CN115408934B - Method for rapidly predicting response of downstream riverway of dam to reservoir outlet water volume and sand volume change - Google Patents

Method for rapidly predicting response of downstream riverway of dam to reservoir outlet water volume and sand volume change Download PDF

Info

Publication number
CN115408934B
CN115408934B CN202210987476.3A CN202210987476A CN115408934B CN 115408934 B CN115408934 B CN 115408934B CN 202210987476 A CN202210987476 A CN 202210987476A CN 115408934 B CN115408934 B CN 115408934B
Authority
CN
China
Prior art keywords
sand
model
amount
change
downstream
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210987476.3A
Other languages
Chinese (zh)
Other versions
CN115408934A (en
Inventor
张磊
黄海
王大宇
关见朝
胡智丹
鲁文
陈伟
刘彧
尹雄锐
王友胜
冯珺
赵莹
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Institute of Water Resources and Hydropower Research
Original Assignee
China Institute of Water Resources and Hydropower Research
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Institute of Water Resources and Hydropower Research filed Critical China Institute of Water Resources and Hydropower Research
Priority to CN202210987476.3A priority Critical patent/CN115408934B/en
Publication of CN115408934A publication Critical patent/CN115408934A/en
Application granted granted Critical
Publication of CN115408934B publication Critical patent/CN115408934B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
    • Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping

Abstract

The invention discloses a method for quickly predicting the response of a downstream riverway of a dam to the change of the reservoir outlet water volume and the sand volume of a reservoir, which comprises the following steps: determining a research river reach, and collecting sample data; constructing a neural network model; training a model; verifying the model; calculating the sand conveying amount and water quantity change of a downstream hydrological station caused by the sand quantity and water quantity change of the reservoir; further calculating the partial derivative relation between the sand conveying amount of the current section and the inflow amount and inflow amount of the upstream section; and judging the response of the sand conveying amount of the current section to the sand incoming amount and the water incoming amount change of the upstream section according to the calculation result. The model constructed based on the artificial neural network can avoid solving the partial derivative function relationship in the water sand mathematical model, and directly determines the partial derivative relationship through big data analysis to obtain parameters required by the model; the calculated amount is reduced, and the calculation efficiency is high.

