A method for calculating the response relationship of an upstream reservoir to the dispatching of a midstream and downstream control station

Abstract

The application belongs to the field of flood control scheduling, and discloses a method for calculating the scheduling response relationship of an upstream reservoir to a midstream and downstream control station, comprising the following steps: generating the river channel hydrodynamic state under the reservoir regulation and storage process by using a Mike model, and adding a measured flood sequence to form an upstream reservoir scheduling-downstream river channel hydrodynamics state sample. The influence period of the upstream discharge flow is determined, the response relationship between the upstream reservoir discharge flow and the downstream control station flow is extracted by using an Elman neural network, and the performance of the model is evaluated. On the basis of the upstream reservoir discharge flow-downstream control station water level data, a flow-water level model for the reverse calculation of the upstream reservoir storage and discharge process is established by using multiple linear regression. The water level or flood discharge flow target of the downstream river channel is determined, the upstream reservoir discharge flow process and the storage process are inversely calculated, and the reverse control of the upstream reservoir is realized. The application can provide a stronger basis for the flood control scheduling of a basin, thereby reducing the scheduling risk.

Description

Technical Field

[0001] The present invention belongs to the field of flood control and dispatching of water storage space, and in particular relates to a method for estimating the dispatching response relationship of an upstream reservoir to a midstream and downstream control station. Background Art

[0002] In recent years, due to factors such as global climate change and human activities, extreme rainstorms have become more frequent, increasing the probability of basin-wide floods. Flood control operations in the upper reaches of a river basin have a direct and critical impact on the hydrological regime in the middle and lower reaches. In particular, the regulation and operation of large reservoirs in the upper reaches of the basin, as well as the activation of flood storage areas, levees, and other water conservancy facilities in the middle and lower reaches, are crucial to the safety of flood control operations throughout the basin.

[0003] In recent years, researchers have conducted in-depth discussions on flood control scheduling in upstream and mid- and downstream areas. Taking into account actual flood control needs, they have established a coordinated flood control scheduling model for upstream and downstream cascade reservoirs. With the goals of minimizing flood control storage capacity usage, minimizing cross-sectional flow, and minimizing maximum reservoir water levels, dynamic programming, genetic algorithms, and other simulations are used to determine the allocation of flood control storage capacity. However, current research has mostly focused on theoretical models and simulations, and further refinement of specific operations and flow process control is needed for actual engineering applications. Focusing on the impact of upstream reservoirs on the scheduling of mid- and downstream control stations, a method for estimating the scheduling response relationship of upstream reservoirs to mid- and downstream control stations is proposed to further refine the flood control compensation scheduling control of large upstream reservoirs on mid- and downstream stations.

[0004] Therefore, the present invention can provide a systematic analysis method for analyzing the scheduling response relationship between upstream reservoirs and downstream areas, provide strong support for the comprehensive water resources management in the middle and lower reaches, and has certain practical significance. Summary of the Invention

[0005] In response to the need for improvement in the existing technology, the present invention provides a method for estimating the dispatching response relationship of an upstream reservoir to a mid- and downstream control station, the purpose of which is to provide a scientific basis for analyzing the dispatching response relationship between an upstream reservoir and a downstream control station.

[0006] To achieve the above objectives, the present invention solves the above technical problems by adopting a technical solution: a method for estimating the dispatch response relationship of an upstream reservoir to a mid- and downstream control station, the method comprising the following steps:

[0007] S1, based on the existing one-dimensional hydrodynamic model, by setting the upstream reservoir operation process and the boundary conditions of different tributaries and intervals, the river hydrodynamic state under different boundaries and reservoir storage processes is calculated, and the flood series actually measured in the basin in recent years are added to form a large number of upstream reservoir operation-downstream river hydrodynamic state samples.

[0008] S2, based on the data samples of S1, determines the impact period of upstream discharge flow, uses the Elman neural network algorithm to extract the response relationship between the upstream reservoir discharge flow and the downstream control station flow, realizes the forward calculation from the upstream reservoir to the downstream station, and uses evaluation indicators to illustrate the performance of the model.

