CN114723177A - Flood disaster prediction and early warning method based on DA-SSL - Google Patents
Flood disaster prediction and early warning method based on DA-SSL Download PDFInfo
- Publication number
- CN114723177A CN114723177A CN202210526493.7A CN202210526493A CN114723177A CN 114723177 A CN114723177 A CN 114723177A CN 202210526493 A CN202210526493 A CN 202210526493A CN 114723177 A CN114723177 A CN 114723177A
- Authority
- CN
- China
- Prior art keywords
- pipe network
- functional
- sample
- waterlogging
- diseases
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 68
- 201000010099 disease Diseases 0.000 claims abstract description 132
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims abstract description 132
- 238000012549 training Methods 0.000 claims abstract description 7
- 238000005259 measurement Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 32
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 23
- 230000008878 coupling Effects 0.000 claims description 16
- 238000010168 coupling process Methods 0.000 claims description 16
- 238000005859 coupling reaction Methods 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 9
- 230000003902 lesion Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 239000012530 fluid Substances 0.000 claims description 3
- 230000008021 deposition Effects 0.000 description 4
- 238000004088 simulation Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000003247 decreasing effect Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- 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
- Y02A50/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE in human health protection, e.g. against extreme weather
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Educational Administration (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a flood disaster prediction and early warning method based on DA-SSL, which comprises the following steps: obtaining an initial sample of the urban pipe network diseases; inputting the initial sample of the urban pipe network diseases into an improved PCWMM model; the improved PCWMM model outputs a pipe network functional disease sample set with an inland inundation result label; carrying out sample set expansion on the initial sample of the urban management network diseases to obtain an expanded labeled sample set with the waterlogging result label; and performing information difference measurement of the labeled/unlabeled sample sets by using cross entropy, randomly generating unlabeled sample sets without the waterlogging result labels, inputting the labeled sample sets and the unlabeled sample sets into the improved PCWMM model, and repeatedly training the improved PCWMM model to obtain urban waterlogging area results under different pipe network functional disease conditions. By adopting the method and the device, the influence of functional diseases on the overflow capacity and flow velocity distribution of the sewer pipe network can be accurately fitted, and the waterlogging early warning result is very accurate and reliable.
Description
Technical Field
The invention relates to the field of flood disaster early warning, in particular to a DA-SSL-based flood disaster prediction early warning method.
Background
In the process of constructing an urban waterlogging early warning and disaster prevention decision system, the conventional flood disaster prediction and early warning method (such as the current mature rainfall flood management model (PCWMM)) can comprehensively consider surface production convergence simulation and pipe network hydrodynamic transmission process under the condition of heavy rain in different recurrence periods, but assumes that a drainage pipe network is in a smooth state and neglects the influence of pipe diseases on the overcurrent capacity and flow rate distribution. However, research shows that functional diseases of drainage pipelines of every kilometer exist in main cities in China, wherein scaling and silting diseases account for more than 70%, so that the accuracy of the conventional flood disaster prediction and early warning method in predicting the overflowing capacity of a drainage pipeline network is seriously insufficient, and a flood disaster prediction and early warning method considering the coupling relation between the pipeline functional diseases and urban inland inundation needs to be constructed urgently.
The existing urban waterlogging early warning model research mainly focuses on refined rainfall flood simulation and low-influence development and layout under heavy rains in different reappearance periods, influences of multiphase flow and multi-field coupling pipe network functional diseases on urban surface underground one-dimensional production convergence are ignored, and the accuracy of waterlogging early warning results is difficult to guarantee.
The urban inland inundation early warning is influenced by multiple parameters such as pipeline gradient, overflowing hydraulic power radius, wet cycle and the like, and the workload of analyzing and calculating the corresponding relation between the pipeline functional diseases and inland inundation areas one by one through a control variable method is large and difficult to realize.
