A non-hysteresis warehouse entry flow backstepping method, device and electronic equipment

Abstract

The application provides a non-hysteresis reservoir inflow flow backstepping method and device and electronic equipment, wherein the method comprises the following steps: collecting original data of a target reservoir; constructing a one-dimensional hydrodynamic model based on the original data of the target reservoir, and verifying the model accuracy of the one-dimensional hydrodynamic model; determining a prior set of the inflow flow of the target reservoir, and obtaining a simulated water level set and flow information propagation delay through the one-dimensional hydrodynamic model; obtaining an observation sample set of the target reservoir, and obtaining a posterior value of the inflow flow through a forward filtering step of the Kalman smoothing method based on the observation sample set; updating the posterior value of the inflow flow at the last time through a backward filtering step of the Kalman smoothing method, and obtaining a smooth inflow flow value without hysteresis. Through the application, the flow propagation delay is fully considered, and the problem of poor estimation accuracy of the inflow flow in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of reservoir scheduling technology, and in particular to a method, apparatus and electronic equipment for back-calculating inflow without time delay. Background Technology

[0002] Calculating inflow is one of the most fundamental tasks in reservoir operation. Inflow data is essential for developing reservoir flood forecasting and operation plans, compiling reservoir operation maps, conducting economic evaluations of reservoir operation, and performing flood regulation calculations. Currently, methods for estimating reservoir inflow include the water balance method and the hydraulic method. Water smoothing methods require long calculation time intervals, limiting inflow calculations to hourly intervals, and are significantly affected by water level observation errors, necessitating post-processing with alternative smoothing algorithms or optimization methods. Hydraulic methods can continuously simulate water surface processes and effectively address the dynamic storage capacity problem of river-type reservoirs, but require the integration of optimization and data assimilation methods.

[0003] However, the results estimated by these existing methods in large river-type reservoirs all have time lag. Since water level stations are generally close to dams, in large river-type reservoirs, the flow entering the river-type reservoir from the upstream reservoir boundary takes a period of time to affect the downstream. During this period, the flow is flattened, and the flow information contained in the water level is dispersed across multiple time points. Relying solely on the water level information at the current time cannot accurately estimate the inflow.

[0004] Chinese invention patent CN119168254A discloses a method for back-calculating inflow based on ensemble Kalman filtering and hydrodynamic simulation. This method fully considers the impact of dynamic reservoir capacity caused by uneven water surface during flood evolution, thus improving the accuracy of inflow calculation. However, this scheme uses ensemble Kalman filtering to calculate inflow, which can only estimate inflow using current information and cannot solve the estimation failure and time lag problems when the observation water level station is too far upstream.

[0005] There is currently no effective solution to the problem of poor accuracy in estimating inbound flow in existing related technologies. Summary of the Invention

[0006] This invention provides a method, apparatus, and electronic device for back-calculating inbound flow without delay, in order to solve the shortcomings of poor accuracy in estimating inbound flow in existing related technologies.

[0007] In a first aspect, the present invention provides a method for back-calculating inbound flow without delay, comprising:

[0008] Collect raw data of the target reservoir; the raw data includes the upstream river topography, water level station locations and monitoring data, and inflow data of the target reservoir;

[0009] Based on the original data of the target reservoir, a one-dimensional hydrodynamic model is constructed, and the accuracy of the one-dimensional hydrodynamic model is verified.

[0010] The prior set of inflow to the target reservoir is determined, and the simulated water level set and the propagation lag of flow information are obtained through the one-dimensional hydrodynamic model.

[0011] Obtain the observation sample set of the target reservoir, and use the forward filtering step of the Kalman smoothing method to obtain the posterior value of the inflow rate based on the observation sample set;

[0012] The backward filtering step based on the Kalman smoothing method updates the posterior value of the inflow rate at the previous time step, resulting in a smooth inflow rate value without lag.

[0013] According to the present invention, a method for back-calculating inflow without time delay is provided, which collects raw data of a target reservoir, including:

[0014] Collect measured cross-sectional data of the river channel from the dam site to the upstream backwater area of ​​the target reservoir;

[0015] Obtain reference values ​​and data ranges for the channel roughness data of the target reservoir;

[0016] Obtain the historical inflow data reported by the target reservoir.

[0017] According to the present invention, a method for back-calculating inflow without time delay is provided, based on the original data of the target reservoir, to construct a one-dimensional hydrodynamic model, including:

[0018] The evolution characteristics of the target reservoir are described by the Saint-Venant equations, which include the continuity equation and the momentum equation.

[0019] The Pressman four-point eccentric implicit scheme is used to solve for the flow and water level at all cross-sections of the target reservoir river.

[0020] The one-dimensional hydrodynamic model is generated by combining the measured cross-sectional data of the target reservoir.

[0021] According to the present invention, a method for back-calculating inflow into a reservoir without delay is provided, which determines the prior set of inflow into the target reservoir and obtains the simulated water level set and the propagation hysteresis of flow information through the one-dimensional hydrodynamic model, including:

[0022] The state transition equation for the inflow of the target reservoir is established using Gaussian random walk, and the prior set of the inflow is generated.

