CN107145965B - River flood prediction method based on similarity matching and extreme learning machine - Google Patents

River flood prediction method based on similarity matching and extreme learning machine Download PDF

Info

Publication number
CN107145965B
CN107145965B CN201710228116.4A CN201710228116A CN107145965B CN 107145965 B CN107145965 B CN 107145965B CN 201710228116 A CN201710228116 A CN 201710228116A CN 107145965 B CN107145965 B CN 107145965B
Authority
CN
China
Prior art keywords
real
flood
time
historical
water flow
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201710228116.4A
Other languages
Chinese (zh)
Other versions
CN107145965A (en
Inventor
李士进
孔俊
朱跃龙
余宇峰
朱小明
冯钧
马凯凯
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hohai University (HHU)
Original Assignee
Hohai University (HHU)
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hohai University (HHU) filed Critical Hohai University (HHU)
Priority to CN201710228116.4A priority Critical patent/CN107145965B/en
Publication of CN107145965A publication Critical patent/CN107145965A/en
Application granted granted Critical
Publication of CN107145965B publication Critical patent/CN107145965B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
    • Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • Development Economics (AREA)
  • Medical Informatics (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Game Theory and Decision Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a river flood prediction method based on similarity matching and an extreme learning machine, which comprises the steps of firstly, obtaining optimal flow prediction models corresponding to historical flood samples respectively based on an extreme learning machine model; then, based on the similarity matching values in descending order, selecting preset previous K historical flood samples as each reference flood sample; and finally, predicting the flow value based on the real-time water flow characteristic and the real-time rainfall characteristic, and judging the flood according to the obtained flow prediction value. Therefore, the river flood prediction method based on the similarity matching and the extreme learning machine can overcome the defects of the prior art and effectively improve the actual forecasting precision of river flood.

Description

River flood prediction method based on similarity matching and extreme learning machine
Technical Field
The invention relates to a river flood prediction method based on similarity matching and an extreme learning machine, and belongs to the technical field of river monitoring.
Background
Flood forecasting is directly used for national economic construction and is an important non-engineering measure for disaster prevention and reduction. The flood forecast is not accurate enough due to the fact that in China, small and medium rivers are numerous, the terrain is complex, the climate types are various, the vegetation types are various, and the rainfall forecast accuracy is not high generally. Natural disasters such as flood, debris flow and the like caused by short-term heavy rainfall cause huge life and property loss to human beings every year, so that the improvement of the flood forecasting precision is of great importance.
A flood forecasting model based on data driving is a physical mechanism without considering hydrological processes, and is a black box method aiming at establishing an optimal mathematical relationship between input data and output data. Data-driven models are most commonly used as regression models, and in recent decades, with the progress of scientific technology, many novel prediction methods, such as artificial neural networks, support vector machines, extreme learning machines, fuzzy mathematics, etc., have been developed rapidly. Literature [ vermilion, lungnan, royal clouds, white Jingyi ] flood water level forecasting model based on artificial neural network [ J ] water conservancy project, 2005, 36 (7): 806-811 obtains good forecasting effect by using the artificial neural network under the condition of selecting reasonable input layer unit data. However, the artificial neural network needs to set a large number of parameters in the training process, and the parameter debugging is also very complicated, so that the artificial neural network is easy to fall into a local optimal solution. Literature [ wanlina, lisean, cheng macro support vector machine-based rainfall-runoff prediction study [ J ]. hydrology, 2009, 29 (1): 13-16, a rainfall-runoff prediction model is constructed by utilizing the support vector, and the rainfall-runoff prediction model obtains higher prediction precision compared with an artificial neural network, but because historical flood data contains a plurality of samples with different data distribution characteristics, a single model cannot well depict all data characteristics at the same time.
Disclosure of Invention
The invention aims to solve the technical problem of providing a river flood prediction method based on similarity matching and an extreme learning machine, which can overcome the defects of the prior art and effectively improve the actual forecasting efficiency of river flood.