In summary, the existing flood disaster prediction and early warning method has the following disadvantages:
(1) the influence of functional diseases (scaling and silting) of the pipeline on the overflowing capacity and flow velocity distribution is ignored, so that the accuracy of the conventional flood disaster prediction early warning method in predicting the overflowing capacity of the drainage pipe network is seriously insufficient;
(2) influence of multiphase flow and multi-field coupling pipe network functional diseases on underground one-dimensional production convergence of the urban ground surface is ignored, and accuracy of waterlogging early warning results is difficult to guarantee;
(3) the control variable method is used for analyzing and calculating the corresponding relation between the functional diseases of the pipeline and the waterlogging area one by one, and the workload is large and difficult to realize.
In order to solve the problems, the inventor of the invention provides a DA-SSL-based flood disaster prediction and early warning method.
Disclosure of Invention
In order to solve the problems, the invention aims to provide a DA-SSL-based flood disaster prediction and early warning method, which can accurately fit functional diseases to influence the overflow capacity and flow rate distribution of a drainage pipe network, and has an accurate and reliable early warning result of the flood.
Based on the above, the invention provides a DA-SSL-based flood disaster prediction and early warning method, which comprises the following steps:
obtaining an initial sample of the urban pipe network functional diseases, wherein the initial sample of the urban pipe network functional diseases comprises the following steps: rainstorm data in different reappearance periods and existing sample sets of pipe network functional diseases;
inputting the initial sample of the functional diseases of the urban pipe network into an improved PCWMM model;
the improved PCWMM model outputs a pipe network functional disease sample set with an inland inundation result label;
carrying out sample set expansion on the initial sample of the urban management network diseases to obtain an expanded labeled sample set with the waterlogging result label;
and performing information difference measurement of the labeled/unlabeled sample set by using cross entropy, randomly generating an unlabeled sample set without an inland inundation result label, inputting the labeled sample set and the unlabeled sample set into the improved PCWMM model, and repeatedly training the improved PCWMM model, so that the improved PCWMM model can obtain the inland inundation area results of different pipe network functional disease conditions.
Wherein, the step of expanding the sample set of the initial sample of the urban pipe network diseases comprises the steps of;
acquiring the existing sample set of the functional diseases of the pipe network and corresponding label values of the waterlogging results, and assigning initial values to the expanded sample set of the functional diseases of the new pipe network;
calculating the loss gradient of a semi-supervised learning inversion loss function to the waterlogging result label value corresponding to the functional damage sample of the pipe network, and acquiring the feature item index value with the maximum loss gradient;
according to a preset assignment rule, reassigning the expanded sample feature items corresponding to the index values;
and continuously adjusting and assigning the new pipe network functional disease sample set with the waterlogging result label by the urban pipe network functional disease initial sample to enable the sample set to be close to the classification boundary, and further obtaining the expanded pipe network functional disease sample set with the label.
The calculating of the loss gradient of the semi-supervised learning inversion loss function to the corresponding waterlogging result label value of the pipe network functional disease sample, and the obtaining of the feature item index value with the maximum loss gradient comprises the following steps:
the calculation method of the semi-supervised learning inversion loss function F comprises the following steps:
wherein M is the number of functional disease classes, yicThe true class representing sample i is equal to c taken 1, otherwise 0, picRepresenting the prediction probability that the sample i belongs to the functional disease category c, wherein N is the number of hidden layers;
the loss gradient is inverted by semi-supervised learning to any extended functional disease sample by a loss function F
Is derived from the direction of (a), i.e.:
wherein L issIndicating the corresponding waterlogging result label value, N, of the original pipe network functional disease sample setsFor the expanded sample set of functional diseases of the new pipe network,
any expanded functional lesion sample;
the index value i of the feature item with the largest loss gradientmaxCalculating a functional disease sample of an inversion loss function F after any expansion
Is solved for the maximum value of (1), i.e.
And n is the maximum row number of the pipe network functional disease sample set, and m is the maximum column number of the pipe network functional disease sample set.
Wherein, according to a preset assignment rule, reassigning the expanded sample feature item corresponding to the index value comprises:
wherein, if left side
For the expanded functional lesion sample, if Right side
Is a functional disease sample before expansion.
