Water pump cooperative scheduling and water supply control method based on multi-stage water tank

Abstract

The application discloses a kind of water pump collaborative scheduling and water supply control method based on multistage water tank, belong to water supply system control technical field.The present application aims at solving the technical problem that existing water supply scheduling method relies on after-the-fact alarm, cannot actively identify risk before fault occurs and adjust operation strategy in advance.This method collects data in real time by using pressure sensor, liquid level sensor and vibration sensor in the water supply system containing multistage water pump unit and water tank, and transmits the data to the digital twin platform preset with virtual twin;the digital twin platform drives the virtual twin to run synchronously according to real-time data and simulates deduction, generates predicted vibration data and its change trend;predicted vibration data and change trend are input into fault identification model, when predicted value reaches first-level alarm threshold and predicted extreme value will reach second-level alarm threshold, identify the target water pump unit with fault risk.

Description

Technical Field

[0001] This invention belongs to the field of water supply system control technology. More specifically, this invention relates to a method for coordinated scheduling of water pumps and water supply control based on multi-stage water tanks. Background Technology

[0002] In municipal water supply, industrial circulating water, and agricultural irrigation, water supply systems consisting of multiple pump units connected in series or parallel and multiple water tanks connected in series are widely used. Currently, the coordinated scheduling of pumps in multi-stage water tank supply systems mainly adopts logic control methods based on programmable logic controllers or frequency conversion regulation methods based on proportional-integral-derivative control, and performs pump unit start-up and shutdown operations and speed regulation according to the water tank level or pipeline pressure.

[0003] However, existing methods have the following problems in practical applications: First, traditional control methods rely on a reactive, real-time alarm mechanism, making it difficult to proactively identify risks before a fault occurs. When early faults such as bearing wear or impeller imbalance occur in a pump unit, the vibration characteristics have already changed, but no obvious pressure or liquid level anomalies have yet been triggered. Existing control methods cannot identify such potential risks. Only when the fault develops to the point of causing a drop in outlet pressure will the system issue an alarm and activate the standby unit. By this time, the pipeline pressure has already fluctuated, potentially causing a brief interruption in water supply. Furthermore, in layouts with multiple pumps operating in parallel or series, the vibration of one pump unit can be transmitted to adjacent units through shared foundations, connecting pipes, and the fluid medium. This causes vibration sensors on multiple units to simultaneously detect similar fault characteristic frequencies, making it difficult to accurately locate the fault source and potentially triggering unnecessary switching between multiple units.

[0004] Secondly, digital twin technology faces challenges in maintaining model accuracy when applied to water supply system simulations. As operating time increases, pump units experience wear, pipe wall scaling, and valve characteristics change, causing the actual physical system's properties to gradually deviate from the initially constructed mechanistic model, leading to inaccurate predictions. Simultaneously, vibration monitoring signals are mixed with multi-source noise, including water flow turbulence, cavitation, vibration from adjacent units, and pipe resonance. Extracting effective fault features under strong noise conditions is difficult, and directly using raw vibration data for judgment can easily lead to false alarms. Furthermore, the simulation and derivation of a high-precision operating mechanism model for the entire system involves enormous computational demands. In large water supply systems, a single simulation can take several minutes, while changes in system operating conditions can occur within seconds. The simulation results lag behind physical reality, failing to meet real-time scheduling requirements.

[0005] Third, the switching operation of water pump units itself will cause disturbances to the system, and the existing dispatching scheme is not adaptable enough to extreme conditions. When a fault risk is identified and a switch is executed, if the rate of decrease in the rotational speed of the deactivated pump does not match the rate of increase in the rotational speed of the activated pump, it will cause instantaneous fluctuations in the pipeline pressure, which may cause the terminal water supply pressure to exceed the allowable range. At the same time, the activation and deactivation of high-power water pump units will impact the local power grid, causing a temporary drop in bus voltage and affecting the normal operation of other sensitive equipment in the same power distribution system. The existing dispatching scheme does not include electrical safety in the optimization conditions. Under extreme conditions such as peak water consumption, when multiple units have reached full capacity and the water tank level is close to the lower limit, a dispatching scheme that simultaneously meets all constraints on water supply pressure and water tank level may not exist, and the system will fall into an unsolvable deadlock, affecting water supply security. Summary of the Invention

[0006] One objective of this invention is to provide a method for coordinated scheduling and water supply control of water pumps based on multi-stage water tanks, applicable to municipal water supply, industrial circulating water, and agricultural irrigation, enabling proactive early warning and adaptive reconfiguration scheduling of water pump units to ensure the continuous and stable operation of the water supply system.

[0007] To achieve these objectives and other advantages of the present invention, the present invention provides a method for coordinated scheduling and water supply control of water pumps based on multi-stage water tanks, comprising the following steps: S1. In a water supply system that includes multiple pump units connected in series or parallel and multiple water tanks connected in series, pressure sensors installed at the outlet pipes of each pump unit, level sensors installed in each water tank, and vibration sensors installed on each pump unit are used to collect the outlet pressure data of each pump unit, the level data of each water tank, and the original vibration data of each pump unit in real time at a preset data acquisition frequency. S2. The real-time collected outlet pressure data, liquid level data, and original vibration data are transmitted to a digital twin platform. The digital twin platform is pre-built with a virtual twin containing the physical attribute information, spatial topology relationship, and operating mechanism model of all water pump units, pipes, valves, and water tanks in the water supply system. The operating mechanism model includes a hydraulic mechanism model and a vibration mechanism model. S3. The digital twin platform drives the virtual twin to operate synchronously based on the received outlet pressure data, liquid level data and original vibration data, and simulates the operation status of the water supply system in a future preset time interval based on the vibration mechanism model in the virtual twin, generating predicted vibration data of each water pump unit and the changing trend of each predicted vibration data. S4. Input the predicted vibration data and the trend of change into the fault identification model. When a certain value in the predicted vibration data reaches the first level alarm threshold, and the predicted extreme value in the second preset time interval in the future, deduced from the trend of change, will reach the second level alarm threshold, the target water pump unit with fault risk is identified. S5. Based on the identified target water pump unit, the real-time liquid level data of each water tank, the valve status of each pipe section, and the current status of the valves associated with the target water pump unit and the standby unit, the digital twin platform generates a reconfiguration scheduling scheme that includes the target water pump unit exiting operation sequence and the standby unit entering operation sequence, provided that the water supply pressure demand of the end user is within the range of 0.14 MPa to 0.40 MPa and the liquid level of each water tank does not exceed the preset maximum liquid level and the preset minimum liquid level. S6. The reconfiguration scheduling scheme is sent to the field programmable logic controller (PLC). The PLC executes the operation of shutting down the target water pump unit and its associated valves and starting the standby unit and its associated valves in sequence according to the timing and parameters in the reconfiguration scheduling scheme.

[0008] In one specific embodiment, the water pump coordinated scheduling and water supply control method based on multi-stage water tanks provided by the present invention includes the following steps.

[0009] Step S1 involves data acquisition. In a water supply system comprising multiple pump units connected in series or parallel and multiple water tanks connected in series, pressure sensors can be installed at the outlet pipes of each pump unit, level sensors can be installed in each water tank, and vibration sensors can be installed on each pump unit. Pressure sensors can be diffused silicon pressure transmitters, typically with stainless steel housings, connected to the pipes via threads or flanges. Level sensors can be submersible level gauges, ultrasonic level gauges, or radar level gauges, with the installation location determined based on the water tank structure and medium characteristics. For example, submersible level gauges need to be inserted and fixed from the top of the water tank, while ultrasonic level gauges are installed on the outer top of the water tank. Vibration sensors can be piezoelectric accelerometers, fixed to vibration-sensitive parts of each pump unit, such as bearing seats or pump bodies, using magnetic mounts or studs. These sensors acquire outlet pressure data, water tank level data, and raw vibration data of the pump units in real time at a pre-set data acquisition frequency. The data acquisition frequency can be set according to the dynamic characteristics of the system. For example, it can be set to 1 Hz when the pressure and water level change slowly, and to 10 kHz when the vibration signal changes rapidly.

[0010] Step S2 involves data transmission and the construction of a digital twin platform. Real-time collected outlet pressure data, liquid level data, and raw vibration data are transmitted to the digital twin platform via industrial Ethernet or fieldbus. This platform can be built on an industrial server, running digital twin software, which includes a pre-built virtual twin. The virtual twin contains the physical properties (such as dimensions, materials, and curves), spatial topology (such as connection sequence and relative position), and operating mechanism models of all pump units, pipes, valves, and water tanks in the water supply system. The operating mechanism models include hydraulic mechanism models, such as differential equations based on the conservation of mass and momentum, and vibration mechanism models, such as rotor dynamics models or bearing failure models. These models can dynamically simulate the system's operating state based on the input data.

[0011] Steps S3 and S4 involve simulation and fault identification. The digital twin platform drives the virtual twin to run synchronously based on the received real-time data, that is, it uses real-time data to update the model's boundary conditions and initial state, ensuring that the virtual model is consistent with the physical system. Subsequently, based on the vibration mechanism model in the virtual twin, the operating state of the water supply system within a preset future time interval is simulated, generating predicted vibration data for each pump unit and the trends of these data. The predicted vibration data can be represented as a time series, and the trend can be a slope or a second derivative. The predicted vibration data and trends are input into the fault identification model. This model has two preset alarm thresholds. For example, the first alarm threshold can be set to 1.2 times the upper limit of the normal vibration amplitude, which is 2.4 mm / s; the second alarm threshold can be set to the allowable dangerous vibration amplitude of the equipment, which is 4.5 mm / s. The second preset time interval can be 30 seconds. When a certain value in the predicted vibration data reaches the first-level alarm threshold, and the predicted extreme value is deduced from the trend to reach the second-level alarm threshold within the next 30 seconds, the fault identification model identifies the pump unit as a target pump unit with a fault risk.