The invention adopts the following technical scheme for solving the technical problems: the invention designs a river flood prediction method based on similarity matching and an extreme learning machine, which comprises the following steps:
step A, respectively carrying out flow value prediction on preset N historical flood samples based on water flow characteristics and rainfall characteristics, selecting an optimal limit learning machine model corresponding to the historical flood sample from preset M limit learning machine models, and using the optimal limit learning machine model as an optimal flow prediction model corresponding to the historical flood sample, so as to obtain optimal flow prediction models corresponding to the historical flood samples;
b, based on similarity matching of the N historical flood samples with respect to the real-time water flow characteristics and the real-time rainfall characteristics, selecting K historical flood samples before presetting as all reference flood samples according to a sequence of similarity matching values from large to small, wherein K is less than N;
and C, according to the similarity matching value of each reference flood sample relative to the real-time water flow characteristic and the real-time rainfall characteristic and the optimal flow prediction model corresponding to each reference flood sample, predicting the flow value based on the real-time water flow characteristic and the real-time rainfall characteristic by adopting a weighted average method, and judging the flood according to the obtained flow prediction value.
As a preferred technical solution of the present invention, the step a is specifically as follows:
respectively aiming at N historical flood samples, respectively adopting preset M limit learning machine models to respectively predict flow values of the historical prediction time points appointed by the historical flood samples according to the water flow characteristics and the rainfall characteristics of the historical flood samples, obtaining M predicted flow values of the historical flood samples corresponding to the appointed historical prediction time points, respectively comparing the M predicted flow values with real flow values of the appointed historical prediction time points corresponding to the historical flood samples, and selecting the limit learning machine model corresponding to the minimum absolute error as an optimal flow prediction model corresponding to the historical flood samples; and further obtaining the optimal flow prediction model corresponding to each historical flood sample.
As a preferred technical solution of the present invention, the step B specifically includes the following steps:
step B1, aiming at N historical flood samples, according to the water flow characteristics and the rainfall characteristics of each appointed historical time point, a water flow characteristic and rainfall characteristic feature matrix X ═ X ═ is constructed1、…、xn、…、xNAnd extracting flow values t of the historical flood samples corresponding to the appointed historical prediction time points respectivelynThen, go to step B2; wherein the content of the first and second substances,
Figure BDA0001265808130000021
n∈{1、…、N}，d∈{1、…、D}，xnrepresenting the water flow characteristic and rainfall characteristic vector corresponding to the nth historical flood sample, D representing 2 times of the number of the appointed historical time points,
Figure BDA0001265808130000022
represents the characteristics of the d corresponding to the n historical flood sample, tnRepresenting the real flow value of the nth historical flood sample corresponding to the appointed historical prediction time point;
b2, respectively aiming at each appointed historical time point, obtaining the average value of all historical flood sample water flow characteristics and rainfall characteristics corresponding to the appointed historical time point
Figure BDA0001265808130000023
And obtaining the flow average value of the flow values of the historical flood samples corresponding to the appointed historical prediction time points respectively
Figure BDA0001265808130000024
Then entering step B3;
Figure BDA0001265808130000025
to represent
Figure BDA0001265808130000026
The average value of all historical flood sample water flow characteristics and rainfall characteristics when the represented characteristics correspond to the designated historical time points;
and B3, aiming at each designated historical time point, respectively, according to the following formula:
Figure BDA0001265808130000031
obtaining coefficients R of water flow characteristics and rainfall characteristics corresponding to each appointed historical time point respectivelydAnd further according to
Figure BDA0001265808130000032
Obtaining the weight omega of each appointed historical time point corresponding to the water flow characteristic and the rainfall characteristic respectivelydThen, go to step B4;
step B4. includes obtaining real-time water flow characteristics and real-time rainfall characteristicsCharacteristic value of each time pointEstablishing water flow characteristics and rainfall characteristic vectors x corresponding to the real-time water flow characteristics and the real-time rainfall characteristicsqThen, go to step B5;
step B5. is performed for each historical flood sample according to the water flow characteristic and rainfall characteristic vector x corresponding to the historical flood samplenAnd real-time water flow characteristics and real-time rainfall characteristic vector xqAccording to the following formula:
Figure BDA0001265808130000034
obtaining a weighted Euclidean distance dist between a historical flood sample and a real-time water flow characteristic and a real-time rainfall characteristicwed(xq,xn) Further, obtaining weighted euclidean distances between each historical flood sample and the real-time water flow characteristics and the real-time rainfall characteristics respectively, taking the weighted euclidean distances as similarity matching values between each historical flood sample and the real-time water flow characteristics and the real-time rainfall characteristics respectively, and then entering step B6;
and B6., selecting historical flood samples corresponding to K preset similarity matching values as each reference flood sample according to the descending order of the similarity matching values, wherein K is less than N.