Wherein the method further comprises: comparing and analyzing the pipe network functional disease input and waterlogging result data set by adopting a semi-supervised learning inversion method and a control variable method, extracting the corresponding relation between the pipe network functional disease and the waterlogging condition of the local urban area, and calibrating various parameters of a loss function;
and adding the prediction result into the initial training set according to a preset updating period, and repeatedly iterating to reduce the error of the loss function.
The method comprises the following steps of adopting a semi-supervised learning inversion method and a control variable method to compare and analyze functional disease input and waterlogging result data sets of a pipe network, extracting the corresponding relation between the functional disease of the pipe network and the waterlogging condition of local urban areas, and calibrating each parameter of a loss function, wherein the parameters comprise:
extracting the corresponding relation between the functional diseases of the pipe network and the waterlogging condition of the local urban area, wherein the corresponding relation is obtained by each parameter of a calibration loss function;
and each parameter of the calibration loss function realizes parameter determination by continuously adjusting and reducing the comprehensive error rate of the input layer, the hidden layer and the output layer of the initial sample set of the urban pipe network functional diseases.
The error rate of the output layer is reduced by adjusting the parameters one by a control variable method. A. theiIs the ith parameter, c is a constant, and n is the total number of the parameters. If the first parameter is adjusted to be (A)1+c,A2,...,An) And when the first parameter is adjusted, adjusting other parameters one by a constant c.
The construction process of the improved PCWMM model comprises the following steps:
the construction process of the improved PCWMM model comprises the following steps: a two-dimensional coupling model considering the functional diseases of the pipe network is engineered to be a functional module and integrated with the conventional PCWMM model in a dynamic library form to obtain an improved PCWMM model considering the functional diseases of the pipe network;
the model of the hydrodynamics coupling relation between functional diseases and waterlogging of the drainage pipe network can be dynamically quantized by a one-dimensional full flow control equation and a two-dimensional shallow water equation, and the method specifically comprises the following steps:
wherein Z represents water level, A represents water passing cross-sectional area, Q represents cross-sectional outlet flow, and Q represents water flowLRepresenting side inflow, g is gravitational acceleration, t and x respectively represent one-dimensional time and space coordinates, a represents wave velocity, SfDecrease in friction ratio, hLFor the local head loss over the unit length, h is the water depth, t is the time, x, y and z are coordinate systems, u and v are the flow velocity components in the x and y directions respectively,
is the average vertical velocity, rho is the density of the fluid, b is the water bottom elevation, SaxAnd SayComponent of the bottom slope term, S, in the x and y directions, respectivelyfxAnd SfyComponent of friction term, τ, in x and y directions, respectivelyzyAnd τzyAre all lateral stresses;
a one-dimensional full-flow control equation and a two-dimensional shallow water equation are engineered into a functional module of the PCWMM to improve the prediction accuracy of the PCWMM on waterlogging.
The invention adopts the numerical simulation and fusion technical means of 'data set expansion (DA) + semi-supervised learning (SSL) + rainfall flood management model (PCWMM)', so as to realize the flood disaster prediction and early warning method based on DA-SSL and carry out flood disaster prediction and early warning under heavy rains in different reappearance periods. The invention aims to realize a DA-SSL-based flood disaster forecasting and early warning method which can meet the requirements of accurately fitting functional diseases to influence the overflowing capacity and flow rate distribution of a drainage pipe network, has accurate early warning results of the waterlogging, is high in efficiency and reliability and overcomes various defects of the current flood disaster forecasting and early warning method based on the coupling relation between the functional diseases and the waterlogging of the drainage pipe network and the design of DA, SSL and PCWMM. Based on this, the invention has the following advantages:
(1) and accurately fitting the functional diseases to influence the flow capacity and flow velocity distribution of the drainage pipe network. The DA-SSL-based flood disaster prediction early warning method utilizes pipeline confluence parameters determined by tests to judge the critical starting conditions of the pipeline functional diseases and the hydrodynamics coupling relation to carry out engineering improvement on the PCWMM, and under the common cooperation of DLL dynamic library link and Arcgis software, the accurate fitting of the overflow capacity and the flow rate distribution of the sewer pipe network influenced by the functional diseases is realized.
(2) And the early warning result of waterlogging is accurate. By means of PCWMM dynamic simulation considering the coupling relation between the functional diseases of the pipe network and the waterlogging hydrodynamics, the urban two-dimensional instantaneous water flow state conversion rule of the coupling relation between the functional diseases of the pipe network and the waterlogging hydrodynamics is considered, the waterlogging indexes such as the submerging water depth, the submerging area and the submerging duration are accurately calculated, and the problem that the influence of the functional diseases of the pipe network on the waterlogging result is not considered in the existing research is solved.
(3) High efficiency. The DA-SSL-based flood disaster prediction and early warning method realizes the high-efficiency output of a flood early warning result by using a deep learning method combining data set expansion and semi-supervised learning.
(4) High reliability. The invention adopts numerical simulation and fusion technical means of 'data set expansion (DA) + semi-supervised learning (SSL) + rainfall flood management model (PCWMM)', so as to realize the flood disaster prediction and early warning method based on DA-SSL. The accurate deep learning input sample set is obtained by the PCWMM considering the coupling relation between the functional diseases of the pipe network and the waterlogging hydrodynamics, and the result output reliability of the DA-SSL-based flood disaster prediction and early warning method can be fully ensured by combining a data set expansion method.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic diagram of a DA-SSL-based flood disaster prediction and early warning method according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic diagram of a DA-SSL-based flood disaster prediction and early warning method provided in an embodiment of the present invention, where the method includes:
s101, obtaining an initial sample of the urban pipe network diseases, wherein the initial sample of the urban pipe network diseases comprises the following steps: rainstorm data in different reappearance periods and existing sample sets of pipe network functional diseases;
the existing sample set for the functional diseases of the pipe network comprises different section pipeline deposition conditions (whether deposition exists), deposition degrees, deposition lengths, scaling conditions (whether scaling bodies exist), scaling degrees and scaling lengths.
S102, inputting the initial sample of the urban pipe network diseases into an improved PCWMM model;
the improved PCWMM model improvement process comprises the following steps: a two-dimensional coupling model considering the functional diseases of the pipe network is engineered into a functional module and integrated with the conventional PCWMM model in a dynamic library mode to obtain the improved PCWMM model considering the functional diseases of the pipe network.
The construction process of the improved PCWMM model comprises the following steps:
the improved PCWMM model improvement process comprises the following steps: a two-dimensional coupling model considering the functional diseases of the pipe network is engineered into a functional module and integrated with the conventional PCWMM model in a dynamic library mode to obtain the improved PCWMM model considering the functional diseases of the pipe network.
The model of the hydrodynamics coupling relation between functional diseases and waterlogging of the drainage pipe network can be dynamically quantized by a one-dimensional full flow control equation and a two-dimensional shallow water equation, and the method specifically comprises the following steps:
wherein Z represents water level, A represents water passing cross-sectional area, Q represents outlet flow of cross-section, and Q represents water levelLRepresenting side inflow, g is gravitational acceleration, t and x represent one-dimensional time and space coordinates, respectively, a represents wave velocity, SfDecrease in friction ratio, hLFor the local head loss over the unit length, h is the water depth, t is the time, x, y and z are coordinate systems, u and v are the flow velocity components in the x and y directions respectively,
is the average vertical velocity, rho is the density of the fluid, b is the water bottom elevation, SaxAnd SayComponent of the bottom slope term, S, in the x and y directions, respectivelyfxAnd SfyComponent of friction term, τ, in x and y directions, respectivelyzyAnd τzyAre all lateral stresses.
The one-dimensional full-flow control equation and the two-dimensional shallow water equation are engineered into a functional module of the PCWMM, so that the prediction accuracy of the PCWMM on waterlogging is improved.
S103, outputting a pipe network functional disease sample set with an inland inundation result label by the improved PCWMM model;
s104, expanding the sample set of the initial sample of the urban management network diseases to obtain an expanded labeled sample set with the waterlogging result label;
because the initial sample of the urban pipe network disease is limited, the initial sample is expanded by a random sample set expansion method, and the improved PCWMM is input to obtain a labeled sample set with a waterlogging result label;
the PCWMM model is obtained by inputting only the network management functional diseases existing sample set in the initial sample of the urban network diseases into the improved PCWMM model.
Wherein, the step of expanding the sample set of the initial sample of the urban pipe network diseases comprises the steps of;
acquiring the existing sample set of the functional diseases of the pipe network and corresponding label values of the waterlogging results, and assigning initial values to the expanded sample set of the functional diseases of the new pipe network;
calculating the loss gradient of a semi-supervised learning inversion loss function to the waterlogging result label value corresponding to the functional damage sample of the pipe network, and acquiring the feature item index value with the maximum loss gradient;
according to a preset assignment rule, reassigning the expanded sample feature items corresponding to the index values;
and continuously adjusting and assigning the new pipe network functional disease sample set with the waterlogging result labels from the initial sample set to enable the new pipe network functional disease sample set to be close to the classification boundary, and further obtaining the expanded labeled sample set.
The calculating of the loss gradient of the semi-supervised learning inversion loss function to the corresponding waterlogging result label value of the pipe network functional disease sample, and the obtaining of the feature item index value with the maximum loss gradient comprises the following steps:
the calculation method of the semi-supervised learning inversion loss function F comprises the following steps:
wherein M is the number of functional disease classes, yicThe true class representing sample i is equal to c taken 1, otherwise 0, p is takenicRepresenting the prediction probability that the sample i belongs to the functional disease category c, wherein N is the number of hidden layers;
the loss gradient is inverted by semi-supervised learning to any extended functional disease sample by a loss function F
Is derived in the direction of (a), i.e.:
wherein L issIndicating the corresponding waterlogging result label value, N, of the original pipe network functional disease sample setsFor the expanded sample set of functional diseases of the new pipe network,
any expanded functional lesion sample;
the index value i of the feature item with the largest loss gradientmaxCalculating a functional disease sample of an inversion loss function F after any expansion
The maximum value of (A) is solved, namely.
And n is the maximum row number of the pipe network functional disease sample set, and m is the maximum column number of the pipe network functional disease sample set.
Wherein, according to a preset assignment rule, reassigning the expanded sample feature item corresponding to the index value comprises:
the principle of sample set expansion of the initial sample of the urban pipe network diseases is as follows: FGSM finds the characteristic item with the maximum gradient value by calculating the gradient of the semi-supervised learning inversion loss function to the input data set, then changes the value of the characteristic item to make the input move towards the direction of increasing the model loss, generates a new sample point, and the new sample point is closer to the decision boundary of the model, and the specific process is as follows:
1) the original pipe network functional disease sample set and the corresponding label values of the waterlogging result are I respectivelyS、Ls；
And n.m dimensional characteristic items are provided, wherein m represents the number of areas for early warning of flooding, and n represents the number of functional disease types of drainage pipelines. Each characteristic item value set is {0,0.25,0.5,0.75,1}, and when the value is 0, the drainage pipeline is indicated to have no functional diseases; when the value is 0.25 to 1, the severity of the functional disease of the drainage pipeline is indicated. Assigning an initial value to the Ns of the expanded new pipe network functional disease sample set;
2) calculating the loss gradient of the semi-supervised learning inversion loss function F to the original waterlogging result label, and finding out the characteristic item index value i with the maximum gradient valuemaxI.e. the following.
3) In order to enable a new sample value of the in-band waterlogging label obtained by data expansion to be closer to the real situation, the index value i is usedmaxAnd readjusting and assigning the corresponding new sample feature items, wherein the assignment rule is as follows:
4) new in-band flooding result label LsPipe network functional disease sample set NsFrom an initial sample set IsAnd continuously adjusting assignment to enable the assignment to be close to the classification boundary, and further obtaining the expanded labeled sample set.
S105, performing information difference measurement of the labeled/unlabeled sample sets by using cross entropy, randomly generating unlabeled sample sets without the waterlogging result labels, inputting the labeled sample sets and the unlabeled sample sets into the improved PCWMM model, and repeatedly training the improved PCWMM model to obtain urban waterlogging area results under different pipe network functional disease conditions.
Wherein the method further comprises: comparing and analyzing the functional disease input and waterlogging result data set of the pipe network by adopting a semi-supervised learning inversion method and a control variable method, extracting the corresponding relation between the functional disease of the pipe network and the waterlogging condition of local urban areas, and calibrating each parameter of a loss function;
and adding the prediction result into the initial training set according to a preset updating period, and repeatedly iterating to reduce the error of the loss function.
Extracting the corresponding relation between the functional diseases of the pipe network and the waterlogging condition of the local urban area, wherein the corresponding relation is obtained by each parameter of a calibration loss function;
and each parameter of the calibration loss function realizes parameter determination by continuously adjusting and reducing the comprehensive error rate of the input layer, the hidden layer and the output layer of the initial sample set of the functional diseases of the urban pipe network.
The error rate of the output layer is reduced by adjusting the parameters one by a control variable method. A. theiIs the ith parameter, c is a constant, and n is the total number of the parameters. If the first parameter is adjusted to be (A)1+c，A2，…，An) If so, the error rate is decreased and the first parameter is increased again, otherwise the first parameter is decreased. After the first parameter adjustment is completed, the other parameters are adjusted one by the constant c.
The flood prediction early warning system adopts a flood prediction early warning system comprising data set expansion (DA), semi-supervised learning (SSL) and a rainfall flood management model (PCWMM), fully considers the influence of the coupling relation between the functional diseases of the underground drainage pipe network and the waterlogging hydrodynamics on the flood result, and simultaneously combines the sample set expansion based on FGSM and the semi-supervised learning inversion technology to improve the output efficiency and accuracy of the flood prediction early warning result. Therefore, the method has the advantages of high prediction efficiency, high accuracy and high reliability. A new output regional flooding result and a pipeline network functional disease input sample form a labeled sample set together based on the DA-SSL flood disaster prediction and early warning method, and error correction is carried out on semi-supervised learning inversion according to a period T. Therefore, the method has the advantage that the prediction early warning accuracy rate is gradually improved along with the increase of the iteration times.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and substitutions can be made without departing from the technical principle of the present invention, and these modifications and substitutions should also be regarded as the protection scope of the present invention.
Claims (7)
1. A DA-SSL-based flood disaster prediction and early warning method is characterized by comprising the following steps:
obtaining an initial sample of the urban pipe network functional diseases, wherein the initial sample of the urban pipe network functional diseases comprises the following steps: rainstorm data in different reappearance periods and existing sample sets of pipe network functional diseases;
inputting the initial sample of the functional diseases of the urban pipe network into an improved PCWMM model;
the improved PCWMM model outputs a pipe network functional disease sample set with an inland inundation result label;
carrying out sample set expansion on the initial sample of the urban management network diseases to obtain an expanded labeled sample set with the waterlogging result label;
and performing information difference measurement of the tagged/untagged sample sets by using cross entropy, randomly generating a non-tagged sample set without a waterlogging result tag, inputting the tagged sample set and the non-tagged sample set into the improved PCWMM, and repeatedly training the improved PCWMM so that the improved PCWMM can obtain urban waterlogging area results of different pipe network functional disease conditions.
2. The DA-SSL-based flood disaster prediction and early warning method of claim 1, wherein the sample set expansion of the initial sample of the urban pipe network disease comprises;
acquiring the existing sample set of the functional diseases of the pipe network and corresponding label values of the waterlogging results, and assigning initial values to the expanded sample set of the functional diseases of the new pipe network;
calculating the loss gradient of a semi-supervised learning inversion loss function to the inland inundation result label value corresponding to the pipe network functional disease sample, and acquiring the feature item index value with the maximum loss gradient;
according to a preset assignment rule, reassigning the expanded sample feature items corresponding to the index values;
and continuously adjusting and assigning the new pipe network functional disease sample set with the waterlogging result label by the urban pipe network functional disease initial sample to enable the sample set to be close to the classification boundary, and further obtaining the expanded pipe network functional disease sample set with the label.
3. The DA-SSL-based flood disaster prediction and early warning method as claimed in claim 2, wherein the calculating a loss gradient of a semi-supervised learning inversion loss function to the corresponding flood result tag value of the pipe network functional damage sample, and obtaining the feature item index value with the maximum loss gradient comprises:
the calculation method of the semi-supervised learning inversion loss function F comprises the following steps:
wherein M is the number of functional disease classes, yicThe true class representing sample i is equal to c taken 1, otherwise 0, p is takenicThe prediction probability that the sample i belongs to the functional disease category c is shown, and N is the number of the hidden layers;
the loss gradient is inverted by semi-supervised learning to any extended functional disease sample by a loss function F
Is derived from the direction of (a), i.e.:
wherein L issTo representCorresponding inland inundation result label value N of original pipe network functional disease sample setsFor the expanded sample set of functional diseases of the new pipe network,
any expanded functional lesion sample;
the index value i of the feature item with the largest loss gradientmaxCalculating a functional disease sample of an inversion loss function F after any expansion
The maximum of (a) is solved, i.e.
And n is the maximum row number of the pipe network functional disease sample set, and m is the maximum column number of the pipe network functional disease sample set.
4. The DA-SSL-based flood disaster prediction and early warning method as claimed in claim 2, wherein the re-assigning the expanded sample feature items corresponding to the index values according to the preset assignment rules comprises:
wherein, if left side
For the expanded functional lesion sample, if Right side
Is a functional disease sample before expansion.
5. The DA-SSL-based flood disaster prediction and early warning method as recited in claim 1, wherein the method further comprises: comparing and analyzing the functional disease input and waterlogging result data set of the pipe network by adopting a semi-supervised learning inversion method and a control variable method, extracting the corresponding relation between the functional disease of the pipe network and the waterlogging condition of local urban areas, and calibrating each parameter of a loss function;
and adding the prediction result into the initial training set according to a preset updating period, and repeatedly iterating to reduce the error of the loss function.
6. The DA-SSL-based flood disaster prediction and early warning method as claimed in claim 3, wherein the method comprises the steps of comparing and analyzing the functional disease input and the flood result data set of the pipe network by adopting a semi-supervised learning inversion method and a control variable method, extracting the corresponding relation between the functional disease of the pipe network and the flood condition of the urban local area, and calibrating each parameter of the loss function, wherein the step of calibrating each parameter of the loss function comprises the following steps:
extracting the corresponding relation between the functional diseases of the pipe network and the waterlogging condition of the local urban area, wherein the corresponding relation is obtained by each parameter of a calibration loss function;
and each parameter of the calibration loss function realizes parameter determination by continuously adjusting and reducing the comprehensive error rate of the input layer, the hidden layer and the output layer of the initial sample set of the urban pipe network functional diseases.
The error rate of the output layer is reduced by adjusting the parameters one by a control variable method. A. theiIs the ith parameter, c is a constant, and n is the total number of parameters. If the first parameter is adjusted to be (A)1+c,A2,...,An) And when the first parameter is adjusted, adjusting other parameters one by a constant c.
7. The DA-SSL-based flood disaster prediction and early warning method as claimed in claim 1, wherein the improved PCSWMM model is constructed by a process of:
the construction process of the improved PCWMM model comprises the following steps: a two-dimensional coupling model considering the functional diseases of the pipe network is engineered into a functional module and is integrated with the conventional PCWMM model in a dynamic library form to obtain an improved PCWMM model considering the functional diseases of the pipe network;
the model of the hydrodynamics coupling relation between functional diseases and waterlogging of the drainage pipe network can be dynamically quantized by a one-dimensional full flow control equation and a two-dimensional shallow water equation, and the method specifically comprises the following steps:
