A method and system for energy-saving scheduling and operation and maintenance of pump units in large-scale water supply systems based on multi-objective optimization.

Abstract

This invention discloses an energy-saving scheduling and operation and maintenance method for pump sets in a large-scale water supply system based on multi-objective optimization, comprising the following steps: S1: operational data perception and water demand prediction; S2: adaptive correction of pump characteristic curves; S3: multi-objective pump set optimization calculation; S4: digital twin monitoring and operation and maintenance. This invention utilizes a convolutional neural network-long short-term memory deep learning model to accurately predict the water demand of the pipeline network, and combines real-time data to adaptively correct the pump characteristic curves; it introduces the BFGS optimization algorithm and approximate dynamic programming to achieve dual optimization solutions for energy consumption and equipment start-up and shutdown frequency while ensuring water supply pressure; simultaneously, the system integrates digital twin technology to achieve three-dimensional visualization monitoring and equipment health prediction for the entire plant; this invention can further reduce the power consumption of pump station operation, significantly extend equipment service life, and achieve unmanned and safe operation of water plants under complex terrain conditions.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of water conservancy engineering and automation control, specifically relating to a method and system for energy-saving scheduling and operation and maintenance of pump sets in large-scale water supply systems based on multi-objective optimization. Background Technology

[0002] In large-scale water supply projects, pumping stations are major energy consumers; existing technologies mainly have the following problems: Single control strategy: Traditional PLC control mostly adopts constant pressure variable frequency logic, which only focuses on achieving the pressure target and ignores the matching problem of high efficiency zone when multiple pumps are running in parallel. This results in the pump being overpowered or running outside the high efficiency zone, leading to serious waste of electricity. Equipment aging and inaccuracy: As operating time increases, the wear of the pump impeller causes the actual performance curve to deviate from the factory curve, and the original control model gradually becomes ineffective, making it impossible to achieve precise scheduling. Frequent start-stop losses: Lack of prediction of future water demand and adjustment based only on current pressure fluctuations can easily lead to frequent start-stop of the pump set, accelerating the aging of the motor and valves; Insufficient operation and maintenance capabilities: Remote areas lack highly skilled technical personnel, making it difficult to carry out refined management and fault prediction. Summary of the Invention

[0003] (1) Technical problems to be solved To address the shortcomings of existing technologies, the present invention aims to provide a method and system for energy-saving scheduling and operation and maintenance of pump sets in large-scale water supply systems based on multi-objective optimization. This method and system are intended to solve the problems of single control strategies, equipment aging and inaccuracy, frequent start-stop losses, and insufficient operation and maintenance capabilities in existing large-scale water supply projects.

[0004] (2) Technical solution To address the aforementioned technical problems, this invention provides a method for energy-saving scheduling and operation and maintenance of pump sets in large-scale water supply systems based on multi-objective optimization, comprising the following steps: S1: Operational data perception and water demand prediction: Using a convolutional neural network-long short-term memory network deep learning model, predict the water demand and target pressure of the pipeline network within a set time window in the future; S2: Adaptive correction of pump characteristic curve: In response to the performance deviation caused by pump aging, the actual characteristic curve of the pump is dynamically fitted and updated by combining real-time collected operating data to obtain the current real operating condition model of a single pump. S3: Multi-objective pump set optimization calculation: Based on the predicted water demand and the actual characteristic curve of a single pump obtained after correction in step S2, the BFGS quasi-Newton method combined with an approximate dynamic programming algorithm is used. Under the constraints of target flow rate and pressure, the optimal pump set combination and the speed ratio of each variable frequency pump are solved with the dual objective functions of minimizing the total energy consumption of the system and minimizing the number of pump set start-ups and shutdowns. S4: Digital Twin Monitoring and Maintenance: Maps the pump unit's operating status to a digital twin system developed based on a 3D engine, performs real-time equipment health diagnosis and fault warning, and automatically issues control commands based on optimization results; The digital twin system in step S4 includes a visual perception and detection mechanism that links the video monitoring system and AI visual analysis module in the pump room to automatically identify abnormal equipment appearances such as leakage and smoke, as well as personnel intrusion behavior. It also combines SCADA data such as current, temperature, and vibration to achieve multi-dimensional health assessment.

[0005] Preferably, step S1 specifically involves: real-time acquisition of pipeline pressure, flow rate, pump current, voltage, and frequency data; using a convolutional neural network-long short-term memory network (CNN-LSTM) deep learning model, based on historical water supply data, meteorological data, and holiday factors, to predict the pipeline water demand and target pressure within a future set time window, transforming passive regulation into feedforward control.

[0006] Furthermore, the real-time operating data collected in step S2 includes the flow rate, head, and power of a single pump.

[0007] Furthermore, step S2 utilizes an online identification algorithm for fitting, thereby periodically correcting the characteristic curves within the system to ensure that the algorithm operates based on the actual device state.

[0008] Furthermore, in the optimization calculation of step S3, a global cost function J that includes energy consumption cost and equipment loss cost is constructed: ; in, P elec For the real-time operating power of the pump set, C switch This is the loss weighting coefficient for a single start-stop cycle. N switch The number of start-stop actions is determined; a rolling optimization is performed in the time domain using an approximate dynamic programming algorithm to avoid shortening the equipment lifespan due to frequent start-stop operations.

[0009] Furthermore, the energy-saving scheduling and operation and maintenance method for pump units of large-scale water supply systems based on multi-objective optimization supports independent deployment and flexible settings. It supports independent deployment of a single module on a PLC or edge server, and can flexibly switch between energy-saving priority mode and voltage stabilization priority mode according to the peak and valley periods of electricity price.

[0010] This invention also provides an energy-saving scheduling and operation and maintenance system for pump sets in large-scale water supply systems based on multi-objective optimization, comprising: The data layer includes historical and real-time databases, used to store pump manufacturing curves, aging calibration curves, and pipeline hydraulic models. The decision-making layer includes a water demand prediction module, a pump set optimization scheduling algorithm module, and an equipment life assessment module. The interaction layer is a web-based 3D digital twin visualization platform that provides functions such as inspection, reverse query of equipment parameters, and remote emergency takeover.

[0011] Beneficial effects Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes a convolutional neural network-long short-term memory deep learning model to accurately predict water demand in the pipeline network, and adaptively corrects the pump characteristic curves based on real-time data. It introduces the BFGS optimization algorithm and approximate dynamic programming to achieve dual optimization of energy consumption and equipment start-up / shutdown frequency while ensuring water supply pressure. Simultaneously, the system integrates digital twin technology to achieve three-dimensional visualization monitoring and equipment health prediction for the entire plant. This invention can reduce power consumption in pump station operation, significantly extend equipment lifespan, and enable unmanned and safe operation of water plants under complex terrain conditions.

[0012] This invention combines the BFGS algorithm and the approximate dynamic programming algorithm to calculate the optimal pump set combination and speed ratio, so that the pump set always operates in the high-efficiency range, thereby fundamentally reducing the operating power consumption. This invention achieves full-condition optimization through the BFGS algorithm and the approximate dynamic programming algorithm. This invention uses the BFGS algorithm to solve for the optimal combination of speed ratios of multiple water pumps to minimize the total power. This invention also uses an approximate dynamic programming algorithm to effectively suppress unnecessary pump start-ups and shutdowns, avoiding the shortened equipment lifespan caused by frequent start-ups and shutdowns, thereby extending the equipment lifespan. This invention enables remote inspection and unmanned operation through a digital twin system, effectively solving the problem of a shortage of professional maintenance personnel; The method described in this application supports independent deployment of a single module and is highly adaptable. Attached Figure Description

[0013] Figure 1 This is a flowchart of the energy-saving scheduling and operation and maintenance method for pump sets in a large-scale water supply system based on multi-objective optimization, as described in this invention. Detailed Implementation

[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. Various modifications or equivalent substitutions made to the present invention by those skilled in the art without departing from the spirit and substance of the present invention should fall within the scope of protection of the present invention.