Description

Method for rapidly predicting response of downstream riverway of dam to reservoir outlet water volume and sand volume change
Technical Field
The invention relates to the technical field of hydraulics and river dynamics, is used for carrying out water and sand transportation calculation, and particularly relates to a method for quickly predicting response of a downstream riverway (each section along the way) of a dam to reservoir delivery water volume and sand volume change.
Background
In natural rivers, the motion of sand-laden water flow is influenced by various factors, and all parameters have complex nonlinear relations, so that at present, a water sand mathematical model is mostly adopted to carry out numerical simulation calculation in order to accurately describe the motion process of water sand. The control equation of the water sand mathematical model is a complete Saint-Venn equation set and a sediment motion basic equation, the physical significance is clear, the calculation precision is high, however, no matter the water sand model is a one-dimensional, two-dimensional or three-dimensional water sand model, or existing mature commercial software is adopted, when the simulation calculation is carried out, complex pretreatment work needs to be carried out, including preparation and processing of topographic data and water sand data, boundary demarcation and grid demarcation, and related parameters need to be continuously adjusted and proper water sand series need to be selected in the process of calibration and verification according to the characteristics of a research area, and the calculation can be carried out on the basis. When the terrain data or the water sand data are missing or incomplete, the application of the water sand mathematical model is greatly limited.
Aiming at the problem of influence of changes of the downstream river reach of the dam on the reservoir outlet water quantity and the sand quantity of the reservoir on the sand conveying quantity of each section of the downstream river reach of the dam, previous researches are mostly carried out based on actual measurement data analysis and water-sand model calculation, the workload is large, and the selection of water-sand series is difficult. In fact, the key to solve this problem is to determine the partial derivative relationship between the sand transport amount of the adjacent downstream section in the river and the runoff and sand amount of the upstream section. In particular, for a single river, each break along the wayAmount of sand transported in dough S i Water flow q of adjacent cross section upstream thereof i-1 And amount of sand S i-1 The correlation is as follows:
S i ＝f(Q i-1 ,S i-1 )i＝0,1,2,… (1)
in the formula, S i Is the sand transportation quantity of the section i, Q i-1 And S i-1 The water and sand transportation of the adjacent section at the upstream of the section i are respectively, and when the subscript i =0, the section corresponds to the inlet section. As shown in (1), for the ith cross section (i)>0) It should be:
Figure BDA0003802376520000011
the formula (2) defines the recursion relation between the changes of the sand transportation amount of each section from the downstream to the upstream in the riverway. For natural river channels, although the correlation of the water and sand conveying amount between adjacent sections is strong, the influence factors are numerous, the explicit functional relationship between the two is difficult to derive, and the reason is that numerical calculation is often performed by adopting a numerical model.
An Artificial Neural Network (ANN) technology is a big data analysis technology, the technology directly analyzes and researches data, potential rules of the data are mined, a fixed mode is not assumed in advance, wrong assumptions made by people in consideration of lack of time or wrong recognition can be avoided, and the obtained model can objectively describe the internal structure of a system. But the method is not applied to the research of the response of the downstream river sediment flux of the reservoir dam to the change of the reservoir outlet water quantity and the sand quantity at present.
Disclosure of Invention
The invention aims to overcome the technical defects and provide a method capable of efficiently and quickly predicting the response of a downstream riverway (each section along the way) of a dam to the change of the water quantity and the sand quantity of the reservoir. The invention provides a method for quickly predicting response of a downstream riverway of a dam to reservoir delivery water volume and sand volume change based on an Artificial Neural Network (ANN), which comprises the following steps:
firstly, determining a research river reach, and acquiring sample data: in order to ensure that the ANN model can reflect the water and sand delivery rule of the reservoir after operation, the upper boundary and the lower boundary of the river reach are determined and researched. And selecting sample data to study the actual measured daily average flow and sand content of the downstream dam of the river reach in n years of m hydrological stations. m is more than or equal to 2,n is more than or equal to 5. The data files are stored in the Excel file in advance according to a specified format.
Secondly, constructing a neural network BP model: in order to determine the recursion relation in the formula (2), based on an artificial neural network technology, a TensorFlow system is utilized, the daily average flow and the sand content of a hydrological station at the upper boundary of a researched river reach are used as input data, the daily average flow and the sand content of a hydrological station at the lower boundary of the river reach are used as output data, and a neural network BP model is respectively constructed for m-1 river reaches;
thirdly, model training: and (3) importing the sample data into a TensorFlow system aiming at a certain river reach, and starting training the model by using 70% of the sample data. In the training process, the structure and the training parameters (including the weight and the offset) of a multi-layer feed-forward (Back Propagation & BP) neural network are set. The interface provides a loss function decay map during training.
