and rotation control strategy; the fuzzy control algorithm takes the flow deviation and deviation change rate as input, and after fuzzification, fuzzy inference and defuzzification, outputs the pump speed adjustment amount; The gate valve of the storage tank is a soft-seal electric gate valve, which is linked with the flow sensor through an electric actuator. It integrates a fault self-check function with a self-check cycle of 1 minute and an emergency shut-off function. Electrical safety components construct a triple safety protection system of leakage current, insulation, and arcing, as shown below: (1) When the residual current operated circuit breaker (RCBO) on the incoming line side is greater than or equal to 30mA, it will instantly disconnect the power supply within 30ms. (2) The DC insulation tester monitors the insulation resistance of the main circuit. When the insulation resistance is less than 1MΩ, it is locked and switched to the standby pump. (3) The arc fault circuit breaker AFCI on the outgoing side extinguishes the arc and prevents fire within 2ms; each component is connected to an independent safety PLC, and the main control fails and the power is forcibly cut off.

[0012] Preferably, the communication and monitoring device includes: network topology, data frame format, monitoring platform, and mobile terminal; The network topology adopts a three-layer architecture of field-edge-cloud, specifically including: (1) Field layer: The RS-485 dual-bus redundant structure is adopted. The main link is Modbus-RTU and the backup link is CANopen. The data refresh cycle is less than or equal to 200ms. The system operation is not affected by the disconnection of any link. (2) Edge layer: The industrial-grade ARM gateway has built-in MQTT, OPC UA and MQTT over TLS protocols to realize protocol conversion and data compression, and is equipped with an AI chip to identify abnormal working conditions in real time and alarm locally; (3) Cloud layer: Ensure data transmission reliability by using 4G / 5G dual SIM dual standby or NB-IoT backup to the cloud; The data frame format adopts a 32-byte fixed-length structure, and hardware verification and dual-channel redundancy comparison ensure that the data transmission error rate is ≤10%. -6 ; The monitoring platform is based on an edge-cloud-device microservice architecture. The industrial gateway has a built-in Redis time-series cache and an SQLite black box to complete data cleaning and millisecond-level closed-loop control. The cloud deploys a time-series database and a digital twin model, which supports synchronous monitoring of web pages, APP and WeChat mini-program, and provides functions such as GIS global liquid level heat map, energy consumption benchmarking and fault tracing. The mobile terminal adopts an integrated monitoring interface and quick control functions based on GIS maps, as detailed below: (1) The homepage GIS map displays liquid level, flow rate, pump status and alarm information in real time, and supports gesture zoom operation; (2) The bottom scene panel has a one-click mode switching function, and the control command is sent to the edge gateway within 3 seconds.

[0013] A method for an intelligent water supply and drainage control system with adaptive flow regulation includes the following steps: Step S1: Collect real-time data from the pipeline network, process the data using an adaptive Kalman filter algorithm, and output valid data after consistency verification. Step S2: Extract the three-dimensional feature vector of the flow-pressure-water flow acoustic signal spectrum features, compare it with the preset feature database, and adaptively update the weights of the model based on the working condition discrimination result and the real-time flow error to complete the iterative optimization of the LSTM model. Step S3: Based on the updated LSTM model, predict the flow rate change trend, calculate the optimal control parameters, and realize the action of the control device. Step S4: The fault-tolerant component uses a three-mode redundancy voting mechanism to verify sensor data and actuator feedback data in real time. When a fault is detected, the faulty component is automatically isolated and switched to the backup channel, and a reduced-order control strategy is invoked. The communication and monitoring device collects system operation data in real time and displays it through the monitoring platform and mobile terminal. Managers can switch the operating mode with one click through the mobile terminal, and control commands are issued and executed within 3 seconds. When a system fault occurs, the monitoring platform and mobile terminal push alarm information in real time, and managers can remotely guide fault handling or activate emergency control strategies.

[0014] Preferably, in step S1, the specific implementation process of signal acquisition and preprocessing is as follows: Step S11: The signal acquisition device collects the raw data of the pipeline network in real time at a sampling frequency of 1-10Hz and transmits it to the signal processing device via RS485 / Modbus-RTU protocol. Step S12: The signal preprocessing element uses an adaptive Kalman filter algorithm to dynamically adjust the filter coefficients and perform noise reduction on the original data. Step S13: Using multi-sensor data fusion technology, perform consistency verification on the same type of data collected by different sensors, remove conflicting data with a deviation of more than 3%, and output valid data. Step S14: The data caching and breakpoint resume element compresses and caches valid data, continuously records it when the network is disconnected, and verifies and resumes transmission after the network is restored.

[0015] Preferably, in step S2, the specific implementation process of working condition discrimination and model update is as follows: Step S21: The online operating condition discrimination element extracts the three-dimensional feature vector of the flow-pressure-water flow acoustic signal spectrum features from the valid data, compares it with the preset feature database, and uses the SVM algorithm to calculate the operating condition confidence level, as shown below: ; in, The output of the discriminant function represents the input feature vector. The results of the working condition classification; For Lagrange multipliers; For sample labels; For kernel functions; The feature vector to be discriminated; The sample vectors in the feature library; For bias terms; It is a symbolic function; Indicates the index of the sample; The total number of samples; Step S22: When the confidence level is greater than or equal to 0.85, output the operating condition judgment result and trigger the control strategy switch; Step S23: Based on the operating condition judgment result and the real-time flow error, the adaptive model updates the weights; wherein, the calculation formula for the operating condition judgment result and the real-time flow error is as follows: ; in, This is the actual measured flow rate; Predict traffic flow for the model; To set the flow rate; when the error value e is greater than 5%, retrain the mini LSTM model and update the model weights; Step S24: The edge node uploads the 24-hour compressed data to the cloud. The cloud retrains the LSTM model using a large-scale dataset and sends the optimized model parameters back to the edge node to complete the model iteration update.