[0015] like Figure 1 As shown, this invention provides a method for energy-saving scheduling and operation and maintenance of pump sets in large-scale water supply systems based on multi-objective optimization, including the following steps: S1: Operational data perception and water demand prediction: Using a convolutional neural network-long short-term memory network deep learning model, predict the water demand and target pressure of the pipeline network within a set time window in the future; Real-time data collection of pipeline pressure, flow rate, pump current, voltage, and frequency; using a convolutional neural network-long short-term memory network (CNN-LSTM) deep learning model, based on historical water supply data, meteorological data, and holiday factors, to predict the pipeline water demand and target pressure within a set time window in the future, transforming passive regulation into feedforward control; The Convolutional Neural Network-Long Short-Term Memory Network (CNN-LSTM) deep learning model for water demand prediction specifically involves: combining convolutional neural networks to extract spatial features and long short-term memory networks to extract time-series features; utilizing historical water supply data, meteorological data, and holiday factors to accurately predict the water demand and target pressure in the short term (e.g., the next hour), transforming passive regulation into feedforward control.

[0016] S2: Adaptive correction of pump characteristic curves: In response to performance deviations caused by pump aging, the system combines real-time operating data of flow rate, head, and power of a single pump with an online identification algorithm for fitting, thereby periodically correcting the characteristic curves within the system. The online identification algorithm can dynamically fit and update the actual characteristic curves of the pumps to obtain the current real operating condition model of a single pump. The water pump curve adaptive correction specifically involves the system calculating the actual operating points Q, H, and P of the water pump in real time, and periodically correcting the internal QH and QP characteristic curves using an online identification algorithm to ensure that the scheduling algorithm always operates based on the actual equipment status.

[0017] S3: Multi-objective pump set optimization calculation: Based on the predicted water demand and the actual characteristic curve of a single pump obtained after correction in step S2, the BFGS quasi-Newton method combined with an approximate dynamic programming algorithm is used. Under the constraints of target flow rate and pressure, the optimal pump set combination and the speed ratio of each variable frequency pump are solved with the dual objective functions of minimizing the total energy consumption of the system and minimizing the number of pump set start-ups and shutdowns. In the optimization calculation of step S3, a global cost function J, which includes energy consumption cost and equipment loss cost, is constructed: ; in, P elec For the real-time operating power of the pump set, C switch This is the loss weighting coefficient for a single start-stop cycle. N switch The number of start-stop actions is determined; rolling optimization in the time domain is performed using an approximate dynamic programming algorithm to avoid shortening the equipment lifespan due to frequent start-stop operations. The dual optimization method using the BFGS quasi-Newton method combined with the approximate dynamic programming algorithm is as follows: Spatial domain optimization: At a certain moment, the BFGS algorithm is used to find the optimal combination of speed ratios of multiple water pumps to minimize the total power. The time-domain optimization adopts an approximate dynamic programming algorithm, while considering the continuity of the time dimension and introducing a start-stop penalty factor to avoid frequent start-stop of water pumps for small energy-saving benefits, and to find the scheduling strategy with the lowest overall cost throughout the day.

[0018] S4: Digital Twin Monitoring and Maintenance: Maps the pump unit's operating status to a digital twin system developed based on a 3D engine, performs real-time equipment health diagnosis and fault warning, and automatically issues control commands based on optimization results; The digital twin system includes a visual perception and detection mechanism that links the video surveillance system and AI visual analysis module in the pump room to automatically identify abnormal equipment appearances such as leaks and smoke, as well as personnel intrusion behavior. It also combines SCADA data such as current, temperature, and vibration to achieve multi-dimensional health assessment. Step S4 realizes a remote inspection and unattended automatic monitoring system that integrates infrastructure construction, data integration, diagnosis and analysis, and intelligent operation and maintenance; Among them: the digital twin system developed based on the 3D engine constitutes the infrastructure; the pump group operation status data integration, equipment health diagnosis and fault early warning constitute diagnosis and analysis; and the automatic issuance of control commands based on the optimization results constitutes intelligent operation and maintenance.

[0019] A specific application example of step S4 is as follows: A virtual factory is built based on a 3D engine and mapped 1:1 to the physical water plant, integrating video surveillance, access control, fire protection, and production data; it detects leaks through AI visual analysis and assesses the health of the units by combining vibration and temperature data, thus enabling predictive maintenance.

[0020] The energy-saving scheduling and operation and maintenance method for pump units in large-scale water supply systems based on multi-objective optimization supports independent deployment and flexible settings. It supports independent deployment of a single module on a PLC or edge server, and can flexibly switch between energy-saving priority mode and voltage stabilization priority mode according to peak and valley electricity prices.

[0021] This invention also provides a multi-objective optimization-based energy-saving scheduling and operation and maintenance system for pump units in large-scale water supply systems, used to implement the aforementioned multi-objective optimization-based energy-saving scheduling and operation and maintenance method for pump units in large-scale water supply systems. The multi-objective optimization-based energy-saving scheduling and operation and maintenance system for pump units in large-scale water supply systems includes: The data layer includes historical and real-time databases, used to store pump manufacturing curves, aging calibration curves, and pipeline hydraulic models. The decision-making layer includes a water demand prediction module, a pump set optimization scheduling algorithm module, and an equipment life assessment module. The interaction layer is a web-based 3D digital twin visualization platform that provides first-person roaming inspection, reverse query of equipment parameters, and remote emergency takeover functions.

[0022] A specific application embodiment of the present invention is as follows: The optimized scheduling of pump sets is as follows: A water plant's pumping station is equipped with 4 large centrifugal pumps (3 in use and 1 on standby).

[0023] Prediction Phase: At 6:00 AM, the Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) deep learning model predicted that the morning peak water demand would be between 7:00 and 8:00 AM, with demand increasing from 5000 m³. 3 / h climbed to 8000 m 3 / h.

[0024] Optimization phase: Traditional logic may directly start the standby pump at the power frequency when the pressure drops; the system calculation in this embodiment found that if the standby pump is started directly, it will cause the pipeline pressure to overshoot and the energy consumption to surge. The optimal solution given by the algorithm is to increase the frequency of the running No. 2 pump to 48Hz 5 minutes before the peak arrives, and smoothly start the No. 3 pump, so that all three pumps can keep operating in the high-efficiency range of 42-45Hz.

[0025] Traditional control group (baseline energy consumption): During the morning peak period of 7:00-8:00, when faced with a drop in pipeline pressure, the traditional control logic usually directly starts the standby pump and makes it run at the power frequency, which will lead to pipeline pressure overshoot and a surge in energy consumption.

[0026] Comparison results: The comprehensive power consumption of the two control methods mentioned above during the same morning peak period (7:00-8:00) was compared. Compared with the traditional power frequency start-stop method, the full-condition collaborative frequency conversion scheduling mode of the present invention achieved an energy saving benefit of about 8% during this period, and the pipeline pressure fluctuation was less than 0.02 MPa.

[0027] The digital twin system is as follows: The central control room's large screen displays a 3D panoramic view of the water plant; the system detects abnormal water stains in the corner of pump room No. 2 through visual perception, which may be due to a seal failure. It immediately pops up an alarm at the corresponding location in the 3D model and retrieves the on-site camera footage for confirmation. The system synchronously analyzed the vibration spectrum data of pump No. 2 and found that the bearing characteristic frequency was abnormal. It automatically generated a maintenance work order and suggested that the machine be shut down and the seals replaced during the off-peak hours at night, thus avoiding a sudden shutdown accident.

[0028] The above embodiments are preferred implementations of the present invention. In addition, the present invention can be implemented in other ways. Any obvious substitutions without departing from the concept of the present technical solution are within the protection scope of the present invention.

Claims

1. A method for energy-saving scheduling and operation and maintenance of pump sets in a large-scale water supply system based on multi-objective optimization, characterized in that, Includes the following steps: S1: Operational data perception and water demand prediction: Using a convolutional neural network-long short-term memory network deep learning model, predict the water demand and target pressure of the pipeline network within a set time window in the future; S2: Adaptive correction of pump characteristic curve: In response to the performance deviation caused by pump aging, the actual characteristic curve of the pump is dynamically fitted and updated by combining real-time collected operating data to obtain the current real operating condition model of a single pump. S3: Multi-objective pump set optimization calculation: Based on the predicted water demand and the actual characteristic curve of a single pump obtained after correction in step S2, the BFGS quasi-Newton method combined with an approximate dynamic programming algorithm is used. Under the constraints of target flow rate and pressure, the optimal pump set combination and the speed ratio of each variable frequency pump are solved with the dual objective functions of minimizing the total energy consumption of the system and minimizing the number of pump set start-ups and shutdowns. S4: Digital Twin Monitoring and Maintenance: Maps the pump unit's operating status to a digital twin system developed based on a 3D engine, performs real-time equipment health diagnosis and fault warning, and automatically issues control commands based on optimization results. The digital twin system in step S4 includes a visual perception and detection mechanism that links the video monitoring system and AI visual analysis module in the pump room to automatically identify abnormal equipment appearances such as leakage and smoke, as well as personnel intrusion behavior. It also combines SCADA data such as current, temperature, and vibration to achieve multi-dimensional health assessment.

2. The method for energy-saving scheduling and operation and maintenance of pump sets in large-scale water supply systems based on multi-objective optimization according to claim 1, characterized in that, Specifically, step S1 involves: real-time acquisition of pipeline pressure, flow rate, pump current, voltage, and frequency data; and using a convolutional neural network-long short-term memory network (CNN-LSTM) deep learning model to predict the pipeline water demand and target pressure within a future set time window based on historical water supply data, meteorological data, and holiday factors, thereby transforming passive regulation into feedforward control.

3. The method for energy-saving scheduling and operation and maintenance of pump sets in large-scale water supply systems based on multi-objective optimization according to claim 2, characterized in that, The real-time operating data collected in step S2 includes the flow rate, head, and power of a single pump.

4. The method for energy-saving scheduling and operation and maintenance of pump sets in large-scale water supply systems based on multi-objective optimization according to claim 3, characterized in that, Step S2 uses an online identification algorithm for fitting, thereby periodically correcting the characteristic curves inside the system to ensure that the algorithm operates based on the actual device state.

5. The method for energy-saving scheduling and operation and maintenance of pump sets in large-scale water supply systems based on multi-objective optimization according to claim 1, characterized in that, In the optimization calculation of step S3, a global cost function J, which includes energy consumption cost and equipment loss cost, is constructed: ; in, P elec For the real-time operating power of the pump set, C swich This is the loss weighting coefficient for a single start-stop cycle. N swich The number of start-stop actions is determined; a rolling optimization is performed in the time domain using an approximate dynamic programming algorithm to avoid shortening the equipment lifespan due to frequent start-stop operations.

6. The method for energy-saving scheduling and operation and maintenance of pump sets in large-scale water supply systems based on multi-objective optimization according to claim 1, characterized in that, The energy-saving scheduling and operation and maintenance method for pump units in large-scale water supply systems based on multi-objective optimization supports independent deployment and flexible settings. It supports independent deployment of a single module on a PLC or edge server, and can flexibly switch between energy-saving priority mode and voltage stabilization priority mode according to peak and valley electricity prices.

7. A multi-objective optimization-based energy-saving scheduling and operation and maintenance system for pump sets in a large-scale water supply system, used to implement the multi-objective optimization-based energy-saving scheduling and operation and maintenance method for pump sets in a large-scale water supply system as described in any one of claims 1-6, characterized in that, include: The data layer includes historical and real-time databases, used to store pump manufacturing curves, aging calibration curves, and pipeline hydraulic models. The decision-making layer includes a water demand prediction module, a pump set optimization scheduling algorithm module, and an equipment life assessment module. The interaction layer is a web-based 3D digital twin visualization platform that provides first-person roaming inspection, reverse query of equipment parameters, and remote emergency takeover functions.

Images