Fourthly, model verification: and starting verification calculation by using the trained model. And inputting the daily average flow and sand content data of the upper boundary in the trained model by using the rest 30% of sample data, outputting the daily average flow and sand content data of the lower boundary, and comparing the obtained calculation result with the actually measured data. If the errors of the calculated values of the daily average flow and the sand content and the measured values are within the acceptable range, the model can reasonably reflect the water and sand conveying relation between adjacent sections on the downstream side of the dam after the reservoir operates, and if the errors do not meet the requirements, the model needs to be trained again in the third step.
Step five, calculating the sand output amount of the reservoir and the sand conveying amount change of a downstream hydrological station caused by the water amount change: after m-1 BP models for building the neural network are trained and verified, calculating based on the neural network modelVariation dS of sand quantity of reservoir 0 Sand conveying amount change dS of each hydrological station along the way of the downstream i (ii) a Calculating reservoir outlet water quantity change dQ based on neural network BP model 0 Sand conveying amount change dS of each hydrological station along the way of the downstream i (ii) a Calculating reservoir delivery water volume change dQ based on neural network model 0 Water delivery rate change dQ of downstream and downstream hydrographic stations i ；
Step six, further calculating the partial derivative relation between the sand transportation amount of the current section and the inflow amount and inflow amount of the upstream section based on the calculation result of the step five: wherein the partial derivative relationship between the sand conveying amount of the front section and the sand coming amount of the upstream section is determined by dS i /dS i-1 To calculate; the partial derivative relation between the sand transportation amount of the current section and the water inflow amount of the upstream section is determined by dS i /dQ i-1 To calculate;
step seven: and judging the response size of the current section sand transportation amount change to the upstream section sand coming amount and the inflow amount change according to the calculation result of the step six.
Further, the expression of the loss function in step three is shown in formula (3):
Figure BDA0003802376520000031
wherein the content of the first and second substances,
Figure BDA0003802376520000032
in the formula: e is a loss function, and N is the number of samples; i represents a node number; x is the number of i Calculating a value for the network of the ith output node training sample; y is i Training the network expectation value of a sample for the ith output node; time t x i Is represented by the formula (5):
Figure BDA0003802376520000033
in the formula:
Figure BDA0003802376520000034
is the flow of the inode at the time t, < >>
Figure BDA0003802376520000035
Is the sand content of the i node at the time t, < >>
Figure BDA0003802376520000036
Is the flow of the i-1 node at the time t->
Figure BDA0003802376520000037
And determining the function relationship f by the weight w in the model for the sand content of the i-1 node at the time t, wherein the weight w is automatically generated, and continuously adjusting according to the change of the loss function E in the data training. Specifically, when x is initially obtained i When the calculated value of (A) causes the loss function E to be larger and exceed the error acceptable range, the weight w of the next step t+1 Will be at w t And adjusting according to the value E, performing the next calculation until the value of the loss function E is gradually reduced to an acceptable degree or reaches a preset learning frequency, and quitting the training.
Further, in the fourth step: the acceptable range of the error between the average daily flow and the measured value is as follows: the calculation error is less than or equal to 0.01; the acceptable range of the error between the sand content and the measured value is as follows: the calculation error is less than or equal to 0.05.
The invention has the beneficial effects that:
the invention solves the technical problem of quickly predicting the response of the downstream river channel of the dam to the change of the reservoir outlet water volume and the sand volume by applying the BP model in the artificial neural network technology in the field of river sediment dynamics. Compared with the prior method, the method has the following advantages and innovations:
1) The model constructed based on the artificial neural network can avoid solving the partial derivative function relationship in the water sand mathematical model, and the partial derivative relationship in the formula (2) can be directly determined through big data analysis to obtain parameters required by the model;
2) Compared with the water-sand mathematical model, the method has the advantages that the calculation amount is obviously reduced, and the calculation efficiency is high.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a graph of loss function decay for a Branch City ANN model training process;
FIG. 3 is a comparison between the calculated value and the measured value of the average daily flow ANN in 2011-2018 of Branch City;
FIG. 4 is a comparison between the calculated value and the measured value of the sand content ANN in the Zhicheng area in 2011-2018 years;
FIG. 5 shows the comparison between the calculated value and the measured value of the average daily flow ANN in 2011-2018 in Sansho;
FIG. 6 shows that the calculated value of the sand content ANN in Sha city in 2011-2018 is compared with the measured value;
FIG. 7 comparison of calculated and actual values of the flow ANN for proctor 2011-2018 year;
FIG. 8 shows the comparison between the calculated value of ANN and the measured value of the sand content in the proctor 2011-2018 every day;
FIG. 9 is a comparison between the calculated value and the measured value of the daily average flow ANN of the spiral hills 2011-2018;