[0009] S3, based on historical measured flood data, uses a multivariate linear regression algorithm to establish a flow-water level model for the downstream control station to reversely infer the storage and discharge process of the upstream reservoir, and calculates the flow coefficient within the cycle.

[0010] S4, determine the water level or flood flow target of the downstream river channel, combine it with the discharge flow constraint of the upstream reservoir, and reversely calculate the storage flow process and discharge process of the upstream reservoir with the goal of minimizing the storage volume of the upstream reservoir or maximizing the discharge flow, so as to achieve reverse control of the storage and discharge process of the upstream reservoir.

[0011] Furthermore, the S1 specifically includes:

[0012] S11, based on the flow conditions of upstream reservoirs in multiple typical years, by artificially adjusting the discharge flow of upstream reservoirs within a fixed time period, a variety of reservoir regulation states can be simulated. The time period is in days and can be set to different lengths such as three months and four months according to actual conditions to reduce the random influence of the data.

[0013] S12, determine the boundaries of the affected basin, including the different tributaries and interval inflows of the basin. In order to consider the impact of different rainfall events and boundary inflows on the effect of the hydrodynamic model, the average rainfall inflow of each boundary in multiple typical years is used as the boundary condition of the model.

[0014] S13, MIKE one-dimensional hydrodynamic model is a hydrological and hydrodynamic simulation software widely used in rivers, channels, estuaries and coastal areas. By clarifying the scope of the river section or channel to be modeled, inputting terrain data, setting parameters such as the riverbed roughness coefficient and boundary conditions, and inputting multiple sets of artificial flow sequences, the hydrodynamic state of the river under different boundaries and reservoir storage processes can be obtained.

[0015] S14, adding the actual measured flow sequences of multiple typical years in the basin, forming a large number of time series samples of upstream reservoir operation-downstream river hydrodynamic state, providing a data basis for subsequent operations.

[0016] Furthermore, the S2 specifically includes:

[0017] S21, determine the impact period of the discharge flow of the upstream reservoir, and define the impact period as the time it takes for the discharge at the downstream station to reach and maintain a stable state by changing the upstream constant flow when the inflow of other tributaries and intervals in the basin is constant and the reservoir flow is a constant sequence. If the impact period of the upstream is N days, the flow from the 1st to the Nth day upstream is used as a scheme to predict the flow change on the N+1th day downstream. The flow from the 2nd to the N+1th day upstream is used as a scheme to predict the flow change on the N+2th day downstream. And so on. A total of DF-N+1 schemes can be generated, where DF is the actual number of flood days upstream.

[0018] S22, the Elman neural network is a typical local regression network. Compared with the feedforward neural network, it adds a connection layer with memory units, which can effectively handle related issues about time series information objects. The actual flood sequence is divided into several training samples according to the upstream impact period N days as the step size. The discharge flow of the upstream reservoir is used as the forecast input of the Elman neural network algorithm according to the step size, and the flow of the downstream control station is used as the output of the network training. It can be used to extract the response relationship between the discharge flow of the upstream reservoir and the flow of the downstream control station.

[0019] S23, when evaluating the performance of the Elman model, multiple indicators are usually used to quantify the quality of the model results. Common evaluation indicators include mean absolute error (MAE), coefficient of certainty (DC), root mean square error (RMSE), and pass rate (QR).

[0020] Furthermore, the evaluation indicators of the Elman neural network method specifically include:

[0021] S2301 mean absolute error MAE:

[0022]

[0023] Where n is the number of samples, y i is the true value of the i-th sample, is the predicted value of the i-th sample, MAE represents the average absolute difference between the predicted value and the true value, and the smaller the value, the smaller the difference.

[0024] S2302 Determinism Coefficient DC:

[0025]

[0026] In the formula is the average value of the samples, DC represents the proportion of the variation explained by the model to the total variation. The closer the value is to 1, the better the fitting effect.

[0027] S2303 root mean square error RMSE:

[0028]

[0029] RMSE is used to measure the square root of the average error between the model prediction value and the true value. The smaller the value, the better the model fit.

[0030] S2304 pass rate QR:

[0031]

[0032] Where n is the number of qualified forecasts, TM is the total number of forecasts, and a larger value means a smaller model fitting error.

[0033] Furthermore, the S3 specifically includes:

[0034] When using the multivariate linear regression algorithm to establish a flow-water level model for the downstream control station to reversely infer the storage and discharge process of the upstream reservoir, the actual flood flow of multiple upstream reservoirs is divided into several scenarios according to the flow cycle of N days. These scenarios are used as independent variables, and the flow or water level series of the corresponding scenarios of the downstream control station are used as dependent variables to determine the model:

[0035]

[0036] Where Y i is the downstream flow or water level predicted by the i-th scenario, α t represents the discharge coefficient on the tth day in the cycle, X it represents the upstream flow on day t in the i-th scenario cycle, where t = 1, 2...30.

[0037] Furthermore, the formula for calculating the flow coefficient using the least squares method specifically includes:

[0038]

[0039] In the formula is the parameter matrix, Y is the output factor matrix, X is the b×(N+1) order input factor matrix, and b is the number of experiments, that is, the number of solutions.

[0040] Furthermore, the S4 specifically includes:

[0041] If the downstream flow requirement is given, the target data type is converted into the water level requirement according to the downstream water level-flow relationship. Based on the known upstream reservoir flow sequence and flow constraints of the first n days in the cycle, with the goal of minimizing the upstream reservoir storage or maximizing the downstream discharge, the model in S3 is used to calculate the storage process for the remaining Nn days, so as to realize the reverse control of the downstream control station on the upstream reservoir storage and discharge process. Otherwise, the reverse process is directly calculated using the water level requirement.

[0042] Compared with the existing technology, the advantages of the present invention are: based on the historical data of the upstream reservoir and the mid- and downstream control stations, the present invention can not only determine the flow impact cycle and realize rapid regulation of the upstream to the downstream through the Elman model; but also use multiple linear regression to calculate the flow coefficient to realize reverse control of the downstream to the upstream, making the regulation process more reasonable and more in line with the actual situation. In this way, it can provide a more powerful basis for flood control scheduling in the basin, thereby reducing scheduling risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of the method proposed in the present invention for estimating the dispatching response relationship between the upstream reservoir and the mid- and downstream control stations.

[0044] Figure 2 Schematic diagram of the locations of key control stations in the middle and lower reaches of the Yangtze River in this embodiment.

[0045] Figure 3 This is a comparison chart of the water level in Chenglingji predicted by Elman in 2017 and the actual water level.

[0046] Figure 4 This is a comparison chart of the Mike model and multiple linear regression predictions from June 1 to September 30, 1954.

[0047] Figure 5 This is a flow chart of reverse calculation of multi-period flow in an embodiment.

[0048] Figure 6 The flow process of the Three Gorges Reservoir is reversed for this example.

[0049] Figure 7 This is a comparison chart of the water level process in Chenglingji, an example. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific examples described here are only used to explain the present invention.

[0051] It is not intended to limit the present invention.

[0052] The following describes a method for estimating the response relationship between upstream reservoirs and downstream control stations, using the water level requirement in the Chenglingji area of ​​the lower Yangtze River to be lowered from 25.49m to 25.00m on July 5, 1954, and the corresponding flow process at the upstream Three Gorges Reservoir from July 2nd to July 4th as an example. The actual flood process at the Three Gorges Reservoir from June 1st to July 4th, 1954, is shown in Table 1, and the actual water level changes at the Chenglingji station are shown in Table 2.

[0053] Table 1 The discharge process of the Three Gorges Reservoir during the 1954 flood

[0054]

[0055] Table 2 Chenglingji water level change process

[0056] Date July 1 July 2 July 3 July 4 July 5 Water level (m) 23.71 24.29 24.75 25.15 25.49

[0057] like Figure 1 The figure shows a flow chart of a method designed by the present invention for estimating the dispatch response relationship between the upstream reservoir and the mid- and downstream control stations. The following is a detailed description with reference to an example:

[0058] S1, based on the existing one-dimensional hydrodynamic model, by setting the upstream reservoir operation process and the boundary conditions of different tributaries and intervals, the river hydrodynamic state under different boundaries and reservoir storage processes is calculated, and the flood series actually measured in the basin in recent years are added to form a large number of upstream reservoir operation-downstream river hydrodynamic state samples.

[0059] S11. This embodiment uses the outflow of the Three Gorges Reservoir in the typical year of 1954 as a basis, selects a fixed period of June 1st to September 30th, and artificially sets 15 discharge flow sequences corresponding to the time period as the flow process of the upstream reservoir.

[0060] S12, according to Figure 2 Schematic diagram of the locations of key control stations in the middle and lower reaches of the Yangtze River. In this example, Shimen Station, Taoyuan Station, Taojiang Station and other stations are determined as boundaries, and the average rainfall runoff at each boundary in five typical years (1954, 2016, 2017, 2020, and 2024) is selected as the boundary conditions of the model.

[0061] S13. In this embodiment, the river section range and terrain data are input into the MIKE model, and parameters such as the riverbed roughness coefficient and boundary conditions are set. A sequence of 14 man-made floods is used as an input forecast to obtain the river hydrodynamic state under different boundaries and reservoir storage processes.

[0062] S14. This embodiment adds actual measured flow sequences of typical years in the Yangtze River Basin, including 1954, 2016, 2017, and 2024, to form a large number of upstream reservoir regulation-downstream river hydrodynamic state time series samples, providing a data basis for subsequent operations.

[0063] S2, based on the data samples of S1, determines the impact period of upstream discharge flow, uses the Elman neural network algorithm to extract the response relationship between the upstream reservoir discharge flow and the downstream control station flow, realizes the forward calculation from the upstream reservoir to the downstream station, and uses evaluation indicators to illustrate the performance of the model.

[0064] S21, in this embodiment, Chenglingji area is a key node for flood control in the Yangtze River Basin. Its flood control dispatch needs to be determined based on the water conditions at the Lianhuatang water level station. The change in the water level at a certain station is obtained by calculating its relationship with the water level at the downstream Luoshan station. Therefore, Luoshan station is used as the conversion station to calculate the water level in Chenglingji area. 3 / s reduced to 33000m 3 / s, while other conditions remain unchanged, is used as an example to determine the flow impact period N of the upstream Three Gorges Reservoir. The flow change process of the downstream Luoshan Station is shown in Table 3. It is found that the flow at the downstream station reaches a stable state on June 30, so the flow impact period N of the upstream Three Gorges Reservoir is determined to be 30 days.

[0065] Table 3. Changes in flow at downstream sites

[0066] Date June 1 June 2 June 3 June 4 June 5 June 6 June 7 June 8 <![CDATA[流量(m 3 / s)]]> 46199.37 46079.92 45758.73 45466.74 45198.64 44957.09 44744.73 44561.58 Date June 9 June 10 June 11 June 12 June 13 June 14 June 15 June 16 <![CDATA[流量(m 3 / s)]]> 44397.00 44251.01 44123.59 44009.45 43916.55 43823.64 43746.66 43680.30 Date June 17 June 18 June 19 June 20 June 21 June 22 June 23 June 24 <![CDATA[流量(m 3 / s)]]> 43621.90 43566.16 43521.03 43481.22 43452.02 43417.51 43390.97 43367.03 Date June 25 June 26 June 27 June 28 June 29 June 30 <![CDATA[流量(m 3 / s)]]> 43343.19 43327.26 43311.33 43295.41 43282.13 43268.86

