Intelligent monitoring system and method for full-flow water network of large power station

Abstract

The invention discloses a large power station full flow water network intelligent monitoring system and method, a system edge calculation layer comprises a data cleaning module and a space-time feature extraction module, a platform layer comprises a dynamic water balance engine, a space-time fusion leakage model and a multi-target optimizer, and an application layer comprises a three-dimensional digital twin interface and an intelligent work order system; the data output end of the ultrasonic flowmeter, the data output end of the multi-parameter water quality sensor and the data output end of the pressure transmitter are connected with the data input end of the data cleaning module, and the output end of the data cleaning module is connected with the input end of the spatial-temporal feature extraction module. The output end of the space-time feature extraction module is connected with the input end of the dynamic water balance engine, the output end of the dynamic water balance engine is connected with the input end of the space-time fusion leakage model, and the output end of the space-time fusion leakage model is connected with the input end of the multi-target optimizer. And the output end of the multi-target optimizer is connected with the input end of the three-dimensional digital twinning interface.

Description

Technical Field

[0001] The present invention belongs to the cross-technical field of smart energy and industrial Internet of Things, and specifically relates to a full-process water network intelligent monitoring system and method for a large-scale power station. Background Art

[0002] At present, the water balance management of power station water network mainly relies on annual manual testing, which has significant defects. Manual testing is affected by instrument calibration errors and instantaneous operating condition fluctuations, and the comprehensive error is generally high. The accuracy of manual testing has always been a problem that plagues the water balance management of power station water network. The influence of instrument calibration errors and instantaneous operating condition fluctuations makes the results of manual testing have great uncertainty, which makes it difficult to meet the high-precision water network management requirements of power stations. The test cycle is as long as 1-2 weeks, and it is impossible to capture sudden leakage or water use abnormalities in time, such as the situation where the cooling tower packing is blocked and the evaporation increases sharply. The long test cycle makes it impossible for manual testing to detect sudden problems in the water network in time, which may lead to more serious consequences. The traditional static model does not consider the influence of variables such as ambient temperature, humidity, and wind speed on evaporation, resulting in a low error correction rate and cannot meet the needs of power station water network management. The influence of environmental factors on evaporation cannot be ignored, but the traditional static model fails to consider these factors, making the water balance calculation inaccurate.

[0003] Existing leakage detection technologies mainly use threshold alarm methods or single-mode sensor analysis, which have obvious shortcomings. The alarm method based on setting thresholds for sudden drops in flow has a high false alarm rate and cannot effectively distinguish between real leakage and water load fluctuations. This makes it difficult for staff to determine whether there is a real leakage problem when facing an alarm, which may lead to unnecessary inspections and repairs, wasting manpower and material resources. The current leakage positioning technology has large errors and it is difficult to accurately determine the location of the leakage, which brings great difficulties to maintenance work, prolongs maintenance time, and increases the waste of water resources.

[0004] The leakage location method that relies on the data of a single pressure sensor has a large positioning error due to the lack of fusion and verification of multi-source data. In the complex water network system of a power station, the data of a single sensor is often difficult to accurately reflect the status of the entire pipe network, resulting in a large deviation in positioning. Acoustic sensors are seriously interfered by environmental noise when detecting leakage, and the probability of failure in complex pipe networks is high. There are various noise sources in the actual operating environment of the power station water network. These noises will mask the acoustic signals generated by the leakage, making it difficult for the acoustic sensor to accurately detect the leakage location, and the complex pipe network structure will also increase the complexity of acoustic signal propagation, further reducing the accuracy of detection. The power station water network involves multiple subsystems such as circulating water, chemical water, desulfurization wastewater, and domestic water. The existing technology has a serious data fragmentation problem.

[0005] Water quality data and equipment operation data are scattered across different platforms, lacking a unified data model, which results in the inability to effectively integrate and share data. Water quality data and equipment operation data are both crucial for the management of the power station's water network. However, due to the lack of a unified data model, these data cannot be comprehensively analyzed and synergistically optimized. For example, when adjusting the pump frequency or valve opening, the impact of water quality changes on equipment operation cannot be considered in a timely manner, which may lead to water quality deterioration or equipment failures.

[0006] Water-saving strategies often come at the expense of water quality, while water quality optimization may increase water treatment costs. In the existing management of power station water networks, water saving and water quality optimization are often regarded as two independent goals, lacking an effective coordination mechanism. When implementing water-saving strategies, the requirements of water quality may be overlooked, leading to a decline in water quality; while when optimizing water quality, it may increase water treatment costs and reduce the water-saving effect. For example, excessive increase in the circulating water concentration ratio can reduce water consumption, but at the same time increase the salt and impurity content in the water, resulting in an increased risk of scaling, which in turn affects the operation efficiency and lifespan of the equipment.

[0007] Based on the above defects, it is urgent to solve the problems of high-precision real-time calculation of dynamic water balance, accurate positioning and rapid response of pipeline leakage, and synergistic optimization of water quality safety and water-saving benefits. Although some improvement schemes in the existing technology use Internet of Things technology to achieve real-time water balance monitoring, they do not introduce a dynamic environmental compensation mechanism, resulting in still relatively high monitoring errors and unable to meet the high-precision requirements. In practical applications, the impact of environmental factors on water balance cannot be ignored, and the lack of a dynamic environmental compensation mechanism will make the monitoring results inaccurate. Some improvement schemes use deep learning to detect leakage, but only rely on single-dimensional flow data and cannot comprehensively reflect the leakage characteristics. In a complex pipeline system, leakage may be affected by various factors, and relying solely on flow data for detection is prone to false alarms with a relatively high false alarm rate. There are also improvement schemes that attempt to integrate water quality and water usage data, but the optimization model only supports linear weighting and is difficult to effectively handle multi-objective optimization problems. In practical applications, there are often complex relationships among goals such as water quality safety, water-saving benefits, and costs, and the linear weighting method is difficult to accurately weigh the importance of each goal and cannot achieve the global optimal water usage strategy. Summary of the Invention

[0008] The purpose of the present invention is to overcome the problems of inaccurate detection and slow response in pipeline leakage detection and positioning, and proposes a large-scale power station full-process water network intelligent monitoring system and method.

[0009] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides an intelligent monitoring system for the entire process water network of a large power station, including a perception layer, an edge computing layer, a platform layer, and an application layer. The edge computing layer is respectively connected to the perception layer and the platform layer. The platform layer is connected to the application layer. The perception layer includes an ultrasonic flowmeter, a multi-parameter water quality sensor, and a pressure transmitter. The edge computing layer includes a data cleaning module and a spatio-temporal feature extraction module. The platform layer includes a dynamic water balance engine, a spatio-temporal fusion leakage model, and a multi-objective optimizer. The application layer includes a three-dimensional digital twin interface and an intelligent work order system; The data output terminals of the ultrasonic flowmeter, the multi-parameter water quality sensor, and the pressure transmitter are respectively connected to the data input terminals of the data cleaning module. The output terminal of the data cleaning module is connected to the input terminal of the spatio-temporal feature extraction module. The output terminal of the spatio-temporal feature extraction module is connected to the input terminal of the dynamic water balance engine. The output terminal of the dynamic water balance engine is connected to the input terminal of the spatio-temporal fusion leakage model. The output terminal of the spatio-temporal fusion leakage model is connected to the input terminal of the multi-objective optimizer. The output terminal of the multi-objective optimizer is connected to the input terminal of the three-dimensional digital twin interface. The output terminal of the three-dimensional digital twin interface is connected to the input terminal of the intelligent work order system; The feedback terminal of the spatio-temporal feature extraction module is connected to the receiving terminal of the data cleaning module. The feedback terminal of the dynamic water balance engine is connected to the receiving terminal of the spatio-temporal feature extraction module. The feedback terminal of the spatio-temporal fusion leakage model is connected to the receiving terminal of the dynamic water balance engine. The feedback terminal of the multi-objective optimizer is connected to the receiving terminal of the spatio-temporal fusion leakage model. The feedback terminal of the intelligent work order system is connected to the receiving terminal of the multi-objective optimizer.

[0010] Further, the installation positions of the ultrasonic flowmeter include the main water supply pipe and the circulating water return pipe. The collected data of the ultrasonic flowmeter includes water flow rate. The ultrasonic flowmeter uses a non-invasive pipe-section ultrasonic flowmeter; The deployment nodes of the multi-parameter water quality sensor include the circulating water inlet, the circulating water outlet, the desulfurized wastewater discharge port, and the chemical water treatment outlet. The collected data of the multi-parameter water quality sensor includes water quality. The monitoring parameters of the water quality include pH, conductivity, and turbidity. The multi-parameter water quality sensor is provided with a self-calibration mechanism; The deployment position of the pressure transmitter includes the outlet of the high-pressure pump. The collected data of the pressure transmitter includes pressure time-series data. The pressure transmitter is redundantly deployed. When the data difference between the two sensors of the redundantly deployed pressure transmitter exceeds the alarm threshold, an alarm is triggered. The pressure transmitter is provided with pressure overload protection.