FIG. 10 is a comparison between the calculated value and the measured value of the sand content ANN of the spiral mountains 2011-2018 in the day;
FIG. 11 is a comparison between the calculated value and the measured value of the average daily flow ANN in Hankou 2011-2018;
FIG. 12 is a comparison of the calculated value and the measured value of the sand content ANN in Hankou 2011-2018 every day;
FIG. 13 comparison of calculated value and measured value of the daily average flow ANN of Jiujiang 2011-2018;
FIG. 14 is a comparison of the calculated value and the measured value of the sand content ANN in Jiujiang 2011-2018;
FIG. 15 comparing the calculated value and the measured value of the average daily flow ANN in Datong 2011-2018;
FIG. 16 shows the comparison between the calculated value and the actual value of the sand content ANN in Datong 2011-2018.
Detailed Description
Example 1:
a method for quickly predicting the response of a downstream riverway of a dam to the change of the reservoir outlet water volume and the sand volume,
firstly, determining a research river reach and determining sample data. Determining an upper boundary and a lower boundary of a research river reach; the sample data comprises the measured daily average flow and sand content within 16 years of 8 hydrological stations on the downstream of the dam after the reservoir is built in the research river reach. In this embodiment, in order to ensure that the ANN model can reflect the water and sand transportation law from yichang to datong since the three gorges reservoir operates in 2003, the sample data adopts 8 hydrological stations 2003-2018 (16 years) of the daily average flow and the sand content in yichang, the branch city, the sandy city, the proclaim, the spiral mountain, the hankou, the jiujiang and the datong. The data files are stored in the Excel file in advance according to a specified format.
Secondly, constructing a neural network BP model: in order to determine the recursion relationship in the formula (2), based on an artificial neural network technology, a neural network model is respectively constructed for 7 river segments (Yichang-Zhicheng, zhicheng-Shashi, shashi-Guoli, gunjiang-Jiujiang, jiujiang-Datong) by taking the daily average flow and sand content of the hydrological station at the upper boundary of the river segment as input factors and the daily average flow and sand content of the hydrological station at the lower boundary of the river segment as output factors;
thirdly, model training: and importing the sample data into a TensorFlow system, and starting training the model by using 70% of the sample data for 7 river reach. In the training process, a structure and training parameters (including weight and bias) of a multi-layer feed-forward (Back Propagation & BP) neural network are set. The interface provides a loss function attenuation graph during the training process, as shown in FIG. 2, which shows the loss function variation graph of the output Branch City station when training the Yichang-Branch City model, wherein the expression of the loss function is
Figure BDA0003802376520000051
Wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003802376520000052
in the formula: e is a loss function, and N is the number of samples; i represents a node number; x is the number of i Calculating a value for the network of the ith output node training sample; y is i Train for ith output nodeNetwork expected values of training samples; time t x i Is represented by the formula (5):
Figure BDA0003802376520000053
in the formula:
Figure BDA0003802376520000054
is the flow of the inode at the time t>
Figure BDA0003802376520000055
Is the sand content of the i node at the time t, < >>
Figure BDA0003802376520000056
Is the flow of the i-1 node at the time t, < >>
Figure BDA0003802376520000057
And determining the function relationship f by the weight w in the model for the sand content of the i-1 node at the time t, wherein the weight w is automatically generated, and continuously adjusting according to the change of the loss function E in the data training. Specifically, when x is initially obtained i When the calculated value of (A) causes the loss function E to be larger and exceed the error acceptable range, the weight w of the next step t+1 Will be at w t And adjusting according to the value E, performing the next calculation until the value of the loss function E is gradually reduced to an acceptable degree or reaches a preset learning frequency, and quitting the training.
In this embodiment, FIG. 1 shows that the loss function has been reduced to 0.04697, which is already within an acceptable range, and therefore training is complete.
Fourthly, model verification: and starting verification calculation by using the trained model. Specifically, the daily average flow rate and the sand content data of the upper boundary are input, and the daily average flow rate and the sand content data of the lower boundary are output, and the results are shown in fig. 3 to 16. Wherein, fig. 3 and fig. 4 represent that the distribution results of the daily average flow and the daily average sand content of the branch city station 2011-2018 obtained by the calculation of the Yichang-branch city river reach are in good accordance with the measured values. By analogy, fig. 5 to 16 represent the comparison results of the average daily flow and sand content of the branch city, the sand city, the proctor, the spiral mountain, the Hankou, the Jiujiang and the Datong obtained by calculation according to the model of the branch city, the sand city, the proctor, the spiral mountain, the Hankou, the Jiujiang and the Jiujiang river, respectively, and the actually measured values, and the results are consistent with each other, which also shows that the model can reasonably reflect the relationship between water and sand transportation between adjacent sections along the route from Yichang to Datong after the three gorges reservoir operates.