[0023] By repeatedly executing the one-dimensional hydrodynamic model, the propagation lag of the simulated water level set and flow information of the target reservoir is obtained.

[0024] According to the present invention, a method for back-calculating inflow without delay is provided, which involves repeatedly executing the one-dimensional hydrodynamic model to obtain the simulated water level set and flow information propagation lag of the target reservoir, including:

[0025] Input the sequence with the same historical inflow data but different current inflow set members into the one-dimensional hydrodynamic model;

[0026] The time lag is estimated by using the one-dimensional hydrodynamic model, and a set of simulated water levels that can reflect the time lag is constructed.

[0027] According to the present invention, a method for back-calculating inflow without lag is provided, which obtains an observation sample set of the target reservoir and uses a forward filtering step of the Kalman smoothing method to obtain the posterior value of the inflow based on the observation sample set, including:

[0028] Based on the observed water level data and white noise of the target reservoir, an observation sample set is generated;

[0029] Based on the state transition equation and the observation equation, the posterior value of the inflow rate is obtained by using the forward filtering step of the Kalman smoothing method; the state transition equation includes the parametric state transition equation and the system state transition equation.

[0030] According to the present invention, a time-delay-free inflow back-calculation method is provided, which obtains the posterior value of the inflow flow based on the state transition equation and the observation equation, and using the forward filtering step of the Kalman smoothing method, including:

[0031] Determine the Kalman gain of the inflow rate to the target reservoir;

[0032] Update the inflow set based on the Kalman gain, the observed sample set, and the simulated water level set;

[0033] The average value of the inbound flow set is used as the posterior value of the inbound flow.

[0034] According to the present invention, a method for back-calculating inbound flow rate without lag is provided, which updates the posterior value of the inbound flow rate at the previous time step based on the backward filtering step of the Kalman smoothing method to obtain a smoothed inbound flow rate value without lag, comprising:

[0035] Based on the covariance matrix of the posterior and prior values ​​of the inflow to the target reservoir, the Kalman smoothing gain of the target reservoir is determined.

[0036] By combining the update window length during backward filtering in the Kalman smoothing method, the lag-free smooth inflow value of the target reservoir is obtained.

[0037] Secondly, the present invention also provides a device for reverse calculation of inbound flow rate without delay, comprising:

[0038] The acquisition module is used to collect raw data of the target reservoir; the raw data includes the upstream river topography, water level station locations and monitoring data, and inflow data of the target reservoir.

[0039] A construction module is used to construct a one-dimensional hydrodynamic model based on the original data of the target reservoir, and to verify the accuracy of the one-dimensional hydrodynamic model.

[0040] The processing module is used to determine the prior set of inflow to the target reservoir and to obtain the simulated water level set and the propagation lag of flow information through the one-dimensional hydrodynamic model.

[0041] The forward filtering module is used to obtain the observation sample set of the target reservoir and to obtain the posterior value of the inflow based on the forward filtering step of the Kalman smoothing method using the observation sample set.

[0042] The backward filtering module is used to update the posterior value of the inflow rate at the previous time step based on the Kalman smoothing method, so as to obtain the smooth inflow rate value without lag.

[0043] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the time-delayed inbound flow reverse calculation method as described in the first aspect above.

[0044] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the time-delayed inbound flow reverse calculation method as described in the first aspect above.

[0045] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the time-delayed inbound traffic reverse calculation method as described in the first aspect above.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The inflow estimation method without lag provided by this invention fully considers the problems of inaccurate inflow estimation caused by flow propagation lag, water surface slope, and observation errors. Compared with existing technologies, this method innovatively utilizes a water level prior set considering lag and the Kalman smoothing method to solve the problem of inaccurate inflow estimation caused by flow propagation lag. This method is applicable to long rivers, large river-type reservoirs, and the need for inflow estimation based on fine time-scale flow data. It uses minute-level measured water level data for data assimilation, fully utilizing data information from multiple time points to dynamically correct the flow, which can well reconstruct the flow process and enhance the real-time performance of flow estimation. The inflow process estimated by this method has the characteristics of small fluctuations, good continuity, and no lag, which can better guide reservoir scheduling and improve forecast accuracy as basic data for forecasting. It achieves a breakthrough from "small rivers" to "large river-type reservoirs," and from "hourly" to "minute-level" inflow estimation, solving the problem of poor accuracy in inflow estimation in existing related technologies. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0049] Figure 1 This is a flowchart of the time-delay-free inbound flow reverse estimation method provided by the present invention;

[0050] Figure 2 This is a schematic diagram of the verification results of the one-dimensional hydrodynamic model in an embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram of the experimental results obtained in July 2020 using the analytical optimization method;

[0052] Figure 4 This is a schematic diagram of the experimental results obtained using this method in July 2020;

[0053] Figure 5 This is a schematic diagram of the experimental results obtained in October 2022 using the analytical optimization method;

[0054] Figure 6 This is a schematic diagram of the experimental results obtained using this method in October 2022;

[0055] Figure 7 This is a structural block diagram of the time-delay-free inbound flow reverse estimation device provided by the present invention;

[0056] Figure 8This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0058] This invention provides a method for back-calculating inbound flow without time delay. Figure 1 This is a flowchart of the time-delay-free inbound flow reverse estimation method provided by the present invention, as follows: Figure 1 As shown, the method includes the following steps:

[0059] Step S101: Collect raw data of the target reservoir; the raw data includes the upstream river topography, water level station locations and monitoring data, and inflow data of the target reservoir.

