CN116562176A - Runoff simulation method based on coupled neural network and hydrologic physical model - Google Patents
Runoff simulation method based on coupled neural network and hydrologic physical model Download PDFInfo
- Publication number
- CN116562176A CN116562176A CN202310841797.7A CN202310841797A CN116562176A CN 116562176 A CN116562176 A CN 116562176A CN 202310841797 A CN202310841797 A CN 202310841797A CN 116562176 A CN116562176 A CN 116562176A
- Authority
- CN
- China
- Prior art keywords
- soil
- data
- runoff
- hydrologic
- river basin
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 76
- 238000004088 simulation Methods 0.000 title claims abstract description 59
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 24
- 230000008569 process Effects 0.000 claims abstract description 27
- 230000008878 coupling Effects 0.000 claims abstract description 11
- 238000010168 coupling process Methods 0.000 claims abstract description 11
- 238000005859 coupling reaction Methods 0.000 claims abstract description 11
- 239000002689 soil Substances 0.000 claims description 177
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 37
- 238000001556 precipitation Methods 0.000 claims description 29
- 229920006395 saturated elastomer Polymers 0.000 claims description 27
- 238000007781 pre-processing Methods 0.000 claims description 16
- 239000011159 matrix material Substances 0.000 claims description 11
- 239000004576 sand Substances 0.000 claims description 10
- 239000002245 particle Substances 0.000 claims description 8
- 239000005416 organic matter Substances 0.000 claims description 7
- 238000005520 cutting process Methods 0.000 claims description 6
- 239000004016 soil organic matter Substances 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 4
- 239000004927 clay Substances 0.000 claims description 4
- 230000008020 evaporation Effects 0.000 claims description 4
- 238000001704 evaporation Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 abstract description 8
- 238000010801 machine learning Methods 0.000 abstract description 6
- 238000012545 processing Methods 0.000 abstract description 6
- 238000011161 development Methods 0.000 abstract description 4
- 230000007246 mechanism Effects 0.000 abstract description 3
- 238000012502 risk assessment Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 5
- 238000009826 distribution Methods 0.000 description 4
- 238000007726 management method Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000002203 pretreatment Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000008595 infiltration Effects 0.000 description 1
- 238000001764 infiltration Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 206010027175 memory impairment Diseases 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000005068 transpiration Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
-
- 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
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention discloses a runoff simulation method based on a coupling neural network and a hydrologic physical model, belongs to the technical field of hydrologic models, and provides the runoff simulation method based on the coupling neural network and the hydrologic physical model. According to the invention, a runoff simulation scheme of coupling the neural network and the hydrologic physical model is established, so that a simulation result is more accurate compared with a single machine learning mode, a processing process is closer to the hydrologic physical mechanism, and the application and development of a runoff simulation technology in early warning and risk assessment of waterlogging disasters can be further promoted.
Description
Technical Field
The invention belongs to the technical field of hydrologic models, and particularly relates to a runoff simulation method based on a coupling neural network and a hydrologic physical model.
Background
The runoff simulation is a key step for realizing the fine management of water resources, and can play an important role in flood prevention and drought resistance decision making, ecological environment protection and the like. Moreover, the runoff simulation is also an important component part for analyzing the waterlogging disaster process, has important significance in revealing and analyzing the characteristics and rules of runoffs, and is an application foundation for realizing scientific planning and reasonable allocation in the waterlogging disasters. The radial flow simulation technology is developed, the formation and conversion process of hydrologic information in space-time can be conceptualized, and scientific understanding of hydrologic process is deepened. With the continuous update of computing power and equipment, the analysis means of runoff simulation have been separated from the traditional experience-statistics category and rapidly developed towards the direction of combining computing simulation. The accurate runoff simulation has strategic significance for the early warning and management of waterlogging disasters, and the simulation accuracy degree is significant for the production, life and economic safety.
High latitude areas such as north China runoff variations are unstable due to changes in water supply caused by marine air transportation or geographical topographical features, and have a significant impact on extreme flood events caused thereby. The non-linearity, randomness and uncertainty of the runoff variation makes the runoff simulation work difficult. Traditional runoff simulation models are mainly based on meteorological physical processes, and the runoff is described by processing space discrete data through mathematical equations. Thus, there are two limitations that affect the accuracy of the runoff simulation: firstly, cognition, modeling and calculation of complex physical processes are high in cost and energy consumption. Secondly, the lack of available data for the hydrologic system leads to inaccuracy in the simulation results. How to effectively collect hydrologic data and develop hydrologic models with low data dependence. At present, the machine learning technology is well applied in the hydrology field, and the machine learning technology and the traditional runoff simulation method are mutually complemented, so that the machine learning technology deserves intensive development and research to form a reliable application technology. Furthermore, runoff simulation is often difficult to achieve or the simulation results are not reliable enough due to the lack of driving variables such as meteorological data. The satellite meteorological products with continuous observation and reliable data are helpful for estimating the hydrologic information of the areas without runoff records.
Therefore, the satellite meteorological-runoff simulation method for coupling the neural network and the hydrophysics is particularly important, and the advantages are complemented by developing the characteristics of the two models at the same time, so that the runoff simulation precision is improved.
Disclosure of Invention
Aiming at the defects in the prior art, the runoff simulation method based on the coupled neural network and the hydrologic physical model solves the problems that the conventional simulation method is inaccurate in long-time-sequence hydrologic prediction and has larger deviation in peak flow simulation.
In order to achieve the aim of the invention, the invention adopts the following technical scheme: a runoff simulation method based on a coupled neural network and a hydrologic physical model comprises the following steps:
s1, collecting satellite meteorological data of a target river basin, site record runoff data and related auxiliary data, and preprocessing the satellite meteorological data, the site record runoff data and the related auxiliary data to obtain a preprocessing result;
the pretreatment result of the satellite meteorological data comprises precipitation data, evapotranspiration data, temperature data and soil humidity data, and the pretreatment result of the site recorded runoff data is specifically observation runoff;
s2, inputting precipitation data, evapotranspiration data and temperature data into an LSTM network to obtain a preliminary runoff result;
s3, calculating the saturated water conductivity of the soil according to the pretreatment result of the related auxiliary data, and further obtaining a soil type CN value of the target river basin;
s4, inputting the soil type CN value, the precipitation data and the soil humidity data of the target river basin into an MMSCS model to obtain the runoff depth of the target river basin;
s5, performing differential feature analysis according to the primary runoff result and the observed runoff, and establishing a regression relation between the primary runoff result with the differential value higher than a set threshold value and the runoff depth of the target river basin and the observed runoff to obtain a final runoff simulation result.
Further: in the step S1, the pretreatment method specifically comprises the following steps:
(1) Image stitching is carried out on the acquired satellite meteorological data, image cutting is carried out according to the target river basin, missing or unreliable pixels in the image are removed, average results of daily precipitation, temperature, soil humidity and evaporation of the target river basin in the image are extracted, and preprocessing results of the satellite meteorological data are obtained;
(2) Extracting flow information of a corresponding target river basin in site-recorded runoff data, and electronically recording according to a daily scale to obtain a preprocessing result of the site-recorded runoff data;
(3) Based on the collected related auxiliary data, extracting the soil texture, the soil organic matter content and the land utilization data of the target river basin in the related auxiliary data through image cutting, and obtaining the preprocessing result of the related auxiliary data.
Further: the step S2 is specifically as follows: inputting precipitation data, evapotranspiration data and temperature data into an LSTM network to obtain a primary runoff result;
the LSTM network comprises a forget gate, an input gate, a unit state and an output gate;
wherein the forgetting doorF t The expression of (2) is specifically:
in the method, in the process of the invention,W f as a first matrix of weights,b f for the first offset vector, σ (·) is the sigmoid function,h t-1 in order to hide the nodes of the network,x t for inputting data, subscriptsfAndtrespectively forgetting and time interval states;
the input doorI t The expression of (2) is specifically:
in the method, in the process of the invention,W i as a second matrix of weights,b i is a second offset vector;
the state of the unitC t The expression of (2) is specifically:
in the method, in the process of the invention,update value for cell state, +.>For the hidden layer state of the previous moment, +.>Performing dot multiplication operation;
the output doorO t The expression of (2) is specifically:
in the method, in the process of the invention,W o as a result of the third weight matrix,b o for the third offset vector, subscript o represents the output state.
Further: the step S3 comprises the following substeps:
s31, calculating the saturated water conductivity of the soil according to the soil texture and the organic matter content of the soil;
s32, obtaining a target river basin soil type CN value according to the saturated water conductivity of the soil and the land utilization data.
Further: in the step S31, the soil texture comprises a clay content and a sand content;
calculating the saturated water conductivity of soilKThe expression of (2) is specifically:
in the method, in the process of the invention,Cin order to achieve a content of the sticky particles,Sfor the sand content of the sand particles,Ois the organic matter content of the soil,z 1 ~z 4 is constant.
Further: the step S32 specifically includes:
classifying hydrologic soil groups according to the saturated water conductivity of the soil to obtain hydrologic soil group types, and looking up a table according to land utilization data and the hydrologic soil group types to obtain a target river basin soil type CN value;
the hydrologic soil group classification comprises A, B, C and D, and the method for carrying out the hydrologic soil group classification specifically comprises the following steps:
obtaining a hydrologic soil group class A when the saturated water conductivity of the soil is greater than 180, obtaining a hydrologic soil group class B when the saturated water conductivity of the soil is within a range of 18-180, obtaining a hydrologic soil group class C when the saturated water conductivity of the soil is within a range of 1.8-18, and obtaining a hydrologic soil group class D when the saturated water conductivity of the soil is less than 1.8;
the land utilization data comprise paddy fields, dry lands, forests, shrubs, grasslands, channels, town lands and bare lands, and the method for acquiring the CN value of the soil type of the target river basin comprises the following steps:
when the land utilization data is paddy fields, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 60, 70, 80 and 80;
when the land utilization data is dry land, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 90, 95 and 95;
when the land utilization data is woodland, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are 35, 60, 70 and 80 respectively;
when the land use data is a shrub land, then the soil type CN values obtained in combination with the hydrologic soil group categories A, B, C and D are 20, 55, 70 and 75, respectively;
when the land use data is grasslands, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are 40, 60, 78 and 80 respectively;
when the land utilization data is a canal, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 80, 85, 90 and 90;
when the land utilization data is urban land, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 90, 95 and 95;
when the land use data is bare land, then the soil type CN values obtained in combination with the hydrologic soil group categories A, B, C and D are 75, 85, 95 and 95, respectively.
Further in the step S4, the soil humidity data includes initial soil humidity, critical soil humidity, potential loss and potential maximum water holding capacity, and the precipitation data is specifically precipitation;
the expression for obtaining the runoff depth Q of the target river basin is specifically:
wherein V is 0 The initial soil humidity, P is precipitation, S a Is critical soil humidity, S is potential loss amount, S b Is the potential maximum water holding capacity.
Further: the expression of the potential loss amount S is specifically as follows:
in the method, in the process of the invention,for different soil type CN values for the target river basin,kis a constant value for different soil types in the target river basin.
The beneficial effects of the invention are as follows:
(1) The runoff simulation method based on the coupling neural network and the hydrologic physical model provided by the invention adopts satellite meteorological products as model driving data, can effectively solve the problems of inaccurate long-time sequence hydrologic prediction and larger deviation in peak flow simulation of the existing simulation method, and can relieve the difficulty of data deficiency faced during simulation work, thereby enhancing the capability of runoff simulation on time sequence capture, improving the processing efficiency and accuracy, and providing support for flood control management and rescue decision.
(2) According to the invention, a runoff simulation scheme of coupling the neural network and the hydrologic physical model is established, so that a simulation result is more accurate compared with a single machine learning mode, a processing process is closer to the hydrologic physical mechanism, and the application and development of a runoff simulation technology in early warning and risk assessment of waterlogging disasters can be further promoted.
Drawings
FIG. 1 is a flow chart of a runoff simulation method based on a coupled neural network and a hydrologic physical model.
FIG. 2 is a schematic diagram of the difference between LSTM and observed runoff according to the present invention.
FIG. 3 is a graph of the spatial distribution results of the soil saturation conductivity of the target river basin according to the present invention and the soil type CN value of the target river basin.
Fig. 4 is a schematic diagram of an application result of the LSTM-MMSCS model of the invention.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and all the inventions which make use of the inventive concept are protected by the spirit and scope of the present invention as defined and defined in the appended claims to those skilled in the art.
Example 1:
as shown in fig. 1, in one embodiment of the present invention, a runoff simulation method based on a coupled neural network and a hydrophysics model includes the following steps:
s1, collecting satellite meteorological data of a target river basin, site record runoff data and related auxiliary data, and preprocessing the satellite meteorological data, the site record runoff data and the related auxiliary data to obtain a preprocessing result;
the pretreatment result of the satellite meteorological data comprises precipitation data, evapotranspiration data, temperature data and soil humidity data, and the pretreatment result of the site recorded runoff data is specifically observation runoff;
s2, inputting precipitation data, evapotranspiration data and temperature data into an LSTM network to obtain a preliminary runoff result;
s3, calculating the saturated water conductivity of the soil according to the pretreatment result of the related auxiliary data, and further obtaining a soil type CN value of the target river basin;
s4, inputting the soil type CN value, the precipitation data and the soil humidity data of the target river basin into an MMSCS model to obtain the runoff depth of the target river basin;
s5, performing differential feature analysis according to the primary runoff result and the observed runoff, and establishing a regression relation between the primary runoff result with the differential value higher than a set threshold value and the runoff depth of the target river basin and the observed runoff to obtain a final runoff simulation result.
In the step S1, the pretreatment method specifically comprises the following steps:
(1) Image stitching is carried out on the acquired satellite meteorological data, image cutting is carried out according to the target river basin, missing or unreliable pixels in the image are removed, average results of daily precipitation, temperature, soil humidity and evaporation of the target river basin in the image are extracted, and preprocessing results of the satellite meteorological data are obtained;
(2) Extracting flow information of a corresponding target river basin in site-recorded runoff data, and electronically recording according to a daily scale to obtain a preprocessing result of the site-recorded runoff data;
(3) Based on the collected related auxiliary data, extracting the soil texture, the soil organic matter content and the land utilization data of the target river basin in the related auxiliary data through image cutting, and obtaining the preprocessing result of the related auxiliary data.
In this embodiment, the satellite meteorological data includes a PERSIANN-CDR precipitation product, MOD11B1 ground surface temperature product, and GLEAM soil humidity and ground surface transpiration data, and the related auxiliary data includes soil texture data, soil organic matter content data, and land utilization data.
In the step S2, the LSTM network comprises a forgetting gate, an input gate, a unit state and an output gate, and aims at the problem that the traditional neural network cannot learn long-term correlation in data, optimizes the hidden layer structure of the cell unit and introduces the working running state of the concept of the conveying belt. This allows the data to be used to determine transport or forget to screen the useful information to control the operation of the overall flow process. Forgetting, input and output information of the LSTM network is determined by the hidden layer output information of the previous moment and the input information of the current moment together, so that the state of the CEC (Constant Error Carousel) unit is updated. The key of LSTM long-term transmission memory information is the cell state, and the information can be stably transmitted through the transmission operation in the cell state and some vector interactive calculation.
The step S2 is specifically as follows: inputting precipitation data, evapotranspiration data and temperature data into an LSTM network to obtain a primary runoff result;
wherein the forgetting doorF t The expression of (2) is specifically:
in the method, in the process of the invention,W f as a first matrix of weights,b f for the first offset vector, σ (·) is the sigmoid function,h t-1 in order to hide the nodes of the network,x t for inputting data, subscriptsfAndtrespectively forgetting and time interval states;
in this embodiment, the forgetting gate is used to control the information discarded from the previous state of the cell unit, for the previous hidden layer state C t-1 Each input value of the range is 0,1]Representing the degree of forgetfulness. If the output value is 1, the memory of the previous state is not deleted, and all the memory is reserved; if the output value is 0, this means the hidden layer state C from the previous time t-1 The corresponding values are all deleted.
The input doorI t The expression of (2) is specifically:
in the method, in the process of the invention,W i as a second matrix of weights,b i is a second offset vector;
the input gate is used to control the cell unit to update the added information, adding new information to the existing state matrix. The sigmoid function determines how much new information is added to the previous state matrix, the hyperbolic tangent function being based on the previous output value h t-1 And the current input value x t-1 To generate a new state, the final updated state corresponds to the weight of the forgotten remaining information and the newly added information.
The state of the unitC t The expression of (2) is specifically:
in the method, in the process of the invention,update value for cell state, +.>For the hidden layer state of the previous moment, +.>Performing dot multiplication operation;
subscript o represents an output state
The output doorO t The expression of (2) is specifically:
in the method, in the process of the invention,W o as a result of the third weight matrix,b o is the third offset vector.
The step S3 comprises the following substeps:
s31, calculating the saturated water conductivity of the soil according to the soil texture and the organic matter content of the soil;
s32, obtaining a target river basin soil type CN value according to the saturated water conductivity of the soil and the land utilization data.
In the step S31, the soil texture comprises a clay content and a sand content;
calculating the saturated water conductivity of soilKThe expression of (2) is specifically:
in the method, in the process of the invention,Cin order to achieve a content of the sticky particles,Sfor the sand content of the sand particles,Ois the organic matter content of the soil,z 1 ~z 4 is constant.
The step S32 specifically includes:
classifying hydrologic soil groups according to the saturated water conductivity of the soil to obtain hydrologic soil group types, and looking up a table according to land utilization data and the hydrologic soil group types to obtain a target river basin soil type CN value;
the hydrologic soil group classification comprises A, B, C and D, and the method for carrying out the hydrologic soil group classification specifically comprises the following steps:
obtaining a hydrologic soil group class A when the saturated water conductivity of the soil is greater than 180, obtaining a hydrologic soil group class B when the saturated water conductivity of the soil is within a range of 18-180, obtaining a hydrologic soil group class C when the saturated water conductivity of the soil is within a range of 1.8-18, and obtaining a hydrologic soil group class D when the saturated water conductivity of the soil is less than 1.8;
in this example, the hydrographic soil group classification was used to demonstrate the general infiltration characteristics of the different underlying surfaces, as specifically shown in table 1.
TABLE 1
The land utilization data comprise paddy fields, dry lands, forests, shrubs, grasslands, channels, town lands and bare lands, and the method for acquiring the CN value of the soil type of the target river basin comprises the following steps:
when the land utilization data is paddy fields, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 60, 70, 80 and 80;
when the land utilization data is dry land, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 90, 95 and 95;
when the land utilization data is woodland, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are 35, 60, 70 and 80 respectively;
when the land use data is a shrub land, then the soil type CN values obtained in combination with the hydrologic soil group categories A, B, C and D are 20, 55, 70 and 75, respectively;
when the land use data is grasslands, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are 40, 60, 78 and 80 respectively;
when the land utilization data is a canal, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 80, 85, 90 and 90;
when the land utilization data is urban land, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 90, 95 and 95;
when the land use data is bare land, then the soil type CN values obtained in combination with the hydrologic soil group categories A, B, C and D are 75, 85, 95 and 95, respectively.
In this embodiment, the land utilization class and its corresponding CN value for the hydrologic soil group may be queried from table 2.
TABLE 2
In the step S4, the soil humidity data comprise initial soil humidity, critical soil humidity, potential loss and potential maximum water holding capacity, and the precipitation data specifically comprise precipitation;
the expression for obtaining the runoff depth Q of the target river basin is specifically:
wherein V is 0 The initial soil humidity, P is precipitation, S a Is critical soil humidity, S is potential loss amount, S b Is the potential maximum water holding capacity.
The expression of the potential loss amount S is specifically as follows:
in the method, in the process of the invention,for different soil type CN values for the target river basin,kfor constant values of different soil types in the target river basin, in generalkThe value is 0.8-1.2, the inventionkThe value takes 1.
In S5, if the difference value is higher than the threshold value, which means that the preliminary runoff result has severe variation with the observed runoff, the preliminary runoff result has errors, so that regression relation needs to be established between the preliminary runoff result with the difference value higher than the set threshold value and the runoff depth and the observed runoff of the target river basin, and the errors are reduced by fitting.
Example 2:
this embodiment is directed to one specific implementation of embodiment 1.
And collecting PERSIANN-CDR, MOD11B1 and GLEAM data of the target river basin and site record runoff information, wherein the time for acquiring the data set is 2010-2015 (test period) and 2017-2018 (verification period) of soil humidity and surface evapotranspiration data. In addition, the acquisition assistance data includes soil texture data, soil organic matter content data, and land use data.
In the LSTM network, rainfall, evaporation and surface temperature data are used as input parameters, and a runoff simulation result is obtained through model training.
As shown in FIG. 2, the present invention counts the difference between LSTM and the observed runoff (in order of difference from large to small), and can be seen at the front endIn the counting sample (the previous 500 counts), the LSTM simulation result is obviously lower than the actual observed runoff, and the maximum difference value reaches 427.75 m 3 S; the result of the latter count (500 counts later) is slightly above the 0 value, i.e. closer to the observed runoff, with relatively small errors.
The soil saturation water conductivity of the target river basin is calculated through the soil texture data (clay particles, sand particles and organic matter content), and the classification of the hydrologic soil group is carried out. On the basis, the land utilization data are acquired, and the target basin soil type CN value distribution information is acquired by a lookup table. As shown in fig. 3, the spatial distribution results of the soil saturation water conductivity and the soil type CN value of the target river basin are shown.
And calculating the radial depth of the research river basin based on the MMSCS model, wherein the radial depth is calculated by soil humidity, precipitation and CN value data. The daily distribution information of the radial flow depth in the study period was counted, and the daily average radial flow depth was 0.78 mm. The runoff depths of different watercourses are distributed differently in time, but are mainly concentrated in September to July, the runoff depths are closely related to frequent precipitation in the months, and month-by-month runoff depth information of the target watercourses can be obtained from Table 3.
TABLE 3 Table 3
And extracting samples with runoff difference values exceeding a threshold value in the test period of the target river basin. For these high value error samples, a regression relationship is established between simulated runoffs and runoff depth results obtained by adopting LSTM and MMSCS models and observed runoffs, and the parameter information of the regression equation established by the target river basin is counted in table 4: LSTM coefficients are 1.37, mmscs coefficients are 0.30, and constant term is-21.12.
TABLE 4 Table 4
As shown in fig. 4, LSTM-MMSCS model comparison work was performed during the validation period (2017 and 2018), and the LSTM-MMSCS model was applied during a certain actual precipitation event.It can be seen that the LSTM-MMSCS model is significantly better than the single LSTM model. By analyzing the fit function between simulated and observed runoffs, the correlation of the LSTM-MMSCS model (R 2 =0.94) higher than LSTM model (R 2 =0.92), closer to 1, indicating that the LSTM-MMSCS model reflects radial flow changes more accurately. In addition, the root mean square error of LSTM-MMSCS is significantly lower than that of LSTM, specifically from 71.06 mm to 31.72 mm. The average observed runoff during the period of precipitation events (2018, 7, 21-8, 4 days) was 1033 m 3 /s, LSTM-MMSCS simulated radial flow average of 1007 m 3 S, average error 2.52%. The observed data of the runoff peak at 7 months and 28 days is 2530 and 2530 m 3 LSTM-MMSCS simulated run-off at 2185 m 3 S, average error 13.64%.
The beneficial effects of the invention are as follows: the runoff simulation method based on the coupling neural network and the hydrologic physical model provided by the invention adopts satellite meteorological products as model driving data, can effectively solve the problems of inaccurate long-time sequence hydrologic prediction and larger deviation in peak flow simulation of the existing simulation method, and can relieve the difficulty of data deficiency faced during simulation work, thereby enhancing the capability of runoff simulation on time sequence capture, improving the processing efficiency and accuracy, and providing support for flood control management and rescue decision.
According to the invention, a runoff simulation scheme of coupling the neural network and the hydrologic physical model is established, so that a simulation result is more accurate compared with a single machine learning mode, a processing process is closer to the hydrologic physical mechanism, and the application and development of a runoff simulation technology in early warning and risk assessment of waterlogging disasters can be further promoted.
In the description of the present invention, it should be understood that the terms "center," "thickness," "upper," "lower," "horizontal," "top," "bottom," "inner," "outer," "radial," and the like indicate or are based on the orientation or positional relationship shown in the drawings, merely to facilitate description of the present invention and to simplify the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be configured and operated in a particular orientation, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be interpreted as indicating or implying a relative importance or number of technical features indicated. Thus, a feature defined as "first," "second," "third," or the like, may explicitly or implicitly include one or more such feature.
Claims (8)
1. The runoff simulation method based on the coupled neural network and the hydrologic physical model is characterized by comprising the following steps of:
s1, collecting satellite meteorological data of a target river basin, site record runoff data and related auxiliary data, and preprocessing the satellite meteorological data, the site record runoff data and the related auxiliary data to obtain a preprocessing result;
the pretreatment result of the satellite meteorological data comprises precipitation data, evapotranspiration data, temperature data and soil humidity data, and the pretreatment result of the site recorded runoff data is specifically observation runoff;
s2, inputting precipitation data, evapotranspiration data and temperature data into an LSTM network to obtain a preliminary runoff result;
s3, calculating the saturated water conductivity of the soil according to the pretreatment result of the related auxiliary data, and further obtaining a soil type CN value of the target river basin;
s4, inputting the soil type CN value, the precipitation data and the soil humidity data of the target river basin into an MMSCS model to obtain the runoff depth of the target river basin;
s5, performing differential feature analysis according to the primary runoff result and the observed runoff, and establishing a regression relation between the primary runoff result with the differential value higher than a set threshold value and the runoff depth of the target river basin and the observed runoff to obtain a final runoff simulation result.
2. The runoff simulation method based on the coupled neural network and the hydrophysics model according to claim 1, wherein in S1, the preprocessing method specifically comprises:
(1) Image stitching is carried out on the acquired satellite meteorological data, image cutting is carried out according to the target river basin, missing or unreliable pixels in the image are removed, average results of daily precipitation, temperature, soil humidity and evaporation of the target river basin in the image are extracted, and preprocessing results of the satellite meteorological data are obtained;
(2) Extracting flow information of a corresponding target river basin in site-recorded runoff data, and electronically recording according to a daily scale to obtain a preprocessing result of the site-recorded runoff data;
(3) Based on the collected related auxiliary data, extracting the soil texture, the soil organic matter content and the land utilization data of the target river basin in the related auxiliary data through image cutting, and obtaining the preprocessing result of the related auxiliary data.
3. The runoff simulation method based on the coupled neural network and the hydrophysics model according to claim 2, wherein the S2 specifically is: inputting precipitation data, evapotranspiration data and temperature data into an LSTM network to obtain a primary runoff result;
the LSTM network comprises a forget gate, an input gate, a unit state and an output gate;
wherein the forgetting doorF t The expression of (2) is specifically:
in the method, in the process of the invention,W f as a first matrix of weights,b f for the first offset vector, σ (·) is the sigmoid function,h t-1 in order to hide the nodes of the network,x t for inputting data, subscriptsfAndtrespectively forgetting and time interval states;
the input doorI t The expression of (2) is specifically:
in the method, in the process of the invention,W i as a second matrix of weights,b i is a second offset vector;
the state of the unitC t The expression of (2) is specifically:
in the method, in the process of the invention,update value for cell state, +.>For the hidden layer state of the previous moment, +.>Performing dot multiplication operation;
the output doorO t The expression of (2) is specifically:
in the method, in the process of the invention,W o as a result of the third weight matrix,b o for the third offset vector, subscript o represents the output state.
4. The runoff simulation method based on the coupling neural network and the hydrophysics model according to claim 2, wherein the step S3 comprises the following sub-steps:
s31, calculating the saturated water conductivity of the soil according to the soil texture and the organic matter content of the soil;
s32, obtaining a target river basin soil type CN value according to the saturated water conductivity of the soil and the land utilization data.
5. The runoff simulation method based on the coupled neural network and the hydrographic physical model according to claim 4, wherein in S31, the soil texture includes a clay content and a sand content;
calculating the saturated water conductivity of soilKThe expression of (2) is specifically:
in the method, in the process of the invention,Cin order to achieve a content of the sticky particles,Sfor the sand content of the sand particles,Ois the organic matter content of the soil,z 1 ~z 4 is constant.
6. The runoff simulation method based on the coupled neural network and the hydrophysics model according to claim 4, wherein the step S32 is specifically:
classifying hydrologic soil groups according to the saturated water conductivity of the soil to obtain hydrologic soil group types, and looking up a table according to land utilization data and the hydrologic soil group types to obtain a target river basin soil type CN value;
the hydrologic soil group classification comprises A, B, C and D, and the method for carrying out the hydrologic soil group classification specifically comprises the following steps:
obtaining a hydrologic soil group class A when the saturated water conductivity of the soil is greater than 180, obtaining a hydrologic soil group class B when the saturated water conductivity of the soil is within a range of 18-180, obtaining a hydrologic soil group class C when the saturated water conductivity of the soil is within a range of 1.8-18, and obtaining a hydrologic soil group class D when the saturated water conductivity of the soil is less than 1.8;
the land utilization data comprise paddy fields, dry lands, forests, shrubs, grasslands, channels, town lands and bare lands, and the method for acquiring the CN value of the soil type of the target river basin comprises the following steps:
when the land utilization data is paddy fields, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 60, 70, 80 and 80;
when the land utilization data is dry land, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 90, 95 and 95;
when the land utilization data is woodland, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are 35, 60, 70 and 80 respectively;
when the land use data is a shrub land, then the soil type CN values obtained in combination with the hydrologic soil group categories A, B, C and D are 20, 55, 70 and 75, respectively;
when the land use data is grasslands, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are 40, 60, 78 and 80 respectively;
when the land utilization data is a canal, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 80, 85, 90 and 90;
when the land utilization data is urban land, the soil type CN values obtained by combining the hydrologic soil group categories A, B, C and D are respectively 90, 95 and 95;
when the land use data is bare land, then the soil type CN values obtained in combination with the hydrologic soil group categories A, B, C and D are 75, 85, 95 and 95, respectively.
7. The runoff simulation method based on the coupling neural network and the hydrologic physical model according to claim 2, wherein in the step S4, the soil humidity data comprises initial soil humidity, critical soil humidity, potential loss and potential maximum water holding capacity, and the precipitation data is specifically precipitation;
the expression for obtaining the runoff depth Q of the target river basin is specifically:
wherein V is 0 The initial soil humidity, P is precipitation, S a Is critical soil humidity, S is potential loss amount, S b Is the potential maximum water holding capacity.
8. The runoff simulation method based on the coupled neural network and the hydrophysics model according to claim 7, wherein the expression of the potential loss S is specifically:
in the method, in the process of the invention,for different soil type CN values for the target river basin,kis a constant value for different soil types in the target river basin.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310841797.7A CN116562176B (en) | 2023-07-11 | 2023-07-11 | Runoff simulation method based on coupled neural network and hydrologic physical model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310841797.7A CN116562176B (en) | 2023-07-11 | 2023-07-11 | Runoff simulation method based on coupled neural network and hydrologic physical model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116562176A true CN116562176A (en) | 2023-08-08 |
| CN116562176B CN116562176B (en) | 2023-10-03 |
Family
ID=87493282
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310841797.7A Active CN116562176B (en) | 2023-07-11 | 2023-07-11 | Runoff simulation method based on coupled neural network and hydrologic physical model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116562176B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102034001A (en) * | 2010-12-16 | 2011-04-27 | 南京大学 | Design method for distributed hydrological model by using grid as analog unit |
| CN110442937A (en) * | 2019-07-24 | 2019-11-12 | 武汉大学 | A kind of valley water security management method merging satellite remote sensing and machine learning techniques |
| CN111079998A (en) * | 2019-12-03 | 2020-04-28 | 华东师范大学 | Flow prediction method based on long and short time sequence correlation attention mechanism model |
| CN112214923A (en) * | 2020-09-09 | 2021-01-12 | 广州地理研究所 | Satellite soil moisture reconstruction method, device and equipment |
| CN112782788A (en) * | 2019-11-06 | 2021-05-11 | 中国科学院国家空间科学中心 | Regional atmosphere hydrologic coupling early warning decision system and method |
| CN113837352A (en) * | 2021-08-06 | 2021-12-24 | 中国地质大学（武汉） | Rainfall-runoff space-time relation simulation method based on long-short term memory neural network |
| CN114118565A (en) * | 2021-11-23 | 2022-03-01 | 南京信息工程大学 | Daily runoff forecasting method based on bidirectional long-and-short-term memory coupling model |
| US20220061236A1 (en) * | 2020-08-25 | 2022-03-03 | The Board Of Trustees Of The University Of Illinois | Accessing agriculture productivity and sustainability |
| CN114462701A (en) * | 2022-01-29 | 2022-05-10 | 武汉大学 | Monthly-scale land vegetation carbon reserve calculation optimization method based on land water reserve |
| CN115544785A (en) * | 2022-10-19 | 2022-12-30 | 中国科学院地理科学与资源研究所 | Data-free cascade reservoir basin hydrological simulation method and system |
| US20230035413A1 (en) * | 2021-07-16 | 2023-02-02 | Climate Llc | Systems and methods for use in application of treatments to crops in fields |
-
2023
- 2023-07-11 CN CN202310841797.7A patent/CN116562176B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102034001A (en) * | 2010-12-16 | 2011-04-27 | 南京大学 | Design method for distributed hydrological model by using grid as analog unit |
| CN110442937A (en) * | 2019-07-24 | 2019-11-12 | 武汉大学 | A kind of valley water security management method merging satellite remote sensing and machine learning techniques |
| CN112782788A (en) * | 2019-11-06 | 2021-05-11 | 中国科学院国家空间科学中心 | Regional atmosphere hydrologic coupling early warning decision system and method |
| CN111079998A (en) * | 2019-12-03 | 2020-04-28 | 华东师范大学 | Flow prediction method based on long and short time sequence correlation attention mechanism model |
| US20220061236A1 (en) * | 2020-08-25 | 2022-03-03 | The Board Of Trustees Of The University Of Illinois | Accessing agriculture productivity and sustainability |
| CN112214923A (en) * | 2020-09-09 | 2021-01-12 | 广州地理研究所 | Satellite soil moisture reconstruction method, device and equipment |
| US20230035413A1 (en) * | 2021-07-16 | 2023-02-02 | Climate Llc | Systems and methods for use in application of treatments to crops in fields |
| CN113837352A (en) * | 2021-08-06 | 2021-12-24 | 中国地质大学（武汉） | Rainfall-runoff space-time relation simulation method based on long-short term memory neural network |
| CN114118565A (en) * | 2021-11-23 | 2022-03-01 | 南京信息工程大学 | Daily runoff forecasting method based on bidirectional long-and-short-term memory coupling model |
| CN114462701A (en) * | 2022-01-29 | 2022-05-10 | 武汉大学 | Monthly-scale land vegetation carbon reserve calculation optimization method based on land water reserve |
| CN115544785A (en) * | 2022-10-19 | 2022-12-30 | 中国科学院地理科学与资源研究所 | Data-free cascade reservoir basin hydrological simulation method and system |
Non-Patent Citations (4)
| Title |
|---|
| S SADEGHI TABAS: "Viriational Bayesian dropout with a Gaussian prior for recurrent neural networks application in rainfall-runoff modeling", 《ENVIRONMENTAL RESEARCH LETTERS》, vol. 17, no. 6, pages 1 - 17 * |
| S. VERMA: "Activation Soil Moisture Accounting(ASMA) for Runoff Estimation using Soil Conservation Service Curve Number(SCS-CN) Method", 《JOURNAL OF HYDROLOGY》, vol. 589, pages 1 - 92 * |
| 郝莹: "气象水文耦合的城市内涝风险多尺度预测与预估研究", 《中国博士学位论文全文数据库（电子期刊）》, no. 5, pages 038 - 19 * |
| 高玉芳: "基于WRF模式的清江流域降雨-径流模拟研究", 《热带气象学报》, vol. 38, no. 5, pages 621 - 630 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116562176B (en) | 2023-10-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jarrah et al. | A review of wind erosion models: Data requirements, processes, and validity | |
| De Roo et al. | Soil erosion modelling using ‘answers’ and geographical information systems | |
| Shi et al. | Evaluating the SWAT model for hydrological modeling in the Xixian watershed and a comparison with the XAJ model | |
| Alatorre et al. | Soil erosion and sediment delivery in a mountain catchment under scenarios of land use change using a spatially distributed numerical model | |
| Nearing | Soil erosion and conservation | |
| CN108536908B (en) | Method for evaluating watershed water environment safety based on non-point source nitrogen and phosphorus loss risk | |
| Cavanagh et al. | Soil phosphate, site boundaries, and change point analysis | |
| Melching et al. | A reliability estimation in modeling watershed runoff with uncertainties | |
| Greve et al. | Quantifying the ability of environmental parameters to predict soil texture fractions using regression-tree model with GIS and LIDAR data: The case study of Denmark | |
| Rockström et al. | On‐farm spatial and temporal variability of soil and water in pearl millet cultivation | |
| Morgan et al. | Erosion modeling | |
| CN108763849A (en) | River pollutant sources computational methods are polluted in conjunction with the basin face source phosphorus of deposit and model | |
| CN113176393B (en) | HASM model-based three-dimensional estimation method and system for soil organic carbon reserves | |
| Bao et al. | Hydrological daily rainfall-runoff simulation with BTOPMC model and comparison with Xin'anjiang model | |
| CN113011993A (en) | Method for measuring and calculating water-entering load of agricultural pollution source based on standard data | |
| Fiener et al. | A compilation and meta-analysis of rainfall simulation data on arable soils | |
| He et al. | Testing distributed soil erosion and sediment delivery models using 137Cs measurements | |
| Mueller et al. | Spatial variability of soil and nutrient characteristics of semi‐arid grasslands and shrublands, Jornada Basin, New Mexico | |
| CN113887064A (en) | Large-scale underground water reserve remote sensing dynamic monitoring and driving factor quantitative splitting method | |
| CN116562176B (en) | Runoff simulation method based on coupled neural network and hydrologic physical model | |
| Moghaddamnia et al. | Evaporation estimation using support vector machines technique | |
| Zhang et al. | Parameter calibration and uncertainty estimation of a simple rainfall-runoff model in two case studies | |
| Persson et al. | Spatial influence of topographical factors on yield of potato (Solanum tuberosum L.) in central Sweden | |
| Hassan et al. | At site and regional frequency analysis for Sudan annual rainfall by using the L-moments and nonlinear regression techniques | |
| Shapiro et al. | Incorporating snowmelt into daily estimates of recharge using a state‐space model of infiltration |
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 |