Fifthly, calculating the sand transportation amount change of the downstream hydrological station caused by the sand amount and water amount change of the reservoir outlet: after 7 models are trained and verified, calculating reservoir ex-warehouse sand quantity change dS based on the neural network model 0 Sand conveying amount change dS of each hydrological station along the way of the downstream i (ii) a Calculating reservoir outlet water quantity change dQ based on neural network BP model 0 Sand conveying amount change dS of each hydrological station along the way of the downstream i (ii) a Calculating reservoir outlet water quantity change dQ based on neural network BP model 0 Water delivery rate change dQ of downstream and downstream hydrographic stations i . In this embodiment, the sand transportation amount of each hydrological station on the downstream side after the change of the sand discharge amount (Yichang station) of the three gorges reservoir is developed based on the model. The effect size is expressed as a dimensionless percentage and the results are shown in table 1. The quantitative change of each section at the downstream to the response of the Yichang station under different amplitude of sand transportation can be found out from the table after the three gorges run.
TABLE 1 changes in the amount of sand transported in the current section due to changes in the amount of sand supplied to the upstream section
Figure BDA0003802376520000061
Figure BDA0003802376520000071
By looking up the table 1, the variation of sand transportation amount of different sections of different hydrological stations at the downstream of the dam can be quickly predicted after the change of sand transportation amount of the three gorges (namely the sand transportation amount of the Yichang station) is brought out of the reservoir.
Similarly, the sand and water transportation amount of each hydrological station at the downstream along the way is changed after the water output amount (Yichang station) of the three gorges reservoir is changed based on the model. The effect size is expressed as a dimensionless percentage and the results are shown in tables 2 and 3. The quantitative changes of the sand and water transportation quantity of each section at the downstream under different amplitude of variation of runoff of the Yichang station after the three gorges run can be found from the table.
TABLE 2 changes in sand transport capacity of the current section due to changes in water inflow of the upstream section
Figure BDA0003802376520000072
By looking up the table 2, the sand transportation amplitude of the cross section of different hydrological stations at the downstream of the dam can be rapidly predicted after the water quantity change of the three gorges (namely the water inflow of the Yichang station) is discharged from the reservoir.
TABLE 3 Water transfer variation in the Current Cross section due to the variation in the inflow of the upstream Cross section
Figure BDA0003802376520000073
Figure BDA0003802376520000081
By looking up the table 3, the water transfer capacity variation of different cross sections of the hydrological station at the downstream of the dam can be quickly predicted after the water volume change of the three gorges (namely the water volume of the Yichang station) is output from the reservoir.
Sixthly, further calculating the partial derivative relation between the sand transportation amount of the current section and the inflow amount and inflow amount of the upstream section based on the calculation result of the step five: wherein the partial derivative relationship between the sand conveying amount of the front section and the sand coming amount of the upstream section is determined by dS i /dS i-1 To calculate; the partial derivative relation between the sand conveying amount of the current section and the water inflow amount of the upstream section is determined by dS i /dQ i-1 To calculate.
In this embodiment, based on tables 1, 2 and 3, the partial derivative relationship between the sand transportation amount of the current section and the inflow amount and inflow amount of the adjacent section at the upstream, that is, the deviation relationship between the sand transportation amount of the current section and the inflow amount and inflow amount of the adjacent section at the upstream, can be further estimatedCan be obtained in the formula (2)
Figure BDA0003802376520000082
And &>
Figure BDA0003802376520000083
Is greater than or equal to, wherein>
Figure BDA0003802376520000084
Is shown in Table 4, is present>
Figure BDA0003802376520000085
The values of (A) are shown in Table 5. Wherein Table 4 is based on Table 1 and Table 5 is based on tables 2 and 3, e.g., the branched City station based on a 50% reduction in sand input>
Figure BDA0003802376520000086
Station in sand market>
Figure BDA0003802376520000087
Figure BDA0003802376520000088
The same calculation process is carried out in other hydrological stations.
TABLE 4 influence of changes in the amount of incoming sand from the upstream section on the changes in the amount of sand transported from the current section
Figure BDA0003802376520000089
Figure BDA0003802376520000091
TABLE 5 influence of inflow water volume change of upstream section on sand transportation volume change of current section
Figure BDA0003802376520000092
Seventh: as can be seen from tables 4 and 5, the change of the sand transportation amount of the current section is more closely related to the change of the sand supply amount of the adjacent upstream section, and the change of the water transportation amount of the upstream section does not greatly contribute to the change of the sand supply amount of the branch and city section. However, in the sand city and the downstream, when the water amount from the upstream section changes, the sand transportation amount of the current section is influenced to a certain extent. From tables 4 and 5, quantitative values of the effects thereof can be found. The achievement of the invention can provide technical support for river regulation, reservoir dispatching optimization and the like.
Finally, it should be noted that the above is only intended to illustrate the technical solution of the present invention and not to limit it, and although the present invention has been described in detail with reference to the preferred arrangement, it should be understood by those skilled in the art that modifications or equivalent substitutions can be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.