[0060] Step S102: Based on the original data of the target reservoir, construct a one-dimensional hydrodynamic model and verify the accuracy of the one-dimensional hydrodynamic model.

[0061] Step S103: Determine the prior set of inflow to the target reservoir, and obtain the simulated water level set and the propagation lag of flow information through a one-dimensional hydrodynamic model;

[0062] Step S104: Obtain the observation sample set of the target reservoir, and obtain the posterior value of the inflow rate based on the forward filtering step of the Kalman smoothing method using the observation sample set.

[0063] Step S105: Update the posterior value of the inflow rate at the previous time step using the backward filtering step based on the Kalman smoothing method to obtain the smoothed inflow rate value without lag.

[0064] For example, firstly, the raw data of the target reservoir is collected and organized. Then, a one-dimensional hydrodynamic model is constructed based on the raw data, using the measured water level data and historical runoff data of the target reservoir as ground truth to verify the model's accuracy. Next, a priori set of inflow is generated, and by executing the one-dimensional hydrodynamic model, the simulated water level set of the target reservoir and the propagation lag of flow information are obtained. Then, the observation sample set of the target reservoir is obtained, and based on the current observation data, the forward filtering step of the Kalman smoothing method is used to obtain the posterior value of the inflow. Finally, the backward filtering step of the Kalman smoothing method is used to update the posterior value of the inflow at the previous moment, obtaining a lag-free smoothed inflow value. In practical applications, steps S103-S105 can be repeated based on newly observed water level data to achieve real-time updates of the inflow.

[0065] This method fully considers the problems of inaccurate inflow estimation caused by flow propagation lag, water surface slope, and observation errors. Compared with existing technologies, this method innovatively utilizes a water level prior set considering lag and the Kalman smoothing method to solve the problem of inaccurate inflow estimation caused by flow propagation lag. This method is applicable to long rivers, large channel-type reservoirs, and the need for back-calculation of flow data at fine time scales. It uses minute-level measured water level data for data assimilation, fully utilizing data information from multiple time points to dynamically correct the flow, which can well reconstruct the flow process and enhance the real-time performance of flow estimation. The inflow process derived by this method has the characteristics of small fluctuations, good continuity, and no lag, which can better guide reservoir operation and improve forecast accuracy as basic data for forecasting. It achieves a breakthrough from "small rivers" to "large channel-type reservoirs," and from "hourly" to "minute-level" inflow estimation, solving the problem of poor accuracy in inflow estimation in existing related technologies.

[0066] Next, we will use the Three Gorges Dam as an example to illustrate the above method. The Three Gorges Reservoir is a typical large river reservoir with a huge dynamic storage capacity. The water level station is far from the upstream inflow section, and its inflow flow can only affect the downstream water level station after being flattened and propagating for several hours.

[0067] In some embodiments, step S101, collecting raw data of the target reservoir, includes: collecting measured cross-sectional data of the river channel from the dam site to the upstream backwater area of ​​the target reservoir; obtaining reference values ​​and data ranges of the river channel roughness data of the target reservoir; and obtaining historical inflow data reported by the target reservoir.

[0068] In this embodiment, the river topographic data refers to underwater topographic data, collected from measured river cross-sections extending from the reservoir dam site to the upstream backwater area, with adjacent river sections spaced 1-3 km apart as one cross-section. Reference values ​​and ranges for river roughness data are obtained. Historical water level data refers to high-temporal-resolution measured water level data obtained from reservoir water level monitoring stations. Inflow data uses historically reported inflow data from the reservoir.

[0069] For example, 287 cross-sections were first collected from the Three Gorges Dam site to the upstream Cuntan section. The distance between adjacent cross-sections was calculated using the center coordinates of the river cross-sections. The data was acquired by the project and included 2-hour inflow and outflow data from 2020 to 2023, measured water level data at 15-minute intervals from 8 water level stations, and the suggested values ​​for river cross-section roughness of 0.03 to 0.08. In cases of missing or incomplete data, linear interpolation was used to extend the data.

[0070] Based on this, in step S102, a one-dimensional hydrodynamic model is constructed based on the original data of the target reservoir, including: describing the evolution characteristics of the target reservoir through the Saint-Venant equations; the Saint-Venant equations include the continuity equation and the momentum equation; using the Pressmann four-point eccentric implicit scheme to solve for the flow and water level of all cross sections of the river in the target reservoir; and combining the measured cross-sectional data of the target reservoir to generate a one-dimensional hydrodynamic model.

[0071] Specifically, the evolution characteristics of river-type reservoir floods are described by the Saint-Venant equations, namely the continuity equation and the momentum equation, as shown in the following formulas:

[0072]

[0073]

[0074] in, A Indicates the cross-sectional area of ​​the water passage. q Indicates river flow. q l This indicates the lateral inflow or outflow per unit length of a river-type reservoir. x and t These represent independent spatial and temporal variables, respectively. g Represents gravitational acceleration. z Indicates the water level at the cross-section. n 1 represents the roughness coefficient. B Indicates the width of the water surface. R Indicates the hydraulic radius, and , Indicates wetted perimeter.

[0075] The Presman four-point eccentric implicit scheme is used to calculate and solve for the flow and water level at all cross-sections of a river in a one-dimensional hydrodynamic model, based on...t At any given time, the flow rate and water level at all cross-sections, and the known boundary conditions ( t The flow rates at the initial and final cross-sections at time +1, i.e., the reservoir flow rates. t (Inbound and outbound flow rates at time +1) are obtained t At time +1, the unknown variables at each cross-section are flow rate and water level. For cases with lateral inflow, a virtual river segment is set up. The basic continuity equation is:

[0076]

[0077] in, Indicates upstream section j The water level at that location Indicates downstream section j Water level at +1 Indicates upstream section j Traffic flow at the location Indicates downstream section j The flow rate at +1 Indicates cross-section j arrive j Side inflow between +1.

[0078] A one-dimensional hydrodynamic model is constructed based on the compiled cross-sectional data of the river section. The roughness values ​​of each section are manually adjusted based on reference values ​​and measured water level data, or calibrated using data assimilation methods or other optimization methods.

[0079] For example, data from the dry season from October 2022 to February 2023 were used to calibrate and validate the one-dimensional hydrodynamic model. The simulation errors of water levels at each section are shown in the figure. Figure 2 , Figure 2 This is a schematic diagram illustrating the verification results of a one-dimensional hydrodynamic model in an embodiment of the present invention. Figure 2 As shown, the average error of the one-dimensional hydrodynamic model is around 0.1, indicating that the results are relatively accurate.

[0080] In some embodiments, step S103, determining the prior set of inflow to the target reservoir and obtaining the simulated water level set and flow information propagation hysteresis through a one-dimensional hydrodynamic model, includes: establishing the state transition equation of the inflow to the target reservoir using a Gaussian random walk and generating the prior set of inflow; executing the one-dimensional hydrodynamic model multiple times to obtain the simulated water level set and flow information propagation hysteresis of the target reservoir.

[0081] In this embodiment, the one-dimensional hydrodynamic model is executed multiple times to obtain the simulated water level set and flow information propagation lag of the target reservoir. This includes: inputting sequences with the same historical inflow data but different current inflow set members into the one-dimensional hydrodynamic model; simulating the time lag through the one-dimensional hydrodynamic model; and constructing a simulated water level set that can reflect the time lag.

[0082] Specifically, taking the inflow rate as a parameter, and assuming the inflow rate is equal at the initial and final times within a short period, the following state transition equation for the inflow rate is obtained using a Gaussian random walk:

[0083]

[0084] in, express t The set at time +1 is the first i Inbound traffic, express t The set mean of inbound flow at time t, with superscript "-" indicating prior value and superscript "+" indicating posterior value. Indicates noise. N Indicates a normal distribution. Indicates the standard deviation of the error. n This represents the size of the set. In the embodiment, the prior set size... n Set to 200, and set the standard deviation of the Gaussian distribution to 200m. 3 / s. If there is no prior value for the inbound traffic at the beginning of the model, an approximate value can be set, and the model will gradually approach the actual value after running.

[0085] Data assimilation aims to correct current state or model parameters by utilizing the difference between observed and simulated values ​​at the current moment. However, due to the inherent time lag and aftereffects in one-dimensional hydrodynamic models, this method is generally not suitable for inflow estimation. After upstream flow enters the channel, the flow signal takes time to propagate to the downstream water level station, and this time lag varies with different flow and water level conditions. Typically, the water level change observed at a certain moment is not caused by the simultaneous inflow, but rather the water level at that section is continuously affected by inflows from multiple previous moments. Therefore, if only the current simulated water level is used to calculate the covariance matrix between the flow set and the simulated water level, the result will be zero.

[0086] To determine the moment when the water level at each cross-section is affected by upstream inflow, two sets of sequences with the same historical inflow data but different current inflow set members are input into a one-dimensional hydrodynamic model. To meet the input requirements of the forward flood evolution model, the inflow at future times is set as the mean of the current inflow set, rather than individual set members, thus ensuring statistical consistency. By comparing the simulated water level sequences at each station, the moment when the water level begins to change can be identified, i.e., the time when the inflow begins to affect that station. The existence of time lag indicates that only historical inflow values ​​can be estimated, because the current inflow signal has not yet reached the station. In this embodiment, before each filtering step, a simulation is performed using the one-dimensional hydrodynamic model to estimate the time lag, and based on this, a simulated water level set reflecting the time lag is constructed as a priori set of the system state for subsequent data assimilation. The specific formula is as follows:

[0087]

[0088]

[0089] in, express t +1 time input the first i The first inbound flow j Simulated water level at each cross section f This represents a one-dimensional hydrodynamic model. m Indicates the number of cross sections. express t Time of the first j Posterior ensemble mean of water level at each cross section t 1 represents the length of the historical inflow. t 2 represents the length of the future inflow; the superscript " s "Indicates the inflow after back-filtering and smoothing; equal ; subscript " future "This refers to hypothetical future inflows."

[0090] For example, based on the generated inflow prior value, using the state of the previous moment in the one-dimensional hydrodynamic model as a warm start, the generated 200 current inflow prior values ​​are used as the current flow rate, and the posterior inflow of the previous moment is used as the flow rate for the next few moments, thus constructing 200 input sequences. Therefore, the one-dimensional hydrodynamic model is executed 200 times to generate simulated water levels. The simulated water levels based on the maximum and minimum inflow prior values ​​are compared. The starting time of the water level at each cross-section is used as the first water level information that can be assimilated, and data from two moments are taken sequentially. A total of three moments of water level data for each cross-section can be used to construct a simulated water level prior set considering lag. This is because the current flow information is distributed across multiple moments at the water level station, but since the flow rate at the next moment is unknown (assumed to be a fixed value), it is not necessarily better to take more future water levels. In this embodiment, the three moments are obtained through comparative testing.

[0091] In some embodiments, step S104, obtaining the observation sample set of the target reservoir and obtaining the posterior value of the inflow based on the forward filtering step of the Kalman smoothing method using the observation sample set, includes: generating the observation sample set based on the observed water level data and white noise of the target reservoir; obtaining the posterior value of the inflow based on the forward filtering step of the Kalman smoothing method using the state transition equation and the observation equation; the state transition equation includes the parametric state transition equation and the system state transition equation.

[0092] In this embodiment, based on the state transition equation and the observation equation, the forward filtering step of the Kalman smoothing method is used to obtain the posterior value of the inflow, including: determining the Kalman gain of the inflow of the target reservoir; updating the inflow set based on the Kalman gain, the observation sample set, and the simulated water level set; and using the average value of the inflow set as the posterior value of the inflow.

[0093] For example, based on the water level data collected from the water level stations, the observation sample set is generated using the following formula:

[0094]

[0095] in, Indicates the first j Water level station t The observed water level at time +1 This represents observed white noise, which follows a Gaussian distribution with a mean of 0. The standard deviation of white noise is represented by a value of 0.01 to 0.03 m. The standard deviation of white noise observed in this embodiment is 0.01 m.

[0096] In the specific operation of Kalman smooth forward filtering, the parameter state transition equation is:

[0097]

[0098] in, express t The set at time +1 is the first i Inbound traffic, express t The average inflow rate at time t. The system state transition equation is:

[0099]

[0100]

[0101] in, express t +1 time input the first i The first inbound flow j Simulated water level at each cross section f This represents a one-dimensional hydrodynamic model. m Indicates the number of cross sections. express t Time of the first j Posterior ensemble mean of water level at each cross section t 1 represents the length of the historical inflow. t 2 represents the length of the future inflow; the superscript " s "Indicates the inflow after back-filtering and smoothing; equal ; subscript " future "This refers to the hypothetical future inflow. The observation equation is:

[0102]

[0103] in, Indicates the first j Water level station t The observed water level at time +1 This represents observed white noise, which follows a Gaussian distribution with a mean of 0. This represents the standard deviation of white noise.

[0104] Based on the state transition equation and observation equation described above, the forward filtering steps of the Kalman smoothing method are as follows:

[0105] First, the Kalman gain of the inflow rate is calculated. The Kalman gain determines the accuracy weight of the observed and simulated values. Using the Kalman gain, the inflow rate set is updated by comparing the observed and simulated water levels. The set average is then taken as the posterior information of the inflow rate. The water level update is the same as that for the inflow rate. The specific formula is as follows:

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112] in, The Kalman gain represents the inflow rate. This represents the covariance matrix between the generated inflow rate and the simulated water level. Represents the covariance matrix of the simulated water level. The covariance matrix representing the observed water level; The Kalman gain represents the water level. Represents the set of elements. i The generated inbound traffic, This represents the average of the inbound flow rate set. Represents the set of elements. i A simulated water level, This represents the ensemble mean of simulated water levels; the superscript "-" indicates a priori value, and the superscript "+" indicates a posterior value. n Indicates the size of the set. express t The observed water level at time +1; express t Simulated water level at time +1.

[0113] In some embodiments, step S105, updating the posterior value of the inflow at the previous time step based on the backward filtering step of the Kalman smoothing method to obtain the lag-free smooth inflow value, includes: determining the Kalman smoothing gain of the target reservoir based on the covariance matrix of the posterior value and the prior value of the inflow of the target reservoir; and obtaining the lag-free smooth inflow value of the target reservoir by combining the update window length during the backward filtering of the Kalman smoothing method.

[0114] For example, firstly, the Kalman smoothing backward update step is used to correct the flow rate at the previous time step. This is achieved by first calculating the covariance matrix using the prior set of the current inflow flow rate and the prior and posterior sets of the inflow flow rate at the previous time step.

[0115]

[0116]

[0117] in, and Let these represent the covariance matrices of the posterior and prior values ​​of the inbound flow, respectively. Represents the set of elements. i The generated inbound traffic, express t The average of the inflow volume at any given time.

[0118] Then, the smoothing gain is calculated. The smoothing gain actually estimates how much of the error in the current flow rate is due to inaccurate flow rates at the previous time step, thereby correcting past flow rates and realizing the transmission of current observation information to the past. The specific formula is as follows:

[0119]

[0120] in, Indicates the Kalman smoothing gain. F t+1 express t The Jacobian matrix of the state transition matrix at time +1. Since the inflow rate in the invention is assumed to follow a Gaussian random walk, therefore... F It can be set as a unit matrix.

[0121] Finally, the Kalman smoothed inflow rate at time step 1 is calculated using the following formula:

[0122]

[0123] in, This represents the smoothed inbound flow rate after backward filtering, and this represents the index of the posterior value of the inbound flow rate in the forward filtering step. At the initial time equals K represents the update window length for backward filtering, which is set to 2 hours. In this embodiment, an inflow traffic is estimated to occur every 15 minutes, and a 2-hour update window means updating 8 historical traffic values.

[0124] When the river channel is relatively long, the inflow from the upstream boundary takes a long time to propagate to the water level station. Furthermore, due to flow flattening, the information is dispersed across multiple time points, which is particularly noticeable when updating at small time scales. By using a backward filtering step, new observation information can be fully utilized to continuously update the historical flow process, eliminating the influence of lag and thus obtaining a lag-free inflow process. It should be noted that this lag varies with flow rate and water level, so setting a fixed value to offset it is inaccurate. The parameter estimate is the desired inflow, and the posterior value is the state data with smaller variance obtained by combining simulation and measured data. The updated inflow is used as the data for the previous time point, and the inflow for the next time point is estimated again. Steps S103-S105 are repeated in a rolling solution to obtain the lag-free inflow data for continuous time points.

[0125] To verify the effectiveness of the above method, two sets of comparative experiments were set up. In the first set, the water level observation station was 60 km away from the upstream inflow section; in the second set, the water level observation station was approximately 320 km away from the upstream inflow section. An analytical optimization method was used as the comparative method. The analytical optimization method is constructed based on the concepts of water balance and flow continuity, and its purpose is to minimize the objective function, which is defined as follows:

[0126]

[0127] in, Indicates the formaldehyde coefficient. express t Inbound flow at time +1 express t Inbound traffic at any given time express t Simulated water level at any given time express t The water level is observed at any given time.

[0128] The concept of flow continuity means that the inflow rate should not change significantly within a short period of time. Using the Lagrange multiplication method, the inflow rate process that minimizes this objective function can be derived. The trade-off between water level error and flow continuity is determined manually by weighting coefficients. However, the results are highly sensitive to data length and parameters, making parsing optimization methods more suitable for post-processing than real-time applications.

[0129] The comparison results are as follows Figure 3-6 As shown, Figure 3 This is a schematic diagram of the experimental results obtained in July 2020 using the analytical optimization method. Figure 4 This is a schematic diagram of the experimental results obtained using this method in July 2020; Figure 5 This is a schematic diagram of the experimental results obtained in October 2022 using the analytical optimization method. Figure 6 This is a schematic diagram of the experimental results obtained using this method in October 2022.

[0130] The analytical optimization method shows significant differences in the estimated inflow process at different distances, exhibiting noticeable lag as the distance increases. In contrast, the results of our proposed method remain largely consistent across different distances, showing no significant lag. Figure 3-6As can be seen, the proposed method considers the propagation time of flood evolution and makes full use of water level data from multiple moments to continuously update the inflow, thus improving the accuracy of inflow calculation. This method uses minute-level measured water level data for data assimilation, which can simultaneously and effectively reconstruct both peak and low-flow processes. The estimated inflow does not vary significantly with the location of the water level station, and it avoids estimation lag even when the water level station is far away. Furthermore, the estimated inflow process exhibits small fluctuations, good continuity, and fine time scale, providing better guidance for reservoir operation and flow forecasting.

[0131] In summary, this method addresses the issue of time lag in the inflow estimation results for large river-type reservoirs with small time scales. It innovatively utilizes a dynamic water level lag matrix and Kalman smoothing to fully leverage all observed water level information and eliminate lag and fluctuations. First, a state transition equation for the inflow is constructed based on a Gaussian random walk. Then, a dynamic water level lag matrix is ​​established by simulating water levels at different inflow scenarios using one-dimensional hydrodynamics. Finally, forward and backward filtering mechanisms of Kalman smoothing are applied to assimilate real-time observed water level data, resulting in minute-level, smooth, accurate, and lag-free reservoir inflow data. This solves the problem of sawtooth fluctuations and systematic errors in reservoir inflow estimation, providing fundamental data support for hydrological model simulation and reservoir scheduling. This method framework is suitable for all large rivers, especially when the estimated flow section is far from the downstream water level station.

[0132] The present invention also provides a device for reverse calculation of inbound flow without delay. The device for reverse calculation of inbound flow without delay provided by the present invention is described below. The device for reverse calculation of inbound flow without delay described below can be referred to in correspondence with the method for reverse calculation of inbound flow without delay described above. Figure 7 This is a structural block diagram of the time-delay-free inbound flow reverse estimation device provided by the present invention, as shown below. Figure 7 As shown, the device includes:

[0133] The acquisition module 701 is used to collect raw data of the target reservoir; the raw data includes the upstream river topography, water level station locations and monitoring data, and inflow data of the target reservoir.

[0134] Module 702 is used to construct a one-dimensional hydrodynamic model based on the original data of the target reservoir and to verify the accuracy of the one-dimensional hydrodynamic model.

[0135] Processing module 703 is used to determine the prior set of inflow to the target reservoir and obtain the simulated water level set and the propagation hysteresis of flow information through a one-dimensional hydrodynamic model;

[0136] The forward filtering module 704 is used to obtain the observation sample set of the target reservoir and obtain the posterior value of the inflow rate based on the forward filtering step of the Kalman smoothing method using the observation sample set.

[0137] The backward filtering module 705 is used to update the posterior value of the inflow rate at the previous time step based on the Kalman smoothing method, so as to obtain the smooth inflow rate value without lag.

[0138] In operation, this device first acquires and organizes the raw data of the target reservoir using module 701. Then, module 702 constructs a one-dimensional hydrodynamic model based on the raw data, using measured water level data and historical runoff data of the target reservoir as ground truth to verify the model's accuracy. Processing module 703 generates a priori set of inflow discharge and, by executing the one-dimensional hydrodynamic model, obtains the simulated water level set of the target reservoir and the propagation hysteresis of discharge information. Next, forward filtering module 704 acquires the observation sample set of the target reservoir and, based on the current observation data, uses the forward filtering step of the Kalman smoothing method to obtain the posterior value of the inflow discharge. Finally, backward filtering module 705 uses the backward filtering step of the Kalman smoothing method to update the posterior value of the inflow discharge at the previous moment, obtaining a smoothed inflow discharge value without hysteresis. In practical applications, processing module 703, forward filtering module 704, and backward filtering module 705 can be repeatedly executed based on newly observed water level data to achieve real-time updates of the inflow discharge.

[0139] This device fully considers the problems of inaccurate inflow estimation caused by flow propagation lag, water surface slope, and observation errors. Compared with existing technologies, this device innovatively utilizes a water level prior set considering lag and the Kalman smoothing method to solve the problem of inaccurate inflow estimation caused by flow propagation lag. This method is applicable to long rivers, large river-type reservoirs, and the need for back-calculation of flow data at fine time scales. It uses minute-level measured water level data for data assimilation, fully utilizing data information from multiple time points to dynamically correct the flow, which can accurately reconstruct the flow process and enhance the real-time performance of flow estimation. The inflow process estimated by this method has the characteristics of small fluctuations, good continuity, and no lag, which can better guide reservoir operation and improve forecast accuracy as basic data. It achieves a breakthrough from "small rivers" to "large river-type reservoirs," and from "hourly" to "minute-level" inflow estimation, solving the problem of poor accuracy in inflow estimation in existing related technologies.

[0140] Figure 8 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 8As shown, the electronic device may include: a processor 801, a communication interface 802, a memory 803, and a communication bus 804. The processor 801, communication interface 802, and memory 803 communicate with each other via the communication bus 804. The processor 801 can call logical instructions in the memory 803 to execute a time-delay-free inbound flow reverse calculation method. This method includes:

[0141] Collect raw data of the target reservoir; the raw data includes the upstream river topography, water level station locations and monitoring data, and inflow data of the target reservoir;

[0142] Based on the original data of the target reservoir, a one-dimensional hydrodynamic model is constructed, and the accuracy of the one-dimensional hydrodynamic model is verified.

[0143] The a priori set of inflows to the target reservoir is determined, and the simulated water level set and the propagation lag of flow information are obtained through a one-dimensional hydrodynamic model.

[0144] Obtain the observation sample set of the target reservoir, and use the forward filtering step of the Kalman smoothing method to obtain the posterior value of the inflow rate based on the observation sample set;

[0145] The backward filtering step based on the Kalman smoothing method updates the posterior value of the inflow rate at the previous time step, resulting in a smooth inflow rate value without lag.

[0146] Furthermore, the logical instructions in the aforementioned memory 803 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0147] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the time-free inbound flow reverse calculation method provided by the above methods, the method including:

[0148] Collect raw data of the target reservoir; the raw data includes the upstream river topography, water level station locations and monitoring data, and inflow data of the target reservoir;

[0149] Based on the original data of the target reservoir, a one-dimensional hydrodynamic model is constructed, and the accuracy of the one-dimensional hydrodynamic model is verified.

[0150] The a priori set of inflows to the target reservoir is determined, and the simulated water level set and the propagation lag of flow information are obtained through a one-dimensional hydrodynamic model.

[0151] Obtain the observation sample set of the target reservoir, and use the forward filtering step of the Kalman smoothing method to obtain the posterior value of the inflow rate based on the observation sample set;

[0152] The backward filtering step based on the Kalman smoothing method updates the posterior value of the inflow rate at the previous time step, resulting in a smooth inflow rate value without lag.

[0153] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the time-free inbound flow reverse calculation method provided by the above methods, the method comprising:

[0154] Collect raw data of the target reservoir; the raw data includes the upstream river topography, water level station locations and monitoring data, and inflow data of the target reservoir;

[0155] Based on the original data of the target reservoir, a one-dimensional hydrodynamic model is constructed, and the accuracy of the one-dimensional hydrodynamic model is verified.

[0156] The a priori set of inflows to the target reservoir is determined, and the simulated water level set and the propagation lag of flow information are obtained through a one-dimensional hydrodynamic model.

[0157] Obtain the observation sample set of the target reservoir, and use the forward filtering step of the Kalman smoothing method to obtain the posterior value of the inflow rate based on the observation sample set;

[0158] The backward filtering step based on the Kalman smoothing method updates the posterior value of the inflow rate at the previous time step, resulting in a smooth inflow rate value without lag.

[0159] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0160] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0161] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A non-hysteresis warehouse entry flow backstepping method, characterized in that, The method comprises the following steps: collecting original data of a target reservoir; the original data comprises upstream river channel topography, water level station location and detection data, and inflow data of the target reservoir; constructing a one-dimensional hydrodynamic model based on the original data of the target reservoir, and verifying the model accuracy of the one-dimensional hydrodynamic model; determining an inflow prior set of the target reservoir, and obtaining a simulated water level set and flow information propagation lag through the one-dimensional hydrodynamic model; obtaining an observation sample set of the target reservoir, and obtaining an inflow posterior value through a forward filtering step of Kalman smoothing method based on the observation sample set; updating the inflow posterior value of the last time based on a backward filtering step of Kalman smoothing method to obtain a smooth inflow value without lag time; determining an inflow prior set of the target reservoir, and obtaining a simulated water level set and flow information propagation lag through the one-dimensional hydrodynamic model, comprising: establishing a state transition equation of the inflow of the target reservoir by using Gaussian random walk, and generating an inflow prior set; executing the one-dimensional hydrodynamic model for multiple times to obtain a simulated water level set and flow information propagation lag of the target reservoir; executing the one-dimensional hydrodynamic model for multiple times to obtain a simulated water level set and flow information propagation lag of the target reservoir, comprising: inputting sequences with the same historical inflow data but different current inflow set members into the one-dimensional hydrodynamic model; simulating through the one-dimensional hydrodynamic model, estimating time lag, and constructing a simulated water level set capable of reflecting time lag; obtaining an observation sample set of the target reservoir, and obtaining an inflow posterior value through a forward filtering step of Kalman smoothing method based on the observation sample set, comprising: generating an observation sample set based on the observed water level data and white noise of the target reservoir; obtaining an inflow posterior value through a forward filtering step of Kalman smoothing method based on a state transition equation and an observation equation; the state transition equation comprises a parameter state transition equation and a system state transition equation; obtaining an inflow posterior value through a forward filtering step of Kalman smoothing method based on a state transition equation and an observation equation, comprising: determining a Kalman gain of the inflow of the target reservoir; updating an inflow set based on the Kalman gain, the observation sample set and the simulated water level set; taking an average value of the inflow set as an inflow posterior value.

2. The no-hysteresis ingress traffic back-calculation method of claim 1, wherein, Collecting original data of a target reservoir, comprising: collecting measured river channel section data of the target reservoir from a dam site to an upstream backwater area; obtaining reference values and data ranges of river channel roughness data of the target reservoir; obtaining historical reported inflow data of the target reservoir.

3. The no-hysteresis ingress traffic back-calculation method of claim 2, wherein, Constructing a one-dimensional hydrodynamic model based on the original data of the target reservoir, comprising: describing evolution characteristics of the target reservoir through Saint-Venant equation set; the Saint-Venant equation set comprises continuity equation and momentum equation; solving flow and water level of all sections of the river of the target reservoir by using Preissmann four-point eccentric implicit scheme; generating the one-dimensional hydrodynamic model in combination with the measured river channel section data of the target reservoir.

4. The no-hysteresis ingress traffic back-calculation method of claim 1, wherein, The backward filtering step of the Kalman smoothing method is used to update the inflow posterior value of the previous time, and a non-delay smoothing inflow value is obtained, including: Based on the covariance matrix of the target reservoir inflow posterior value and the inflow prior value, the Kalman smoothing gain of the target reservoir is determined. Combined with the update window length of the backward filtering of the Kalman smoothing method, the non-delay smoothing inflow value of the target reservoir is obtained.

5. A device for non-hysteretic warehouse inflow rate backstepping, for implementing the non-hysteretic warehouse inflow rate backstepping method of any one of claims 1-4, characterized in that, It includes: The acquisition module is used to collect the original data of the target reservoir; the original data includes the upstream river channel topography, water level station position and detection data and inflow data of the target reservoir; The construction module is used to construct a one-dimensional hydrodynamic model based on the original data of the target reservoir, and verify the model accuracy of the one-dimensional hydrodynamic model; The processing module is used to determine the inflow prior set of the target reservoir, and obtain the simulated water level set and flow information propagation delay through the one-dimensional hydrodynamic model; The forward filtering module is used to obtain the observation sample set of the target reservoir, and obtain the inflow posterior value by using the forward filtering step of the Kalman smoothing method based on the observation sample set; The backward filtering module is used to update the inflow posterior value of the previous time based on the backward filtering step of the Kalman smoothing method, and obtain a non-delay smoothing inflow value.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to realize the non-delay inflow backstepping method of any one of claims 1 to 4.

Images