[0067] S22. This embodiment utilizes the Elman neural network algorithm to divide the actual floods from 2006 to 2024 (excluding the four floods in 2016, 2017, 2020, and 2024) into training samples with a step size of N = 30 days. The flow rate sequences of the 15th man-made flood and the four-year measured floods are used as the forecast inputs of the neural network, and the flow rate at Luoshan Station is used as the output of the network training. The flow rate at Luoshan Station is then calculated. The water level-flow relationship at Luoshan Station and the water level amplitude correlation between Luoshan Station and Chenglingji Station are then used to convert the flow rate at Luoshan Station to the water level at Lianhuatang Hydrological Station, and the water level change at Chenglingji Station can be calculated.

[0068] S23, this embodiment obtains the error index of the Chenglingji water level for the 15th simulated flood and the four years of 2016, 2017, 2020, and 2024, as shown in Table 4:

[0069] Table 4 Error index results of Chenglingji water level simulation in different typical years

[0070] Year 2016 2017 2020 2024 RMSE 1.43 1.67 1.45 1.44 MRE 0.04 0.05 0.04 0.05 DC 0.87 0.79 0.89 0.99 QR 0.99 0.99 0.99 0.73

[0071] It can be seen that the water level of Chenglingji is highly consistent with the simulation results of MIKE, the measured floods in 2016, 2017 and 2020, the positive calculation model has high accuracy, however, compared with the actual water level of Chenglingji in 2024, the qualified rate is only about 73%, this deviation is mainly due to the fact that when constructing the Elman neural network model, the inflow of the tributary between the Three Gorges and Luoshan Station and the uncontrolled area is not taken as a variable, but the average value of the actual flood boundary flow sequence in 1954, 2016, 2017, 2020 and 2024 is used as a fixed boundary, which deviates from the actual flood boundary, in addition, the flood sequence from October to December in 2024 is also missing, which may also cause errors. Taking 2017 as an example, the contrast chart of Elman predicted water level of Chenglingji and actual water level is shown in Figure 3 .

[0072] S3, based on historical measured flood data, a flow-water level model for reverse calculation of the storage and discharge process of the upstream reservoir at the downstream control station is established using a multiple linear regression algorithm, and the flow coefficient in the period is calculated. This embodiment uses 19 measured flood data of the Yangtze River Basin from 2006 to 2024 as the basis, divides the actual flood flow of the Three Gorges Reservoir into several schemes according to the flow period N = 30 days, and takes these schemes as independent variables and the water level sequence of the corresponding scheme at the downstream control station as dependent variables to establish a reverse regression model:

[0073]

[0074] where Y i is the predicted water level of Chenglingji at the downstream, T t represents the flow coefficient on the t-th day in the period, X it represents the discharge flow of the upstream on the t-th day in the i-th scheme period, and m is a constant, where t = 1, 2...30. The discharge flow coefficient of the Three Gorges is solved by the least square method as shown in Table 5:

[0075] Table 5 Flow coefficient of the Three Gorges Reservoir

[0076]

[0077]

[0078] To verify the accuracy of the model coefficients, the data from June 1, 1954 to September 30, 1954 of the Three Gorges Reservoir are used for simulation comparison, which is shown in Figure 4 , the results show that the fitting effect is good.

[0079] S4, determine the water level or flood flow target of the downstream river, combine it with the discharge flow constraint of the upstream reservoir, take the minimum storage capacity of the upstream reservoir or the maximum discharge flow as the goal, reverse the storage flow process and discharge process of the upstream reservoir, and realize the reverse control of the storage and discharge process of the upstream reservoir. This embodiment determines that the water level requirement of Chenglingji Station on July 5 is reduced from 25.49m to 25.00m, and the goal is to plan and solve the flow of the upstream Three Gorges Reservoir from July 2 to July 4. The Three Gorges flow process corresponding to the water level of Chenglingji Station on July 5 is June 5-July 4, of which the discharge flow of the Three Gorges from June 5 to July 1 is shown in Table 1, and the flow flow chart for solving the Three Gorges from July 2 to July 4 is shown in Table 1. Figure 5 , the specific model is as follows:

[0080]

[0081] Where, α t represents the discharge q of the Three Gorges Reservoir on day t t The corresponding flow coefficient is, where q is divided by 28 ,q 29 ,q 30 Except for the unknown, the other parameters are known and can be transformed into the following equation:

[0082] α 28 q 28 +α 29 q 29 +α 30 q 30 =1.283

[0083] Obviously, when the flow rate corresponding to the days with small flow coefficient is larger, the outflow of the Three Gorges Reservoir in these three days is the largest, that is, when α 30 When the discharge flow of the Three Gorges Reservoir is the smallest, the discharge flow of the Three Gorges Reservoir on the 30th day is adjusted in a linear increasing manner. If the equation constraint is still not satisfied after the adjustment, the discharge flow of the Three Gorges Reservoir on the 29th day is increased in the same way, and so on, until the maximum discharge flow is reached. Using the above calculation process, the flow process of the Three Gorges Reservoir and the water level process of Chenglingji can be obtained as shown in the following figure: Figure 6 、 Figure 7 By planning the Three Gorges Dam's outflow from July 2nd to 4th, particularly reducing the outflow on July 2nd, the water level at Chenglingji was effectively lowered over the next three days. Furthermore, the Three Gorges Dam's outflow on July 3rd and 4th was greater than the typical flood runoff, ensuring the minimum impoundment required by the Three Gorges Reservoir.

[0084] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A method for estimating the dispatch response relationship between an upstream reservoir and a mid- and downstream control station, characterized in that: These include: S1, based on the existing one-dimensional hydrodynamic model, by setting the upstream reservoir operation process and the boundary conditions of different tributaries and intervals, the river hydrodynamic state under different boundaries and reservoir storage processes is calculated. The actual flood series measured in the basin are then added to form a large number of upstream reservoir operation-downstream river hydrodynamic state samples; S2, based on the data samples of S1, determines the impact period of upstream discharge flow, uses the Elman neural network algorithm to extract the response relationship between the discharge flow of the upstream reservoir and the flow of the downstream control station, realizes the forward calculation from the upstream reservoir to the downstream station, and uses evaluation indicators to illustrate the performance of the model; Specifically include: S21, determine the impact period of the upstream reservoir discharge, which is defined as the time it takes for the downstream station discharge to reach and maintain a stable state by changing the upstream constant discharge when the inflow of other tributaries and intervals in the basin is constant and the reservoir discharge is a constant sequence. If the upstream impact period is N days, the discharge from the upstream day 1 to the Nth day is used as one scenario to predict the downstream discharge change on the N+1th day. The discharge from the upstream day 2 to the N+1th day is used as another scenario to predict the downstream discharge change on the N+2th day. And so on. A total of DF-N+1 scenarios are generated, where DF is the actual number of flood days upstream. S22, the Elman neural network is a typical local regression network. Compared with the feedforward neural network, it adds a connection layer with memory units. It can effectively handle related problems about time series information objects. The actual flood series is divided into several training samples according to the step length of the upstream impact period N days. The discharge of the upstream reservoir is used as the forecast input of the Elman neural network algorithm according to the step length, and the discharge of the downstream control station is used as the output of the network training. It is used to extract the response relationship between the discharge of the upstream reservoir and the discharge of the downstream control station. S23, when evaluating the performance of the Elman model, multiple indicators are used to quantify the quality of the model results, including mean absolute error MAE, coefficient of certainty DC, root mean square error RMSE and pass rate QR; S3, based on historical flood data, uses a multivariate linear regression algorithm to establish a flow-water level model for the downstream control station to reversely infer the storage and discharge process of the upstream reservoir, and calculates the flow coefficient within the cycle; S4, determine the water level or flood flow requirements of the downstream river channel, combine it with the discharge flow constraints of the upstream reservoir, and reversely calculate the storage flow process and discharge process of the upstream reservoir with the goal of minimizing the upstream reservoir's storage volume or maximizing the discharge flow, so as to achieve reverse control of the upstream reservoir's storage and discharge process.