Claims (3)

1. A method for rapidly predicting response of a downstream riverway of a dam to changes of reservoir outlet water volume and sand volume is characterized by comprising the following steps: the method comprises the following steps:
the method comprises the following steps: determining a research river reach, and collecting and finishing sample data: determining an upper boundary and a lower boundary of a research river reach; the sample data comprises actual measured daily average flow and sand content within n years of m hydrological stations on the downstream of the dam along the course of building the reservoir at the research river reach, wherein m is more than or equal to 2,n and more than or equal to 5;
step two: constructing a neural network BP model: based on an artificial neural network technology, a TensorFlow system is utilized, the daily average flow and the sand content of a hydrological station at the upper boundary of a researched river reach are used as input data, the daily average flow and the sand content of the hydrological station at the lower boundary of the river reach are used as output data, and a neural network BP model is respectively constructed for m-1 river reaches;
step three: model training: importing the sample data into a TensorFlow system aiming at each river reach, and starting training the neural network BP model constructed in the second step by utilizing 70% of the sample data; in the training process, setting a structure and training parameters of a multi-layer feed-forward type back propagation neural network; the interface provides a loss function attenuation map in the training process;
step four: and (3) model verification: inputting the daily average flow and sand content data of the upper boundary in the neural network BP model trained in the third step by using the rest 30 percent of sample data, outputting the daily average flow and sand content data of the lower boundary, and comparing the obtained calculation result with the actually measured data; if the errors of the calculated values of the daily average flow and the sand content and the measured values are within the acceptable range, the model can reasonably reflect the water and sand conveying relation between adjacent sections of the downstream side of the dam along the way after the reservoir operates, and if the errors do not meet the requirements, the model returns to the third step to retrain the model;
step five, calculating the sand output amount of the reservoir and the sand conveying amount change of a downstream hydrological station caused by the water amount change: after m-1 neural network BP model construction training and verification are finished, calculating reservoir ex-warehouse sand change dS based on the neural network BP model 0 Sand conveying amount change dS of each hydrological station along the way of the downstream i (ii) a Calculating the change dQ of the water quantity of the reservoir discharged from the reservoir based on the neural network BP model 0 Sand conveying amount change dS of each hydrological station along the way of downstream i (ii) a Calculating reservoir outlet water quantity change dQ based on neural network BP model 0 Water delivery rate change dQ of each hydrological station downstream and downstream i ；
Step six, further calculating the partial derivative relation between the sand transportation amount of the current section and the inflow amount and inflow amount of the upstream section based on the calculation result of the step five: wherein the partial derivative relationship between the sand conveying amount of the front section and the sand coming amount of the upstream section is determined by dS i /dS i-1 To calculate; the partial derivative relation between the sand conveying amount of the current section and the water inflow amount of the upstream section is determined by dS i /dQ i-1 To calculate;
step seven: and judging the response size of the current section sand transportation amount change to the upstream section sand coming amount and the inflow amount change according to the calculation result of the step six.
2. The method for rapidly predicting the response of the riverway downstream of the dam to the change of the reservoir outlet water volume and the sand volume according to claim 1, wherein the method comprises the following steps: the expression of the loss function in step three is shown in formula (3):
Figure QLYQS_1
wherein the content of the first and second substances,
Figure QLYQS_2
in the formula: e is a loss function, and N is the number of samples; i represents a node number; x is the number of i Calculating a value for the network of the ith output node training sample; y is i Training the network expectation value of a sample for the ith output node; time t x i Is expressed by formula (5):
Figure QLYQS_3
in the formula:
Figure QLYQS_4
is the flow of the inode at the time t, < >>
Figure QLYQS_5
Is the sand content of the i node at the time t, < >>
Figure QLYQS_6
Is the flow of the i-1 node at the time t, < >>
Figure QLYQS_7
And determining the function relationship f by the weight w in the model for the sand content of the i-1 node at the time t, wherein the weight w is automatically generated, and continuously adjusting according to the change of the loss function E in the data training.
3. The method for rapidly predicting the response of the riverway downstream of the dam to the change of the reservoir outlet water volume and the sand volume according to claim 1, wherein the method comprises the following steps: in the fourth step: the acceptable range of the error between the average daily flow and the measured value is as follows: the calculation error is less than or equal to 0.01; the acceptable range of the error between the sand content and the measured value is as follows: the calculation error is less than or equal to 0.05.
CN202210987476.3A 2022-08-17 2022-08-17 Method for rapidly predicting response of downstream riverway of dam to reservoir outlet water volume and sand volume change Active CN115408934B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210987476.3A CN115408934B (en) 2022-08-17 2022-08-17 Method for rapidly predicting response of downstream riverway of dam to reservoir outlet water volume and sand volume change