wherein Z represents water level, A represents water passing cross-sectional area, Q represents outlet flow of cross-section, and Q represents water levelLRepresenting side inflow, g is gravitational acceleration, t and x represent one-dimensional time and space coordinates, respectively, a represents wave velocity, SfDecrease in friction ratio, hLFor the local head loss over the unit length, h is the water depth, t is the time, x, y and z are coordinate systems, u and v are the flow velocity components in the x and y directions respectively,
is the average vertical velocity, rho is the density of the fluid, b is the water bottom elevation, SaxAnd SayComponent of the bottom slope term, S, in the x and y directions, respectivelyfxAnd SfyComponent of friction term, τ, in x and y directions, respectivelyzyAnd τzyAre all lateral stresses;
a one-dimensional open-full flow control equation and a two-dimensional shallow water equation are engineered into a functional module of the PCWMM to improve the prediction accuracy of the PCWMM on waterlogging.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210526493.7A CN114723177A (en) | 2022-05-16 | 2022-05-16 | Flood disaster prediction and early warning method based on DA-SSL |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210526493.7A CN114723177A (en) | 2022-05-16 | 2022-05-16 | Flood disaster prediction and early warning method based on DA-SSL |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114723177A true CN114723177A (en) | 2022-07-08 |
Family
ID=82232115
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210526493.7A Pending CN114723177A (en) | 2022-05-16 | 2022-05-16 | Flood disaster prediction and early warning method based on DA-SSL |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114723177A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115268937A (en) * | 2022-09-27 | 2022-11-01 | 中国空气动力研究与发展中心计算空气动力研究所 | Multi-phase flow field initialization method, device, equipment and medium |
-
2022
- 2022-05-16 CN CN202210526493.7A patent/CN114723177A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115268937A (en) * | 2022-09-27 | 2022-11-01 | 中国空气动力研究与发展中心计算空气动力研究所 | Multi-phase flow field initialization method, device, equipment and medium |
| CN115268937B (en) * | 2022-09-27 | 2023-02-17 | 中国空气动力研究与发展中心计算空气动力研究所 | Multi-phase flow field initialization method, device, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20230195979A1 (en) | Intelligent decision-making method and system for maintaining urban underground sewer network | |
| Jia et al. | Surrogate modeling for peak or time-dependent storm surge prediction over an extended coastal region using an existing database of synthetic storms | |
| CN110738355B (en) | Urban waterlogging prediction method based on neural network | |
| CN108921944B (en) | Calculation method of hydrological response unit outflow process based on dynamic channel | |
| Yazdi | Rehabilitation of urban drainage systems using a resilience-based approach | |
| Park et al. | Optimal design of stormwater detention basin using the genetic algorithm | |
| CN117236673B (en) | Urban water network multi-scale flood control and drainage combined optimization scheduling method and system | |
| CN113807008B (en) | Urban storm waterlogging simulation method based on deep learning | |
| KR102584619B1 (en) | Water pollution accident prediction system based on hydraulic hydrology analysis with distributed model applied | |
| Hao et al. | Coupling the Xinanjiang model with geomorphologic instantaneous unit hydrograph for flood forecasting in northeast China | |
| KR101670918B1 (en) | Realtime analysis method for ground and underground flooding | |
| CN114936505B (en) | Method for rapidly forecasting multi-point water depth of urban rainwater well | |
| CN114723177A (en) | Flood disaster prediction and early warning method based on DA-SSL | |
| CN111199298A (en) | Flood forecasting method and system based on neural network | |
| Zhou et al. | Prediction and early warning method of inundation process at waterlogging points based on Bayesian model average and data-driven | |
| CN114648617A (en) | Water system extraction method based on digital elevation model DEM | |
| CN113627817B (en) | Road connectivity analysis method based on urban inland inundation condition | |
| Kazemian-Kale-Kale et al. | The uncertainty of the Shannon entropy model for shear stress distribution in circular channels | |
| CN117688844A (en) | Urban waterlogging real-time simulation method and system based on deep neural network | |
| Jhong et al. | Optimization of LSTM parameters for flash flood forecasting using genetic algorithm | |
| Ekmekcioğlu et al. | Exploring the practical application of genetic programming for stormwater drain inlet hydraulic efficiency estimation | |
| Dang et al. | Application of machine learning-based surrogate models for urban flood depth modeling in Ho Chi Minh City, Vietnam | |
| Jin et al. | Integration of an improved transformer with physical models for the spatiotemporal simulation of urban flooding depths | |
| Hsu et al. | Study on Flood Para-Tank Model Parameters with Particle Swarm Optimization. | |
| Thakur et al. | Exploring CCHE2D and its sediment modelling capabilities |
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 |