[0012] Steps S5 and S6 involve the generation and execution of a reconfiguration scheduling scheme. Based on the identified target pump unit, real-time water level data of each tank, valve status of each pipe section, and the current status of the valves associated with both the target and standby pump units, the digital twin platform generates a reconfiguration scheduling scheme that includes a target pump unit exit sequence and a standby pump unit activation sequence, provided that the end-user's water supply pressure requirement is within the range of 0.14 MPa to 0.40 MPa and the water tank levels do not exceed the preset maximum and minimum levels. For example, the exit sequence may include gradually closing the target unit's outlet valve and reducing its operating frequency; the activation sequence may include opening the standby pump unit's outlet valve, increasing its operating frequency, and ensuring minimal pressure fluctuations during the switching process. The generated scheme is distributed to the field programmable logic controller (PLC) via the industrial network. The PLC can be a Siemens S7-1200 or similar model, installed in the pump station control cabinet, and executes the operations sequentially according to the timing and parameters in the scheme: closing the target pump unit and its associated valves, starting the standby pump unit and its associated valves, and completing the switching.

[0013] Through the above implementation methods, this approach can utilize vibration prediction technology to identify early faults in water pump units, avoiding water supply interruptions caused by sudden shutdowns. Real-time synchronization and simulation using a digital twin platform ensure the accuracy of the predictions. The reconfigured scheduling scheme is generated under pressure and water level constraints, guaranteeing hydraulic stability during the switching process. The precise execution of the field-programmable logic controller (FPGA) ensures the reliable implementation of the scheme. This effectively improves the operational safety and water supply continuity of the multi-stage water tank supply system.

[0014] Preferably, the method further includes the following steps: S701. In the digital twin platform, a model calibration period T is preset; S702. When each model calibration cycle T arrives, acquire the historical outlet pressure data P collected in real time by the pressure sensor and the liquid level sensor during that cycle. h (t) and historical liquid level data L h (t); S703, Transfer the historical outlet pressure data P h (t) and the historical liquid level data L h (t) Input the virtual twin and drive it to perform a simulation of historical operating conditions to obtain historical simulation pressure data P corresponding to the historical outlet pressure data and the historical liquid level data. s (t) and historical simulation liquid level data L s (t); S704. Calculate the pressure deviation sequence ΔP(t) between the historical outlet pressure data and the historical simulated pressure data.h (t)-P s (t), and the sequence of liquid level deviation values ​​ΔL(t)=L between the historical liquid level data and the historical simulated liquid level data. h (t)-L s (t); S705, when the root mean square error (RMSE) of the pressure deviation value sequence ΔP(t) is... P Greater than 0.022 MPa, or the root mean square error (RMSE) of the liquid level deviation sequence ΔL(t) L When the liquid level deviation exceeds the preset threshold range of 0.14 m to 0.16 m, with a maximum value of 0.16 m, model correction is triggered. S706. After triggering model correction, a parameter identification algorithm is used, based on the historical outlet pressure data P. h (t) and the historical liquid level data L h (t) is used as the objective to iteratively optimize at least one parameter to be corrected in the operating mechanism model in the virtual twin until the root mean square error (RMSE) between the simulation data output by the updated operating mechanism model and the historical outlet pressure data and the historical liquid level data is found. P The pressure deviation threshold is less than the minimum value of 0.018 MPa, and the root mean square error (RMSE) between the simulated liquid level data output by the updated operating mechanism model and the historical liquid level data is less than the minimum value of the pressure deviation threshold. L The minimum value of the liquid level deviation threshold range is 0.14 m; S707. Update the parameters of the corrected operating mechanism model to the virtual twin for simulation and deduction in subsequent steps.

[0015] In one specific implementation, to achieve online self-calibration of the virtual twin, the following steps can be performed. First, a model calibration period T can be preset in the digital twin platform, for example, T can be set to 24 hours. At the end of each model calibration period T, the platform can acquire historical outlet pressure data P collected in real time by pressure sensors and level sensors during that period. h (t) and historical liquid level data L h (t). This data can be stored in an industrial real-time database. Subsequently, historical export pressure data P h (t) and historical liquid level data L h (t) Input the virtual twin and drive it to simulate historical operating conditions to obtain the corresponding historical simulation pressure data P. s (t) and historical simulation liquid level data L s(t). Next, calculate the pressure deviation sequence ΔP(t) = P between historical outlet pressure data and historical simulated pressure data. h (t)-P s (t), and the sequence of liquid level deviation values ​​ΔL(t)=L between historical liquid level data and historical simulated liquid level data. h (t)-L s (t). The root mean square error (RMSE) of these deviation sequences can be calculated using numerical calculation software. P and RMSE L When RMSE P Greater than 0.022 MPa, or RMSE L When the value is greater than 0.16 m, model correction is triggered. After correction is triggered, a parameter identification algorithm is used, such as the least squares method or particle swarm optimization algorithm, based on historical outlet pressure data P. h (t) and historical liquid level data L h (t) is used as the objective to iteratively optimize at least one parameter of the operating mechanism model in the virtual twin that needs to be corrected. The parameter to be corrected can be a pipe roughness coefficient or a pump characteristic curve correction coefficient. The iterative process continues until the RMSE between the simulation data output by the updated operating mechanism model and the historical outlet pressure data is found. P Less than 0.018 MPa, and the RMSE between simulated liquid level data and historical liquid level data. L The value is less than 0.14 m. Finally, the parameters of the corrected operating mechanism model are updated in the virtual twin for simulation in subsequent steps. This process ensures that the virtual twin maintains consistency with the physical system over the long term.

[0016] By setting a model calibration cycle and periodically calibrating the virtual twin using historical operating data, the model mismatch problem caused by equipment wear, aging, or replacement can be effectively solved. This ensures that simulation and fault identification are always based on an accurate model, avoiding false alarms or missed alarms caused by model deviations, and improving the long-term stability and reliability of the system.

[0017] Preferably, step S1 further includes the following sub-steps: S101, The original vibration data V of each water pump unit is collected in real time by the vibration sensors installed on each water pump unit. raw (t); S102, regarding the original vibration data V raw (t) Perform wavelet decomposition to obtain wavelet coefficients containing high-frequency and low-frequency components; S103. Using a pre-set threshold rule, the high-frequency components in the wavelet coefficients are subjected to threshold quantization to obtain the corrected high-frequency wavelet coefficients. S104. Perform wavelet reconstruction on the low-frequency component and the corrected high-frequency wavelet coefficients to obtain the original vibration data V. raw (t) Corresponding denoised vibration data V den (t); S105, the denoised vibration data V den (t) serves as the reference data for model calibration during simulation and deduction of the digital twin platform described in step S3.

[0018] Step S1 may also include a refined operation for vibration signal denoising. Specifically, this involves real-time acquisition of the raw vibration data V of each pump unit using vibration sensors installed on each pump unit. raw After (t), the original vibration data V can be processed. raw (t) Wavelet decomposition is performed. The Daubechies wavelet series can be used for wavelet decomposition. The number of decomposition levels can be determined based on the sampling frequency and signal characteristics; for example, five levels can be used to obtain wavelet coefficients for approximate and detail components. Then, a pre-defined threshold rule is used to perform threshold quantization on the wavelet coefficients of the high-frequency components. A soft thresholding method can be used, and the threshold value can be determined based on noise level estimation, for example, based on the standard deviation of the high-frequency coefficients at each level. After obtaining the corrected high-frequency wavelet coefficients, the low-frequency components are reconstructed using wavelet reconstruction with the corrected high-frequency wavelet coefficients to obtain the original vibration data V. raw (t) Corresponding denoised vibration data V den (t). Finally, the denoised vibration data V den (t) serves as the benchmark data for model calibration during subsequent simulation and deduction on the digital twin platform.

[0019] By performing wavelet denoising on the raw vibration data, environmental noise such as water flow turbulence, cavitation, and pipeline resonance can be effectively filtered out, and feature signals that more accurately reflect the health status of the equipment can be extracted. This provides a high-quality data foundation for subsequent model calibration and fault identification, and reduces the risk of false alarms.

[0020] Preferably, step S3 further includes the following step: S301. When generating predicted vibration data for each water pump unit based on the virtual twin, generate the denoised vibration data V collected by the vibration sensor. den (t) Predicted vibration data V at the same time pre (t); S302, Calculate the vibration deviation value ΔV(t) between the denoised vibration data and the predicted vibration data. den (t)-V pre (t); S303. Calculate the root mean square error (RMSE) of the vibration deviation value ΔV(t) for a preset number of N consecutive sampling times. v When RMSE v If the vibration deviation exceeds the preset threshold range of 0.45 mm / s to 0.55 mm / s, with a maximum value of 0.55 mm / s, and occurs consecutively for 3 to 5 times, the mechanism model parameters related to the vibration characteristics in the virtual twin will be corrected. S304. Employ a parameter identification algorithm using the denoised vibration data V den (t) is used as the objective to iteratively optimize the parameters to be corrected in the vibration mechanism model of the virtual twin, so that the predicted vibration data V output by the updated vibration mechanism model is... pre (t) and the denoised vibration data V den The deviation between (t) satisfies RMSE v The minimum value of the vibration deviation threshold range is 0.45 mm / s; S305. Update the parameters of the corrected vibration mechanism model to the virtual twin for simulation and deduction in subsequent steps.

[0021] Step S3 may further include an online calibration step for the vibration mechanism model. Specifically, when generating predicted vibration data for each pump unit based on a virtual twin, denoised vibration data V collected by vibration sensors can be generated. den (t) Predicted vibration data V at the same time pre (t). Then, calculate the vibration deviation ΔV(t) = V between the denoised vibration data and the predicted vibration data. den (t)-V pre (t). The root mean square error (RMSE) can be calculated by selecting the vibration deviation value ΔV(t) at a preset number of consecutive sampling times N. v For example, N can be 10. When RMSE v When the vibration speed is greater than 0.55 mm / s and occurs consecutively for 3 to 5 times, the mechanism model parameters related to the vibration characteristics in the virtual twin are triggered for correction. The correction also employs a parameter identification algorithm, using denoised vibration data V... den (t) serves as the objective, iteratively optimizing the parameters to be corrected in the vibration mechanism model. These parameters may include the bearing stiffness coefficient. The iteration continues until the predicted vibration data V output by the updated vibration mechanism model is obtained. pre (t) and denoised vibration data V den RMSE between (t) v The velocity is less than 0.45 mm / s. Finally, the parameters of the corrected vibration mechanism model are updated in the virtual twin for subsequent simulations.

[0022] By comparing the deviation between the denoised vibration data and the predicted vibration data in real time, and triggering online correction of the vibration mechanism model when the deviation continues to exceed the standard, it can be ensured that the predicted vibration data always keeps up with the changes in the actual vibration characteristics of the equipment, thereby improving the accuracy and timeliness of fault prediction and avoiding the failure of early warning due to model aging.

[0023] Preferably, the step of distributing the reconfiguration scheduling scheme to the field-programmable logic controller and executing the operation specifically includes the following sub-steps: S601. After the field programmable logic controller receives the reconfiguration scheduling scheme, it acquires the first outlet pressure data P1(t) collected in real time by the first pressure sensor set at the outlet pipe of the target water pump unit, and the second outlet pressure data P2(t) collected in real time by the second pressure sensor set at the outlet pipe of the standby unit. S602, Based on the first outlet pressure data P1(t) and the preset target switching pressure value P set The first deviation between them is e1(t) = P set -P1(t), using a proportional-integral-derivative control algorithm, calculates the rate of speed decrease r of the target pump unit during the shutdown process. down (t); the target switching pressure value P set The value is determined by the digital twin platform to be a constant value in the range of 0.14 MPa to 0.40 MPa after simulation optimization based on the steady-state operating data of the target pump unit before it is decommissioned and the expected operating conditions after the standby unit is put into operation. S603, Based on the second outlet pressure data P2(t) and the target switching pressure value P set The second deviation between them is e2(t)=P set -P2(t), using a proportional-integral-derivative control algorithm, calculates the rate of increase in speed r of the standby unit during the commissioning process. up (t); S604. Control the target water pump unit to decrease at the speed reduction rate r. down (t) linearly reduces its operating frequency f1(t), while controlling the standby unit to increase its speed according to the speed increase rate r. up (t) linearly increases its operating frequency f2(t), and when the operating frequency f1(t) of the target water pump unit decreases to the preset no-load frequency range of 20 Hz to 30 Hz, the operation of shutting down the target water pump unit and its associated valves is performed; S605. During the entire process of shutting down the target water pump unit and its associated valves and starting the standby unit and its associated valves, the difference ΔP(t) = |P1(t)-P2(t)| between the first outlet pressure data P1(t) and the second outlet pressure data P2(t) is monitored in real time. The opening rate of the electric valves associated with the target water pump unit and the standby unit is adjusted so that ΔP(t) is always less than the minimum value of 0.045 MPa in the preset pressure difference threshold range of 0.045 MPa to 0.055 MPa.

[0024] In practical implementation, when the field programmable logic controller (PLC) receives the reconfiguration scheduling scheme, it first acquires the first outlet pressure data P1(t) collected in real time by the first pressure sensor located at the outlet pipe of the target pump unit, and the second outlet pressure data P2(t) collected in real time by the second pressure sensor located at the outlet pipe of the standby unit. Then, based on the first outlet pressure data P1(t) and the preset target switching pressure value P... set The first deviation between them is e1(t) = P set -P1(t), the rate of speed decrease r of the target pump unit during the shutdown process is calculated using the proportional-integral-derivative control algorithm. down (t). The parameters of the proportional-integral-derivative (PID) controller can be tuned based on the system response characteristics. Target switching pressure value P set The pressure is determined by the digital twin platform based on steady-state operating data before the target pump unit is decommissioned and simulated under expected operating conditions after the standby unit is put into operation. The pressure is a constant value within the range of 0.14 MPa to 0.40 MPa, for example, 0.28 MPa. Simultaneously, based on the second outlet pressure data P2(t) and the target switching pressure value P... set The second deviation between them is e2(t)=P set -P2(t), using the same proportional-integral-derivative control algorithm, calculates the rate of increase in speed r of the standby unit during the start-up process. up (t). Subsequently, the target water pump unit is controlled according to the speed decrease rate r. down (t) linearly reduces its operating frequency f1(t), while controlling the standby unit according to the speed increase rate r up(t) Linearly increase its operating frequency f2(t). When the operating frequency f1(t) of the target pump unit decreases to the preset no-load frequency range of 20 Hz to 30 Hz (e.g., 25 Hz), the operation of shutting down the target pump unit and its associated valves is performed. Throughout the switching process, the difference ΔP(t) = |P1(t) - P2(t)| between the first outlet pressure data P1(t) and the second outlet pressure data P2(t) is monitored in real time, and the opening rate of the electric valves associated with the target pump unit and the standby unit is adjusted to ensure that ΔP(t) is always less than 0.045 MPa, thereby ensuring stable pipeline pressure during the switching process.

[0025] By precisely coordinating the speed changes of the decommissioned and commissioned units through proportional-integral-derivative control, and dynamically adjusting the valve opening by monitoring the pressure difference in real time, seamless switching between the target unit and the standby unit is achieved. This effectively suppresses pressure fluctuations during the switching process, ensures the stability of the water supply pressure for end users, and improves user experience and system reliability.

[0026] Preferably, the step of simulating and extrapolating the operating status of the water supply system within a preset future time interval based on a virtual twin specifically includes the following sub-steps: S311. Before performing a simulation of the water supply system's operation within a future preset time interval based on the virtual twin, record the actual computation time T of this simulation. calc ; S312, the actual calculation time T calc The maximum value T within the preset maximum allowable time threshold range of 0.5 seconds to 2 seconds. max =2 seconds for comparison; S313, when T calc Greater than T max When the model order reduction mode is triggered, the intrinsic orthogonal decomposition method is used to reduce the order of the full-order operating mechanism model in the virtual twin, generating a reduced-order operating mechanism model with a lower order than the full-order operating mechanism model. The reduced-order operating mechanism model is generated based on the latest corrected full-order operating mechanism model. S315. Using the reduced-order operation mechanism model to replace the full-order operation mechanism model, simulate and extrapolate the operation state of the water supply system within a future preset time interval, and generate predicted vibration data V for each water pump unit. pre,red (t) and the changing trends of each predicted vibration data; S316. After completing this simulation, record the predicted vibration data V output by the reduced-order operation mechanism model. pre,red(t) and the actual vibration data V collected at the same time by the vibration sensors installed on each water pump unit. act The absolute value of the error between (t) |V pre,red (t)-V act (t)|; S317. When the absolute value of the error is greater than the maximum value of 0.55 mm / s within the preset error allowable range of 0.45 mm / s to 0.55 mm / s for three consecutive times, exit the model reduction mode, resume the simulation and deduction using the full-order operation mechanism model, and when the next model correction cycle T arrives, use historical data to correct the parameters of the full-order operation mechanism model, and regenerate the reduced-order operation mechanism model based on the corrected full-order operation mechanism model.

[0027] In practice, before performing a simulation based on a virtual twin for a future preset time interval, the actual computation time T of the simulation can be recorded first. calc T calc The maximum value T of the preset maximum allowable time threshold range max =2 seconds for comparison. When T calc When the time exceeds 2 seconds, the model order reduction mode is triggered. In this mode, the intrinsic orthogonal decomposition method can be used to reduce the order of the full-order operating mechanism model in the virtual twin. The intrinsic orthogonal decomposition method first collects snapshots of the transient response data of the full-order model under typical operating conditions, then extracts the main modes through singular value decomposition, generating a reduced-order operating mechanism model with an order much lower than the full-order model. This reduced-order model is generated based on the latest corrected full-order model. Then, the reduced-order model is used to replace the full-order model for simulation, generating predicted vibration data V for each pump unit. pre,red (t) and its changing trend. After completing this simulation, record the predicted vibration data V output by the reduced-order model. pre,red (t) and the actual vibration data V collected by the vibration sensor at the same time. act The absolute value of the error between (t) is taken. When this absolute value of error is greater than 0.55 mm / s three times consecutively, the model reduction mode is exited, and the full-order operating mechanism model is used for simulation. At the same time, when the next model calibration period T arrives, the parameters of the full-order model are calibrated using historical data, and the reduced-order model is regenerated based on the calibrated full-order model.

[0028] By monitoring the simulation calculation time in real time and automatically switching to a reduced-order model when the timeout occurs, the problem of excessively long simulation time for full-order models in large-scale water supply systems is effectively solved, ensuring the real-time performance of the simulation and enabling the predicted data to keep up with changes in operating conditions. At the same time, error monitoring and regular updates ensure the accuracy of the reduced-order model, achieving a balance between real-time performance and accuracy.

[0029] Preferably, the method further includes the following steps: S801. A first absolute vibration sensor and a second absolute vibration sensor are respectively installed on the bearing housing and pump body of the target water pump unit and the standby unit, and a first eddy current displacement sensor and a second eddy current displacement sensor are respectively installed on the rotor shaft of the target water pump unit and the standby unit near the bearing. S802, using a preset synchronous sampling frequency f s The absolute vibration data sequence A1(t) and relative vibration data sequence R1(t) of the target water pump unit are simultaneously acquired by the first absolute vibration sensor and the first eddy current displacement sensor, and the absolute vibration data sequence A2(t) and relative vibration data sequence R2(t) of the standby unit are simultaneously acquired by the second absolute vibration sensor and the second eddy current displacement sensor. S803. Perform a fast Fourier transform on the absolute vibration data sequence and the relative vibration data sequence to obtain the absolute vibration amplitude spectrum A(f) and the relative vibration amplitude spectrum R(f) on each frequency component f. S804. Calculate the ratio T between the relative vibration amplitude spectrum R(f) and the absolute vibration amplitude spectrum A(f) at each frequency component f. R / A (f)=R(f) / A(f), generating the magnitude transferability matrix M(n,Q)=[T] associated with the operating conditions of speed n and flow rate Q. R / A (f)]; S805. Using a multivariate nonlinear regression method, with the real-time speed data n and real-time flow data Q of the target pump unit and the standby unit as input variables, and each element in the amplitude transfer rate matrix M(n,Q) as output variables, a dynamic correction model of the transfer function is established. S806. In step S1, raw vibration data V is collected by the vibration sensors installed on each water pump unit. raw (t) after which the original vibration data V raw (t) Perform the wavelet denoising process described above sequentially to obtain V den (t), and then, based on the current rotational speed n(t) and flow rate Q(t), the dynamic correction model of the transfer function is called to calculate the amplitude transfer rate matrix M(n(t),Q(t)) under the current operating condition, and V den (t) Equivalent relative vibration data V converted to the rotor shaft system reference frame eq (t)= V den (t)·M(n(t),Q(t)), and the equivalent relative vibration data V eq (t) serves as the input for subsequent simulation and fault identification.

[0030] In practical implementation, a first absolute vibration sensor and a second absolute vibration sensor can be respectively installed on the bearing housing and pump body of the target pump unit and the standby pump unit. These sensors can be piezoelectric accelerometers. Simultaneously, a first eddy current displacement sensor and a second eddy current displacement sensor are respectively installed on the rotor shaft of the target pump unit and the standby pump unit, near the bearing, to directly measure the vibration displacement of the rotor relative to the bearing. The sampling frequency is set to a pre-defined synchronous sampling frequency f. s (e.g., 10 kHz) Simultaneously acquire the absolute vibration data sequence A1(t) and relative vibration data sequence R1(t) of the target water pump unit, and the absolute vibration data sequence A2(t) and relative vibration data sequence R2(t) of the standby unit. Perform a Fast Fourier Transform on the acquired data to obtain the absolute vibration amplitude spectrum A(f) and relative vibration amplitude spectrum R(f) at each frequency component f. Calculate the ratio T between the relative vibration amplitude spectrum R(f) and the absolute vibration amplitude spectrum A(f) at each frequency component f. R / A (f)=R(f) / A(f), thus generating the amplitude transferability matrix M(n,Q)=[T] associated with the operating conditions of rotational speed n and flow rate Q. R / A (f)]. Rotational speed data can be obtained from the frequency converter, and flow rate data can be obtained from the flow meter on the pipeline. A multivariate nonlinear regression method is used, such as support vector regression, with real-time rotational speed n and flow rate Q as input variables, and each element in the amplitude transferability matrix M(n,Q) as the output variable, to establish a dynamic correction model for the transfer function. In step S1, raw vibration data V is collected using a vibration sensor. raw After (t), first treat V raw (t) Perform wavelet denoising processing sequentially to obtain V den (t), and then, based on the current rotational speed n(t) and flow rate Q(t), the transfer function dynamic correction model is called to calculate the amplitude transfer rate matrix M(n(t),Q(t)) under the current operating condition, and then V den (t) Equivalent relative vibration data V converted to the rotor shaft system reference frame eq (t)=V den (t)·M(n(t),Q(t)), and finally V eq (t) serves as the input for subsequent simulation and fault identification.

[0031] By synchronously collecting absolute and relative vibration data and establishing a transfer function model related to operating conditions, the easily measurable shell vibration was successfully converted into an equivalent relative vibration that can more directly reflect the rotor's health status. This solved the problem that shell vibration could not accurately characterize shaft system faults in traditional methods, and significantly improved the accuracy of fault diagnosis and early warning capabilities.

[0032] Preferably, before inputting the predicted vibration data into the fault identification model in step S4, the following sub-steps are also included: S401, The original vibration data V is collected in real time by the vibration sensors installed on each water pump unit. raw (t), and the original vibration data V raw (t) is subjected to fast Fourier transform to obtain the real-time vibration spectrum X(f) of each water pump unit; S402, extract the power frequency f0 and its harmonics kf0 components of each water pump unit, as well as the characteristic frequency f of the rolling bearing, from the real-time vibration spectrum X(f). bearing The components constitute the multi-unit characteristic frequency vector F; S403. Using the independent component analysis algorithm, with the multi-unit characteristic frequency vector F as input, solve the unmixing matrix W, and decompose the multi-unit characteristic frequency vector F into source signal components S=W·F corresponding to each independent vibration source. S404, Calculate each of the source signal components S i The correlation coefficient ρ between the frequency and the preset fault characteristic frequency library i When a certain source signal component S i The fault characteristic frequency f of the target water pump unit fault correlation coefficient ρ i The correlation coefficient is greater than the preset threshold range of 0.7 to 0.9, and the source signal component S i Projected energy percentage η on the target water pump unit i When the energy percentage exceeds the preset threshold range of 60% to 80%, the target water pump unit is confirmed as the actual source of failure. S405. Input the confirmed fault source information into the fault identification model, which is used to identify the target water pump unit with fault risk in step S4.

[0033] In practical implementation, before inputting the predicted vibration data into the fault identification model in step S4, the raw vibration data V can be collected in real time by vibration sensors installed on each water pump unit. raw (t), and the original vibration data V raw (t) Perform a Fast Fourier Transform to obtain the real-time vibration spectrum X(f) of each pump unit. Extract the power frequency f0 and its harmonics kf0 of each pump unit, as well as the characteristic frequency f of the rolling bearing, from the real-time vibration spectrum X(f). bearing The components are used to form a multi-unit characteristic frequency vector F. Then, using the multi-unit characteristic frequency vector F as input, the independent component analysis algorithm is employed to solve for the unmixing matrix W, decomposing F into source signal components S = W·F corresponding to each independent vibration source. Next, the source signal components S are calculated. iThe correlation coefficient ρ between the frequency and the preset fault characteristic frequency library i When a certain source signal component S i The fault characteristic frequency f of the target water pump unit fault correlation coefficient ρ i Greater than 0.7 to 0.9 (e.g., 0.8), and the source signal component S i The proportion of projected energy on the target pump unit η i When the percentage of projected energy exceeds 60% to 80% (e.g., 70%), the target pump unit can be confirmed as a genuine source of failure. The projected energy percentage can be calculated by projecting the source signal components onto the direction of the unit's sensor. Finally, the confirmed failure source information is input into the failure identification model for use in step S4 to identify target pump units with potential failure risks.

[0034] By using independent component analysis to separate the source signals of coupled vibrations from multiple units, and combining the dual criteria of correlation coefficient and energy proportion, interference vibrations from adjacent pumps transmitted through foundations, pipelines, etc. are effectively eliminated, accurately pinpointing the true source of the fault. This avoids misjudgment and unnecessary switching of multiple units due to vibration coupling, improves the accuracy of fault handling, and reduces operating costs.

[0035] Preferably, when generating the reconfiguration scheduling scheme that includes the target pump unit exit sequence and the standby unit entry sequence in step S5, the following constraints are also included: S501. Obtain the bus voltage data U(t) and feeder current data I(t) collected in real time by the voltage transformer and current transformer installed at the bus of the distribution network where the water supply system is located. S502. Based on the bus voltage data U(t) and the feeder current data I(t), calculate the current distribution network bus voltage amplitude |U| and the short-circuit capacity S of the feeder where the target pump unit and the standby unit are located. sc ; S503, the bus voltage amplitude |U| during the switching process is not lower than the preset allowable lower limit of voltage sag, ranging from 0.85 pu to 0.90 pu, and the feeder short-circuit capacity S sc The instantaneous rate of change dS during the process of the target pump unit being deactivated and the standby unit being activated. sc / dt does not exceed the preset capacity change rate threshold range of 10% to 15% per second as an additional constraint. S504. When the reconfiguration scheduling scheme does not meet the additional constraints, adjust the exit sequence of the target pump unit and the start-up sequence of the standby unit, or adjust the speed reduction rate r during the exit process of the target pump unit. down (t) and the rate of increase in rotational speed r during the process of activating the standby unit.up (t), generating a modified reconfiguration scheduling scheme that satisfies electrical safety constraints.

[0036] In practical implementation, when generating the reconfiguration scheduling scheme in step S5, the following additional constraints may also be included. First, obtain the bus voltage data U(t) and feeder current data I(t) collected in real time by voltage transformers and current transformers located at the bus of the distribution network where the water supply system is located. Electromagnetic voltage transformers can be selected, and through-type current transformers can be selected. Based on U(t) and I(t), the current bus voltage amplitude |U| of the distribution network and the short-circuit capacity S of the feeders where the target pump unit and standby unit are located can be calculated. sc Then, the bus voltage amplitude |U| during the switching process is not lower than the preset allowable lower limit of voltage sag, ranging from 0.85 pu to 0.90 pu (e.g., 0.87 pu), and the feeder short-circuit capacity S sc Instantaneous rate of change dS during the process of target pump unit shutdown and standby unit startup. sc An additional constraint is that the rate of change of capacity ( / dt) should not exceed a preset threshold range of 10% to 15% per second (e.g., 12% per second). When generating a reconfiguration scheduling scheme, the digital twin platform must simultaneously satisfy both hydraulic and electrical constraints. If the generated scheme does not meet the above additional constraints, it is necessary to adjust the decommissioning timing of the target pump unit and the commissioning timing of the standby unit, or adjust the rate of speed decrease (r) during the decommissioning process of the target pump unit. down (t) and the rate of increase in speed r during the standby unit's activation process. up (t), and re-optimize until a modified reconfiguration scheduling scheme that satisfies electrical safety constraints is generated.

[0037] By introducing electrical safety constraints such as bus voltage sag and short-circuit capacity change rate into the reconfiguration scheduling scheme, the coordinated optimization of hydraulic scheduling and electrical safety was achieved. This effectively avoided the impact on the local power grid during the switching of high-power water pump units, ensured the normal operation of other sensitive equipment in the power distribution system, and improved the electrical safety level of the entire water supply system.

[0038] Preferably, when generating the reconfiguration scheduling scheme that includes the target pump unit exit sequence and the standby unit entry sequence in step S5, the following sub-steps are also included: S511. A constraint relaxation priority list is pre-set, and the constraint relaxation priority list is arranged in order from low to high as follows: upper limit constraint of water tank level, lower limit constraint of water tank level, lower limit constraint of end user water supply pressure, and upper limit constraint of end user water supply pressure. S512, when the water supply pressure requirement of the end user is within the range of 0.14 MPa to 0.40 MPa and the liquid level of each water tank does not exceed the preset maximum liquid level H. max With preset minimum liquid level H min Under the condition that no feasible solution exists, the constraint relaxation mechanism is triggered; S513. According to the order of the constraint relaxation priority list, relax the restriction range of the current priority constraint in turn, wherein the upper limit constraint of the water tank liquid level is relaxed to the preset highest liquid level H. max The lower limit constraint of the water tank level is relaxed to the preset minimum level H, ranging from 105% to 110%. min For 90% to 95% of end users, the lower limit constraint on water supply pressure is relaxed to 0.10 MPa to 0.12 MPa, and the upper limit constraint on water supply pressure is relaxed to 0.42 MPa to 0.45 MPa. S514. After relaxing one priority constraint each time, the reconfiguration scheduling scheme is solved again until a feasible solution exists; S515. When a feasible solution exists, the relaxed constraints and the reconfigured scheduling scheme are output as an emergency scheduling scheme, and an emergency scheduling early warning signal is issued at the same time as it is sent to the field programmable logic controller.

[0039] In practical implementation, when generating the reconfiguration scheduling scheme in step S5, a constraint relaxation priority list can be pre-set. This list, arranged from low to high, includes: upper limit constraint of water tank level, lower limit constraint of water tank level, lower limit constraint of end-user water supply pressure, and upper limit constraint of end-user water supply pressure. During the scheduling scheme solution process, if it is found that the end-user water supply pressure demand is within the range of 0.14 MPa to 0.40 MPa and the water tank level does not exceed the preset maximum level H, the constraint will be relaxed. max With preset minimum liquid level H min If no feasible solution exists under the given conditions, the constraint relaxation mechanism is triggered. Then, according to the priority list, the constraint range of the current priority condition is relaxed in turn. The specific relaxation range can be set as follows: the upper limit constraint of the water tank level is relaxed to the preset maximum level H. max The lower limit constraint of the water tank level is relaxed to the preset minimum level H, ranging from 105% to 110% (e.g., 108%). minFor each priority constraint, the lower limit constraint of water supply pressure for end users is relaxed to 0.10 MPa to 0.12 MPa (e.g., 0.11 MPa), and the upper limit constraint is relaxed to 0.42 MPa to 0.45 MPa (e.g., 0.43 MPa). After relaxing one priority constraint at a time, the reconstructed scheduling scheme is resolved until a feasible solution exists. Finally, the relaxed constraints and reconstructed scheduling scheme corresponding to the existence of a feasible solution are output as the emergency scheduling scheme, and an emergency scheduling early warning signal is issued simultaneously with the distribution to the field programmable logic controller.

[0040] By presetting the priority and relaxation range of constraint relaxation, the relaxation mechanism is automatically triggered when there is no feasible solution under extreme conditions. The secondary constraints are relaxed step by step, ensuring that the system can generate emergency scheduling plans and maintain basic operation. This avoids scheduling deadlock and water supply interruption caused by no solution. At the same time, it issues early warning signals to facilitate manual intervention and improves the survivability of the water supply system under extreme conditions.

[0041] The present invention has at least the following beneficial effects: The water pump coordinated scheduling and water supply control method based on multi-stage water tanks described in this invention achieves comprehensive optimization of water pump coordinated scheduling in multi-stage water tank water supply systems through the following technical means: At the data acquisition level, by installing pressure sensors at the outlet pipes of each water pump unit, level sensors in each water tank, and vibration sensors on each water pump unit, a comprehensive perception of the operating status of the water supply system was achieved, providing a multi-dimensional real-time data foundation for subsequent simulation and fault identification. In the digital twin platform, a virtual twin containing physical attribute information, spatial topology relationships, and operating mechanism models of all equipment was pre-built, enabling the simulation to truly reflect the dynamic characteristics of the system.

[0042] At the fault prediction level, the digital twin platform drives the virtual twin to operate synchronously based on real-time collected outlet pressure data, liquid level data, and raw vibration data, and simulates and extrapolates the operating status within a preset time interval in the future, generating predicted vibration data and its changing trends for each pump unit. The predicted vibration data and changing trends are input into the fault identification model. When the predicted value reaches the first-level alarm threshold and the predicted extreme value is about to reach the second-level alarm threshold, the target pump unit with fault risk is identified, realizing the leap from passive response to active early warning, and identifying potential risks before the fault actually occurs.

[0043] At the scheduling decision level, the digital twin platform generates a reconfiguration scheduling scheme based on the identified target water pump unit, the real-time liquid level data of each water tank, the valve status of each pipe section, and the current status of the valves associated with the target water pump unit and the standby unit. Under the premise of meeting the end user's water supply pressure requirements within the range of 0.14 MPa to 0.40 MPa and ensuring that the liquid level of each water tank does not exceed the preset maximum and minimum liquid levels, the platform generates a reconfiguration scheduling scheme that includes the target water pump unit's exit sequence and the standby unit's entry sequence. This ensures that the scheduling scheme achieves the safe removal of risky equipment while meeting system constraints.

[0044] At the execution control level, the reconfiguration scheduling plan is sent to the field programmable logic controller (PLC). The PLC then executes the operation of shutting down the target water pump unit and its associated valves and starting the standby unit and its associated valves in sequence according to the timing and parameters in the plan. This achieves precise execution of the scheduling decision and ensures a smooth and orderly switching process.

[0045] Through the synergistic effect of the above-mentioned technical means, the present invention effectively solves the technical problems of existing water supply scheduling methods that rely on post-event alarms and cannot proactively identify risks and adjust operating strategies in advance before a fault occurs. It significantly improves the continuity and stability of the water supply system and reduces the risk of water supply interruption caused by sudden faults.

[0046] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Detailed Implementation

[0047] The present invention will now be described in further detail with reference to specific embodiments, so that those skilled in the art can implement it based on the description.

[0048] It should be understood that terms such as “having,” “comprising,” and “including” as used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

[0049] It should be noted that, unless otherwise specified, the experimental methods described in the following implementation plan are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified.

[0050] Example 1 In a municipal water supply system, the method provided by this invention is used to coordinate and control a three-stage booster pump station and an elevated water tank. The water supply system comprises six pump units connected in parallel and three water tanks connected in series. In specific implementation, diffused silicon pressure transmitters are installed at the outlet pipes of each pump unit as pressure sensors, submersible level gauges are installed in each water tank as level sensors, and piezoelectric accelerometers are installed at the bearing seats of each pump unit as vibration sensors. These sensors collect outlet pressure and level data at a frequency of 1 Hz and raw vibration data at a frequency of 10 kHz. The collected data is transmitted in real-time via industrial Ethernet to a digital twin platform, which is built on an industrial server and runs digital twin software. The platform pre-constructs a virtual twin containing the geometric dimensions, material properties, spatial connections, hydraulic mechanism model (based on mass and momentum conservation equations), and vibration mechanism model (based on rotor dynamics model) of all pump units, pipes, valves, and water tanks.

[0051] The digital twin platform drives the synchronous operation of the virtual twin based on received real-time data, and simulates the operating status for the next 30 seconds, generating predicted vibration data and trends for each pump unit. This predicted vibration data and trends are then input into a fault identification model. The model has two preset alarm thresholds: a first threshold of 2.4 mm / s and a second threshold of 4.5 mm / s. When the predicted vibration data of a pump unit reaches 2.4 mm / s, and the predicted extreme value for the next 30 seconds, based on the trend, is predicted to reach 4.5 mm / s, that pump unit is identified as a target pump unit with a potential fault risk.

[0052] Based on the identified target pump unit, current water tank level data, valve status of each pipe section, and the current status of valves associated with the target and standby pump units, the digital twin platform generates a reconfiguration scheduling scheme that includes a sequence for the target pump unit to exit operation and a sequence for the standby pump unit to start operation, provided that the water supply pressure demand of the end users is within the range of 0.14 MPa to 0.40 MPa and the water tank levels do not exceed the preset maximum and minimum levels. This scheme is then distributed to the field programmable logic controller (PLC) via industrial Ethernet. The PLC executes the operations of closing the target pump unit and its associated valves, and starting the standby pump unit and its associated valves, according to the timing and parameters in the scheme.

[0053] Example 2 Building upon Example 1, to ensure the long-term accuracy of the digital twin model, a model calibration period T of 24 hours can be set in the digital twin platform. At the end of each calibration period, historical outlet pressure data P collected by the pressure sensor and level sensor within that period is acquired. h(t) and historical liquid level data L h (t). Historical data is input into a virtual twin to simulate historical operating conditions, resulting in historical simulation pressure data P. s (t) and historical simulation liquid level data L s (t). Calculate the pressure deviation value sequence ΔP(t) = P h (t)-P s (t) and the sequence of liquid level deviation values ​​ΔL(t)=L h (t)-L s RMSE of (t) P and RMSE L When RMSE P Greater than 0.022 MPa or RMSE L When the value is greater than 0.16 m, model correction is triggered. The least squares method is used to iteratively optimize the parameters to be corrected (such as the pipe roughness coefficient) until RMSE is achieved. P Less than 0.018 MPa and RMSE L If the value is less than 0.14 m, then the corrected parameters are updated to the virtual twin.

[0054] Simultaneously, to eliminate environmental noise in the vibration signal, wavelet denoising processing can be performed on the original vibration data in step S1. Specifically, the acquired original vibration data V raw (t) Perform 5-level wavelet decomposition, apply soft thresholding rules to threshold quantize high-frequency components, and then perform wavelet reconstruction to obtain denoised vibration data V. den (t). V den (t) serves as the baseline data for model calibration during subsequent simulations.

[0055] Based on this, to ensure the accuracy of the vibration mechanism model, the denoised vibration data and the predicted vibration data can be compared in real time in step S3. (Generation and V) den (t) Predicted vibration data V at the same time pre (t), calculate the vibration deviation value ΔV(t) = V den (t)-V pre (t). Calculate the root mean square error (RMSE) using ΔV(t) at 10 consecutive sampling times. v When RMSEv is greater than 0.55 mm / s and occurs four times consecutively, vibration mechanism model correction is triggered. A particle swarm optimization algorithm is used to iteratively optimize the parameters to be corrected in the vibration mechanism model (such as bearing stiffness coefficient) until RMSEv is reached. v The speed is less than 0.45 mm / s, and then the corrected parameters are updated to the virtual twin.

[0056] Example 3 Based on Example 2, the following extended implementation method can be adopted to further improve system performance.

[0057] In terms of switching control, after the field programmable logic controller (PLC) receives the reconfiguration scheduling scheme, it acquires the first pressure data P1(t) at the outlet of the target pump unit and the second pressure data P2(t) at the outlet of the standby unit. Based on P1(t) and the target switching pressure value P... set The deviation e1(t) = P (determined to be 0.28 MPa by simulation optimization using a digital twin platform) set -P1(t), the PID control algorithm is used to calculate the target unit's speed decrease rate rdown(t). Based on P2(t) and P set The deviation e2(t) = P set -P2(t), calculate the rate of increase of the standby unit's rotational speed r up (t). Control the target unit according to r down (t) Linearly reduce the operating frequency, while controlling the standby unit according to r up (t) The operating frequency is linearly increased. When the operating frequency of the target unit drops to 25 Hz, a shutdown operation is performed. Throughout the process, the pressure difference ΔP(t) = |P1(t) - P2(t)| is monitored in real time, and the opening rate of the electric valve is adjusted to ensure that ΔP(t) is always less than 0.045 MPa, thus achieving a smooth switching.

[0058] Regarding simulation computation efficiency, the actual computation time T can be recorded before conducting the simulation. calc When T calc When the time exceeds 2 seconds, the model order reduction mode is triggered. The intrinsic orthogonal decomposition method is used to reduce the order of the full-order operating mechanism model, generating a lower-order reduced model. Simulations are then performed using the reduced model to generate predicted vibration data V. pre,red (t). After completion, record the output of the reduced-order model and the actual vibration data V. act The absolute value of the error (t) is taken as the error. When the error is greater than 0.55 mm / s three times in a row, the model is taken out of the reduced-order mode and the full-order model is used again. The reduced-order model is then updated after the full-order model is recalibrated in the next calibration cycle.

[0059] For vibration mapping, absolute vibration sensors can be installed on the bearing housings and pump bodies of both the target and standby units, while eddy current displacement sensors can be installed on the rotor shaft near the bearings. Absolute vibration data A1(t), A2(t) and relative vibration data R1(t), R2(t) are acquired at a synchronous sampling frequency of 10 kHz. Fast Fourier transform is performed on the data to obtain amplitude spectra A(f) and R(f), and the ratio T of each frequency component is calculated. R / A(f)=R(f) / A(f), generating the magnitude transferability matrix M(n,Q)=[T] related to the rotational speed n and flow rate Q. R / A (f)]. A dynamic correction model for the transfer function is established using support vector regression. This is based on the collected raw vibration data V. raw After (t), wavelet denoising is performed to obtain V. den (t), and then based on the current rotational speed n(t) and flow rate Q(t), the corrected model is called to obtain M(n(t),Q(t)), and V den (t) is converted into equivalent relative vibration data V eq (t)= V den (t)·M(n(t),Q(t)), is used for subsequent simulation and fault identification.

[0060] Regarding fault source separation, before inputting the predicted vibration data into the fault identification model, a fast Fourier transform is performed on the original vibration data to obtain the real-time vibration spectrum X(f), and the power frequency f0 and its harmonics kf0 components, as well as the characteristic frequency f of the rolling bearing, are extracted. bearing The components constitute a multi-unit characteristic frequency vector F. The unmixing matrix W is solved using independent component analysis (ICA), decomposing F into source signal components S = W·F. The correlation coefficient ρ between each source signal component Si and a pre-defined fault characteristic frequency database is then calculated. i When a certain S i With the target unit fault characteristic frequency f fault correlation coefficient ρ i Greater than 0.8, and the S i Projected energy percentage η on the target unit i When the failure rate is greater than 70%, the true source of the fault is identified, and the identified information is input into the fault identification model.

[0061] Regarding electrical safety constraints, when generating the reconfiguration scheduling scheme, the bus voltage U(t) and feeder current I(t) collected in real time by voltage transformers and current transformers at the distribution network bus are obtained, and the bus voltage amplitude |U| and feeder short-circuit capacity S are calculated. sc The |U| value during the switching process should not be lower than 0.87 pu, and the dS value should be... sc / dt not exceeding 12% per second is used as an additional constraint. When the scheme does not meet the constraints, the unit shutdown sequence, startup sequence, or speed change rate are adjusted to generate a modified scheme that meets the electrical safety constraints.

[0062] For handling extreme operating conditions, a pre-set constraint relaxation priority list is configured in the following order: upper limit of water tank level, lower limit of water tank level, lower limit of terminal water supply pressure, and upper limit of terminal water supply pressure. When no feasible solution is found, the relaxation mechanism is triggered, and the constraints are relaxed sequentially: the upper limit of water tank level is relaxed to the preset maximum level H. max108%, the lower limit of the water tank level is relaxed to the preset minimum level H. min The lower pressure limit is relaxed to 0.11 MPa, and the upper pressure limit is relaxed to 0.43 MPa. Each time a constraint is relaxed, the solution is recalculated until a feasible solution exists. The resulting solution is then output as an emergency dispatch plan, and an early warning signal is issued.

[0063] Comparative Example 1 In a municipal water supply system, a traditional PLC logic control method is used to schedule and control a three-stage booster pump station and elevated water tanks. The system comprises six pump units and three series-connected water tanks. Pressure sensors are installed at the outlet pipes of each pump, and level sensors are installed in each water tank. The control strategy is based on preset water level thresholds: when the water tank level is below the lower limit, the pumps are started sequentially; when the level is above the upper limit, the pumps are stopped sequentially. The start and stop of all pumps are directly controlled by the PLC based on the water level signals, without vibration monitoring or prediction functions. During one operation, one pump experienced progressively increased vibration due to bearing wear, but no alarm was triggered. Only when the bearing failed, causing a sudden drop in outlet pressure, did the PLC alarm due to low pressure and urgently start the backup pump. However, during the switchover, the network pressure fluctuated violently, and the water supply pressure to end users instantly dropped from 0.38 MPa to 0.09 MPa, exceeding the allowable range, causing a brief water outage in some areas. The comparison shows that traditional logic control methods rely on post-event alarms, cannot identify potential equipment problems in advance, and the switching process is abrupt, making it difficult to ensure the continuity and stability of water supply.

[0064] Comparative Example 2 In another water supply system, a traditional PID variable frequency control method was used to maintain constant pipeline pressure. The system was also equipped with pressure and level sensors, and the controller adjusted the pump speed based on the outlet pressure deviation. The start and stop of all pumps were controlled sequentially by a PLC, without vibration monitoring or model correction functions. After a period of operation, due to pipe scaling and pump wear, the system's hydraulic characteristics changed, but the PID controller parameters were not adjusted, leading to increased oscillations during the adjustment process. The pressure fluctuation range expanded to 0.12 MPa to 0.45 MPa, exceeding the end-user's needs at certain times. Simultaneously, due to the lack of vibration signal processing, environmental noise (such as water turbulence and vibration from adjacent units) frequently triggered false alarms. The PLC misjudged faults and repeatedly activated standby pumps, increasing equipment wear and energy consumption. This comparative example shows that traditional variable frequency control lacks model adaptation capabilities and signal processing methods, resulting in performance degradation, high misjudgment rate, and poor system stability after long-term operation.

[0065] Comparative Example 3 In the third scenario, a traditional unit switching method was used: when a running pump malfunctions or requires maintenance, the operator manually or via PLC starts the standby pump and stops the faulty pump in a fixed sequence. The switching process lacks pressure coordination, directly starting and stopping at rated speed, causing significant instantaneous fluctuations in pipeline pressure. During one switching operation, the simultaneous operation of two high-power pumps caused the distribution bus voltage to temporarily drop to 0.75 pu, triggering the undervoltage protection of other frequency converters on the same bus and causing multiple devices to shut down. Furthermore, during peak water usage periods, when all running pumps are operating at full capacity and the water tank level is near its lower limit, the PLC cannot generate a switching plan that satisfies all constraints, causing the system to enter a dead loop and unable to perform any operations. Only emergency manual intervention was required to prevent a water outage. This comparative example illustrates that the traditional switching method lacks precise control, electrical safety considerations, and emergency strategies under extreme conditions, posing serious safety hazards and operational risks.

[0066] Test methods and comparison results Comparative tests were conducted on Examples 1-3 and Comparative Examples 1-3 in the same water supply system experimental platform or simulation environment. The test system included 6 parallel water pump units, 3 series water tanks, and related pipes and valves. The water supply pressure requirement for end users was set to 0.14 MPa to 0.40 MPa, and the maximum liquid level H in the water tanks was set to... max =10 m, lowest liquid level H min =2 m. All tests were run for 30 days under the same operating conditions, and early failures such as bearing wear and impeller imbalance were simulated by a preset fault injection method.

[0067] Test metrics and methods 1. Fault warning lead time: Inject a gradually developing bearing wear fault into the system, record the time interval from the appearance of fault characteristics (vibration amplitude begins to rise) to the actual occurrence of the fault (leading to pressure drop or equipment shutdown), as well as the moment when each method first issues a warning, and calculate the warning lead time.

[0068] 2. Maximum pressure fluctuation during switching process: When unit switching is required, record the lowest or highest water supply pressure of the end user during the switching process, and calculate the maximum deviation from the set pressure range boundary.

[0069] 3. Model prediction error: The root mean square error between the predicted pressure and liquid level output by the virtual twin and the actual collected data is calculated every 24 hours, and the average value over 30 days is taken.

[0070] 4. Number of false alarms: Count the number of non-real fault alarms caused by environmental noise or interference within 30 days.

[0071] 5. Simulation calculation time: Record the average time of a single full-system simulation, as well as the processing capacity in case of calculation timeout.

[0072] 6. Electrical sag amplitude: Record the lowest value of the distribution network bus voltage amplitude during unit switching.

[0073] 7. Extreme operating condition scheme generation rate: During peak water usage periods (all units are operating at full capacity and water tank levels are close to the lower limit), simulate scenarios requiring switching and calculate the percentage of scheduling schemes that can be successfully generated.

[0074] The test results are shown in Table 1.

[0075] Table 1 Note: "-" indicates that the comparison scale does not have the corresponding function or has not been measured.

[0076] Comparative Example 1 uses traditional PLC logic control, which only triggers an alarm after a fault occurs. The warning lead time is 0, and the pressure fluctuation during the switching process is as high as 0.31 MPa, far exceeding the allowable range. It also lacks functions such as model prediction and false alarm statistics, and cannot generate solutions under extreme working conditions.

[0077] Comparative Example 2 uses PID variable frequency control, which can maintain pressure but has no early warning capability. The switching fluctuation is still 0.28 MPa, the model prediction error is large (0.035 MPa), and because no signal denoising is performed, there are as many as 12 false alarms, resulting in unnecessary switching.

[0078] Comparative Example 3 uses traditional manual or sequential switching. The pressure fluctuation during the switching process is the largest (0.33 MPa), and it causes the electrical voltage to drop to 0.75 pu, affecting other equipment. Under extreme conditions, the solution generation rate is 0.

[0079] In comparison, Example 1 achieved a 15-25 minute advance warning through vibration prediction, reducing switching fluctuation to 0.08 MPa, initially demonstrating the advantages of the present invention. Example 2, after introducing model correction and signal denoising, extended the warning time to 20-30 minutes, reduced the model error to 0.008 MPa, significantly reduced the number of false alarms to 1, and further reduced switching fluctuation to 0.06 MPa. Example 3 comprehensively applied all optimization methods, achieving a warning time of 25-35 minutes, a switching fluctuation of only 0.04 MPa, a model error of 0.006 MPa, zero false alarms, reduced simulation time to 0.8 seconds through a reduced-order model, controlled electrical sag above 0.88 pu, and achieved a 100% generation rate for extreme operating condition scenarios.

[0080] Test results show that the present invention has achieved significant improvements over traditional methods in terms of fault warning timeliness, switching process smoothness, model accuracy, anti-interference ability, real-time performance, electrical safety, and adaptability to extreme working conditions.

[0081] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and embodiments shown and described herein.

Claims

1. A method for coordinated scheduling of water pumps and water supply control based on multi-stage water tanks, characterized in that, Includes the following steps: S1. In a water supply system that includes multiple pump units connected in series or parallel and multiple water tanks connected in series, pressure sensors installed at the outlet pipes of each pump unit, level sensors installed in each water tank, and vibration sensors installed on each pump unit are used to collect the outlet pressure data of each pump unit, the level data of each water tank, and the original vibration data of each pump unit in real time at a preset data acquisition frequency. S2. The real-time collected outlet pressure data, liquid level data, and original vibration data are transmitted to a digital twin platform. The digital twin platform is pre-built with a virtual twin containing the physical attribute information, spatial topology relationship, and operating mechanism model of all water pump units, pipes, valves, and water tanks in the water supply system. The operating mechanism model includes a hydraulic mechanism model and a vibration mechanism model. S3. The digital twin platform drives the virtual twin to operate synchronously based on the received outlet pressure data, liquid level data and original vibration data, and simulates the operation status of the water supply system in a future preset time interval based on the vibration mechanism model in the virtual twin, generating predicted vibration data of each water pump unit and the changing trend of each predicted vibration data. S4. Input the predicted vibration data and the trend of change into the fault identification model. When a certain value in the predicted vibration data reaches the first level alarm threshold, and the predicted extreme value in the second preset time interval in the future, deduced from the trend of change, will reach the second level alarm threshold, the target water pump unit with fault risk is identified. S5. Based on the identified target water pump unit, the real-time liquid level data of each water tank, the valve status of each pipe section, and the current status of the valves associated with the target water pump unit and the standby unit, the digital twin platform generates a reconfiguration scheduling scheme that includes the target water pump unit exiting operation sequence and the standby unit entering operation sequence, provided that the water supply pressure demand of the end user is within the range of 0.14 MPa to 0.40 MPa and the liquid level of each water tank does not exceed the preset maximum liquid level and the preset minimum liquid level. S6. The reconfiguration scheduling scheme is sent to the field programmable logic controller (PLC). The PLC executes the operation of shutting down the target water pump unit and its associated valves and starting the standby unit and its associated valves in sequence according to the timing and parameters in the reconfiguration scheduling scheme.

2. The method for coordinated scheduling and water supply control of water pumps based on multi-stage water tanks as described in claim 1, characterized in that, It also includes the following steps: S701. In the digital twin platform, a model calibration period T is preset; S702. When each model calibration cycle T arrives, acquire the historical outlet pressure data P collected in real time by the pressure sensor and the liquid level sensor during that cycle. h (t) and historical liquid level data L h (t); S703, Transfer the historical outlet pressure data P h (t) and the historical liquid level data L h (t) Input the virtual twin and drive the virtual twin to perform a simulation of historical operating conditions to obtain historical simulation pressure data P corresponding to the historical outlet pressure data and the historical liquid level data. s (t) and historical simulation liquid level data L s (t); S704. Calculate the pressure deviation sequence ΔP(t) between the historical outlet pressure data and the historical simulated pressure data. h (t)-P s (t), and the sequence of liquid level deviation values ​​ΔL(t)=L between the historical liquid level data and the historical simulated liquid level data. h (t)-L s (t); S705, when the root mean square error (RMSE) of the pressure deviation value sequence ΔP(t) is... P Greater than 0.022 MPa, or the root mean square error (RMSE) of the liquid level deviation sequence ΔL(t) L When the liquid level deviation exceeds the preset threshold range of 0.14 m to 0.16 m, with a maximum value of 0.16 m, model correction is triggered. S706. After triggering model correction, a parameter identification algorithm is used, based on the historical outlet pressure data P. h (t) and the historical liquid level data L h (t) is used as the objective to iteratively optimize at least one parameter to be corrected in the operating mechanism model in the virtual twin until the root mean square error (RMSE) between the simulation data output by the updated operating mechanism model and the historical outlet pressure data and the historical liquid level data is found. P The pressure deviation threshold is less than the minimum value of 0.018 MPa, and the root mean square error (RMSE) between the simulated liquid level data output by the updated operating mechanism model and the historical liquid level data is less than the minimum value of the pressure deviation threshold. L The minimum value of the liquid level deviation threshold range is 0.14 m; S707. Update the parameters of the corrected operating mechanism model to the virtual twin for simulation and deduction in subsequent steps.

3. The water pump coordinated scheduling and water supply control method based on multi-stage water tanks as described in claim 1, characterized in that, Step S1 also includes the following sub-steps: S101, The original vibration data V of each water pump unit is collected in real time by the vibration sensors installed on each water pump unit. raw (t); S102, regarding the original vibration data V raw (t) Perform wavelet decomposition to obtain wavelet coefficients containing high-frequency and low-frequency components; S103. Using a pre-set threshold rule, the high-frequency components in the wavelet coefficients are subjected to threshold quantization to obtain the corrected high-frequency wavelet coefficients. S104. Perform wavelet reconstruction on the low-frequency component and the corrected high-frequency wavelet coefficients to obtain the original vibration data V. raw (t) Corresponding denoised vibration data V den (t); S105, the denoised vibration data V den (t) serves as the reference data for model calibration during simulation and deduction of the digital twin platform described in step S3.

4. The water pump coordinated scheduling and water supply control method based on multi-stage water tanks as described in claim 3, characterized in that, Step S3 also includes the following steps: S301. When generating predicted vibration data for each water pump unit based on the virtual twin, generate the denoised vibration data V collected by the vibration sensor. den (t) Predicted vibration data V at the same time pre (t); S302, Calculate the vibration deviation value ΔV(t) between the denoised vibration data and the predicted vibration data. den (t)-V pre (t); S303. Calculate the root mean square error (RMSE) of the vibration deviation value ΔV(t) for a preset number of N consecutive sampling times. v When RMSE v If the vibration deviation exceeds the preset threshold range of 0.45 mm / s to 0.55 mm / s, with a maximum value of 0.55 mm / s, and occurs consecutively for 3 to 5 times, the mechanism model parameters related to the vibration characteristics in the virtual twin are triggered for correction. S304. Employ a parameter identification algorithm using the denoised vibration data V den (t) is used as the objective to iteratively optimize the parameters to be corrected in the vibration mechanism model of the virtual twin, so that the predicted vibration data V output by the updated vibration mechanism model is... pre (t) and the denoised vibration data V den The deviation between (t) satisfies RMSE v The minimum value of the vibration deviation threshold range is 0.45 mm / s; S305. Update the parameters of the corrected vibration mechanism model to the virtual twin for simulation and deduction in subsequent steps.

5. The method for coordinated scheduling and water supply control of water pumps based on multi-stage water tanks as described in claim 1, characterized in that, The step of distributing the reconfiguration scheduling scheme to the field-programmable logic controller and executing the operation specifically includes the following sub-steps: S601. After the field programmable logic controller receives the reconfiguration scheduling scheme, it acquires the first outlet pressure data P1(t) collected in real time by the first pressure sensor set at the outlet pipe of the target water pump unit, and the second outlet pressure data P2(t) collected in real time by the second pressure sensor set at the outlet pipe of the standby unit. S602, Based on the first outlet pressure data P1(t) and the preset target switching pressure value P set The first deviation between them is e1(t) = P set -P1(t), using a proportional-integral-derivative control algorithm, calculates the rate of speed decrease r of the target pump unit during the shutdown process. down (t); the target switching pressure value P set The value is determined by the digital twin platform to be a constant value in the range of 0.14 MPa to 0.40 MPa after simulation optimization based on the steady-state operating data of the target pump unit before it is decommissioned and the expected operating conditions after the standby unit is put into operation. S603, Based on the second outlet pressure data P2(t) and the target switching pressure value P set The second deviation between them is e2(t)=P set -P2(t), using a proportional-integral-derivative control algorithm, calculates the rate of increase in speed r of the standby unit during the commissioning process. up (t); S604. Control the target water pump unit to decrease at the speed reduction rate r. down (t) linearly reduces its operating frequency f1(t), while controlling the standby unit to increase its speed according to the speed increase rate r. up (t) linearly increases its operating frequency f2(t), and when the operating frequency f1(t) of the target water pump unit decreases to the preset no-load frequency range of 20 Hz to 30 Hz, the operation of shutting down the target water pump unit and its associated valves is performed; S605. During the entire process of shutting down the target water pump unit and its associated valves and starting the standby unit and its associated valves, the difference ΔP(t) = |P1(t)-P2(t)| between the first outlet pressure data P1(t) and the second outlet pressure data P2(t) is monitored in real time. The opening rate of the electric valves associated with the target water pump unit and the standby unit is adjusted so that ΔP(t) is always less than the minimum value of 0.045 MPa in the preset pressure difference threshold range of 0.045 MPa to 0.055 MPa.

6. The method for coordinated scheduling and water supply control of water pumps based on multi-stage water tanks as described in claim 1, characterized in that, The step of simulating and extrapolating the operating status of the water supply system within a preset future time interval based on a virtual twin specifically includes the following sub-steps: S311. Before performing a simulation of the water supply system's operation within a future preset time interval based on the virtual twin, record the actual computation time T of this simulation. calc ; S312, the actual calculation time T calc The maximum value T within the preset maximum allowable time threshold range of 0.5 seconds to 2 seconds. max =2 seconds for comparison; S313, when T calc Greater than T max When the model order reduction mode is triggered, the intrinsic orthogonal decomposition method is used to reduce the order of the full-order operating mechanism model in the virtual twin, generating a reduced-order operating mechanism model with a lower order than the full-order operating mechanism model. The reduced-order operating mechanism model is generated based on the latest corrected full-order operating mechanism model. S315. Using the reduced-order operation mechanism model to replace the full-order operation mechanism model, simulate and extrapolate the operation state of the water supply system within a future preset time interval, and generate predicted vibration data V for each water pump unit. pre,red (t) and the changing trends of each predicted vibration data; S316. After completing this simulation, record the predicted vibration data V output by the reduced-order operation mechanism model. pre,red (t) and the actual vibration data V collected at the same time by the vibration sensors installed on each water pump unit. act The absolute value of the error between (t) |V pre,red (t)-V act (t)|; S317. When the absolute value of the error is greater than the maximum value of 0.55 mm / s within the preset error allowable range of 0.45 mm / s to 0.55 mm / s for three consecutive times, exit the model reduction mode, resume the simulation and deduction using the full-order operation mechanism model, and when the next model correction cycle T arrives, use historical data to correct the parameters of the full-order operation mechanism model, and regenerate the reduced-order operation mechanism model based on the corrected full-order operation mechanism model.

7. The method for coordinated scheduling and water supply control of pumps based on multi-stage water tanks as described in claim 3, characterized in that, It also includes the following steps: S801. A first absolute vibration sensor and a second absolute vibration sensor are respectively installed on the bearing housing and pump body of the target water pump unit and the standby unit, and a first eddy current displacement sensor and a second eddy current displacement sensor are respectively installed on the rotor shaft of the target water pump unit and the standby unit near the bearing. S802, using a preset synchronous sampling frequency f s The absolute vibration data sequence A1(t) and relative vibration data sequence R1(t) of the target water pump unit are simultaneously acquired by the first absolute vibration sensor and the first eddy current displacement sensor, and the absolute vibration data sequence A2(t) and relative vibration data sequence R2(t) of the standby unit are simultaneously acquired by the second absolute vibration sensor and the second eddy current displacement sensor. S803. Perform a fast Fourier transform on the absolute vibration data sequence and the relative vibration data sequence to obtain the absolute vibration amplitude spectrum A(f) and the relative vibration amplitude spectrum R(f) on each frequency component f. S804. Calculate the ratio T between the relative vibration amplitude spectrum R(f) and the absolute vibration amplitude spectrum A(f) at each frequency component f. R / A (f)=R(f) / A(f), generating the amplitude transferability matrix M(n,Q)=[T] associated with the operating conditions of speed n and flow rate Q. R / A (f)]; S805. Using a multivariate nonlinear regression method, with the real-time speed data n and real-time flow data Q of the target pump unit and the standby unit as input variables, and each element in the amplitude transfer rate matrix M(n,Q) as output variables, a dynamic correction model of the transfer function is established. S806. In step S1, raw vibration data V is collected by the vibration sensors installed on each water pump unit. raw (t) after which the original vibration data V raw (t) Perform the wavelet denoising process described above sequentially to obtain V den (t), and then, based on the current rotational speed n(t) and flow rate Q(t), the dynamic correction model of the transfer function is called to calculate the amplitude transfer rate matrix M(n(t),Q(t)) under the current operating condition, and V den (t) Equivalent relative vibration data V converted to the rotor shaft system reference frame eq (t)=V den (t)·M(n(t),Q(t)), and the equivalent relative vibration data V eq (t) serves as the input for subsequent simulation and fault identification.

8. The method for coordinated scheduling and water supply control of water pumps based on multi-stage water tanks as described in claim 1, characterized in that, Before inputting the predicted vibration data into the fault identification model as described in step S4, the following sub-steps are also included: S401, The original vibration data V is collected in real time by the vibration sensors installed on each water pump unit. raw (t), and the original vibration data V raw (t) is subjected to fast Fourier transform to obtain the real-time vibration spectrum X(f) of each water pump unit; S402, extract the power frequency f0 and its harmonics kf0 components of each water pump unit, as well as the characteristic frequency f of the rolling bearing, from the real-time vibration spectrum X(f). bearing The components constitute the multi-unit characteristic frequency vector F; S403. Using the independent component analysis algorithm, with the multi-unit characteristic frequency vector F as input, solve the unmixing matrix W, and decompose the multi-unit characteristic frequency vector F into source signal components S=W·F corresponding to each independent vibration source. S404, Calculate each of the source signal components S i The correlation coefficient ρ between the frequency and the preset fault characteristic frequency library i When a certain source signal component S i The fault characteristic frequency f of the target water pump unit fault correlation coefficient ρ i The correlation coefficient is greater than the preset threshold range of 0.7 to 0.9, and the source signal component S i Projected energy percentage η on the target water pump unit i When the energy percentage exceeds the preset threshold range of 60% to 80%, the target water pump unit is confirmed as the actual source of failure. S405. Input the confirmed fault source information into the fault identification model, which is used to identify the target water pump unit with fault risk in step S4.

9. The method for coordinated scheduling and water supply control of water pumps based on multi-stage water tanks as described in claim 1, characterized in that, When generating the reconfiguration scheduling scheme that includes the target pump unit exit sequence and the standby unit entry sequence in step S5, the following constraints are also included: S501. Obtain the bus voltage data U(t) and feeder current data I(t) collected in real time by the voltage transformer and current transformer installed at the bus of the distribution network where the water supply system is located. S502. Based on the bus voltage data U(t) and the feeder current data I(t), calculate the current distribution network bus voltage amplitude |U| and the short-circuit capacity S of the feeder where the target pump unit and the standby unit are located. sc ; S503, the bus voltage amplitude |U| during the switching process is not lower than the preset allowable lower limit of voltage sag, ranging from 0.85 pu to 0.90 pu, and the feeder short-circuit capacity S sc The instantaneous rate of change dS during the process of the target pump unit being deactivated and the standby unit being activated. sc / dt does not exceed the preset capacity change rate threshold range of 10% to 15% per second as an additional constraint. S504. When the reconfiguration scheduling scheme does not meet the additional constraints, adjust the exit sequence of the target pump unit and the start-up sequence of the standby unit, or adjust the speed reduction rate r during the exit process of the target pump unit. down (t) and the rate of increase in rotational speed r during the process of activating the standby unit. up (t), generating a modified reconfiguration scheduling scheme that satisfies electrical safety constraints.

10. The method for coordinated scheduling and water supply control of water pumps based on multi-stage water tanks as described in claim 1, characterized in that, When generating the reconfiguration scheduling scheme that includes the target pump unit exit sequence and the standby unit entry sequence in step S5, the following sub-steps are also included: S511. A constraint relaxation priority list is pre-set, and the constraint relaxation priority list is arranged in order from low to high as follows: upper limit constraint of water tank level, lower limit constraint of water tank level, lower limit constraint of end user water supply pressure, and upper limit constraint of end user water supply pressure. S512, when the water supply pressure requirement of the end user is within the range of 0.14 MPa to 0.40 MPa and the liquid level of each water tank does not exceed the preset maximum liquid level H. max With preset minimum liquid level H min Under the condition that no feasible solution exists, the constraint relaxation mechanism is triggered; S513. According to the order of the constraint relaxation priority list, relax the restriction range of the current priority constraint in turn, wherein the upper limit constraint of the water tank liquid level is relaxed to the preset highest liquid level H. max The lower limit constraint of the water tank level is relaxed to the preset minimum level H, ranging from 105% to 110%. min For 90% to 95% of end users, the lower limit constraint on water supply pressure is relaxed to 0.10 MPa to 0.12 MPa, and the upper limit constraint on water supply pressure is relaxed to 0.42 MPa to 0.45 MPa. S514. After relaxing one priority constraint each time, the reconfiguration scheduling scheme is solved again until a feasible solution exists; S515. When a feasible solution exists, the relaxed constraints and the reconfigured scheduling scheme are output as an emergency scheduling scheme, and an emergency scheduling early warning signal is issued at the same time as it is sent to the field programmable logic controller.

Images