[0016] Preferably, in step S3, the specific implementation process of control strategy generation and execution is as follows: Step S31: The model predictive control element, based on the updated LSTM model, predicts the flow rate change trend. Combining this with system constraints, the MPC algorithm is used to minimize the objective function and calculate the optimal control parameters, as shown below: ; in, For prediction in the time domain; For time steps; To control the weighting coefficients; The control input at time t; This is the control input at time t-1; Predict traffic flow for the model; To set the flow rate; Step S32: The intelligent control device converts the control parameters into control commands, including pump speed adjustment commands and valve opening adjustment commands, and transmits them to the execution and adjustment device through the communication and monitoring device. Step S33: The regulating device performs the following actions according to the control command: The electric regulating valve adjusts its opening with an accuracy of 0.1mm. The variable frequency centrifugal pump adjusts its speed according to fuzzy control output; The gate valve of the regulating reservoir is linked to a flow sensor to adjust the opening degree, ensuring that the measured flow rate tracks the set flow rate, with an adjustment error of less than or equal to ±3%. Step S34: The execution adjustment device feeds back the actual execution results to the communication and monitoring device to form a closed-loop control.

[0017] Therefore, the present invention employs the above-mentioned intelligent water supply and drainage control system and method with adaptive flow regulation, and the beneficial effects are as follows: (1) Improved the accuracy and real-time performance of signal acquisition. Through the adaptive filtering algorithm of the signal processing module, noise such as electromagnetic interference and mechanical vibration is dynamically removed. Compared with the traditional fixed filtering method, the signal acquisition accuracy is improved by 15%~25%, and the data transmission delay is reduced by 30%~40%, providing a highly reliable data foundation for the formulation of control strategies.

[0018] (2) Enhanced system adaptability and control precision. Relying on intelligent algorithms such as fuzzy control and LSTM neural network, the control strategy is dynamically adjusted. Compared with existing frequency conversion control technology, the flow regulation precision is improved by 20%~30%, the system response speed is increased by 40%~50%, and it can accurately adapt to complex and ever-changing water supply demands.

[0019] (3) Improved system stability and reliability. Through precise adjustment of the actuator and fault tolerance mechanism, compared with traditional control systems, the flow fluctuation amplitude is reduced by 50%~60%, the pressure imbalance rate is reduced by 70%~80%, and the mean time between failures (MTBF) is extended by 1.5 times~2 times, effectively avoiding problems such as system oscillation.

[0020] (4) The system has achieved intelligent management and remote monitoring. With the help of the "edge-cloud-end" communication architecture, it can achieve unattended operation and maintenance around the clock. Compared with the traditional manual inspection mode, the fault identification response time is shortened by 50% to 70%, and the manual maintenance cost is reduced by 40% to 60%, which greatly reduces the difficulty of maintenance and manpower input.

[0021] (5) It achieves the effects of energy saving, consumption reduction and equipment life extension. By adaptive flow regulation, the inefficient operation and frequent start-stop of the water pump are avoided. Compared with the traditional fixed speed water pump system, energy consumption is reduced by 20% to 30%, and the service life of core equipment such as water pumps and valves is extended by 30% to 50%, taking into account both energy saving benefits and equipment economy.

[0022] In summary, this invention achieves intelligent and adaptive control of the water supply and drainage system through the coordinated operation of various modules, significantly improving the system's control accuracy, response speed, stability, and reliability, and achieving the technical effects of energy saving, intelligent management, and efficient operation.

[0023] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0024] Figure 1 This is an overall structural block diagram of an intelligent water supply and drainage control system with adaptive flow regulation according to the present invention. Figure 2 This is a flowchart of an intelligent water supply and drainage control method with adaptive flow regulation according to the present invention. Detailed Implementation

[0025] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] like Figure 1 As shown, the present invention discloses an intelligent water supply and drainage control system with adaptive flow regulation, comprising a signal acquisition device, a signal processing device, an intelligent control device, an execution regulation device, and a communication and monitoring device. The devices work together to achieve adaptive flow regulation of the water supply and drainage system.

[0027] 1. Signal acquisition device, used to collect key operating parameters of the water supply and drainage system in real time, providing raw data support for subsequent signal processing and control decisions.

[0028] The signal acquisition device includes: flow rate acquisition element, water level acquisition element, and auxiliary acquisition element.

[0029] Flow acquisition element: High-precision flow sensor is used and deployed at the inlet and outlet of the storage tank and key nodes of the pipeline network for online calculation of pump efficiency and total flow measurement, with an acquisition accuracy of ≤±0.5%FS.

[0030] Water level acquisition element: An anti-interference water level sensor is used, deployed in the regulating reservoir and pump station forebay, for real-time monitoring of water level data and triggering low / high water level alarms. The measurement range is 0-10m, and the resolution is 0.1cm. The conditions for triggering low / high water level alarms are as follows: (1) Triggering conditions for high water level alarm: When the real-time water level is greater than or equal to 90% of the pool depth, an audible and visual alarm will be triggered immediately and pushed remotely to prevent the storage tank from overflowing; (2) Low water level alarm triggering conditions: When the real-time water level is less than or equal to 10%, an alarm is triggered and the pump set is controlled to reduce the load to prevent the water pump from running dry and being damaged.

[0031] Meanwhile, the system will record the duration of the water level exceeding the limit: if the exceeding state lasts for more than 30 seconds, the emergency plan will be automatically activated, the backup storage tank will be switched, and the operation and maintenance personnel will be notified.

[0032] Auxiliary acquisition components include a pressure sensor and a water flow acoustic signal spectrum characteristic sensor, which respectively collect pipeline pressure data and water flow acoustic signals to provide multi-dimensional data support for operating condition judgment; among them, the pressure sensor has a measurement range of 0-1.6MPa and an accuracy of ±0.2%FS.

[0033] All acquisition elements support the RS485 / Modbus-RTU protocol, and the data sampling frequency can be configured from 1 to 10 Hz to ensure real-time data acquisition.

[0034] II. The signal processing device employs adaptive filtering and edge computing technology to perform noise reduction, fusion, and buffering on the original acquired signals, outputting highly reliable and low-redundancy effective data.

[0035] The signal processing device includes: edge computing node elements, signal preprocessing elements, and data buffering and breakpoint resume elements.

[0036] Edge computing node components: Based on industrial-grade ARM chips, lightweight data acquisition programs, preprocessing algorithm kernels and local control logic are deployed in the chip, integrating data acquisition, preprocessing and local control functions to achieve offline autonomy and multi-node collaboration, ensuring that the system can still perform normal traffic regulation when the network is disconnected.

[0037] Signal preprocessing component: Relying on the computing power of the ARM chip in the edge computing node, an adaptive Kalman filter algorithm is adopted, combined with multi-sensor data fusion technology, to dynamically adjust the filter coefficients, remove noise such as electromagnetic interference and mechanical vibration, and transform the noisy raw signal into effective data; at the same time, through data consistency verification, conflicting data is eliminated to ensure data credibility ≥99.9%; the filter coefficient adjustment range is 0.01-0.1, and the adjustment step size is 0.001.

[0038] Data caching and breakpoint resume components: A circular caching mechanism is adopted to cache compressed data for 72 hours; data is continuously recorded when the network is interrupted, and the data is resumed according to the data sequence number after the network is restored, ensuring that the integrity rate of cloud model data is not less than 99.9%; at the same time, it supports seamless connection of control logic to avoid traffic overshoot caused by restarting after communication interruption; the data compression ratio is 10:1.

[0039] Third, the intelligent control device, based on the processed effective data, uses intelligent algorithms to realize operating condition identification, control strategy optimization and fault tolerance, and is the core of the system's adaptive adjustment.

[0040] The intelligent control device includes: online condition discrimination element, adaptive model update element, model predictive control element, cloud-edge collaborative element, and fault-tolerant element.

[0041] Online operating condition discrimination element: Construct a three-dimensional feature database of flow-pressure-water flow acoustic signal spectrum characteristics, and use the support vector machine (SVM) algorithm to compare the collected data with the feature database data in real time, extract frequency domain feature vectors of faults such as leakage, blockage, and pump surge, and output the operating condition discrimination result and confidence level within 50ms, triggering the control strategy switch; wherein, the dimension of the frequency domain feature vector is greater than or equal to 128 dimensions; the confidence level threshold is greater than or equal to 0.85.

[0042] When using the Support Vector Machine (SVM) algorithm, the radial basis function (RBF) kernel function should be selected. As shown below: ; in, The feature vector of the working condition to be determined; The sample vectors in the feature library; Indicates the index of the sample; The square of the Euclidean distance; This represents an exponential function with base e; The kernel function bandwidth parameter, with a value between 0.01 and 0.1, is used to adapt to the nonlinear distribution of the acoustic spectrum characteristics of flow-pressure-water flow, thereby improving the accuracy of fault condition discrimination. The larger the kernel size, the stronger the "locality" of the kernel function, meaning it only applies to samples. Close to It has a significant response and is suitable for distinguishing working conditions with large differences in details, such as "minor blockage" and "severe blockage"; The smaller the kernel function, the stronger its "globality" and the wider the range of feature vectors it can cover. It is suitable for distinguishing between different operating conditions, such as "normal operation" and "pipe rupture and leakage".

[0043] Adaptive model update component: An integrated miniature Long Short-Term Memory (LSTM) network model is used to retrain model weights locally based on the operating condition judgment results and real-time flow error, enabling dynamic updates to the flow prediction model to adapt to changes in pipeline characteristics (such as pipeline aging and seasonal water usage fluctuations). Specifically, the error threshold is less than or equal to 5%; the number of model training iterations is less than or equal to 100, with a learning rate of 0.001-0.01; pipeline characteristic changes include pipeline aging and seasonal water usage fluctuations.

[0044] The built-in miniature Long Short-Term Memory (LSTM) network model employs a 2-layer hidden layer structure, as detailed below: The first hidden layer has 64 neurons, and the activation function used is ReLU. The second hidden layer has 32 neurons, and the activation function used is tanh. The output layer is a fully connected layer with 1 neuron, used to output traffic prediction values. The total number of model parameters is controlled within 10,000, which is adapted to the computing power of ARM chips in edge computing nodes.

[0045] Model predictive control element: Based on the updated LSTM model, the flow rate change trend is predicted for the next 5-30 seconds. Combined with system constraints, the model predictive control (MPC) algorithm is used to calculate the optimal control parameters, with a control period of ≤1 second. The system constraints include the rated speed of the pump and the maximum opening of the valve, etc. The control parameters include the pump speed adjustment and the valve opening adjustment, etc.

[0046] Cloud-edge collaborative components: Based on the model parameter synchronization mechanism between edge nodes and cloud servers, edge nodes upload 24-hour compressed data to the cloud. The cloud retrains the LSTM model using a large-scale dataset and sends the optimized model parameters back to the edge nodes, achieving closed-loop control with fast local response (≤100ms response time) and fine-tuning in the cloud. The data compression format is differential coding + LZ77 algorithm; the LSTM model training batch size is 32, and the number of iterations is 500.

[0047] Fault-tolerant components: A three-mode redundancy voting mechanism is adopted to perform redundancy verification on sensor data and actuator feedback data; when a sensor or actuator failure is detected, the faulty component is automatically isolated within 10ms, switched to the backup channel, and the reduced-order control table is called, with a preset reduced-order control strategy of greater than or equal to 5 levels to ensure that the flow regulation error is less than or equal to 10%.

[0048] Fourth, the actuator adjusts the action of the actuator according to the control command output by the intelligent control device to achieve precise control of flow and pressure.

[0049] The regulating devices include: electric regulating valves, variable frequency centrifugal pumps, storage tank gate valves, and electrical safety components.

[0050] Electric regulating valve: It adopts a high-precision electric actuator with a stroke adjustment accuracy of 0.1mm and a command refresh frequency of 1Hz. It adjusts the opening degree in real time according to the control command, and the opening degree adjustment range is 0-100%. The opening degree error is ≤±0.5% through Hall closed-loop feedback.

[0051] Variable frequency centrifugal pump: adopts a PLC + fuzzy control architecture, and the pump group adopts a multi-pump parallel and rotation control strategy; the fuzzy control algorithm takes the flow deviation (E) and deviation change rate (EC) as inputs, and after fuzzification, fuzzy inference and defuzzification, outputs the water pump speed adjustment, with a speed adjustment range of 500-3000 rpm, avoiding frequent water pump start-stop, with a start-stop interval of greater than or equal to 5 minutes and water hammer effect; among them, the fuzzification membership function adopts the triangular function; the fuzzy inference adopts the Mamdani inference method; and the defuzzification adopts the centroid method.

[0052] Storage tank gate valve: adopts soft-seal electric gate valve, which is linked with the flow sensor through electric actuator. The opening adjustment range is 0-100%, the adjustment accuracy is ±1%, and it integrates fault self-check function with a self-check cycle of 1 minute and emergency shut-off function with a shut-off response time of less than or equal to 2 seconds. It is suitable for extreme water levels and rainstorm conditions.

[0053] Electrical safety components construct a triple safety protection system of "leakage current - insulation - arc", as shown below: (1) When the residual current operated circuit breaker (RCBO) on the incoming line side is greater than or equal to 30mA, it will instantly disconnect the power supply within 30ms; (2) The DC insulation tester monitors the insulation resistance of the main circuit. When the insulation resistance is less than 1MΩ, it is locked and switched to the standby pump. (3) The arc fault circuit interrupter (AFCI) on the outgoing side extinguishes the arc and prevents fire within 2ms; each component is connected to an independent safety PLC, which can force power off when the main control fails.

[0054] V. Communication and monitoring devices enable remote transmission of system data, status monitoring and remote control, and support intelligent management and maintenance.

[0055] Communication and monitoring equipment includes: network topology, data frame format, monitoring platform, and mobile terminal.

[0056] The network topology adopts a three-layer architecture of field-edge-cloud, specifically including: (1) Field layer: The RS-485 dual-bus redundant structure is adopted. The main link is Modbus-RTU and the backup link is CANopen. The data refresh cycle is less than or equal to 200ms. The system operation is not affected by the disconnection of any link.

[0057] (2) Edge layer: The industrial-grade ARM gateway has built-in MQTT, OPC UA and MQTT over TLS protocols to realize protocol conversion and data compression with a compression efficiency of greater than or equal to 80%. It is equipped with an AI chip to identify 10 typical abnormal working conditions in real time and alarm locally. Among them, abnormal working conditions include: water level rise, pipe burst and motor abnormality, etc.

[0058] (3) Cloud layer: Ensure data transmission reliability ≥99.9% by using 4G / 5G dual SIM dual standby or NB-IoT backup to the cloud.

[0059] The data frame format adopts a 32-byte fixed-length structure, specifically: frame header (2B) + timestamp (4B) + node ID (2B) + data type (2B) + differential compressed payload (20B) + CRC16 checksum (2B) + frame trailer (2B). Hardware verification and dual-channel redundancy comparison ensure a data transmission error rate ≤10%. -6 .

[0060] The monitoring platform is based on an edge-cloud-device microservice architecture. The industrial gateway has a built-in Redis time-series cache and an SQLite black box to perform data cleaning and millisecond-level closed-loop control on-site. The cloud deploys a time-series database and a digital twin model, which supports synchronous monitoring via web pages, apps and WeChat mini programs, and provides functions such as GIS global liquid level heat map, energy consumption benchmarking and fault tracing.

[0061] The mobile terminal adopts an integrated monitoring interface and quick control functions based on GIS maps, as detailed below: (1) The homepage GIS map displays liquid level, flow rate, pump status and alarm information in real time, and supports gesture zoom operation; (2) The bottom scene panel can be set up to switch between peak supply guarantee, rainy day overflow, and nighttime voltage reduction modes with one click. Control commands are sent to the edge gateway within 3 seconds.

[0062] like Figure 2 As shown, based on the above control system, this invention also proposes an intelligent water supply and drainage control method with adaptive flow regulation, comprising the following steps: Step S1, Signal Acquisition and Preprocessing: Acquire real-time data from the pipeline network and process the data using an adaptive Kalman filter algorithm. After consistency verification, output highly reliable and valid data.

[0063] Step S11: The signal acquisition device collects raw data such as flow rate, pressure, water level, and water flow acoustic signal spectrum characteristics of the pipeline network in real time at a sampling frequency of 1-10Hz, and transmits them to the signal processing device through RS485 / Modbus-RTU protocol.

[0064] Step S12: The signal preprocessing element uses an adaptive Kalman filter algorithm to dynamically adjust the filter coefficients and perform noise reduction on the original data, as shown below: ; ; ; ; ; = + ( -H ); ; in, This is the system state vector (including flow rate, pressure, and water level). This is the system state vector at time k-1; It is a 3×3 dimensional state transition matrix in diagonal matrix form, describing the "inheritance coefficient" of the previous state to the current state. It is calibrated based on the hydraulic characteristics of the pipeline network (pipeline resistance, storage tank volume) and has a value range of 0.95~0.99 (the closer to 1, the stronger the state continuity). For control input; The control input matrix is ​​3×2 dimensional, describing the weight of each control input on each state. This is process noise; For observation vectors (data acquired by sensors); It is a 3×3 dimension observation matrix in identity matrix form; To observe noise; This is the prior state estimate at time k; This is the posterior state estimate at time k-1; The prior covariance matrix; Let be the posterior covariance matrix at time k-1; The process noise covariance matrix; Kalman gain; To observe the noise covariance matrix; This is the posterior state estimate at time k; The posterior covariance matrix; It is the identity matrix; This is for the transpose operation.

[0065] Step S13: Using multi-sensor data fusion technology, perform consistency verification on the same type of data collected by different sensors, remove conflicting data with a deviation of more than 3%, and output highly reliable and valid data.

[0066] Step S14: The data caching and breakpoint resume element compresses and caches valid data, continuously records it when the network is disconnected, and verifies and resumes transmission after the network is restored.

[0067] Step S2, Operating Condition Judgment and Model Update: Extract the three-dimensional feature vector of the flow-pressure-water flow acoustic signal spectrum features, compare it with the preset feature database, and adaptively update the model weights based on the operating condition judgment results and real-time flow error to complete the LSTM model iterative optimization.

[0068] Step S21: The online operating condition discrimination element extracts the three-dimensional feature vector of the flow-pressure-water flow acoustic signal spectrum features from the valid data, compares it with the preset feature database, and uses the SVM algorithm to calculate the operating condition confidence level, as shown below: ; in, The output of the discriminant function represents the input feature vector. The results of the working condition classification; For Lagrange multipliers; For sample labels; The kernel function (using the RBF kernel function); The feature vector to be discriminated; The sample vectors in the feature library; For bias terms; It is a symbolic function; Indicates the index of the sample; The total number of samples.

[0069] Step S22: When the confidence level is greater than or equal to 0.85, output the operating condition judgment result and trigger the control strategy switch; the judgment result includes: normal, leakage, blockage and pumping.

[0070] Step S23: Based on the operating condition judgment result and the real-time flow error, the adaptive model updates the weights; wherein, the operating condition judgment result and the real-time flow error... The calculation formula is as follows: ; in, This is the actual measured flow rate; Predict traffic flow for the model; To set the flow rate; when the error value e is greater than 5%, retrain the mini LSTM model and update the model weights.

[0071] Step S24: The edge node uploads the 24-hour compressed data to the cloud. The cloud retrains the LSTM model using a large-scale dataset and sends the optimized model parameters back to the edge node to complete the model iteration update.

[0072] Step S3, Control Strategy Generation and Execution: Based on the updated LSTM model, predict the flow rate change trend, calculate the optimal control parameters, and implement the action of the control device.

[0073] Step S31: The model predictive control element, based on the updated LSTM model, predicts the flow rate change trend for the next 5-30 seconds. Combined with system constraints (pump speed 500-3000 rpm, valve opening 0-100%), the MPC algorithm is used to minimize the objective function. The optimal control parameters are calculated as follows: ; in, For prediction in the time domain; For time steps; To control the weighting coefficients; The control input at time t; This is the control input at time t-1; Predict traffic flow for the model; To set the flow rate.

[0074] Step S32: The intelligent control device converts the control parameters into control commands (pump speed adjustment command, valve opening adjustment command), and transmits them to the execution and regulation device through the communication and monitoring device.

[0075] Step S33: The regulating device performs the following actions according to the control command: The electric regulating valve adjusts its opening with an accuracy of 0.1mm. The variable frequency centrifugal pump adjusts its speed according to fuzzy control output; The gate valve of the regulating reservoir is linked to the flow sensor to adjust the opening degree, ensuring that the measured flow rate tracks the set flow rate, and the adjustment error is less than or equal to ±3%.

[0076] Step S34: The execution regulating device feeds back the actual execution results (measured flow rate, pressure, speed, etc.) to the communication and monitoring device to form a closed-loop control.

[0077] Step S4, Fault Tolerance and Remote Monitoring: Through a three-mode redundancy voting mechanism, sensor data and actuator feedback data are verified in real time, and real-time monitoring is visualized through a monitoring platform and mobile terminal.

[0078] Step S41: The fault-tolerant component uses a three-modal redundancy voting mechanism to perform real-time verification of sensor data and actuator feedback data. When a fault is detected, the faulty component is automatically isolated and switched to the backup channel, and the downgraded control strategy is invoked.

[0079] Step S42: The communication and monitoring device collects system operation data (flow rate, pressure, water level, equipment status, etc.) in real time and displays it through the monitoring platform and mobile terminal, supporting functions such as GIS global liquid level heat map, energy consumption benchmarking, and fault alarm.

[0080] Step S43: Managers can switch operating modes (peak supply guarantee / rain overflow / nighttime voltage reduction, etc.) with one click via mobile terminal, and control commands are issued and executed within 3 seconds; when a system failure occurs, the monitoring platform and mobile terminal push alarm information in real time (including fault type, location, and handling suggestions), and managers can remotely guide fault handling or activate emergency control strategies.

[0081] Example 1 This embodiment is applied to the water supply and drainage system of a residential community in a city. The community has a total of 1,000 households and a designed daily water consumption of 150 cubic meters. 3 The pipeline is 2km long and has a diameter of DN50-DN200.

[0082] (a) System configuration.

[0083] Signal acquisition devices: 10 flow sensors, model LDG-MIK, with an accuracy of ±0.5%FS; 8 water level sensors, model submersible level gauges, with a measurement range of 0-10m; 12 pressure sensors, model MPM480, with an accuracy of ±0.2%FS; and 6 water flow acoustic signal spectrum characteristic sensors, model customized underwater acoustic sensors.

[0084] Signal processing unit: adopts an industrial-grade ARM gateway, model RK3588, with built-in adaptive Kalman filter algorithm and data buffer module.

[0085] Intelligent control device: The edge node adopts a PLC, model S7-1500, with a built-in micro LSTM model and MPC algorithm; the cloud adopts Alibaba Cloud server, deploying time series database and digital twin model.

[0086] The control system includes: 4 electric regulating valves, model ZDLP-16C, with a stroke accuracy of 0.1mm; 6 variable frequency centrifugal pumps, model ISG100-160, with a rated speed of 2900rpm; 2 rainwater storage tank gate valves, model Z941H-16C, with a soft-seal structure; and several electrical safety components, including RCBO, DC insulation tester, and AFCI.

[0087] Communication and monitoring equipment: The field layer adopts an RS-485 dual-bus redundant structure, with Modbus-RTU as the main link and CANopen as the backup link. The edge layer adopts an industrial-grade ARM gateway, and the cloud layer adopts 4G / 5G dual-SIM dual-standby cloud access. The monitoring platform adopts a web page + APP + WeChat mini-program architecture.

[0088] (II) Implementation process.

[0089] System initialization: Set the water flow rate for the community, with a peak flow rate of 10m³ / h. 3 / h, off-peak hours 5m 3 / h, 3m during nighttime. 3 / h; Configure the sensor sampling frequency to 5Hz, the initial value of the filter coefficient to 0.05, the initial weights of the LSTM model, the MPC algorithm prediction time domain N=10, and the control weight coefficient λ=0.1.

[0090] Signal acquisition and preprocessing: The signal acquisition device collects real-time data on the spectral characteristics of flow rate, pressure, water level, and water flow acoustic signals. The signal processing device reduces noise through adaptive Kalman filtering, eliminates conflicting data through multi-sensor data fusion, and caches 72 hours of compressed data using a data caching module.

[0091] Operating condition discrimination and model update: The online operating condition discrimination element compares feature vectors in real time. In a certain operation, it detects a pipeline leakage condition with a confidence level of 0.92, triggering a switch in the leakage control strategy; the adaptive model update element retrains the LSTM model on-site and updates the weights due to a real-time flow error of 6.2%; the edge node uploads 24-hour data to the cloud, and after retraining in the cloud, it sends back the optimized parameters.

[0092] Control strategy execution: The model predictive control element predicts flow changes based on the updated model and calculates the optimal control command using the MPC algorithm. The electric regulating valve closes its opening by 30%, the variable frequency centrifugal pump reduces its speed to 1800 rpm, and the stormwater storage tank gate valve closes its opening by 20%. Within 10 seconds, the measured flow rate decreases from 8.5 m³ / h. 3 / h decreased to the set value of 6m 3 / h, adjustment error 2.1%.

[0093] Fault handling and monitoring: During a certain operation, the pressure sensor failed. The fault-tolerant component switched to the backup sensor within 8ms and called the reduced-order control strategy, maintaining the flow regulation error at 8.3%. The monitoring platform pushed fault alarms in real time, and the management personnel confirmed and arranged maintenance remotely through mobile terminals, reducing the fault handling time to 2 hours.

[0094] (III) Implementation Results.

[0095] After running for 6 months, the system performance in this embodiment is as follows: Flow rate regulation accuracy: average regulation error 2.8%, meeting design requirements (≤±3%); Energy consumption performance: Compared with traditional constant speed water pump systems, energy consumption is reduced by 26.7%, saving approximately 3,200 yuan in electricity costs per month; Stability: The system's MTBF reached 8920 hours, and no water usage complaints were received due to flow fluctuations or pressure imbalances; Maintenance efficiency: Automatic fault alarm accuracy rate is 98.5%, fault handling response time is reduced by an average of 55%, and manual maintenance costs are reduced by 40%.

[0096] Example 2 This embodiment tested the operating condition discrimination time under different pipeline operating conditions, and the test results are shown in Table 1.

[0097] Table 1. Test results of operating condition discrimination time under different pipeline network conditions.

[0098] The test environment in this embodiment simulates a real residential community's water supply and drainage network, with pipe diameters ranging from DN50 to DN200 and a network length of 2 km, equipped with the intelligent control system of this invention. Each test simultaneously collects three-dimensional feature data of flow rate, pressure, and water flow acoustic signal spectrum characteristics, with a feature vector dimension of 128.

[0099] The discrimination time was defined as the time interval between the completion of feature vector extraction and the output of the working condition result and confidence level, all of which were collected using a high-precision timer. Results verified that the maximum discrimination time (46.8 ms) for all test conditions was less than 50 ms, and the average time (20.5 ms - 33.5 ms) was far below the threshold, validating the rapid response capability of the online working condition discrimination element.

[0100] Therefore, the present invention adopts the above-mentioned intelligent water supply and drainage control system and method with adaptive flow regulation. By integrating technologies such as adaptive filtering, fuzzy control, neural network, model predictive control and cloud-edge collaboration, it realizes dynamic removal of signal noise, real-time optimization of control strategy, online identification of operating conditions and fault tolerance, and finally achieves high-precision flow regulation, low-energy operation, intelligent management and high reliability assurance of water supply and drainage system.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent water supply and drainage control system with adaptive flow regulation, characterized in that, It includes signal acquisition devices, signal processing devices, intelligent control devices, execution and regulation devices, and communication and monitoring devices. These devices work together to achieve adaptive flow regulation of the water supply and drainage system. Signal acquisition device, used to collect operating parameters of water supply and drainage system in real time; The signal processing device employs adaptive filtering and edge computing techniques to perform noise reduction, fusion, and buffering on the original acquired signals. The intelligent control device, based on the processed data, uses intelligent algorithms to achieve operating condition identification, control strategy optimization, and fault tolerance. The actuator adjusts the action of the actuator according to the control commands output by the intelligent control device, thereby controlling the flow and pressure; Communication and monitoring devices enable remote transmission of system data, status monitoring and remote control, and support intelligent management and maintenance.

2. The intelligent water supply and drainage control system with adaptive flow regulation according to claim 1, characterized in that, The signal acquisition device includes: flow rate acquisition element, water level acquisition element, and auxiliary acquisition element; Flow acquisition elements: Flow sensors are used and deployed at the inlet and outlet of the storage tank and at pipeline nodes for online calculation of pump efficiency and total flow measurement. Water level acquisition element: An anti-interference water level sensor is used, deployed in the regulating reservoir and pump station forebay, to monitor water level data in real time and trigger water level alarms; the conditions for triggering water level alarms are as follows: (1) Triggering conditions for high water level alarm: When the real-time water level is greater than or equal to 90% of the pool depth, an audible and visual alarm will be triggered immediately and pushed remotely to prevent the storage tank from overflowing; (2) Low water level alarm triggering conditions: When the real-time water level is less than or equal to 10%, an alarm is triggered and the pump set is controlled to reduce the load to prevent the water pump from running dry and being damaged. Meanwhile, the system will record the duration of the water level exceeding the limit: if the exceeding state lasts for more than 30 seconds, the emergency plan will be automatically activated, the backup storage tank will be switched, and the operation and maintenance personnel will be notified. Auxiliary acquisition components include pressure sensors and water flow acoustic signal spectrum characteristic sensors, which respectively collect pipeline pressure data and water flow acoustic signals to provide multi-dimensional data support for operating condition judgment.

3. The intelligent water supply and drainage control system with adaptive flow regulation according to claim 1, characterized in that, The signal processing device includes: edge computing node components, signal preprocessing components, and data buffering and breakpoint resume components; Edge computing node components: Based on industrial-grade ARM chips, lightweight data acquisition programs, preprocessing algorithm kernels and local control logic are deployed in the chip, integrating data acquisition, preprocessing and local control functions to achieve offline autonomy and multi-node collaboration, ensuring that the system can perform normal traffic regulation when the network is disconnected; Signal preprocessing components: Relying on the computing power of the ARM chip in the edge computing node, an adaptive Kalman filter algorithm is adopted, combined with multi-sensor data fusion technology, to dynamically adjust the filter coefficients to remove noise and transform the noisy raw signal into effective data; at the same time, through data consistency verification, conflicting data is eliminated to ensure data credibility. Data caching and breakpoint resume components: A circular caching mechanism is adopted to cache compressed data for 72 hours; data is continuously recorded when the network is interrupted, and the data is resumed according to the data sequence number after the network is restored, ensuring that the integrity rate of cloud model data is not less than 99.9%; at the same time, it supports seamless connection of control logic to avoid traffic overshoot caused by restarting after communication interruption.

4. The intelligent water supply and drainage control system with adaptive flow regulation according to claim 1, characterized in that, The intelligent control device includes: online condition discrimination element, adaptive model update element, model predictive control element, cloud-edge collaborative element, and fault-tolerant element; Online operating condition discrimination element: Construct a three-dimensional feature database of flow-pressure-water flow acoustic signal spectrum features, adopt a support vector machine algorithm, compare the collected data with the feature database data in real time, extract the fault frequency domain feature vector, and output the operating condition discrimination result and confidence level within 50ms, triggering the control strategy switching; When using the Support Vector Machine (SVM) algorithm, the radial basis function (RBF) kernel function should be selected as the kernel type. As shown below: ； in, The feature vector of the working condition to be determined; The sample vectors in the feature library; Indicates the index of the sample; The square of the Euclidean distance; The kernel function bandwidth parameter, with a value between 0.01 and 0.1, is used to adapt to the nonlinear distribution of the acoustic spectrum characteristics of flow-pressure-water flow, thereby improving the accuracy of fault condition discrimination. This represents an exponential function with base e; Adaptive model update component: The built-in miniature long short-term memory network LSTM model retrains the model weights based on the working condition discrimination results and real-time flow error, realizing the dynamic update of the flow prediction model and adapting to changes in pipeline characteristics. The built-in miniature Long Short-Term Memory (LSTM) network model employs a two-layer hidden layer structure, as detailed below: The first hidden layer has 64 neurons, and the activation function used is ReLU. The second hidden layer has 32 neurons, and the activation function used is tanh. The output layer is a fully connected layer with 1 neuron, used to output the predicted flow rate. Model predictive control element: Based on the updated LSTM model, the flow rate change trend is predicted. Combined with system constraints, the model predictive control algorithm is used to calculate the optimal control parameters. Cloud-edge collaborative components: Based on the model parameter synchronization mechanism between edge nodes and cloud servers, edge nodes upload 24-hour compressed data to the cloud, and the cloud retrains the LSTM model with a large-scale dataset and sends the optimized model parameters back to the edge nodes, realizing local response and cloud closed-loop control; Fault-tolerant components: A three-modal redundancy voting mechanism is adopted to perform redundancy verification on sensor data and actuator feedback data; when a sensor or actuator failure is detected, the faulty component is automatically isolated within 10ms, switched to the backup channel, and the downgraded control table is invoked.

5. The intelligent water supply and drainage control system with adaptive flow regulation according to claim 1, characterized in that, The regulating devices include: electric regulating valves, variable frequency centrifugal pumps, storage tank gate valves, and electrical safety components; Electric regulating valve: It adopts an electric actuator to adjust the opening degree in real time according to control commands; Variable frequency centrifugal pump: adopts PLC + fuzzy control architecture, and the pump group adopts a multi-pump parallel and rotation control strategy; the fuzzy control algorithm takes the flow deviation and deviation change rate as input, and after fuzzification, fuzzy inference and defuzzification, outputs the pump speed adjustment amount; The gate valve of the storage tank is a soft-seal electric gate valve, which is linked with the flow sensor through an electric actuator. It integrates a fault self-check function with a self-check cycle of 1 minute and an emergency shut-off function. Electrical safety components construct a triple safety protection system of leakage current, insulation, and arcing, as shown below: (1) When the residual current operated circuit breaker (RCBO) on the incoming line side is greater than or equal to 30mA, it will instantly disconnect the power supply within 30ms. (2) The DC insulation tester monitors the insulation resistance of the main circuit. When the insulation resistance is less than 1MΩ, it is locked and switched to the standby pump. (3) The arc fault circuit breaker AFCI on the outgoing side extinguishes the arc within 20ms and prevents fires, effectively preventing electrical fires; each component is connected to an independent safety PLC, and the main control fails to force power off.

6. The intelligent water supply and drainage control system with adaptive flow regulation according to claim 1, characterized in that, Communication and monitoring equipment includes: network topology, data frame format, monitoring platform, and mobile terminal; The network topology adopts a three-layer architecture of field-edge-cloud, specifically including: (1) Field layer: The RS-485 dual-bus redundant structure is adopted. The main link is Modbus-RTU and the backup link is CANopen. The data refresh cycle is less than or equal to 200ms. The system operation is not affected by the disconnection of any link. (2) Edge layer: The industrial-grade ARM gateway has built-in MQTT, OPC UA and MQTT over TLS protocols to realize protocol conversion and data compression, and is equipped with an AI chip to identify abnormal working conditions in real time and alarm locally; (3) Cloud layer: Ensure data transmission reliability by using 4G / 5G dual SIM dual standby or NB-IoT backup to the cloud; The data frame format adopts a 32-byte fixed-length structure, and hardware verification and dual-channel redundancy comparison ensure that the data transmission error rate is ≤10%. -6 ; The monitoring platform is based on an edge-cloud-device microservice architecture. The industrial gateway has a built-in Redis time-series cache and an SQLite black box to complete data cleaning and millisecond-level closed-loop control. The cloud deploys a time-series database and a digital twin model, which supports synchronous monitoring of web pages, APP and WeChat mini-program, and provides functions such as GIS global liquid level heat map, energy consumption benchmarking and fault tracing. The mobile terminal adopts an integrated monitoring interface and quick control functions based on GIS maps, as detailed below: (1) The homepage GIS map displays liquid level, flow rate, pump status and alarm information in real time, and supports gesture zoom operation; (2) The bottom scene panel has a one-click mode switching function, and the control command is sent to the edge gateway within 3 seconds.

7. A method for an adaptive flow regulation intelligent water supply and drainage control system according to any one of claims 1-6, characterized in that, Includes the following steps: Step S1: Collect real-time data from the pipeline network, process the data using an adaptive Kalman filter algorithm, and output valid data after consistency verification. Step S2: Extract the three-dimensional feature vector of the flow-pressure-water flow acoustic signal spectrum features, compare it with the preset feature database, and adaptively update the weights of the model based on the working condition discrimination result and the real-time flow error to complete the iterative optimization of the LSTM model. Step S3: Based on the updated LSTM model, predict the flow rate change trend, calculate the optimal control parameters, and realize the action of the control device. Step S4: The fault-tolerant component uses a three-mode redundancy voting mechanism to verify the sensor data and actuator feedback data in real time. When a fault is detected, the faulty component is automatically isolated and switched to the backup channel, and the downgraded control strategy is invoked. The communication and monitoring devices collect system operation data in real time and display it through the monitoring platform and mobile terminals; Administrators can switch operating modes with a single click via mobile terminal, and control commands can be issued and executed within 3 seconds. When a system malfunctions, the monitoring platform and mobile terminals push alarm information in real time, and managers can remotely guide the troubleshooting or activate emergency control strategies.

8. The method for an intelligent water supply and drainage control system with adaptive flow regulation according to claim 7, characterized in that, In step S1, the specific implementation process of signal acquisition and preprocessing is as follows: Step S11: The signal acquisition device collects the raw data of the pipeline network in real time at a sampling frequency of 1-10Hz and transmits it to the signal processing device via RS485 / Modbus-RTU protocol. Step S12: The signal preprocessing element uses an adaptive Kalman filter algorithm to dynamically adjust the filter coefficients and perform noise reduction on the original data. Step S13: Using multi-sensor data fusion technology, perform consistency verification on the same type of data collected by different sensors, remove conflicting data with a deviation of more than 3%, and output valid data. Step S14: The data caching and breakpoint resume element compresses and caches valid data, continuously records it when the network is disconnected, and verifies and resumes transmission after the network is restored.

9. The method for an intelligent water supply and drainage control system with adaptive flow regulation according to claim 7, characterized in that, In step S2, the specific implementation process of working condition discrimination and model update is as follows: Step S21: The online operating condition discrimination element extracts the three-dimensional feature vector of the flow-pressure-water flow acoustic signal spectrum features from the valid data, compares it with the preset feature database, and uses the SVM algorithm to calculate the operating condition confidence level, as shown below: ； in, The output of the discriminant function represents the input feature vector. The results of the working condition classification; For Lagrange multipliers; For sample labels; For kernel functions; The feature vector to be discriminated; The sample vectors in the feature library; For bias terms; It is a symbolic function; Indicates the index of the sample; The total number of samples; Step S22: When the confidence level is greater than or equal to 0.85, output the operating condition judgment result and trigger the control strategy switch; Step S23: Based on the operating condition judgment result and the real-time flow error, the adaptive model updates the weights; wherein, the calculation formula for the operating condition judgment result and the real-time flow error is as follows: ； in, This is the actual measured flow rate; Predict traffic flow for the model; To set the flow rate; when the error value e is greater than 5%, retrain the mini LSTM model and update the model weights; Step S24: The edge node uploads the 24-hour compressed data to the cloud. The cloud retrains the LSTM model using a large-scale dataset and sends the optimized model parameters back to the edge node to complete the model iteration update.

10. The method for an intelligent water supply and drainage control system with adaptive flow regulation according to claim 7, characterized in that, In step S3, the specific implementation process of control strategy generation and execution is as follows: Step S31: The model predictive control element, based on the updated LSTM model, predicts the flow rate change trend. Combining this with system constraints, the MPC algorithm is used to minimize the objective function and calculate the optimal control parameters, as shown below: ； in, For prediction in the time domain; For time steps; To control the weighting coefficients; The control input at time t; This is the control input at time t-1; Predict traffic flow for the model; To set the flow rate; Step S32: The intelligent control device converts the control parameters into control commands, including pump speed adjustment commands and valve opening adjustment commands, and transmits them to the execution and adjustment device through the communication and monitoring device. Step S33: The regulating device performs the following actions according to the control command: The electric regulating valve adjusts its opening with an accuracy of 0.1mm. The variable frequency centrifugal pump adjusts its speed according to fuzzy control output; The gate valve of the regulating reservoir is linked to a flow sensor to adjust the opening degree, ensuring that the measured flow rate tracks the set flow rate, with an adjustment error of less than or equal to ±3%. Step S34: The execution adjustment device feeds back the actual execution results to the communication and monitoring device to form a closed-loop control.

Images