As a preferred technical solution of the present invention, the step C includes the steps of:
c1, obtaining weights of the reference flood samples corresponding to the real-time water flow characteristics and the real-time rainfall characteristics respectively according to similarity matching values of the reference flood samples corresponding to the real-time water flow characteristics and the real-time rainfall characteristics respectively;
step C2., adopting optimal flow prediction models corresponding to the reference flood samples respectively, and predicting flow values according to the real-time water flow characteristics and the real-time rainfall characteristics respectively to obtain K flow value prediction results corresponding to the real-time water flow characteristics and the real-time rainfall characteristics;
step C3., weighting the K flow value prediction results based on the weights of the real-time water flow characteristics and the real-time rainfall characteristics corresponding to the reference flood samples, respectively, to obtain flow prediction values of the real-time water flow characteristics and the real-time rainfall characteristics, and performing flood judgment according to the obtained flow prediction values.
As a preferred technical solution of the present invention, in the step C1, according to the similarity matching values dist of the respective reference flood samples with respect to the real-time water flow characteristics and the real-time rainfall characteristicskK is equal to {1, …, K }, and the following formula is adopted:
Figure BDA0001265808130000041
obtaining weights w of each reference flood sample corresponding to the real-time water flow characteristic and the real-time rainfall characteristic respectivelyk，wkAnd representing the weight of the K-th reference flood sample in the K reference flood samples corresponding to the real-time water flow characteristic and the real-time rainfall characteristic.
As a preferred technical solution of the present invention, in the step C3, based on the weights of the real-time water flow characteristics and the real-time rainfall characteristics respectively corresponding to each reference flood sample, weighting is performed on the K flow value prediction results, and the following formula is adopted:
Figure BDA0001265808130000042
obtaining a flow predicted value Q of the real-time water flow characteristic and the real-time rainfall characteristic, and judging flood according to the obtained flow predicted value; where K ∈ {1, …, K }, wkRepresenting the weight Q of the kth reference flood sample in the K reference flood samples corresponding to the real-time water flow characteristic and the real-time rainfall characteristickAnd representing an optimal flow prediction model corresponding to the kth reference flood sample in the K reference flood samples, predicting flow values according to the real-time water flow characteristics and the real-time rainfall characteristics, and obtaining a flow value prediction result.
Compared with the prior art, the river flood prediction method based on the similarity matching and the extreme learning machine has the following technical effects: the river flood prediction method based on the similarity matching and the extreme learning machine, which is designed by the invention, can overcome the defects of the prior art, can select a proper extreme learning mechanism to build and integrate by utilizing the similar characteristics of hydrologic phenomena, and effectively improve the actual prediction accuracy of river flood.
Drawings
FIG. 1 is a flow chart of a river flood prediction method based on similarity matching and extreme learning machine designed by the present invention;
fig. 2 is a schematic diagram of a variation relationship between flow rates corresponding to different technical solutions in the embodiment and time.
Detailed Description
The following description will explain embodiments of the present invention in further detail with reference to the accompanying drawings.
In the learning process of an extreme learning machine model (ELM model), an input weight value and a bias value of a hidden layer of a network are randomly generated, and only the number of nodes of the hidden layer of the network is needed to be set to generate a unique optimal solution, so that the method has the advantages of high learning speed, no local optimization and good generalization performance. However, the stability of the single extreme learning machine model (ELM model) is insufficient, and inappropriate parameters can result in poor forecasting effect. Therefore, in the technical scheme designed by the invention, M preset extreme learning machine models (ELM models) are adopted for design, and forecasting is carried out by utilizing an extreme learning machine integration method, so that higher forecasting precision and better stability are obtained.
As shown in fig. 1, the invention designs a river flood prediction method based on similarity matching and extreme learning machine, which comprises the following steps in practical application:
respectively aiming at N historical flood samples, predicting flow values of the historical prediction time points appointed by the historical flood samples by adopting preset M limit learning machine models (ELM models) according to the water flow characteristics and the rainfall characteristics of the historical flood samples at the appointed historical time points, obtaining M predicted flow values of the historical flood samples corresponding to the appointed historical prediction time points, respectively comparing the M predicted flow values with real flow values of the historical prediction time points corresponding to the historical flood samples, and selecting the limit learning machine model (ELM model) corresponding to the minimum absolute error as an optimal flow prediction model corresponding to the historical flood samples; and further obtaining the optimal flow prediction model corresponding to each historical flood sample.
In many applications, the similarity between two sequences is usually measured by some distance function, and the similarity degree is inversely proportional to the distance, and the smaller the distance, the more similar the distance is. Due to the similar characteristics of the hydrological phenomenon and considering that the importance of different characteristic values in a sample space is different, the weighted Euclidean distance is selected as a sample similarity measurement method.