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210987476.3A CN115408934B (en) 2022-08-17 2022-08-17 Method for rapidly predicting response of downstream riverway of dam to reservoir outlet water volume and sand volume change

Publications (2)

Publication Number Publication Date
CN115408934A CN115408934A (en) 2022-11-29
CN115408934B true CN115408934B (en) 2023-04-18

Family

ID=84159188

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210987476.3A Active CN115408934B (en) 2022-08-17 2022-08-17 Method for rapidly predicting response of downstream riverway of dam to reservoir outlet water volume and sand volume change

Country Status (1)

Country Link
CN (1) CN115408934B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117556184B (en) * 2024-01-12 2024-03-26 长江水利委员会水文局 Method, system and medium for reducing sediment transport amount of downstream river channel of reservoir in sandy area

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107451682A (en) * 2017-07-13 2017-12-08 中国水利水电科学研究院 A kind of city tidal reach Water Requirement Forecasting Methodology based on neutral net
CN107558432A (en) * 2016-09-06 2018-01-09 长江水利委员会长江科学院 A kind of method for quick predicting of Three Gorges Reservoir flood season outbound sand peak silt content
CN108052761A (en) * 2017-12-25 2018-05-18 贵州东方世纪科技股份有限公司 A kind of Prediction of Landslide

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9433843B2 (en) * 2011-11-18 2016-09-06 Tomtom International B.V. Method and apparatus for creating cost data for use in generating a route across an electronic map

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107558432A (en) * 2016-09-06 2018-01-09 长江水利委员会长江科学院 A kind of method for quick predicting of Three Gorges Reservoir flood season outbound sand peak silt content
CN107451682A (en) * 2017-07-13 2017-12-08 中国水利水电科学研究院 A kind of city tidal reach Water Requirement Forecasting Methodology based on neutral net
CN108052761A (en) * 2017-12-25 2018-05-18 贵州东方世纪科技股份有限公司 A kind of Prediction of Landslide

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
胡光伟 等.三峡工程运行对洞庭湖与荆江三口关系的影响分析.《海洋与湖泊》.2014,第第45卷卷(第第3期期),全文. *
陈震宇.澜沧江水沙关系与趋势预测.《中国优秀硕士毕业论文集》.2018,全文. *

Also Published As

Publication number Publication date
CN115408934A (en) 2022-11-29

Similar Documents

Publication Publication Date Title
CN111125969A (en) Cross-reservoir basin river runoff calculation method and system
CN107885958B (en) Method for calculating dirt holding capacity of plain tidal river network area
Hipel et al. Intervention analysis in water resources
CN111126847B (en) Cascade reservoir short-term optimization scheduling method and system coupled with riverway water power process
CN107091911B (en) river water quality prediction method
CN115408934B (en) Method for rapidly predicting response of downstream riverway of dam to reservoir outlet water volume and sand volume change
CN110705171A (en) Method for treating water environment of tidal river network based on MIKE model
CN113762618B (en) Lake water level forecasting method based on multi-factor similarity analysis
CN113449993B (en) Urban water source water supply scheduling method
CN113569438B (en) Urban flood model construction method based on multisource rainfall fusion and real-time correction
CN111553394A (en) Reservoir water level prediction method based on cyclic neural network and attention mechanism
CN115983483A (en) Method for predicting river water level under reservoir building future water regime scene by inverse calculation of relation between coupled cyclic neural network and river water level flow
CN106383997A (en) Calculation method for carrying out inverse estimation on Three Gorges Reservoir interval inflow process
CN114357868A (en) Multi-target cooperative scheduling method and device for complex flood control system
CN109326332B (en) Tidal water network area water environment capacity calculation method
Salmasi et al. Genetic algorithms for optimizing stepped spillways to maximize energy dissipation
CN115730739A (en) Urban waterlogging prediction method based on LSTM neural network
Song et al. Research on the influence of diversion flow ratio on the water environment in Nanjing
CN112529247B (en) Dry-flow sand discharge optimal scheduling method and system based on combined water supplementing of tributary reservoir
Yeh et al. Central arizona project: operations model
CN106383935A (en) Reduction method of natural average daily water and sediment process of reservoir dam site
CN106228277A (en) A kind of reservoir operation forecast information effective accuracy recognition methods based on data mining
Papalaskaris et al. Artificial neural network for bed load transport rate in Nestos River, Greece
CN110706126A (en) Method and device for establishing flow time-lag matching model of upstream and downstream hydropower stations
CN115017727B (en) Sewage collection simulation method based on Ma Sijing root method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant