Digital twin hydraulic modeling method and system for long-distance water transmission system

Abstract

The digital twin hydraulic modeling method and system of the long-distance water delivery system of the present invention monitors the operation status of the project in real time and accurately. The method includes: step 1, establishing a physical model according to the geographic information and set size of the actual project; step 2, constructing a hydraulic model, and continuously loading real-time observation data into the hydraulic model for the operation and correction of the hydraulic model to form a self-correcting hydraulic model; regularly correcting the upstream water level simulated by the model through the real-time observed downstream water level, and then correcting the gate flow; adding the hydraulic model operation logic of the self-correction link; step 3, using the observation data to correct the gate flow coefficient and roughness of the self-correcting hydraulic model, and comparing the model output value with the observation data to adjust the parameters online to form a digital twin hydraulic model; step 4: comparing the root mean square error of the simulated water level and flow of each channel pool, identifying abnormal conditions under the operating state, and locating the fault from the two dimensions of time and space.

Description

Technical Field

[0001] The present invention belongs to the technical field of automated dispatching of water conservancy systems, and in particular relates to a digital twin hydraulic modeling method and system for a long-distance water transfer system. Background Art

[0002] Water resources are strategic resources. Water shortage will restrict regional development and further affect local economy, science and technology, life, etc. In view of the current problems, it is urgent to improve the level of water resources management and promote the rational and efficient allocation of water resources.

[0003] The construction of long-distance water diversion projects, especially inter-basin water diversion, is an effective engineering measure to solve the problems of regional shortage of water resources and uneven resource distribution. Therefore, long-distance water diversion projects are conducive to the development of agriculture, industry, etc., and alleviate water use conflicts. Since large-scale water diversion projects benefit the general public, the safety and efficiency of water supply have attracted much attention. Long-distance water transfer systems generally include structures such as open channels for water transfer, control gates, and water-passing buildings. The hydraulic process is complex and the system is large in scale. In addition, the water transfer system supplies different water users. As the span of the channel increases, the number of users along the channel system increases, and the water intake needs are different. Therefore, the difficulty of channel operation and scheduling management continues to increase.

[0004] Current channel management is mostly based on the reflection of sensor observation data and scheduling plans formulated by manual experience. However, the number of sensors is limited and the collection interval is fixed, making it difficult to reflect the real-time overall status of channel operation in the time and space dimensions. Therefore, to achieve real-time monitoring and prediction of the channel system, it is necessary to rely on reliable high-precision model simulation, and real-time data assimilation is required to reduce the accumulation of simulation errors. In addition, without model updates and corrections, the uncertainty of the model will mislead the system state estimation. Therefore, operational problems may remain undetected until they develop into catastrophic accidents. In order to avoid the above problems, it is advocated to apply digital twin technology (Digital Twin, DT) in water conservancy projects.

[0005] For the construction of digital twin hydraulic models of long-distance water diversion projects, the technical problem that technical personnel in this field need to urgently solve is to propose an integrated digital twin hydraulic modeling framework for long-distance water diversion systems. The framework needs to meet the following requirements: 1) It includes various components such as control gates, open channels, sensors, and establishes connections and interactions between different components; 2) Considering the uncertainties of hydraulic changes, model parameters and measurement errors, a high-precision real-time modeling and updating simulation method is proposed; 3) The important links in the framework are clarified, and the steps and applications of each link are provided based on water conservancy projects and hydraulic processes, so as to provide guidance and reference for the construction of digital twin platforms for water diversion projects. Summary of the invention

[0006] In order to solve the above problems, the present invention provides a digital twin hydraulic modeling method and system for long-distance water supply systems, which can monitor the operation status of the project in real time and accurately, and comprehensively reflect the actual operation status of the channel.

[0007] The present invention adopts the following scheme:

[0008] <Method>

[0009] The present invention provides a digital twin hydraulic modeling method for a long-distance water delivery system, comprising the following steps:

[0010] Step 1: Establish a physical model based on the geographic information and assembly dimensions of the actual project;

[0011] Step 2: construct a hydraulic model based on the physical model, and continuously load the real-time observation data into the hydraulic model for operation and correction of the hydraulic model to form a self-correcting hydraulic model;

[0012] The upstream water level simulated by the model is regularly corrected by the real-time observed downstream water level, thereby correcting the flow through the gate; the correction interval should be equal to the time interval for collecting observation data and greater than the simulation time step; the hydraulic model operation logic with the addition of the self-correction link is shown in the following formula:

[0013] F(Z d ,q,E,Δt,ΔT,Δs)=(Z u , Q g ) (Formula 1)

[0014] Where: F(·) is the operation logic of the self-correcting hydraulic model; Z d is the downstream water level of each channel pool; q is the water intake flow of the water diversion port; E is the gate opening; Δt is the simulation time step; ΔT is the self-correction time interval; Δs is the simulation space step; Z u is the upstream water level of each channel pool; Q g The flow rate through the upstream control gate;

[0015] Step 3: Model parameter adjustment: Continue to use the observed data to adjust the gate flow coefficient C of the self-calibrated hydraulic model. d The model output value is compared with the observed data to adjust the parameters online and form a digital twin hydraulic model.

[0016] Step 4: Monitoring and prediction:

[0017] Based on the observed data and the output values ​​of the digital twin hydraulic model, the root mean square errors of the simulated water level and flow of each channel pool are compared, abnormal conditions under the operating state are identified, and the faults are located in both time and space dimensions.

[0018] Preferably, in the digital twin hydraulic modeling method for the long-distance water transfer system provided by the present invention, in step 2, the boundary conditions are set as follows: the upstream boundary condition is the water level-flow relationship [Q=f(Z)] considering the gate action, and the downstream boundary condition is the dynamic water level boundary [Z(t)]; the observed gate openings and the downstream water levels of each channel pool are input into the model to simulate and correct the upstream water levels of each channel pool and the through-gate flow of each gate; the water diversion flow of each diversion outlet is the average value of the observed values ​​during the simulation duration; if the relevant observation data is missing, the total water intake is the flow difference between the two adjacent gates, and is distributed to each diversion gate according to the design flow of each diversion outlet.

[0019] Preferably, in the digital twin hydraulic modeling method for a long-distance water delivery system provided by the present invention, in step 3, based on the self-correcting hydraulic model, C d and n joint parameter adjustment, for the parameter combination (C d ,n) to search for the best, the model simulation error is the minimum as the optimization direction; the parameter combination is searched from the upstream to the downstream of the channel;

[0020]

[0021] δ target =min{δ1,δ2,…δ k ,…δ j} (Formula 3)

[0022]

[0023] Where: k is the model error corresponding to the kth parameter combination; δ target is the optimization target; MAPE is the relative error percentage; subscripts H and Q represent water level and flow rate, respectively; (C dk ,n k ) is the kth parameter combination; j is the total number of parameter combinations that satisfy the three constraints at the end of the calibration phase; N is the total number of observations or simulations; x o,i is the ith observation, x s,i is the ith simulation value.

[0024] Three constraint optimization conditions: 1) limit the average relative error percentage range of water level and flow; 2) limit the average absolute error of water level; 3) limit the average absolute error of flow. The parameter combination that meets the constraints enters the optimization range, and the number of optimization iterations and termination time depend on the computing space and parameter optimization period during the project operation.

[0025] Preferably, in the digital twin hydraulic modeling method for the long-distance water transfer system provided by the present invention, in step 4, the parameter combination finally obtained by optimal search in the parameter adjustment section is substituted, and the hydraulic changes are simulated using the digital twin hydraulic model; the physical project is monitored based on the water level and flow changes simulated by the model; and according to the hydrodynamic changes of any monitored channels, pools and gates, the project scheduling situation is determined through data analysis and possible abnormal situations are indicated.

[0026] Preferably, in the digital twin hydraulic modeling method for a long-distance water transfer system provided by the present invention, in step 4, first, the root mean square error of the simulated water level and flow of each channel pool is calculated (two variables for abnormal channel pool detection), and the channel pool where the root mean square error of the water level or flow (either variable) exceeds the first threshold is judged as a channel pool where an abnormality may exist, and the first threshold is set to 4-5% of the water level and flow value (corresponding to the variable to be judged) in the operating state; then, the time of occurrence is detected according to the error variance of the water level and flow (two variables for abnormal section detection), and the time period when the error variance of the water level or flow (either variable) exceeds the second threshold is the abnormal section, and the second threshold is set to 3-6 times the average value of the error variance of the water level and flow (corresponding to the variable to be judged); according to the consistency of the water level and flow variance, the abnormal situation is divided into sensor failure and gate misoperation. If the water level and flow have a synchronous error exceeding the second threshold, it is judged as a gate misoperation. If only one of the variables exceeds the second threshold, it is judged as a sensor failure.

[0027] Preferably, in the digital twin hydraulic modeling method for a long-distance water transfer system provided by the present invention, in step 4, abnormal section detection is performed only for the channel pool judged to be likely to have abnormalities. Preferably, in the digital twin hydraulic modeling method for a long-distance water transfer system provided by the present invention, in step 4, the first threshold is set to 5% of the water level and flow value under the operating state, and the second threshold is set to 5 times the average value of the water level and flow.

[0028] Preferably, the digital twin hydraulic modeling method for the long-distance water transmission system provided by the present invention adopts the 3σ criterion and the dimensionless flow formula to jointly detect the outliers of the observed data, and removes and interpolates the outliers before using them in the model.

[0029] <System>

[0030] Furthermore, the present invention also provides a full flow simulation system, which can automatically implement the above <method>, including:

[0031] The pre-processing department builds a physical model based on the geographic information and set size of the actual project;

[0032] The self-correction unit builds a hydraulic model based on the physical model and continuously loads real-time observation data into the hydraulic model for operation and correction of the hydraulic model to form a self-correction hydraulic model;

[0033] The upstream water level simulated by the model is regularly corrected by the real-time observed downstream water level, and then the simulation of the flow through the gate is corrected; the correction interval should be equal to the time interval of observation data collection and greater than the simulation time step; the hydraulic model operation logic with added self-correction link is as follows:

[0034] F(Z d ,q,E,Δt,ΔT,Δs)=(Z u , Q g ) (Formula 1)

[0035] Where: F(·) is the operation logic of the self-correcting hydraulic model; Z d is the downstream water level of each channel pool; q is the water intake flow of the water diversion port; E is the gate opening; Δt is the simulation time step; ΔT is the self-correction time interval; Δs is the simulation space step; Z u is the upstream water level of each channel pool; Q g The flow rate through the upstream control gate;

[0036] The digital twin hydraulic model construction department uses observation data to adjust the gate flow coefficient C of the self-correcting hydraulic model. d The model output value is compared with the observed data to adjust the parameters online and form a digital twin hydraulic model.

[0037] The monitoring and prediction department compares the root mean square error of the simulated water level and flow of each channel pool based on the observed data and the output value of the digital twin hydraulic model, identifies abnormal conditions under the operating state, and locates faults in both time and space dimensions;

[0038] The control unit is communicatively connected with the preprocessing unit, the self-correction unit, the digital twin hydraulic model construction unit, and the monitoring and prediction unit to control their operations.

[0039] Preferably, the open-flow simulation system provided by the present invention may also include: a scheduling unit, which is communicated with the control unit and generates a channel operation scheduling plan based on the simulation results of the digital twin hydraulic model; an early warning unit, which is communicated with the control unit and generates corresponding warning information to prompt the user based on the abnormal conditions and fault locations identified by the monitoring and prediction unit.

[0040] Preferably, the bright full flow simulation system provided by the present invention may further include: an input display unit, which is communicatively connected with the control unit and is used to allow the user to input operation instructions and perform corresponding displays.

[0041] Preferably, the open full flow simulation system provided by the present invention, in the digital twin hydraulic model construction unit, based on the self-correcting hydraulic model, converts C d and n joint parameter adjustment, for the parameter combination (C d ,n) to search for the best, the model simulation error is the minimum as the optimization direction; the parameter combination is searched from the upstream to the downstream of the channel;

[0042]

[0043] δ target =min{δ1,δ2,…δ k ,…δ j} (Formula 3)

[0044]

[0045] Where: k is the model error corresponding to the kth parameter combination; δ target is the optimization target; MAPE is the relative error percentage; subscripts H and Q represent water level and flow rate, respectively; (C dk ,n k ) is the kth parameter combination; j is the total number of parameter combinations that satisfy the constraints at the end of the calibration phase; N is the total number of observations or simulations; x o,i is the i-th observation value; x s,i is the ith simulation value.

[0046] Functions and Effects of the Invention

[0047] (1) The present invention constructs a digital twin hydraulic model for the hydrodynamic process. The upstream water level simulated by the model is regularly corrected by the real-time observation of the downstream water level, and then the flow rate through the gate is corrected. This makes up for the deficiency of sensor data collection in terms of time and space density, and can monitor the operation status of the project in real time and accurately, and comprehensively reflect the actual operation status of the channel.

[0048] (2) The present invention can indicate abnormal conditions in a timely manner, locate channels or gates where failures may occur, and further locate the failure period. This is conducive to improving the management level and dispatch response speed of long-distance water supply systems, and provides preliminary exploration and reference for the construction and application of digital twin watersheds.

[0049] (3) The present invention can reflect the current real state of the system and predict the changing trend of the system in the short term. It can be used for comprehensive monitoring and prediction of water delivery systems, such as simulation of water conditions along the line, testing of scheduling schemes, inversion of hydraulic transition processes, etc. It can also be used to indicate abnormal conditions under real-time operation and avoid false alarms based on observation data alone, such as sensor failure, gate misoperation, etc. It can also provide reference for other large-scale water diversion projects with a lot of observation data, and guide the parallel development of virtual models and actual projects to ensure the safety and efficiency of daily scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a schematic diagram of the digital twin hydraulic modeling process involved in the first embodiment of the present invention;

[0051] Figure 2 for Figure 1 A magnified view of the preparation stage in middle I;

[0052] Figure 3 for Figure 1 Enlarged view of the II preheating section;

[0053] Figure 4 for Figure 1 A magnified picture of the online parameter adjustment of the middle III;

[0054] Figure 5 This is a parameter optimization flow chart involved in the first embodiment of the present invention;

[0055] Figure 6 This is a flowchart of abnormal situation identification involved in the first embodiment of the present invention;

[0056] Figure 7 for Figure 1 A zoomed-in diagram comparing the measured and predicted values ​​of the water level and flow rate change lines in the middle IV;

[0057] Figure 8 for Figure 1 A magnified view of the abnormal channel pool located (identified) in the middle IV;

[0058] Fig. 9 for Figure 1 A magnified view of the abnormal period of IV positioning (identification). DETAILED DESCRIPTION

[0059] The specific implementation scheme of the digital twin hydraulic modeling method and system for the long-distance water delivery system involved in the present invention is described in detail below with reference to the accompanying drawings.

[0060] <Example 1>

[0061] like Figure 1 As shown, the digital twin hydraulic modeling method provided in this embodiment is specifically as follows:

[0062] 1. Data Preprocessing and Physical Modeling

[0063] Physical modeling involves open channel cross-sectional dimensions, channel bottom elevation, gate dimensions and structure, water-passing structures and their connections. The raw data collected by the sensors are preprocessed based on statistical methods and hydraulic relationships.

[0064] like Figure 2 As shown in the figure, the raw data collected by the sensor is first preprocessed. Data preprocessing includes three steps: deleting outliers, filling missing values, and denoising. For the identification of outliers, the 3σ criterion and the dimensionless flow formula are used to jointly detect the outliers of the water level and flow observations, and the outliers are removed and interpolated. For the processing of outliers jointly judged by formulas (1) and (2), the hydraulic characteristics of the channel need to be considered to avoid misjudgment.

[0065] The 3σ criterion requires calculating the standard deviation (σ) of the data sample, removing the samples outside [x-3σ, x+3σ]. The calculation formula for σ is as follows:

[0066]

[0067] Where N is the total number of data points; x i is the i-th data; is the average value of i data samples.

[0068] At the same time, the dimensionless flow formula is used to measure the relationship between the flow and water depth at each control gate, and the value that deviates from the relationship curve by more than ±5% is considered an outlier. The dimensionless flow formula is as follows:

[0069]

[0070] Where K is the critical water depth, m; E is the gate opening, m; H E is the cross-sectional energy head, m; a and c are constants in the formula, which need to be obtained by fitting the observed data.

[0071] Based on the continuity of the hydraulic transition process, this scheme proposes: 1) If the outlier lasts for more than 12 hours, it will be retained and regarded as an abnormal event in this scheme; 2) If the outlier lasts for less than 12 hours and is an isolated point, the deleted outlier is interpolated with two adjacent data points. 3) If the outlier lasts for less than 12 hours and is a number of consecutive points, the deleted outlier is linearly interpolated according to the nearest observed data points.

[0072] After removing and supplementing outliers, the data should be denoised. There are many mature denoising methods, such as filters, time series decomposition, etc., which can be selected according to actual conditions. The processed data is used for subsequent model simulation, calibration and comparative analysis.

[0073] In addition, when physically modeling long-distance water supply systems, the size of the model and the setting of calculation sections and nodes should be fully considered. This will significantly affect the calculation accuracy and operation speed. It is recommended to select obvious section changes and large buildings as nodes. When modeling the system, their detailed structures and hydraulic transitions can be ignored, and local head losses can be applied to focus on reflecting the changes in the overall water surface line and the flow through the gate. In addition, if the span is large, in order to reduce the error accumulation caused by the mutual influence between channels and pools, a segmented modeling method can be adopted, based on the flow scale or geographical area, to establish information transmission between different channel sections and improve the model accuracy of the overall system.

[0074] 2. Simulation calculation of self-correction model

[0075] As a core step in the twin framework, the model-related simulation settings are more important. The boundary conditions are set as follows: the upstream boundary condition is the water level-flow relationship considering the gate action [Q=f(Z)], and the downstream boundary condition is the dynamic water level boundary [Z(t)]. The observed gate openings and the downstream water levels of each channel pool are input into the model to simulate and correct the upstream water levels of each channel pool and the flow through each gate. The diversion flow of each diversion outlet is the average of the observed values ​​during the simulation duration. If the relevant observation data is missing, the total water intake can be taken as the flow difference between the two adjacent gates and distributed to each diversion gate according to the design flow of each diversion outlet.

[0076] like Figure 3 As shown, for the calculation of the flow rate through the gate, considering the gate structure and velocity head, the flow formula used is as follows:

[0077]

[0078] In the formula, C d is the gate flow coefficient; L is the gate width, m; H u and H d are the water depths of the sections before and after the gate, m; A u is the water flow area of ​​the section before the gate, m 2 .

[0079] For the calculation of the water level in the open channel section, the one-dimensional Saint-Venant equations are discretized using the four-point difference implicit scheme and solved using the Preissmann method. The Saint-Venant equations are shown in formulas (4) and (5):

[0080]

[0081]

[0082] Where B is the width of the water surface, m; Z is the water level of each section, m; Q is the flow rate of each section, m 3 / s; C is the coefficient of gravity; g is the acceleration due to gravity, m / s 2 ; A is the cross-sectional water flow area, m 2 ; R is the hydraulic radius, m.

[0083] The Xie Cai coefficient C is calculated by the following formula:

[0084]

[0085] Where n is the channel roughness.

[0086] The upstream water level simulated by the model is regularly corrected using the real-time observed downstream water level, thereby correcting the simulation of the flow through the gate. The correction interval should be equal to the time interval of sensor acquisition and greater than the simulation time step, so as to correct the simulation value in time while ensuring numerical stability. The self-correction approach can reduce the simulation error caused by the mutual influence between channels and pools, and correct the change trend. The hydraulic model operation logic with the addition of the self-correction link is shown in the following formula:

[0087] F(Z d ,q,E,Δt,ΔT,Δs)=(Z u , Q g ) Formula (7)

[0088] Where, F(·) is the operation logic of the self-correcting hydraulic model; Z d is the downstream water level of each channel pool, m; q is the water intake flow rate of the water diversion port, m 3 / s; E is the gate opening, m; Δt is the simulation time step, min; ΔT is the self-correction time interval, min; Δs is the simulation space step, m; Z u is the upstream water level of each channel pool, m; Q g is the flow rate through the upstream control gate, m 3 / s;

[0089] From this stage, real-time data is continuously loaded into the model for model operation and correction. Due to the oscillation problem in numerical simulation (especially in the initial stage), the warm-up stage is conducive to the establishment of the model and the stability of the calculation results. The more important work in this stage is to determine the input and output of the hydraulic model and set the model simulation logic.

[0090] 3. Model parameter optimization process

[0091] Two important parameters in the model related to the control gate and open channel need to be adjusted in real time, namely, the gate flow coefficient C d and roughness n. Since most actual projects lack observation data of the middle sections along the channel, the energy loss and hydraulic changes of the local structure can be included in the estimation of the n value when implementing this scheme, that is, the adjusted roughness n includes the head loss along the channel and locally.

[0092] Based on the self-correction model in step 2, the flow coefficient has the same effect on the upstream water level and gate flow, while the roughness has the opposite effect on the upstream water level and gate flow. There are 4 pairs of corresponding relationships between the two coefficients and the two variables, among which, except for the negative correlation between the roughness and the flow, the rest are positively correlated. This scheme proposes to change C d and n joint parameter adjustment, for the parameter combination (C d ,n) to optimize, the model simulation error is the minimum optimization direction. The parameter combination is also searched from the upstream to the downstream of the channel.

[0093] like Figure 4 As shown in the figure, this scheme treats the two parameters as a parameter combination for optimization, with the minimum water level error plus flow error as the optimization goal. The optimization range should meet three constraints: 1) limit the average relative error percentage range of water level and flow (such as 5%); 2) limit the average absolute error of water level (such as 10cm); 3) limit the average absolute error of flow (such as 4m 3 / s). The total number of combinations that meet the requirements is determined by the computing power and tuning time. The parameter combinations that meet the limited conditions enter the optimization range, while the number of optimization iterations and the termination time depend on the computing space and parameter optimization period during the project operation. Due to the coupling effect of the series channel pools, the adjustment of the parameters in a single channel pool will have a slight impact on the hydraulic variables of the adjacent pools. Since the actual tuning duration is limited and the impact is small, this scheme does not need to consider the problem of repeated and over-tuning caused by the coupling effect.

[0094] Guided by the minimum error of water level and flow, C d The simulation error is measured by the relative percentage error (MAPE).

[0095]

[0096] δ target =min{δ1,δ2,…δ k ,…δ j} Formula (9)

[0097]

[0098] In the formula, (C dk ,n k ) is the kth parameter combination; δk is the model error corresponding to the kth parameter combination; j is the total number of parameter combinations that meet the above three conditions at the end of the calibration phase, N is the total number of observations or simulations (the total number of observations = the total number of simulations), x o,i is the ith observation, x s,i is the ith simulation value.

[0099] Figure 5 The flowchart of adjusting two parameters at the same time is shown. In the figure, the subscript o represents the observed value and the subscript s represents the simulated value. The detailed optimization steps are as follows: 1) Set the initial optimization parameter combination (C d ,n) and input it into the set hydraulic model. 2) Compare the simulated water level and flow. If the simulated water level is lower than the observed water level, increase C d and n; otherwise, reduce their values; if the simulated flow rate is less than the observed flow rate, increase C d And reduce n; otherwise reduce C d And increase n. Note that the above steps do not have a clear order of precedence and can be coordinated and carried out simultaneously. 3) Repeat step 2) until the parameter adjustment phase ends. 4) Output all parameter combinations and their corresponding water level and flow errors. The combination with the minimum error is the optimization result, which can be substituted into the model simulation.

[0100] 4. Model Monitoring and Abnormal Situation Identification

[0101] Substituting the parameter combination finally optimized in the parameter adjustment section, the model can start to simulate hydraulic changes. The simulated water level and flow changes can realize the monitoring of the physical project on the virtual platform. The hydrodynamic changes of any channel pool and gate can be presented through simulation data, and data analysis can guide project scheduling and indicate possible abnormal situations.

[0102] like Figures 6 to 9 As shown in the figure, by comparing the simulation data with the observation data and calculating the simulation error of the model, the abnormal situation under the running state can be identified and the fault can be located from the two dimensions of time and space.

[0103] First, if Figures 7-8 As shown in the figure, the root mean square error of the simulated water level and flow of each channel pool is calculated, and the channel pool with a large error is diagnosed as having a possible abnormality. Fig. 9 As shown in the figure, the time of occurrence is detected according to the error variance of water level and flow. The time period with obviously large variance is the abnormal period, and according to the consistency of water level and flow variance, the abnormal situation can be divided into sensor failure, gate misoperation, etc. This method uses the corresponding hydraulic relationship between water level and flow through the gate to detect faults according to the consistency of the changes of the two variables. The calculation formula of the root mean square error is as follows:

[0104]

[0105] Where: x o is the observed data; x s For simulation data.

[0106] In this embodiment, the first threshold is set to 5% of the water level and flow value under the operating state. If the root mean square error of the water level or flow (any variable) exceeds the first threshold, it is judged that there may be an abnormal channel pool; the second threshold is set to 5 times the average value of the water level and flow error variance. The time period when the error variance of the water level or flow (any variable) exceeds the second threshold is an abnormal period. If the error variance of the water level and flow exceeds the second threshold during the same period, it is judged as a gate misoperation. If only one of the variables exceeds the second threshold, it is judged as a sensor failure.

[0107] Figure 6 The model is shown to identify and locate abnormal situations that may occur in the channel system in real time. The specific steps are as follows:

[0108] (1) Calculate the RMSE between the simulated data set and the observed data set in each channel pool, and compare the RMSE of the same period. Channel pools with significantly larger RMSE of water level and gate flow (exceeding the first threshold) are identified as channels where abnormal events may have occurred. On the contrary, if the RMSE is within a relatively stable and low range (not exceeding the first threshold), the entire system can be considered normal.

[0109] (2) The identified abnormal channels and pools were analyzed on the water level and gate flow observation data sets, and the change rate of their observation data within 2 hours was calculated. The possible fault problems were divided into three situations: 1) If one variable in the flow or water level changes abnormally quickly (exceeding the second threshold), it is considered as a sensor acquisition failure; 2) If both variables change abnormally quickly at the same time, it is considered as improper gate operation; 3) If none of the above abnormal events occurs, the large RMSE may be caused by model generalization, and the hydraulic model can be further optimized, and the problem can be eliminated in the real system.

[0110] (3) After locating and diagnosing the abnormal channel pool, the time when the abnormal event occurred can be determined by calculating the error variance between the predicted value and the observed value. The time period with a particularly large value (exceeding the second threshold) is the time period when the abnormal situation occurs. In this embodiment, the search frequency (such as 6h) is selected to calculate the variance of the simulation error.

[0111] After determining the model parameters, hydraulic simulation can be performed according to the control equations, as shown in formulas 4 to 6. In order to measure the fidelity of the model, some indicators are needed to measure the model simulation accuracy. Here, absolute error (AE), absolute percentage error (APE), and mean absolute percentage error (MAPE) can be used for measurement. The calculation formula of MAPE has been mentioned in the previous article, as shown in formula 10. The calculation formulas for AE and APE are as follows:

[0112] AE=x o,i -x s,i Formula (12)

[0113]

[0114] In order to express the difference between the observed value of the real physical system and the simulated value of the twin model, this scheme proposes a new metric: overall similarity (S). This metric is based on the MAPE of water level and flow, and reflects the model similarity in percentage form. The calculation formula is as follows:

[0115]

[0116]

[0117] Where: α, β represent the weights of the two variables of water level and flow, respectively, and α = β = 0.5 is generally taken. m is the channel number; M is the total number of channels.

[0118] If S is greater than 95%, it means that the twin model is highly similar to the real-time state of the system and can continue to run; if S is less than 95%, it means that there is still room for improvement in the fit between the twin model and the real-time state, and you should return to the parameter adjustment stage to re-optimize the parameters before entering the monitoring and prediction stage.

[0119] The digital twin hydraulic framework for long-distance water transfer systems proposed in the present invention effectively combines observation data with hydraulic models to establish a self-correcting twin model that runs synchronously in real time. Based on this invention, the observation data (water level, flow) and operating process (gate opening) are integrated into the model, focusing on simulating real hydraulic processes, and updating the model in real time, which preliminarily realizes the interaction between the virtual model and the real object. The present invention can serve the construction of digital twin watersheds, provide reference ideas for the construction of digital twin hydraulic models, and be applied to the "four predictions": forecast, warning, rehearsal, and plan functions.

[0120] <Example 2>

[0121] Furthermore, in the second embodiment, a digital twin hydraulic system of a long-distance water supply system capable of automatically implementing the above method of the present invention is provided, the system comprising a preprocessing unit, a self-correction unit, a digital twin hydraulic model construction unit, a monitoring and prediction unit, a scheduling unit, an early warning unit, an input display unit, and a control unit.

[0122] The preprocessing part is used to execute the contents described in step 1 above and establish a physical model according to the geographic information and set size of the actual project.

[0123] The self-correction unit is used to execute the contents described in step 2 above, construct a hydraulic model based on the physical model, and continuously load real-time observation data into the hydraulic model for operation and correction of the hydraulic model to form a self-correction hydraulic model.

[0124] The digital twin hydraulic model construction unit is used to execute the content described in step 3 above, using the observation data to calculate the gate flow coefficient C of the self-correcting hydraulic model. d The model output value is compared with the observed data to adjust the parameters online to form a digital twin hydraulic model.

[0125] The monitoring and prediction unit is used to execute the contents described in step 4 above. Based on the observed data and the output values ​​of the digital twin hydraulic model, it compares the root mean square error of the simulated water level and flow of each channel pool, identifies abnormal conditions under the operating state, and locates the fault in both time and space dimensions.

[0126] The dispatching department generates a canal operation dispatching plan based on the simulation results of the digital twin hydraulic model.

[0127] The early warning department generates corresponding warning information and prompts users based on the abnormal conditions and fault locations identified by the monitoring and prediction department.

[0128] The input display unit is used to allow the user to input operating instructions, and display the input, output and intermediate processing data of the corresponding unit in the form of numerical values, tables, graphs, etc. according to the operating instructions, and can display the corresponding data statically or dynamically on a two-dimensional or three-dimensional model.

[0129] The control unit is communicatively connected with the preprocessing unit, self-correction unit, digital twin hydraulic model construction unit, monitoring and prediction unit, scheduling unit, early warning unit, and input and display unit to control their operations.

[0130] The above embodiments are merely examples of the technical solutions of the present invention. The digital twin hydraulic modeling method and system of the long-distance water supply system involved in the present invention are not limited to the contents described in the above embodiments, but are subject to the scope defined by the claims. Any modification, supplement or equivalent replacement made by technicians in the field of the present invention on the basis of this embodiment is within the scope of protection required by the claims of the present invention.

Claims

1. A digital twin hydraulic modeling method for a long-distance water delivery system, characterized in that: include: Step 1: Establish a physical model based on the geographic information and assembly dimensions of the actual project; Step 2: construct a hydraulic model based on the physical model, and continuously load the real-time observation data into the hydraulic model for operation and correction of the hydraulic model to form a self-correcting hydraulic model; The upstream water level simulated by the model is regularly corrected by the real-time observed downstream water level, thereby correcting the flow through the gate; the correction interval should be equal to the time interval for collecting observation data and greater than the simulation time step; the hydraulic model operation logic with the addition of the self-correction link is shown in the following formula: F(Z d , q, E, Δt, ΔT, Δs) = (Z u , Q g ) (Formula 1) Where: F(·) is the operation logic of the self-correcting hydraulic model; Z d is the downstream water level of each channel pool; q is the water intake flow of the water diversion port; E is the gate opening; Δt is the simulation time step; ΔT is the self-correction time interval; Δs is the simulation space step; Z u is the upstream water level of each channel pool; Q g The flow rate through the upstream control gate; Step 3: Model parameter adjustment: Continue to use the observed data to adjust the gate flow coefficient C of the self-calibrated hydraulic model. d The model output value is compared with the observed data to adjust the parameters online and form a digital twin hydraulic model. Step 4: Monitoring and prediction: Based on the observed data and the output values ​​of the digital twin hydraulic model, the root mean square errors of the simulated water level and flow of each channel pool are compared, abnormal conditions under the operating state are identified, and the faults are located in both time and space dimensions.

2. The digital twin hydraulic modeling method for a long-distance water delivery system according to claim 1, characterized in that: in, In step 2, the upstream boundary condition is the water level-flow relationship considering the gate action [Q=f(Z)], and the downstream boundary condition is the dynamic water level boundary [Z(t)]; the observed gate openings and downstream water levels of each channel pool are input into the model to simulate and correct the upstream water levels of each channel pool and the flow through each gate; the diversion flow of each diversion outlet is the average value of the observed values ​​during the simulation duration; if relevant observation data are missing, the total water intake is taken as the flow difference between the two adjacent gates, and is distributed to each diversion gate according to the design flow of each diversion outlet.

3. The digital twin hydraulic modeling method for a long-distance water delivery system according to claim 1, characterized in that: in, In step 3, based on the self-correcting hydraulic model, C d and n joint parameter adjustment, for the parameter combination (C d ,n) to search for the best, the model simulation error is the minimum as the optimization direction; the parameter combination is searched from the upstream to the downstream of the channel; δ target = min{δ1, δ2, … δ k , … δ j}(Formula 3) Where: k is the model error corresponding to the kth parameter combination; δ target is the optimization target; MAPE is the relative error percentage; subscripts H and Q represent water level and flow rate, respectively; (C dk ,n k ) is the kth parameter combination; j is the total number of parameter combinations that satisfy the constraints at the end of the calibration phase; N is the total number of observations or simulations; x o,i is the i-th observation value; x s,i is the ith simulation value.

4. The digital twin hydraulic modeling method for a long-distance water delivery system according to claim 1, characterized in that: in, In step 4, the parameter combination finally obtained by optimization is substituted into the parameter adjustment section, and the hydraulic changes are simulated by the digital twin hydraulic model; the physical project is monitored based on the water level and flow changes simulated by the model; according to the hydrodynamic changes of any monitored channels, pools and gates, the project scheduling situation is determined through data analysis and possible abnormal situations are indicated.

5. The digital twin hydraulic modeling method for a long-distance water delivery system according to claim 1, characterized in that: in, In step 4, first, the root mean square error of the simulated water level and flow of each channel pool is calculated, and the channel pool with a water level or flow root mean square error exceeding the first threshold is judged as a possible abnormality. The first threshold is set to 4-5% of the water level and flow value under the operating state; then, the time of occurrence is detected according to the error variance of the water level and flow, and the time period when the error variance of the water level or flow exceeds the second threshold is the abnormal period, and the second threshold is set to 3-6 times the average value of the water level and flow error variance; according to the consistency of the water level and flow variance, the abnormal situation is divided into sensor failure and gate misoperation. If the error variance of the water level and flow exceeds the second threshold during the same period, it is judged as a gate misoperation. If only one of the variables exceeds the second threshold, it is judged as a sensor failure.

6. The digital twin hydraulic modeling method for a long-distance water delivery system according to claim 4, characterized in that: in, In step 4, the first threshold is set to 5% of the water level and flow value in the operating state, and the second threshold is set to 5 times the average value of the water level and flow.

7. The digital twin hydraulic modeling method for a long-distance water delivery system according to claim 1, characterized in that: in, For the observed data, the 3σ criterion and the dimensionless flow formula are used to jointly detect the outliers of the observed values, and the outliers are eliminated and interpolated before being used in the model.

8. Digital twin hydraulic system for long-distance water delivery system, characterized by: include: The pre-processing department builds a physical model based on the geographic information and set size of the actual project; The self-correction unit builds a hydraulic model based on the physical model and continuously loads real-time observation data into the hydraulic model for operation and correction of the hydraulic model to form a self-correction hydraulic model; The upstream water level simulated by the model is regularly corrected by the real-time observed downstream water level, and then the simulation of the flow through the gate is corrected; the correction interval should be equal to the time interval of observation data collection and greater than the simulation time step; the hydraulic model operation logic with added self-correction link is as follows: F(Z d , q, E, Δt, ΔT, Δs) = (Z u , Q g ) (Formula 1) Where: F(·) is the operation logic of the self-correcting hydraulic model; Z d is the downstream water level of each channel pool; q is the water intake flow of the water diversion port; E is the gate opening; Δt is the simulation time step; ΔT is the self-correction time interval; Δs is the simulation space step; Z u is the upstream water level of each channel pool; Q g The flow rate through the upstream control gate; The digital twin hydraulic model construction department uses observation data to adjust the gate flow coefficient C of the self-correcting hydraulic model. d The model output value is compared with the observed data to adjust the parameters online and form a digital twin hydraulic model. The monitoring and prediction department compares the root mean square error of the simulated water level and flow of each channel pool based on the observed data and the output value of the digital twin hydraulic model, identifies abnormal conditions under the operating state, and locates faults in both time and space dimensions; The control unit is communicatively connected with the preprocessing unit, the self-correction unit, the digital twin hydraulic model construction unit, and the monitoring and prediction unit to control their operations.

9. The digital twin hydraulic system of the long-distance water delivery system according to claim 8, characterized in that: Also includes: The dispatching department is connected to the control department in communication and generates a canal operation dispatching plan based on the simulation results of the digital twin hydraulic model; The early warning unit is connected to the control unit in communication, and generates corresponding warning information to prompt the user based on the abnormal conditions and fault locations identified by the monitoring and prediction unit; The input display unit is connected to the control unit for communication and is used to allow the user to input operation instructions and display them accordingly.

10. The digital twin hydraulic system of the long-distance water delivery system according to claim 8, characterized in that: in, In the digital twin hydraulic model construction department, based on the self-correcting hydraulic model, C d and n joint parameter adjustment, for the parameter combination (C d ,n) to search for the best, the model simulation error is the minimum as the optimization direction; the parameter combination is searched from the upstream to the downstream of the channel; δ target = min{δ1, δ2, … δ k , … δ j}} (Formula 3) Where: k is the model error corresponding to the kth parameter combination; δ target is the optimization target; MAPE is the relative error percentage; subscripts H and Q represent water level and flow rate, respectively; (C dk ,n k ) is the kth parameter combination; j is the total number of parameter combinations that satisfy the constraints at the end of the calibration phase; N is the total number of observations or simulations; x o,i is the i-th observation value; x s,i is the ith simulation value.

Images