2. The method for estimating the dispatch response relationship of an upstream reservoir to a midstream or downstream control station according to claim 1, characterized in that: Said S1 specifically includes: S11, based on the flow conditions of upstream reservoirs in multiple typical years, simulates a variety of different reservoir storage states by artificially adjusting the discharge flow of upstream reservoirs within a fixed period. The period is in days and can be set to different lengths according to actual conditions to reduce the randomness of the data; S12, determine the boundaries of the affected basin, including the different tributaries and interval inflows of the basin. To consider the impact of different rainfall events and boundary inflows on the hydrodynamic model, the average rainfall flow of each boundary in multiple typical years is used as the boundary condition of the model; S13, MIKE one-dimensional hydrodynamic model is a hydrological and hydrodynamic simulation software widely used in rivers, channels, estuaries and coastal areas. By defining the modeled river section or channel range, inputting terrain data, setting the riverbed roughness coefficient and boundary condition parameters, and inputting multiple sets of artificial flow sequence steps, the hydrodynamic state of the river channel under different boundaries and reservoir regulation processes is obtained; S14, adding the actual measured flow sequences of multiple typical years in the basin, forming a large number of time series samples of upstream reservoir operation-downstream river hydrodynamic state, providing a data basis for subsequent operations.

3. The method for estimating the dispatch response relationship between the upstream reservoir and the mid- and downstream control stations according to claim 1 is characterized in that: The evaluation indicators specifically include: S2301 mean absolute error MAE: Where n is the number of samples, y i is the true value of the i-th sample, is the predicted value of the i-th sample, MAE represents the average absolute difference between the predicted value and the true value, and the smaller the value, the smaller the difference; S2302 Determinism Coefficient DC: In the formula is the mean value of the samples, DC represents the proportion of the variation explained by the model to the total variation, and the closer the value is to 1, the better the fitting effect; S2303 root mean square error RMSE: RMSE is used to measure the square root of the average error between the model prediction value and the true value. The smaller the value, the better the model fit. S2304 pass rate QR: Where n is the number of qualified forecasts, TM is the total number of forecasts, and a larger value means a smaller model fitting error.

4. The method for estimating the dispatch response relationship between an upstream reservoir and a mid- and downstream control station according to claim 1, characterized in that: In S3, when using the multivariate linear regression algorithm to establish a flow-water level model for the downstream control station to reversely infer the storage and discharge process of the upstream reservoir, the actual flood flow of multiple upstream reservoirs is divided into several scenarios according to the flow cycle of N days. These scenarios are used as independent variables, and the water level sequence of the downstream control station corresponding to the scenario is used as the dependent variable to determine the model: Where Y i is the downstream water level predicted by the i-th scenario, α t represents the discharge coefficient on the tth day in the cycle, X it represents the upstream flow on day t in the i-th scenario cycle, where t = 1, 2...

30.

5. The method for estimating the dispatch response relationship of an upstream reservoir to a mid- and downstream control station according to claim 4, characterized in that: The least squares method is used to derive the flow coefficient. The specific formula is as follows: In the formula is the parameter matrix, Y is the output factor matrix, X is the b×(N+1) order input factor matrix, and b is the number of experiments, that is, the number of solutions.

6. The method for estimating the dispatch response relationship of an upstream reservoir to a mid- and downstream control station according to claim 1, characterized in that: The S4 specifically includes: if a downstream flow requirement is given, the target data type is converted into a water level requirement according to the downstream water level-flow relationship; based on the known upstream reservoir flow sequence and flow constraints of the first n days in the cycle, with the goal of minimizing the upstream reservoir storage capacity or maximizing the downstream discharge flow, the model in S3 is used to calculate the storage process for the remaining Nn days, so as to realize the reverse control of the upstream control station over the storage and discharge process of the upstream reservoir; otherwise, the reverse process is directly calculated using the water level requirement.

Images