[0011] Further, the data cleaning module includes a sliding window cleaner, which includes an anomaly filtering logic and data compression. The anomaly filtering calculates the mean and standard deviation through a sliding window, eliminates data points exceeding the threshold, and fills in the missing values through linear interpolation to obtain the cleaned data; the data compression uses a lossy compression algorithm; The spatio-temporal feature extraction module expands the pressure gradient and flow ratio into a 128-dimensional vector feature matrix and inputs the 128-dimensional vector feature matrix into the platform layer; the spatio-temporal feature extraction module uses a 32-bit cyclic redundancy check algorithm to perform self-check on the data cleaning module.

[0012] Further, the dynamic water balance engine includes a three-level verification hierarchy and a dynamic compensation formula. The three-level verification hierarchy includes a whole-plant-level verification, a unit-level verification, and a unit-level verification; the three-level verification hierarchy samples and adjusts the spatio-temporal feature extraction module. The three-level verification hierarchy receives the feature matrix extracted by the spatio-temporal feature extraction module, combines the feature matrix with the dynamic compensation formula to convert it into a deviation heat map, and then inputs it into the spatio-temporal fusion leakage model; the technical indicators of the dynamic water balance engine include calculation delay, data integrity rate, and update period; The spatio-temporal fusion leakage model includes an ARIMA-LSTM dual-stream network structure and a fusion decision logic, a time synchronization unit, and an AI preprocessing unit. The fusion decision logic includes a thermal resistance threshold and a pressure gradient threshold. The spatio-temporal fusion leakage model uses the AI preprocessing unit to perform a preliminary screening of sudden flow drops on the deviation heat map, and then locates the leakage coordinates of the deviation heat map through the ARIMA-LSTM dual-stream network structure and the fusion decision logic, and inputs the deviation heat map and the leakage coordinates into the multi-objective optimizer; the spatio-temporal fusion leakage model performs clock calibration feedback on the dynamic water balance engine through the time synchronization unit, and the time synchronization unit includes an NTP protocol and a timestamp alignment process; the configuration of the spatio-temporal fusion leakage model includes the ARIMA difference order and the LSTM sliding window; The multi-objective optimizer includes an NSGA-III algorithm process and a Pareto solution set output. After receiving the deviation heat map and the leakage coordinates, the multi-objective optimizer uses the NSGA-III algorithm to combine the constraint conditions to obtain the Pareto solution set, and then outputs the deviation heat map and the Pareto solution set to the application layer. The Pareto solution set is an optimized parameter set; the multi-objective optimizer feeds back and updates the spatio-temporal fusion leakage model. The configuration of the multi-objective optimizer includes the population size of the NSGA-III algorithm, and the constraint conditions of the NSGA-III algorithm include the concentration ratio and COD.

[0013] Further, the dynamic compensation formula is as follows:

[0014] Where, where \(E\) is the evaporation rate and \(\alpha\) is the dynamic compensation factor, \(\Delta T\) is the temperature difference between the inlet and outlet, and 1.2 is the experimental fitting parameter.

[0015] Furthermore, the three-dimensional digital twin interface includes a heat map, a leakage error circle, and a water-saving strategy simulation module; the three-dimensional digital twin interface receives the deviation heat map and the optimized parameter set, performs heat map rendering and water-saving strategy simulation, generates a heat map, a leakage error circle, and a simulated water-saving strategy, and at the same time inputs the control instructions into the intelligent work order system; The intelligent work order management system includes an automatic dispatching logic and a closed-loop verification process. The intelligent work order management system automatically manages dispatching through the automatic dispatching logic and the closed-loop verification process according to the heat map, the leakage error circle, and the simulated water-saving strategy generated by the three-dimensional digital twin interface; the intelligent work order management system feeds back the dynamic compensation factor to the multi-objective optimizer for self-learning.

[0016] Furthermore, the real-time data flow is, in sequence, the perception layer, the edge layer, the platform layer, and the application layer; the feedback direction of the optimization instructions is, in sequence, the application layer, the platform layer, and the edge computing layer.

[0017] In a second aspect, the present invention provides an operation method for an intelligent monitoring system of a full-process water network in a large power station, using the intelligent monitoring system of a full-process water network in a large power station as described above, including the following steps: The ultrasonic flowmeter, multi-parameter water quality sensor, and pressure transmitter in the perception layer collect water network data and transmit it to the edge computing layer; After removing the outliers of the water network data through the data cleaning module in the edge computing layer, the spatio-temporal features of the cleaned water network data are extracted by the spatio-temporal feature extraction module and transmitted to the platform layer; The platform layer generates a deviation heat map through the dynamic water balance engine according to the spatio-temporal features of the cleaned water network data, inputs the deviation heat map into the spatio-temporal fusion leakage model to locate the leakage coordinates, and uses the deviation heat map and the leakage coordinates as the input of the multi-objective optimizer to obtain an optimized water network; the dynamic water balance engine adopts a three-level verification hierarchy; The optimized water network is input into the application layer, and the optimal strategy is converted into control instructions through the three-dimensional digital twin interface and input into the intelligent work order management system. The intelligent work order management system combines the leakage coordinates and the priority to perform automatic repair dispatching; The intelligent work order management system generates a dynamic compensation factor and feeds it back to optimize the multi-objective optimizer, and the three-dimensional digital twin interface real-time simulates and displays the optimized water network.

[0018] Furthermore, the three-level verification hierarchy includes a whole-plant-level verification, a unit-level verification, and a unit-level verification; The overall plant-level verification calculates the total water intake and the total dynamic water consumption, obtains the balance verification threshold by combining the total water intake and the total dynamic water consumption, and determines whether to trigger the overall plant-level calibration according to the balance verification threshold; The unit-level verification modifies the associated equipment according to the sub-unit water balance, and determines whether to perform equipment self-inspection or manual inspection according to the corrected error; The unit-level verification matches the water volume of the unit equipment, determines whether to perform maintenance according to the error limit, and outputs a list of abnormal units.

[0019] Furthermore, the feedback mechanism of the three-level verification hierarchy is that the failure of the unit-level verification triggers feedback to the unit level, and the failure of the unit-level verification review triggers feedback to the overall plant-level calibration to complete the closed-loop correction.

[0020] Compared with the prior art, the present invention has the following beneficial technical effects: A full-process water network intelligent monitoring system for large-scale power plants proposed by the present invention is applicable to large-scale power plants such as thermal power plants, nuclear power plants, hydropower plants, and solar thermal power plants. By integrating edge computing, multi-source data fusion, digital twin, and intelligent diagnosis technologies, it effectively solves the core problems of large water balance errors, fuzzy leakage location, and insufficient multi-index collaborative optimization in traditional water network management. A multi-objective optimization model is established, comprehensively considering multiple factors such as water saving rate, water quality stability, and cost. Through reasonable weight allocation and constraint condition setting, the collaborative optimization of multiple objectives is achieved. A multi-objective optimization algorithm can be used to balance and coordinate different optimization objectives. For example, on the premise of ensuring water quality stability, the water saving rate is increased as much as possible while reducing the water treatment cost. The water resource linkage scheduling across subsystems (such as circulating water and desulfurization wastewater) is realized to achieve the global optimal water use strategy. A unified water resource management platform needs to be established to integrate the water resource data and equipment operation data of each subsystem. By reasonably allocating and scheduling the water resources between different subsystems, the global optimal water use strategy is realized. For example, according to the water use requirements and water quality requirements of each subsystem, a dynamic water resource allocation plan is formulated to achieve the efficient utilization of water resources. Description of the Drawings

[0021] The drawings described herein are for illustrative purposes only and are not intended to limit the scope of the disclosure of the present invention in any way. In addition, the shapes and proportional dimensions of the components in the drawings are only schematic and are used to assist in understanding the present invention, rather than specifically limiting the shapes and proportional dimensions of the components of the present invention. In the drawings: Figure 1 It is a structural schematic diagram of a full-process water network intelligent monitoring system for a large-scale power plant.

[0022] Figure 2 It is a flow schematic diagram of the three-level verification hierarchy of the operation method of a full-process water network intelligent monitoring system for a large-scale power plant.

[0023] Figure 3 This is an optimization scheme and constraint boundary for an intelligent monitoring system of the full-process water network in a large power station in an embodiment of the present invention. Detailed implementation manners

[0024] In order to enable those skilled in the art of this technology to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0025] It should be noted that when an element is referred to as being "disposed on" another element, it can be directly on the other element or there may also be a middle element. When an element is considered to be "connected" to another element, it can be directly connected to the other element or there may be a middle element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are only for the purpose of illustration and do not represent the only embodiments.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the description of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0027] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0028] Embodiment 1 An intelligent monitoring system for the full-process water network of a large power station, including a sensing layer, an edge computing layer, a platform layer, an application layer, and a data stream, specifically including; see Figure 1 , specifically as follows: (1)The perception layer includes ultrasonic flowmeters, multi-parameter water quality sensors, and pressure transmitters; The installation locations of the ultrasonic flowmeters are on the main water supply pipe and the circulating water return pipe, with pipe diameters such as DN300 and DN200. The ultrasonic flowmeters play a crucial role in the water network system. By accurately measuring the water flow, they provide basic data for the intelligent monitoring and management of the water network. Information including the installation locations and pipe diameters helps users better understand the distribution of the flowmeters, facilitating maintenance and management.

[0029] The multi-parameter water quality sensors include monitoring parameters (pH, conductivity, turbidity) and deployment nodes (such as the outlet of chemical water treatment). The multi-parameter water quality sensors can monitor the water quality in real time, providing important basis for water quality management and optimization. Information including the monitoring parameters and deployment nodes enables people to clearly understand the functions and scope of the sensors, ensuring the comprehensiveness and accuracy of water quality monitoring.

[0030] The pressure transmitters include measuring ranges (0 - 2.5 MPa) and redundant deployment locations (such as the outlet of high-pressure pumps). The pressure transmitters are used to measure the pressure in the water network, which is crucial for ensuring the safe operation of the water network. Information including the measuring ranges and redundant deployment locations can improve the reliability and accuracy of pressure measurement, promptly detect abnormal pressure conditions, and take corresponding measures for handling.

[0031] A hardware device of a large power station's full-process water network intelligent monitoring system includes flowmeters, water quality sensors, and pressure transmitters. The flowmeters are non-intrusive pipe-section ultrasonic flowmeters, with measuring ranges between 0.5 and 15 m / s and an accuracy as high as ±0.5%. This accuracy level can provide extremely accurate data support for the monitoring of water network flow in practical applications. At the same time, this flowmeter supports the Modbus TCP / IP protocol, having high compatibility and stability when communicating data with other devices. In addition, the protection level reaches IP68, enabling it to work reliably under harsh environmental conditions. Whether it is in a humid, dusty environment or under a certain degree of water immersion, it will not affect its normal operation. This high protection level characteristic makes it particularly suitable for pipes with diameters ranging from DN50 to DN2000, covering most of the common pipe size ranges in the power station water network, providing a solid hardware foundation for comprehensively monitoring the water network flow.

[0032] Install relevant monitoring equipment every 200 meters on the main pipeline and at key nodes of the branch pipelines (such as the pump outlet, before and after the valves). Such a deployment method can form a flow monitoring grid for the entire plant's water network. Installing every 200 meters on the main pipeline can ensure relatively intensive monitoring of the main water flow paths in the water network, promptly detecting changes and abnormalities in the flow rate. Installing at key nodes of the branch pipelines, such as the pump outlet and before and after the valves, can focus on these parts where flow fluctuations are likely to occur and faults may happen. Through such a deployment strategy, the flow situation of the entire plant's water network can be comprehensively and meticulously grasped, providing accurate data support for the intelligent monitoring and management of the water network.

[0033] The sampling frequency of the non-intrusive pipe ultrasonic flowmeter is 1Hz, and it converges to the edge gateway through the RS485 bus. The sampling frequency of 1Hz can relatively quickly obtain the flow data in the water network, ensuring the real-time nature of the data. As a commonly used industrial communication bus, the RS485 bus has advantages such as long transmission distance and strong anti-interference ability. Converging the data to the edge gateway through the RS485 bus can achieve centralized processing and analysis of the data, improving the efficiency and accuracy of data processing.

[0034] The water quality sensor adopts a multi-parameter water quality sensor that can measure pH (measurement range 0 - 14, accuracy ±0.1), conductivity (measurement range 0 - 2000 μS / cm, accuracy ±1%), and turbidity (measurement range 0 - 1000 NTU, accuracy ±2%), and the sampling period is 10s. The accurate measurement of the pH value is crucial for evaluating the corrosion and scaling effects of water on equipment, and its accuracy range can relatively precisely reflect the acidity and alkalinity of the water. Conductivity measurement can effectively reflect the ion concentration in water, helping to judge the purity and salt content of the water quality, etc. The conductivity accuracy of ±1% can provide reliable data for water quality analysis. Turbidity measurement is used to measure the turbidity of water, and the turbidity accuracy of ±2% can better reflect the quality and usage effect of the water. In addition, this sensor integrates a temperature compensation function, which can accurately measure water quality parameters under different temperature conditions, avoiding the interference of temperature changes on the measurement results.

[0035] The multi-parameter water quality sensor is deployed at key positions such as the inlet / outlet of the circulating water, the desulfurization wastewater discharge outlet, and the outlet of the chemical water treatment system. Deploying at the inlet and outlet of the circulating water can monitor the water quality changes of the circulating water in real time and promptly detect potential problems such as corrosion and scaling; monitoring at the desulfurization wastewater discharge outlet can ensure that the discharged wastewater meets environmental protection requirements; monitoring at the outlet of the chemical water treatment system can ensure that the water entering the water network meets the usage standards. These deployment positions help to comprehensively monitor the water quality situation of the water network, providing key data support for water quality management and optimization.

[0036] The multi-parameter water quality sensor has a self-calibration mechanism and automatically performs zero calibration every 24 hours. This mechanism can effectively prevent the sensor from drifting due to various factors during long-term use, avoid inaccurate measurement results, thus greatly improving the stability and service life of the sensor, and at the same time reducing the cost and workload of manual maintenance.

[0037] The pressure transmitter has a range of 0 - 2.5 MPa, which can meet the pressure measurement requirements of most parts of the power station water network. Its pressure overload protection reaches 150%, which can effectively protect the transmitter from being damaged when the pressure suddenly rises, improving the reliability and safety of the equipment. It adopts a 4 - 20 mA analog output mode, which is a common industrial signal output form with the advantages of long transmission distance and strong anti-interference ability. The temperature tolerance range is -40 - 85 °C, which can adapt to the temperature changes in different parts of the power station water network and ensure normal operation in a relatively harsh temperature environment.

[0038] Dual-sensor redundant deployment is adopted at key nodes (such as the outlet of the high-pressure feed water pump). When the data difference between the two sensors exceeds the alarm threshold (5%), an alarm is triggered. This design can timely remind the management personnel of possible problems, such as sensor failures and abnormal pressure fluctuations, effectively avoiding monitoring errors caused by a single sensor failure, significantly improving the reliability and accuracy of the water network pressure measurement, and thus ensuring the safe and stable operation of the water network.

[0039] (2) The edge computing layer includes a data cleaning module, a time synchronization module, and an AI preprocessing unit; The data cleaning module includes a sliding window (60 seconds) and an anomaly filtering logic (±5σ rejection). The data cleaning module can remove outliers from the data, improving the quality and reliability of the data. The inclusion of a sliding window and anomaly filtering logic enables people to understand the process and method of data cleaning, ensuring that the cleaned data can accurately reflect the actual operation of the water network.

[0040] Anomaly filtering calculates the mean and standard deviation through a 60-second sliding window, rejects data points outside ±5σ, and fills them in by linear interpolation. This can effectively identify outliers in the data, remove obvious abnormal data, and improve the accuracy and reliability of the data. At the same time, filling in the rejected data points ensures the continuity and integrity of the data, avoiding the impact of data loss on subsequent analysis and processing. After cleaning, the outlier rate of the data is <0.3%.

[0041] The spatio-temporal feature extraction module expands the pressure gradient and flow ratio into a 128-dimensional vector feature matrix and inputs the 128-dimensional vector feature matrix into the dynamic water balance engine in the platform layer; the spatio-temporal feature extraction module uses a 32-bit cyclic redundancy check algorithm to self-check the data cleaning module.

[0042] Data compression uses a lossy compression algorithm (retaining key trend features), reducing the data volume by 70%. During data transmission and storage, costs can be reduced. Retaining key trend features ensures that the compressed data still reflects the main change trends, providing valuable information for subsequent analysis and processing.

[0043] (3) The platform layer includes a dynamic water balance engine, a spatio-temporal fusion leakage model, and a multi-objective optimizer; The dynamic water balance engine includes a three-level verification hierarchy (plant → unit → cell) and a dynamic compensation formula. The dynamic water balance engine is one of the core technologies of the present invention. Through three-level verification and dynamic compensation, high-precision water balance calculations are achieved, which helps users understand the process and method of water balance calculations and improve the scientificity and rationality of water network management. The three-level warning signals are graded as red, yellow, and green. The technical indicators of the dynamic water balance engine include calculation delay, data integrity rate, and update cycle. The calculation delay < 50 ms / node, the data integrity rate > 99.7%, and the update cycle < 30 min. The three-level verification hierarchy samples and adjusts the spatio-temporal feature extraction module, and the sampling frequency is adjustable from 1 to 10 Hz; the dynamic compensation formula is:

[0044] Among them, is the evaporation rate, α is the dynamic compensation factor, is the temperature difference between the inlet and outlet, and 1.2 is the experimental fitting parameter; The spatio-temporal fusion leakage model includes an ARIMA-LSTM dual-stream network structure and a fusion decision logic (R_t (thermal resistance) > 3σ (three times the standard deviation) and ΔP / ΔL (pressure gradient) mutation > 20%), a time synchronization module, and an AI preprocessing unit. The spatio-temporal fusion leakage model combines time series prediction and spatial topology feature analysis to achieve precise positioning and rapid response of pipeline network leakage, enabling people to understand the technical principles and implementation methods of leakage detection and improving the accuracy and reliability of leakage detection. The configuration of the spatio-temporal fusion leakage model includes the ARIMA (traditional statistical model) differencing order and the LSTM (deep learning model) sliding window. The differencing order d = 1, and the LSTM sliding window is 10 minutes. The ARIMA branch processes linear trends, seasonality, and short-term dependencies. Time series prediction (ARIMA) is trained through historical data, in hours, to predict the flow trend for the next 24 hours. The differencing order is selected by determining the d value through the ACF / PACF graph (usually 1 or 2), and 1 is adopted in this example. Anomaly detection includes marking as potential leakage when the deviation between the predicted value and the actual value > 3σ. The LSTM branch captures non-linear patterns, long-range dependencies, and complex dynamics. Pipes, valves, and equipment can be abstracted as graph structure nodes (Node) and edges (Edge) for digital modeling of the pipeline network. Pressure, flow rate, and equipment type are used as node features to construct a feature matrix, and pipe diameter, length, and roughness are used as edge features. An example of the work order dispatching logic is that work orders are dispatched according to the leakage volume priority > 10 m³ / h > 5 m³ / h > 2 m³ / h, and the response time of the work order: ≤ 2 hours, which can be based on the path planning of the leakage point from the maintenance station; The time synchronization module includes the NTP protocol and the timestamp alignment process. The time synchronization module ensures the timestamp consistency of multi-source data, providing a reliable basis for data analysis and comparison. Including the NTP protocol and the timestamp alignment process helps users understand the implementation method of time synchronization and improves the accuracy and efficiency of data processing. The NTPv4 protocol achieves millisecond-level synchronization (±1 ms), supports GPS / Beidou dual-mode time service, and ensures the timestamp consistency of multi-source data. The NTPv4 protocol is a commonly used network time synchronization protocol that can achieve millisecond-level time synchronization accuracy. Supporting GPS / Beidou dual-mode time service improves the reliability and stability of time synchronization, ensuring accurate acquisition of time information under different environmental conditions. The consistency of multi-source data timestamps is very important for the intelligent monitoring and analysis of the water network. Only in this way can accurate data analysis and comparison be carried out, providing a reliable basis for the management and decision-making of the water network.

[0045] The AI preprocessing unit includes a sudden drop in flow rate initial screening threshold (a drop of ≥ 20% in 10 minutes). The AI preprocessing unit quickly discovers possible leakage situations through the preliminary screening of flow rate data. Including the sudden drop in flow rate initial screening threshold enables people to understand the triggering conditions of leakage detection and improves the timeliness and accuracy of leakage detection; After the AI preprocessing unit of the spatio-temporal fusion leakage model performs a preliminary screening of sudden flow drops on the deviation heat map, it locates the leakage coordinates of the deviation heat map through the ARIMA-LSTM dual-stream network structure and the fusion decision logic, and inputs the deviation heat map and the leakage coordinates into the multi-objective optimizer, with the confidence level of the leakage coordinates ≥ 95%.

[0046] The multi-objective optimizer includes the NSGA-III algorithm process and the output of the Pareto solution set. By comprehensively considering multiple factors such as water saving rate, water quality stability, and cost, the multi-objective optimizer realizes the multi-objective optimization of water network management. Including the algorithm process and the output of the Pareto solution set helps users understand the optimization process and results, providing a scientific and reasonable decision-making basis for water network management. The population size of the NSGA-III algorithm is 200. The multi-objective optimizer feeds back and updates the spatio-temporal fusion leakage model with a learning rate of 0.001. The constraint conditions of the NSGA-III algorithm include the concentration ratio and COD (chemical oxygen demand), with the concentration ratio < 6.0 or COD > 800 mg / L. The constraint conditions can also include the data difference of the pressure transmitter ≤ 5%; the NSGA-III algorithm is an evolutionary algorithm designed for high-dimensional multi-objective optimization problems (≥ 3 objectives). By introducing a reference point mechanism to maintain the diversity of solutions, the Pareto solution set is screened from the final population of the NSGA-III algorithm and the output value is applied to the application layer; the Pareto solution set is a set of optimized parameters, and the water saving rate is 12% - 19%.

[0047] The NSGA-III algorithm process includes: population initialization: randomly generate 100 groups of parameters (such as pump frequency, valve opening); non-dominated sorting: screen the optimal solution set according to the Pareto front; crowding degree calculation: maintain the diversity of the solution set; output scheme: generate 3 - 5 groups of Pareto optimal solutions for manual selection; the closed-loop verification of the dynamic adjustment mechanism is to compare the actual water saving rate with the simulated value after the work order is executed, and update the optimizer weight. Model self-learning retrains the LSTM network every 24 hours and adjusts the spatio-temporal feature extraction weight.

[0048] (4)The application layer includes a three-dimensional digital twin interface and an intelligent work order system The three-dimensional digital twin interface includes a heat map (red / yellow / green warning), a leakage error circle (radius ≤ 50 meters, a visual marker with a radius of 50 meters centered on the predicted coordinates), and a water saving strategy simulation module. The three-dimensional digital twin interface provides users with an intuitive and vivid display of the water network operation status, enabling people to quickly understand the operation situation and potential problems of the water network, and providing visual decision-making support for water network management. The warning response delay of the three-dimensional digital twin interface < 500 ms.

[0049] The intelligent work order management system includes an automatic work order dispatching logic (based on location and team load) and a closed-loop verification process. The intelligent work order management system can automatically dispatch work orders, improve the maintenance efficiency and management level, enable people to understand the working principle and management process of the work order management system, and ensure the timely and accurate completion of maintenance work. The intelligent work order management system conducts automatic management dispatching based on the heat map, leakage error circle, and simulated water-saving strategies generated by the three-dimensional digital twin interface. The automatic work order dispatching logic is as follows: leakage volume > 10 m³ / h → emergency work order (response within 1 hour); 5 m³ / h < leakage volume ≤ 10 m³ / h → regular work order (response within 2 hours). The control instructions are encrypted by SHA-256, and the intelligent work order management system feeds the dynamic compensation factor parameters back to the multi-objective optimizer for self-learning. The feedback of the dynamic compensation factor parameters is △α = ±0.05 / 5℃, and the success rate of automatic work order dispatching is ≥98% and it supports mobile synchronization. The closed-loop verification indicators are as follows: the leakage volume after repair decreases by ≥80%, and the pressure fluctuation of surrounding equipment is ≤5%.

[0050] (5) The data flow includes Solid arrow: The real-time data flow from the perception layer → edge layer → platform layer → application layer. The solid arrow indicates the transmission direction of the real-time data flow. The data collected from the perception layer is processed and analyzed by the edge computing layer, transmitted to the platform layer for comprehensive calculation and optimization, and finally provides visual decision support for users at the application layer.

[0051] Dashed arrow: The feedback of the optimization instruction from the application layer → platform layer (such as adjusting the concentration ratio). The dashed arrow indicates the feedback direction of the optimization instruction. The application layer sends an optimization instruction to the platform layer according to the user's needs and actual situation, and the platform layer makes corresponding adjustments and optimizations according to the instruction to achieve the intelligentization and automation of water network management.

[0052] The technology of the present invention has high generality and adaptability and can be widely applied to the water system management in different fields. In the urban water supply network, with the help of real-time monitoring and precise control, the stability and reliability of water supply can be improved, and water resource waste can be reduced. In the water system of the petrochemical park, it can effectively manage complex process flows, ensuring the coordinated optimization of water quality safety and water-saving benefits. Achieving full-chain intelligent monitoring of "source-network-load" helps to improve the utilization efficiency of water resources, reduce operating costs, and contribute to the sustainable development of cities and industries.

[0053] Promoted in arid regions to support precise water resource scheduling (e.g., the water saving rate of a certain solar thermal power station in a certain region increased by 21%). Water resources are scarce in arid regions, and precise water resource scheduling is crucial. The successful application of the present invention in a solar thermal power station in Xinjiang fully demonstrates its great potential in arid regions. Through intelligent monitoring and optimized control, efficient utilization of water resources can be achieved, the water saving rate can be increased, providing strong support for the economic development and ecological protection in arid regions. Further, by integrating digital twin and VR technologies, water network fault simulation and emergency drills can be realized. The integration of digital twin and VR technologies will bring a brand-new experience to water network management. By constructing a digital twin model, the operating status of the water network can be simulated in real time, and potential fault risks can be predicted. Combining VR technology for water network fault simulation and emergency drills can improve the response ability and decision-making level of staff. This helps to discover problems in advance, formulate effective emergency plans, and ensure the safe and stable operation of the water network. Further, an AI self-learning model can be developed to continuously optimize the dynamic compensation factor and diagnostic threshold. With the continuous accumulation of data and the continuous progress of technology, developing an AI self-learning model has become an inevitable trend. By learning and analyzing a large amount of historical data, the AI self-learning model can continuously optimize the dynamic compensation factor and diagnostic threshold, improving the accuracy of water balance calculation and the precision of leakage detection. This will continuously improve the performance of the present invention and better adapt to different application scenarios and requirements.

[0054] The multi-level dynamic water balance model is a progressive water balance framework (plant → unit → cell) that is corrected in real time based on environmental parameters (temperature, humidity, wind speed), breaking through the limitation of traditional static water balance relying on manual testing. The real-time correction of environmental parameters makes the water balance model more accurately reflect the actual operation of the water network. The progressive water balance framework, from the plant to the unit and then to the cell, is gradually refined, enabling a more precise grasp of the dynamic changes in the water network.

[0055] The spatio-temporal dual-stream leakage diagnosis algorithm combines time series prediction and spatial topology feature analysis to achieve precise positioning (error ≤ 50 meters) and rapid response (≤ 2 hours) of pipeline network leakage. The combination of time series prediction and spatial topology feature analysis provides a more comprehensive perspective for leakage diagnosis. Precise positioning and rapid response can promptly detect and handle leakage problems, reducing water resource waste.

[0056] The multi-objective collaborative optimization engine integrates multi-dimensional indicators such as water saving rate, water quality stability, and economy, and generates dynamic optimization strategies through algorithms to meet the refined water use and environmental protection compliance requirements of the power station. The integration of multi-dimensional indicators enables the optimization engine to comprehensively consider multiple factors such as water saving, water quality, and economy, providing a more scientific and reasonable water use plan for the power station. The application of relevant algorithms ensures the efficiency and reliability of the optimization strategy.

[0057] Dynamic three - level verification flowchart for water balance, including input data, plant - level verification, unit - level verification, cell - level verification, and their optimization schemes and constraint boundaries; participate in Figure 2 , and the specific steps are as follows: (1) Input data: environmental parameters (temperature, humidity, wind speed), equipment operating status (pump / valve opening), real - time flow / quality data. These input data are the basis for the dynamic three - level verification of water balance. By real - time monitoring and analyzing these data, the water balance of the water network can be accurately calculated.

[0058] (2) Plant - level verification: Step 1: Calculate the total water intake

[0059] In plant - level verification, the total water intake is calculated first. By accumulating the water intakes from various water sources, the total water intake of the water network is obtained. This step provides the basic data for subsequent unit - level and cell - level verifications.

[0060] Step 2: Dynamically compensate the total water consumption

[0061] In this process, the evaporation water volume, drainage water volume, and loss water volume are summed up to obtain the dynamically compensated total water consumption. Among them, the evaporation water volume is affected by environmental parameters such as temperature, humidity, and wind speed. Through real - time monitoring and dynamic compensation, this value can be calculated more accurately. The drainage water volume comes from the normal drainage operations of the water network system, such as regular wastewater discharge. The loss water volume covers possible leakage and undetected small - flow losses, etc. Accurately calculating and summarizing these different types of water consumption is crucial for ensuring the accuracy of the dynamic three - level verification of water balance.

[0062] Step 3: Verify the balance condition

[0063] If it exceeds the limit, calibration is triggered. This step strictly verifies the water balance of the water network system. Calculate the difference between the total water intake and the total water consumption, and judge whether the absolute value is less than or equal to 2% of the total water intake. If the condition is met, it means that the water balance of the water network system is within an acceptable range; if it exceeds the range, the calibration mechanism is triggered. The calibration mechanism may include re - checking the accuracy of data acquisition equipment, checking the correctness of environmental parameter monitoring, and adjusting the calculation model, etc. Through strict verification and timely calibration, the accuracy and reliability of the water balance calculation results of the water network system can be ensured, providing strong support for the intelligent management of the water network.

[0064] (3) Unit - level verification: Step 4: Calculate the water balance of the circulating water, chemical water, and desulfurization water subsystems for each unit. In this step, the water network system is divided according to different units, and the water balances of subsystems such as circulating water, chemical water, and desulfurization water are calculated separately. Different subsystems have different water usage characteristics and requirements, so they need to be calculated and analyzed separately. For example, the circulating water subsystem is mainly used for cooling equipment, with a large water consumption and obvious water temperature changes; the chemical water subsystem is used for treating and supplying chemical water, with high requirements for water quality; the desulfurization water subsystem is mainly used for treating sulfur dioxide in flue gas, and it also has specific requirements for water quality and quantity. By calculating the water balance of the subsystems for each unit, the operation of the water network system can be understood more carefully, providing a more accurate basis for subsequent error correction and optimization.

[0065] Step 5: Correct the error by correlating equipment status (such as pump efficiency), with a tolerance ≤ 1.5%. Consider the impact of equipment status on the water balance, and correct the error in the water balance calculation by correlating equipment status parameters such as pump efficiency. The change in pump efficiency will directly affect the water flow and pressure, and thus affect the calculation result of the water balance. By real-time monitoring of equipment status parameters such as pump efficiency and incorporating them into the water balance calculation for correction, the accuracy of the water balance calculation can be improved. At the same time, set a tolerance ≤ 1.5% to ensure that the corrected error is within an acceptable range. If the corrected error still exceeds the tolerance, it is necessary to further check whether the equipment status is normal and whether the data collection is accurate, etc., to ensure the reliability of the water balance calculation.

[0066] (4) Unit-level verification: Step 6: Match the input / output water volumes of key equipment (such as closed coolers), with an error ≤ 1%. Match the input and output water volumes of key equipment such as closed coolers to ensure the water balance of the water network system at the unit level. The closed cooler is one of the important equipment in the water network system, and the matching situation of its input and output water volumes directly reflects the operation status of the equipment and the local water balance of the water network system. By real-time monitoring and matching calculation of the input and output water volumes of key equipment such as closed coolers, and setting a requirement of error ≤ 1%, problems in equipment operation such as leakage and blockage can be detected in a timely manner, so as to take corresponding measures for maintenance and optimization to ensure the stable operation of the water network system.

[0067] Step 7: Output the list of abnormal units (such as cooling towers with makeup water deviation > 5%). Analyze each unit in the water network system and output the list of abnormal units. Abnormal units usually refer to those with large makeup water deviation and the results of water balance calculation exceeding the acceptable range, such as cooling towers with makeup water deviation greater than 5%. By outputting the list of abnormal units, problems in the water network system can be detected in a timely manner, providing specific targets for maintenance and optimization. At the same time, through the analysis of abnormal units, the root causes of problems can be found, such as equipment failures, pipeline leaks, improper operations, etc., and targeted measures can be taken to solve them, improving the overall operation efficiency and reliability of the water network system.

[0068] Feedback mechanism: If the unit-level verification fails, the unit-level data review is automatically triggered; if the unit-level fails, the plant-level system calibration is triggered. This feedback mechanism is an important guarantee for ensuring the accuracy and reliability of the water balance calculation of the water network system. If the unit-level verification fails, it indicates that there are problems at the unit level, which may be caused by abnormal operating states of key equipment, inaccurate data collection, etc. At this time, the unit-level data review is automatically triggered to recheck and analyze the water balance calculation results at the unit level to determine whether the problem lies at the unit level. If the unit-level verification also fails, it means that the problem may be more serious and the plant-level system calibration needs to be triggered to comprehensively check and adjust the water balance calculation model, data collection equipment, environmental parameter monitoring, etc. of the entire water network system to ensure the accuracy and reliability of the water balance calculation results of the water network system.

[0069] (5) Refer to Figure 3 , optimization plan and constraint boundary: Point A: Water-saving priority plan ((W = 158,000) tons, (S = 85), (C = 1,050,000) yuan). This plan focuses on reducing the total water consumption and realizing the efficient utilization of water resources on the premise of ensuring a certain water quality stability and water treatment cost. In specific implementation, the total water consumption can be reduced by optimizing the water use process and strengthening the recycling of water resources. For example, technical transformation can be carried out on high-water-consumption links to improve the reuse rate of water resources; the treatment and reuse of wastewater can be strengthened to reduce the dependence on fresh water resources. At the same time, by reasonably adjusting the water treatment process, the water treatment cost can be reduced on the premise of ensuring water quality stability. In this way, both the water-saving goal can be achieved and the normal operation and water quality requirements of the water system can be ensured.

[0070] Point B: Water Quality Priority Plan (W = 182,000 tons, (S = 95), (C = 1.32 million yuan)). This plan prioritizes water quality stability, moderately increasing the total water consumption and water treatment cost to ensure the high-quality operation of the water system. To achieve the water quality priority goal, more advanced water treatment technologies and equipment can be adopted to improve the effect and stability of water quality treatment. For example, increasing the density of water quality monitoring points to monitor water quality changes in real-time and adjusting the water treatment process in a timely manner; adopting efficient filtration and disinfection equipment to ensure that the water quality meets high-standard requirements. Although this plan will increase the total water consumption and water treatment cost, it can provide strong guarantee for the high-quality operation of the water system, especially for some application scenarios with high water quality requirements, such as nuclear power plants, high-end manufacturing industries, etc., which is of great significance.

[0071] Point C: Cost Priority Plan (W = 175,000 tons, (S = 88), (C = 920,000 yuan)). This plan gives priority to reducing the water treatment cost while taking into account the total water consumption and water quality stability. In the cost priority plan, the water treatment cost can be reduced by optimizing the water treatment process and reducing the equipment operation cost. For example, adopting energy-saving water treatment equipment to reduce energy consumption; reasonably adjusting the dosage of water treatment chemicals to reduce chemical costs. At the same time, by strengthening water use management and reasonably controlling the total water consumption, the water quality stability can be ensured within an acceptable range. This plan is applicable to some application scenarios with high cost control requirements, such as small and medium-sized enterprises, urban water supply, etc., which can reduce the operation cost on the premise of meeting the basic water use needs.

[0072] Grey Area: Infeasible Solution Area (such as concentration ratio < 6.0 or COD > 800 mg / L). This area represents the solution space that does not meet the actual application requirements and defines the limitation conditions for system optimization. The concentration ratio refers to the multiple of the reduction in the volume of wastewater during the wastewater treatment process through technologies such as evaporation, reverse osmosis, and membrane separation; COD is an index measuring the degree of organic pollution in water bodies, indicating the amount of oxygen required to oxidize the reducing substances in 1 L of water (mg / L). The higher the COD, the more organic pollutants there are and the greater the treatment difficulty. The determination of the infeasible solution area is based on the experience and technical requirements in actual applications. For example, when the concentration ratio is too low, it may lead to excessive salt accumulation in the water system, affecting the normal operation of equipment; while when the COD is too high, it indicates that the organic matter content in the water is too high, which may cause water quality deterioration and affect the stability of the water system. Therefore, clarifying the infeasible solution area can help us avoid falling into the ineffective solution space during the system optimization process and improve the optimization efficiency.

[0073] Green dotted line: Equipment safety pressure boundary (±15% of the rated value). This constraint ensures that the water system will not exceed the safe bearing range of the equipment during operation, guaranteeing the stability and safety of the system. The equipment safety pressure boundary is set to protect the equipment in the water system from damage caused by excessive pressure. During the operation of the water system, pressure changes may affect the structure and performance of the equipment. If the pressure exceeds the safe bearing range of the equipment, a series of serious consequences may occur, such as equipment damage and leakage problems, which will undoubtedly have a great impact on the normal operation of the water system. Therefore, setting the equipment safety pressure boundary is particularly important as it can effectively control the pressure changes in the water system and ensure the safe operation of the equipment.

[0074] Dynamic environment compensation breaks through the limitations of traditional static models that rely on manual calibration. Real-time correction of environmental parameters reduces the water balance error by 40%. The spatio-temporal dual-stream fusion combines time series and spatial topology analysis, improving the leakage location accuracy by 60% (from 200 meters to 50 meters). The multi-objective collaborative optimization Pareto solution set covers water conservation, water quality, and economy, supporting power plants to flexibly select strategies and reducing the comprehensive cost by 18%. The closed-loop feedback mechanism conducts step-by-step verification from the unit level to the plant level, shortening the fault response time by 50% (from 4 hours to 2 hours).

[0075] The application scenarios of the present invention not only cover the conventional power plant water network (sub-systems such as circulating water, chemical water, and desulfurization wastewater), but can also be extended to scenarios such as the integrated water system in industrial parks and the urban distributed water supply network. It provides key technical support for building intelligent infrastructure for efficient water resource utilization, contributing to the construction of water-saving power plants and the green energy transformation under the "dual carbon" goal. The wide range of application scenarios gives the present invention great market potential and social value. Whether it is the power plant water network, the integrated water system in industrial parks, or the urban distributed water supply network, they can all benefit from the technical advantages of the present invention and achieve efficient water resource utilization.

[0076] The hardware devices of an intelligent monitoring system for the entire process water network of a large power station include flow meters, water quality sensors, and pressure transmitters. The flow meter uses a non-intrusive pipe-section ultrasonic flow meter, with a measurement range between 0.5 and 15 m / s and an accuracy as high as ±0.5%. This accuracy level can provide extremely accurate data support for the monitoring of water network flow in practical applications. At the same time, this flow meter supports the Modbus TCP / IP protocol and has high compatibility and stability when communicating with other devices. In addition, its protection level reaches IP68, enabling it to work reliably under harsh environmental conditions. Whether it is wet, dusty, or subjected to a certain degree of water immersion, it will not affect its normal operation. This feature of high protection level makes it particularly suitable for pipes with diameters ranging from DN50 to DN2000, covering most of the common pipe size ranges in the power station water network, providing a solid hardware foundation for comprehensively monitoring the water network flow.

[0077] Install relevant monitoring devices at every 200 meters on the main pipeline and at key nodes of the branch pipelines (such as the pump outlet, before and after the valve). Such a deployment method can form a flow monitoring grid for the entire plant's water network. Installing at intervals of 200 meters on the main pipeline can ensure relatively intensive monitoring of the main water flow paths in the water network, promptly detecting changes and abnormalities in the flow rate. Installing at the key nodes of the branch pipelines, such as the pump outlet and before and after the valve, can focus on these parts where flow fluctuations are likely to occur and faults may happen. Through such a deployment strategy, the flow conditions of the entire plant's water network can be comprehensively and meticulously grasped, providing accurate data support for the intelligent monitoring and management of the water network.

[0078] The sampling frequency of the non-intrusive pipe-section ultrasonic flow meter is 1 Hz, and the data is aggregated to the edge gateway through the RS485 bus. The sampling frequency of 1 Hz can relatively quickly obtain the flow data in the water network, ensuring the real-time nature of the data. The RS485 bus, as a commonly used industrial communication bus, has advantages such as long transmission distance and strong anti-interference ability. Aggregating the data to the edge gateway through the RS485 bus can achieve centralized processing and analysis of the data, improving the efficiency and accuracy of data processing.

[0079] The water quality sensor uses a multi-parameter water quality sensor that can measure pH (measurement range 0 - 14, accuracy ±0.1), conductivity (measurement range 0 - 2000 μS / cm, accuracy ±1%), and turbidity (measurement range 0 - 1000 NTU, accuracy ±2%). The accurate measurement of pH value is crucial for evaluating the corrosion and scaling effects of water on equipment, and its accuracy range can more precisely reflect the acidity and alkalinity of water. Conductivity measurement can effectively reflect the ion concentration in water, helping to judge the purity and salt content of water quality, etc. The conductivity accuracy of ±1% can provide reliable data for water quality analysis. Turbidity measurement is used to measure the turbidity of water, and the turbidity accuracy of ±2% can better reflect the quality and usage effect of water. In addition, the sensor integrates a temperature compensation function, which can accurately measure water quality parameters under different temperature conditions and avoid the interference of temperature changes on the measurement results.

[0080] The multi-parameter water quality sensor is deployed at key locations such as the inlet / outlet of circulating water, the discharge port of desulfurization wastewater, and the outlet of the chemical water treatment system. Deployed at the inlet and outlet of circulating water, it can monitor the water quality changes of circulating water in real time and detect potential problems such as corrosion and scaling in a timely manner; the monitoring of the desulfurization wastewater discharge port can ensure that the discharged wastewater meets environmental protection requirements; the monitoring of the outlet of the chemical water treatment system can ensure that the water entering the water network meets the usage standards. These deployment locations help to comprehensively monitor the water quality of the water network and provide key data support for water quality management and optimization.

[0081] The multi-parameter water quality sensor has a self-calibration mechanism and automatically performs zero calibration every 24 hours. This mechanism can effectively prevent the sensor from drifting due to various factors during long-term use, avoid inaccurate measurement results, thus greatly improving the stability and service life of the sensor, and at the same time reducing the cost and workload of manual maintenance.

[0082] The pressure transmitter has a range of 0 - 2.5 MPa, which can meet most pressure measurement requirements in the power station water network. Its overload protection reaches 150%, which can effectively protect the transmitter from being damaged when the pressure suddenly rises, improving the reliability and safety of the equipment. It adopts a 4 - 20 mA analog output method, which is a commonly used industrial signal output form and has the advantages of long transmission distance and strong anti-interference ability. The temperature tolerance range is -40 - 85 °C, which can adapt to the temperature changes in different parts of the power station water network and ensure normal operation in a relatively harsh temperature environment.

[0083] Dual-sensor redundant deployment is adopted at key nodes (such as the outlet of the high-pressure feed water pump). When the data difference between the two sensors exceeds 5%, an alarm is triggered. This design can timely remind the management personnel of possible problems such as sensor failures and abnormal pressure fluctuations, effectively avoiding monitoring errors caused by a single sensor failure, significantly improving the reliability and accuracy of water network pressure measurement, and thus ensuring the safe and stable operation of the water network.

[0084] The present invention constructs a dynamic compensation model for environmental parameters (temperature, humidity, wind speed), reducing the water balance error from a traditional relatively high level to a relatively low level. This requires real-time monitoring and analysis of the environment where the power plant water network is located, collecting data such as temperature, humidity, and wind speed, and fusing this data with the operation data of the water network. By establishing a mathematical model to accurately describe the influence of environmental parameters on factors such as evaporation, dynamic compensation of the water balance is then achieved. For example, using machine learning algorithms to train historical data, constructing a relationship model between temperature, humidity, wind speed and evaporation, and then adjusting the water balance according to real-time environmental parameters to reduce errors.

[0085] Implement online progressive verification of the water balance at the plant, unit, and component levels, meeting the real-time requirements (short data update cycle) under complex working conditions. This requires establishing an efficient data acquisition and processing system to ensure that the water balance data at all levels can be updated and verified in a timely manner. Distributed sensor networks can be used to collect data such as flow rate and pressure in the water network in real time, and the data is transmitted to the central control system through a high-speed communication network for analysis and verification. At the same time, optimize algorithms and calculation models to improve the efficiency and accuracy of verification. For example, using parallel computing technology to distribute the verification tasks to multiple computing nodes for simultaneous execution, shortening the calculation time. Fuse time series prediction (flow trend) and spatial topological features (pipe network pressure gradient), and comprehensively use a variety of data analysis techniques to deeply analyze the flow rate and pressure data in the pipe network. Time series prediction can predict future flow trends by analyzing historical flow rate data and detect abnormal changes; spatial topological feature analysis uses the structural information of the pipe network to determine possible leakage locations. By combining these two methods, the accuracy of leakage location is improved. For example, establishing a deep learning-based model that takes into account both the time series characteristics of the flow rate and the spatial topological structure of the pipe network to achieve precise positioning of the leakage location.

[0086] Reduce the false alarm rate and shorten the response time, improving the accuracy and reliability of the leakage detection algorithm. Methods such as multi-modal data fusion and intelligent algorithm optimization are adopted. For example, combining data from various sensors such as flow rate, pressure, and acoustics for comprehensive analysis to reduce the possibility of false alarms. At the same time, establish a fast alarm and processing mechanism to ensure that once a leakage is detected, measures can be taken in a timely manner. For example, using a real-time monitoring and early warning system that immediately issues an alarm when an abnormal situation is detected and activates an emergency plan to shorten the response time.

[0087] Embodiment 2 An operation method of an intelligent monitoring system for the entire process water network of a large power station follows a closed-loop process of "perception - cleaning - analysis - decision - feedback", combined with a multi-level collaborative optimization and dynamic self-learning mechanism. The following is the specific step-by-step process of its core operation logic: (1)Data acquisition and preprocessing (perception layer → edge computing layer) The present invention is multi-source perception, including an ultrasonic flowmeter, which is non-invasively installed on the main water supply pipe and the circulating water return pipe to collect water flow data in real time. A multi-parameter water quality sensor is deployed at the inlet / outlet of the circulating water, the desulfurization wastewater discharge port, and the outlet of chemical water treatment to monitor pH, conductivity, and turbidity, and has a self-calibration function. A pressure transmitter is redundantly deployed at the outlet of the high-pressure pump to collect pressure time-series data. When the difference between the dual-sensor data exceeds the threshold, an alarm is triggered, and pressure overload protection is started.

[0088] (2)Data cleaning and feature extraction The sliding window cleaner calculates the mean and standard deviation of the data within the window, eliminates outliers (such as instantaneous noise), fills in missing data using linear interpolation, and reduces the data transmission volume through a lossy compression algorithm.

[0089] The spatio-temporal feature extraction module expands the original data such as pressure gradient and flow ratio into a 128-dimensional vector feature matrix, and performs self-check on the data cleaning module through a 32-bit CRC check algorithm to ensure data integrity.

[0090] (2)Dynamic water balance analysis and leakage location (edge computing layer → platform layer) The dynamic water balance engine adopts a three-level verification hierarchy, specifically including the whole plant level to verify the macroscopic balance of the water input and output of the whole plant. The unit level analyzes the rationality of the water volume distribution of a single unit. The unit level focuses on the water volume abnormality of specific equipment (such as heat exchangers). The dynamic compensation formula combines the spatio-temporal feature matrix to generate a deviation heat map (reflecting the regional distribution of water volume / quality deviation), and inputs it into the spatio-temporal fusion leakage model.

[0091] In the spatio-temporal fusion leakage model, the AI preprocessing unit performs a preliminary screening of sudden flow drops on the deviation heat map to quickly locate potential leakage areas. The ARIMA in the ARIMA-LSTM dual-stream network captures the long-term trend of the time series (such as periodic water consumption changes). LSTM extracts spatial features (such as the topological association of the pipe network), and combines the fusion decision logic (thermal resistance threshold + pressure gradient threshold) to accurately locate the leakage coordinates.

[0092] At the same time, the time synchronization unit calibrates the clock of the dynamic water balance engine through the NTP protocol to ensure the alignment of spatio-temporal data.

[0093] (3)Multi-objective optimization and strategy generation (platform layer → application layer) The NSGA-III algorithm of the multi-objective optimizer takes the deviation heat map and leakage coordinates as inputs, combines constraint conditions (such as concentration ratio, COD threshold), and generates a Pareto solution set (optimized parameter combinations, such as valve opening, pump speed adjustment). Feedback self-learning dynamically adjusts the population size and constraint condition weights of the optimizer according to the closed-loop verification results of the intelligent work order system to improve the optimization efficiency.

[0094] The heat map rendering of the three-dimensional digital twin interface real-time displays the water volume / water quality deviation area and overlays the leakage error circle (positioning accuracy range). The water-saving strategy simulation is based on the Pareto solution set to simulate the effects of different optimization schemes (such as reducing circulating water loss, reducing the usage amount of chemical agents). The control instruction generation converts the optimal strategy into specific control instructions (such as adjusting the pump frequency, opening the standby pipeline) and inputs them into the intelligent work order system.

[0095] (4) Work order management and closed-loop feedback (application layer → platform layer / edge computing layer) The automatic work order assignment logic of the intelligent work order system automatically assigns maintenance tasks to the operation and maintenance team according to the leakage coordinates and priorities (such as leakage volume, influence range). The closed-loop verification process compares the actual data after maintenance with the simulated predicted values, generates a dynamic compensation factor (such as leakage repair rate), and feeds it back to the multi-objective optimizer for model update.

[0096] System-level feedback includes: pressure transmitter → data cleaning module and dynamic water balance engine → spatio-temporal feature extraction module. For the pressure transmitter → data cleaning module, after the redundant sensor data difference triggers an alarm, the cleaning module adjusts the sliding window parameters to enhance the filtering ability for transient pressure. For the dynamic water balance engine → spatio-temporal feature extraction module, if the heat map deviation continuously exceeds the threshold, it triggers the reconstruction of the feature matrix and optimizes the spatio-temporal feature extraction algorithm.

[0097] The end-to-end latency of the data of the present invention from the perception layer to the application layer is ≤1 second (meeting the dynamic adjustment requirements). The leakage location error is ≤0.5% of the pipeline network length, and the water-saving efficiency of the optimization strategy is increased by 15%-20%. Through redundant sensors, self-calibration mechanisms, and self-learning feedback, it can adapt to the complex working conditions of power plants (such as load fluctuations and water quality mutations). The present invention realizes the real-time perception, dynamic balance analysis, and intelligent decision-making of the water network state through the "end-edge-cloud-application" collaborative architecture, forming a closed-loop management of "monitoring-diagnosis-optimization-verification". Combining spatio-temporal AI models with multi-objective optimization reduces unnecessary water losses. Through three-dimensional visual simulation, the operation and maintenance efficiency is improved. Continuously optimizing the model based on closed-loop feedback can adapt to system drift during long-term operation. The present invention brings a new revolution to the management of the power plant water network, effectively improving the operation efficiency and reliability of the water network. In today's energy field, the integration of smart energy and industrial Internet of Things has become an inevitable trend of development. The present invention conforms to this trend, organically combines a variety of advanced technologies, and provides strong support for the intelligent management of the power plant water network. The full-process water network intelligent monitoring system and method for large-scale power plants cover various types of power plants such as thermal power plants, nuclear power plants, hydropower plants, and solar thermal power plants, and have wide applicability.

[0098] Upon reading the above description, many embodiments and many applications other than the provided examples will be obvious to those skilled in the art. Therefore, the scope of this teaching should not be determined with reference to the above description, but should be determined with reference to the full scope of the foregoing claims and the equivalents of those claims. For the sake of completeness, all articles and references, including patent applications and published publications, are incorporated herein by reference. The omission of any aspect of the subject matter disclosed herein in the foregoing claims is not intended to abandon such subject matter, nor should the applicant be regarded as not considering such subject matter as part of the disclosed inventive subject matter.

[0099] The above content is a further detailed description of the present invention. It cannot be determined that the specific implementation manner of the present invention is limited to this. For those of ordinary skill in the technical field to which the present invention belongs, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should all be regarded as belonging to the scope determined by the present invention as submitted.

Claims

1. A large-scale power station full-process water network intelligent monitoring system, characterized in that: It includes a perception layer, an edge computing layer, a platform layer and an application layer. The edge computing layer is connected to the perception layer and the platform layer respectively. The platform layer is connected to the application layer. The perception layer includes an ultrasonic flow meter, a multi-parameter water quality sensor and a pressure transmitter. The edge computing layer includes a data cleaning module and a spatiotemporal feature extraction module. The platform layer includes a dynamic water balance engine, a spatiotemporal fusion leakage model and a multi-objective optimizer. The application layer includes a three-dimensional digital twin interface and an intelligent work order system. The data output end of the ultrasonic flowmeter, the data output end of the multi-parameter water quality sensor and the data output end of the pressure transmitter are respectively connected to the data input end of the data cleaning module, the output end of the data cleaning module is connected to the input end of the spatiotemporal feature extraction module, the output end of the spatiotemporal feature extraction module is connected to the input end of the dynamic water balance engine, the output end of the dynamic water balance engine is connected to the input end of the spatiotemporal fusion leakage model, the output end of the spatiotemporal fusion leakage model is connected to the input end of the multi-objective optimizer, the output end of the multi-objective optimizer is connected to the input end of the three-dimensional digital twin interface, and the output end of the three-dimensional digital twin interface is connected to the input end of the intelligent work order system; The feedback end of the spatiotemporal feature extraction module is connected to the receiving end of the data cleaning module, the feedback end of the dynamic water balance engine is connected to the receiving end of the spatiotemporal feature extraction module, the feedback end of the spatiotemporal fusion leakage model is connected to the receiving end of the dynamic water balance engine, the feedback end of the multi-objective optimizer is connected to the receiving end of the spatiotemporal fusion leakage model, and the feedback end of the intelligent work order system is connected to the receiving end of the multi-objective optimizer.

2. According to claim 1, a large-scale power station full-process water network intelligent monitoring system is characterized in that: The installation position of the ultrasonic flowmeter includes the main water supply pipe and the circulating water return pipe. The collected data of the ultrasonic flowmeter includes the water flow rate. The ultrasonic flowmeter adopts a non-intrusive pipe section ultrasonic flowmeter; The deployment nodes of the multi-parameter water quality sensor include a circulating water inlet, a circulating water outlet, a desulfurization wastewater discharge outlet, and a chemical water treatment outlet. The collected data of the multi-parameter water quality sensor include water quality. The monitoring parameters of the water quality include pH, conductivity, and turbidity. The multi-parameter water quality sensor is provided with a self-calibration mechanism. The deployment position of the pressure transmitter includes the high-pressure pump outlet, the collected data of the pressure transmitter includes pressure timing data, the pressure transmitter is redundantly deployed, and an alarm is triggered when the data difference between the two sensors of the redundantly deployed pressure transmitter exceeds the alarm threshold, and the pressure transmitter is set with pressure overload protection.

3. According to claim 1, a large-scale power station full-process water network intelligent monitoring system is characterized in that: The data cleaning module includes a sliding window cleaner, which includes anomaly filtering logic and data compression. The anomaly filtering calculates the mean and standard deviation through a sliding window, removes data points that exceed a threshold, and obtains cleaned data through linear interpolation. The data compression adopts a lossy compression algorithm; The spatiotemporal feature extraction module expands the pressure gradient and flow ratio into a 128-dimensional vector feature matrix, and inputs the 128-dimensional vector feature matrix into the platform layer; the spatiotemporal feature extraction module uses a 32-bit cyclic redundancy check algorithm to perform self-check on the data cleaning module.

4. According to claim 1, a large-scale power station full-process water network intelligent monitoring system is characterized in that: The dynamic water balance engine includes three levels of verification and dynamic compensation formulas, the three levels of verification include plant-level verification, unit-level verification and unit-level verification; the three-level verification sampling and adjustment spatiotemporal feature extraction module, the three-level verification receiving the feature matrix extracted by the spatiotemporal feature extraction module, converting the feature matrix into a deviation heat map in combination with the dynamic compensation formula, and then inputting the spatiotemporal fusion leakage model; the technical indicators of the dynamic water balance engine include calculation delay, data integrity rate and update cycle; The spatiotemporal fusion leakage model includes an ARIMA-LSTM dual-stream network structure and fusion decision logic, a time synchronization unit and an AI preprocessing unit. The fusion decision logic includes a thermal resistance threshold and a pressure gradient threshold. The spatiotemporal fusion leakage model uses an AI preprocessing unit to perform a preliminary screening of a sudden drop in flow on the deviation thermal map, and then uses the ARIMA-LSTM dual-stream network structure and fusion decision logic to locate the leakage coordinates of the deviation thermal map, and inputs the deviation thermal map and the leakage coordinates into a multi-objective optimizer; the spatiotemporal fusion leakage model performs clock calibration feedback on the dynamic water balance engine through a time synchronization unit, and the time synchronization unit includes an NTP protocol and a timestamp alignment process; The configuration of the spatiotemporal fusion leakage model includes ARIMA difference order and LSTM sliding window; The multi-objective optimizer includes an NSGA-III algorithm process and a Pareto solution set output. After receiving the deviation heat map and the leakage coordinates, the multi-objective optimizer uses the NSGA-III algorithm in combination with the constraints to obtain the Pareto solution set, and then outputs the deviation heat map and the Pareto solution set to the application layer. The Pareto solution set is an optimized parameter set. The multi-objective optimizer feeds back to update the spatiotemporal fusion leakage model. The configuration of the multi-objective optimizer includes the population size of the NSGA-III algorithm, and the constraints of the NSGA-III algorithm include the concentration ratio and COD.

5. According to claim 4, a large-scale power station full-process water network intelligent monitoring system is characterized in that: The dynamic compensation formula is: in, is the evaporation amount, α is the dynamic compensation factor, is the inlet and outlet temperature difference, and 1.2 is the experimental fitting parameter.

6. According to claim 1, a large-scale power station full-process water network intelligent monitoring system is characterized in that: The three-dimensional digital twin interface includes a thermal map, a leakage error circle, and a water-saving strategy simulation module; the three-dimensional digital twin interface receives the deviation thermal map and the optimized parameter set, performs thermal map rendering and water-saving strategy simulation, generates a thermal map, a leakage error circle, and a simulated water-saving strategy, and inputs control instructions into the intelligent work order system; The intelligent work order management system includes automatic dispatching logic and a closed-loop verification process. The intelligent work order management system automatically manages dispatching through automatic dispatching logic and a closed-loop verification process based on the heat map, leakage error circle and simulated water-saving strategy generated by the three-dimensional digital twin interface; the intelligent work order management system feeds back the dynamic compensation factor to the multi-objective optimizer for self-learning.

7. According to claim 1, a large-scale power station full-process water network intelligent monitoring system is characterized in that: The real-time flow of data is in the order of perception layer, edge layer, platform layer and application layer; the feedback direction of optimization instructions is in the order of application layer, platform layer and edge computing layer.

8. A method for operating a large-scale power station full-process water network intelligent monitoring system, using a large-scale power station full-process water network intelligent monitoring system as claimed in any one of claims 1 to 7, characterized in that: The following steps are involved: The ultrasonic flowmeter, multi-parameter water quality sensor and pressure transmitter in the perception layer collect water network data and transmit it to the edge computing layer; After the edge computing layer removes the outliers of the water network data through the data cleaning module, the spatiotemporal features of the cleaned water network data are extracted by the spatiotemporal feature extraction module and transmitted to the platform layer; The platform layer generates a deviation heat map through the dynamic water balance engine according to the spatiotemporal characteristics of the cleaned water network data, inputs the deviation heat map into the spatiotemporal fusion leakage model to locate the leakage coordinates, and uses the deviation heat map and leakage coordinates as the input of the multi-objective optimizer to obtain the optimized water network; the dynamic water balance engine adopts a three-level verification hierarchy; The water network input application layer uses a three-dimensional digital twin interface to convert the optimal strategy into control instructions and input them into the intelligent work order management system. The intelligent work order management system automatically dispatches repair orders based on the leakage coordinates and priorities. The intelligent work order management system generates a dynamic compensation factor feedback optimization multi-objective optimizer, and the three-dimensional digital twin interface simulates and displays the optimized water network in real time.

9. The method for operating a large-scale power station full-process water network intelligent monitoring system according to claim 8, characterized in that: The three-level verification level includes plant-level verification, unit-level verification and unit-level verification; The plant-wide verification calculates the total water intake and the dynamic total water consumption, combines the total water intake and the dynamic total water consumption to obtain a balance verification threshold, and determines whether to trigger a plant-wide calibration based on the balance verification threshold; The unit-level verification is performed based on the sub-unit horizontal balance correction of the associated equipment, and based on the corrected error, it is determined whether the equipment is self-checked or manually inspected; The unit-level verification matches the water volume of the unit equipment, determines whether to perform maintenance based on the error limit, and outputs a list of abnormal units.

10. The method for operating a large-scale power station full-process water network intelligent monitoring system according to claim 9, characterized in that: The feedback mechanism of the three-level verification hierarchy is that verification failure at the unit level triggers feedback to the unit level, and verification review failure at the unit level triggers feedback to the plant-wide calibration to complete the closed-loop correction.

Images