And B, based on similarity matching of the N historical flood samples with respect to the real-time water flow characteristics and the real-time rainfall characteristics, selecting K historical flood samples before presetting as all reference flood samples according to a descending order of similarity matching values, wherein K is less than N.
In the practical application process, the step B specifically includes the following steps:
step B1, aiming at N historical flood samples, according to the water flow characteristics and the rainfall characteristics of each appointed historical time point, a water flow characteristic and rainfall characteristic feature matrix X ═ X ═ is constructed1、…、xn、…、xNAnd extracting flow values t of the historical flood samples corresponding to the appointed historical prediction time points respectivelynThen, go to step B2; wherein the content of the first and second substances,n∈{1、…、N}，d∈{1、…、D}，xnrepresenting the water flow characteristic and rainfall characteristic vector corresponding to the nth historical flood sample, D representing 2 times of the number of the appointed historical time points,
Figure BDA0001265808130000052
represents the d-th bit corresponding to the n-th historical flood sampleSymbol, tnAnd the real flow value of the nth historical flood sample corresponding to the appointed historical prediction time point is represented.
B2, respectively aiming at each appointed historical time point, obtaining the average value of all historical flood sample water flow characteristics and rainfall characteristics corresponding to the appointed historical time point
Figure BDA0001265808130000061
And obtaining the flow average value of the flow values of the historical flood samples corresponding to the appointed historical prediction time points respectively
Figure BDA0001265808130000062
Then entering step B3;
Figure BDA0001265808130000063
to represent
Figure BDA0001265808130000064
And when the represented characteristics correspond to the designated historical time points, the average value of all historical flood sample water flow characteristics and rainfall characteristics.
And B3, aiming at each designated historical time point, respectively, according to the following formula:
Figure BDA0001265808130000065
obtaining coefficients R of water flow characteristics and rainfall characteristics corresponding to each appointed historical time point respectivelydAnd further according to
Figure BDA0001265808130000066
Obtaining the weight omega of each appointed historical time point corresponding to the water flow characteristic and the rainfall characteristic respectivelydThen, the process proceeds to step B4.
Step B4. is to obtain the characteristic values of the real-time water flow characteristic and the real-time rainfall characteristic at each time point
Figure BDA0001265808130000067
Establishing real-time water flow characteristicsAnd water flow characteristic and rainfall characteristic vector x corresponding to real-time rainfall characteristicqThen, the process proceeds to step B5.
Step B5. is performed for each historical flood sample according to the water flow characteristic and rainfall characteristic vector x corresponding to the historical flood samplenAnd real-time water flow characteristics and real-time rainfall characteristic vector xqAccording to the following formula:
obtaining a weighted Euclidean distance dist between a historical flood sample and a real-time water flow characteristic and a real-time rainfall characteristicwed(xq,xn) And further obtaining weighted Euclidean distances between each historical flood sample and the real-time water flow characteristic and the real-time rainfall characteristic respectively, wherein the weighted Euclidean distances are used as similarity matching values between each historical flood sample and the real-time water flow characteristic and the real-time rainfall characteristic respectively, and then entering step B6.
And B6., selecting historical flood samples corresponding to K preset similarity matching values as each reference flood sample according to the descending order of the similarity matching values, wherein K is less than N.
And C, according to the similarity matching value of each reference flood sample relative to the real-time water flow characteristic and the real-time rainfall characteristic and the optimal flow prediction model corresponding to each reference flood sample, predicting the flow value based on the real-time water flow characteristic and the real-time rainfall characteristic by adopting a weighted average method, and judging the flood according to the obtained flow prediction value.
In the practical application process, the step C specifically includes the following steps:
c1, according to similarity matching values dist of the reference flood samples relative to the real-time water flow characteristics and the real-time rainfall characteristicskK is equal to {1, …, K }, and the following formula is adopted:
Figure BDA0001265808130000071
obtaining weights w of each reference flood sample corresponding to the real-time water flow characteristic and the real-time rainfall characteristic respectivelyk，wkAnd representing the weight of the K-th reference flood sample in the K reference flood samples corresponding to the real-time water flow characteristic and the real-time rainfall characteristic.
And C2., predicting flow values respectively according to the real-time water flow characteristics and the real-time rainfall characteristics by adopting the optimal flow prediction models respectively corresponding to the reference flood samples, and obtaining K flow value prediction results corresponding to the real-time water flow characteristics and the real-time rainfall characteristics.
Step C3. weights the K flow value prediction results based on the weights of the real-time water flow characteristics and the real-time rainfall characteristics corresponding to each reference flood sample, and adopts the following formula:
Figure BDA0001265808130000072
obtaining a flow predicted value Q of the real-time water flow characteristic and the real-time rainfall characteristic, and judging flood according to the obtained flow predicted value; where K ∈ {1, …, K }, wkRepresenting the weight Q of the kth reference flood sample in the K reference flood samples corresponding to the real-time water flow characteristic and the real-time rainfall characteristickAnd representing an optimal flow prediction model corresponding to the kth reference flood sample in the K reference flood samples, predicting flow values according to the real-time water flow characteristics and the real-time rainfall characteristics, and obtaining a flow value prediction result.
In order to verify the effect of the technical scheme designed by the invention, Zhejiang Chang flood season data is selected as a research object, the forecast period is set to be 4 hours, 1998-2010 annual flood season flood data is selected, the data time interval is 1 hour, wherein 7109 data in 1998-2008 are used as training samples, and 1679 data in 2009-2010 are used as test samples. Three methods, namely prediction by adopting a single extreme learning machine model, prediction by Bagging integration and river flood prediction method based on similarity matching and extreme learning machine, are respectively specifically applied to the research objects, and the specific practical application results are shown in the following table 1.
Model selection Deterministic Coefficient (DC) Mean Square Error (MSE)
Single extreme learning machine model 0.78 5196
Bagging integration 0.82 4242
Similarity matched ELM integration 0.85 3498
TABLE 1
Based on the practical application results shown in table 1, it is shown that: the forecasting accuracy can be well improved by the extreme learning machine integrated forecasting based on similarity matching, because the real-time flood samples and the historical flood samples have certain similarity, the optimal ELM models corresponding to the historical flood samples which are most similar to the real-time flood samples are found through the similarity matching, and the models are integrated, so that the forecasting effect is improved. As shown in fig. 2, it can be found that the flood peak error predicted by a single extreme learning machine model (ELM model) is large, the flood peak error is slightly reduced after prediction by a Bagging integration method, and the flood peak error is obviously reduced after prediction by a river flood prediction method based on similarity matching and an extreme learning machine, which is designed by the invention, and the overall prediction accuracy of flood in a field is greatly improved.
The embodiments of the present invention have been described in detail with reference to the drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.

Claims (5)

1. A river flood prediction method based on similarity matching and an extreme learning machine is characterized by comprising the following steps:
step A, respectively carrying out flow value prediction on preset N historical flood samples based on water flow characteristics and rainfall characteristics, selecting an optimal limit learning machine model corresponding to the historical flood sample from preset M limit learning machine models, and using the optimal limit learning machine model as an optimal flow prediction model corresponding to the historical flood sample, so as to obtain optimal flow prediction models corresponding to the historical flood samples;
b, based on similarity matching of the N historical flood samples with respect to the real-time water flow characteristics and the real-time rainfall characteristics, selecting K historical flood samples before presetting as all reference flood samples according to a sequence of similarity matching values from large to small, wherein K is less than N;
c, according to similarity matching values of the reference flood samples relative to the real-time water flow characteristics and the real-time rainfall characteristics respectively and optimal flow prediction models corresponding to the reference flood samples respectively, adopting a weighted average method to predict flow values based on the real-time water flow characteristics and the real-time rainfall characteristics, and judging flood according to the obtained flow prediction values; the step C comprises the following steps:
c1, obtaining weights of the reference flood samples corresponding to the real-time water flow characteristics and the real-time rainfall characteristics respectively according to similarity matching values of the reference flood samples corresponding to the real-time water flow characteristics and the real-time rainfall characteristics respectively;
step C2., adopting optimal flow prediction models corresponding to the reference flood samples respectively, and predicting flow values according to the real-time water flow characteristics and the real-time rainfall characteristics respectively to obtain K flow value prediction results corresponding to the real-time water flow characteristics and the real-time rainfall characteristics;
step C3., weighting the K flow value prediction results based on the weights of the real-time water flow characteristics and the real-time rainfall characteristics corresponding to the reference flood samples, respectively, to obtain flow prediction values of the real-time water flow characteristics and the real-time rainfall characteristics, and performing flood judgment according to the obtained flow prediction values.
2. The river flood prediction method based on the similarity matching and the extreme learning machine according to claim 1, wherein the step A is specifically as follows:
respectively aiming at N historical flood samples, respectively adopting preset M limit learning machine models to respectively predict flow values of the historical prediction time points appointed by the historical flood samples according to the water flow characteristics and the rainfall characteristics of the historical flood samples, obtaining M predicted flow values of the historical flood samples corresponding to the appointed historical prediction time points, respectively comparing the M predicted flow values with real flow values of the appointed historical prediction time points corresponding to the historical flood samples, and selecting the limit learning machine model corresponding to the minimum absolute error as an optimal flow prediction model corresponding to the historical flood samples; and further obtaining the optimal flow prediction model corresponding to each historical flood sample.
3. The river flood prediction method based on the similarity matching and the extreme learning machine according to claim 1, wherein the step B specifically comprises the following steps:
step B1, aiming at N historical flood samples, according to the water flow characteristics and the rainfall characteristics of each appointed historical time point, a water flow characteristic and rainfall characteristic feature matrix X ═ X ═ is constructed1、…、xn、…、xNAnd extracting each historical flood sample to respectively correspond to the appointmentsFlow value t at historical prediction time pointnThen, go to step B2; wherein the content of the first and second substances,
Figure FDA0002245603130000021
xnrepresenting the water flow characteristic and rainfall characteristic vector corresponding to the nth historical flood sample, D representing 2 times of the number of the appointed historical time points,
Figure FDA0002245603130000022
represents the characteristics of the d corresponding to the n historical flood sample, tnRepresenting the real flow value of the nth historical flood sample corresponding to the appointed historical prediction time point;
b2, respectively aiming at each appointed historical time point, obtaining the average value of all historical flood sample water flow characteristics and rainfall characteristics corresponding to the appointed historical time point
Figure FDA0002245603130000023
And obtaining the flow average value of the flow values of the historical flood samples corresponding to the appointed historical prediction time points respectivelyThen entering step B3;
Figure FDA0002245603130000025
to represent
Figure FDA0002245603130000026
The average value of all historical flood sample water flow characteristics and rainfall characteristics when the represented characteristics correspond to the designated historical time points;
and B3, aiming at each designated historical time point, respectively, according to the following formula:
Figure FDA0002245603130000027
obtaining water flows respectively corresponding to all the appointed historical time pointsCoefficient R of quantity characteristic and rainfall characteristicdAnd further according to
Figure FDA0002245603130000028
Obtaining the weight omega of each appointed historical time point corresponding to the water flow characteristic and the rainfall characteristic respectivelydThen, go to step B4;
step B4. is to obtain the characteristic values of the real-time water flow characteristic and the real-time rainfall characteristic at each time point
Figure FDA0002245603130000029
Establishing water flow characteristics and rainfall characteristic vectors x corresponding to the real-time water flow characteristics and the real-time rainfall characteristicsqThen, go to step B5;
step B5. is performed for each historical flood sample according to the water flow characteristic and rainfall characteristic vector x corresponding to the historical flood samplenAnd real-time water flow characteristics and real-time rainfall characteristic vector xqAccording to the following formula:
Figure FDA00022456031300000210
obtaining a weighted Euclidean distance dist between a historical flood sample and a real-time water flow characteristic and a real-time rainfall characteristicwed(xq,xn) Further, obtaining weighted euclidean distances between each historical flood sample and the real-time water flow characteristics and the real-time rainfall characteristics respectively, taking the weighted euclidean distances as similarity matching values between each historical flood sample and the real-time water flow characteristics and the real-time rainfall characteristics respectively, and then entering step B6;
and B6., selecting historical flood samples corresponding to K preset similarity matching values as each reference flood sample according to the descending order of the similarity matching values, wherein K is less than N.
4. The method of claim 1, wherein the river flood prediction method based on the similarity matching and the extreme learning machine is characterized in thatIn the step C1, according to the similarity matching value dist of each reference flood sample with respect to the real-time water flow characteristic and the real-time rainfall characteristickK is equal to {1, …, K }, and the following formula is adopted:
obtaining weights w of each reference flood sample corresponding to the real-time water flow characteristic and the real-time rainfall characteristic respectivelyk，wkAnd representing the weight of the K-th reference flood sample in the K reference flood samples corresponding to the real-time water flow characteristic and the real-time rainfall characteristic.
5. The method according to claim 1, wherein in the step C3, weighting is performed on the K flow value prediction results based on the weights of the real-time water flow characteristics and the real-time rainfall characteristics corresponding to the reference flood samples, respectively, and the following formula is adopted:
obtaining a flow predicted value Q of the real-time water flow characteristic and the real-time rainfall characteristic, and judging flood according to the obtained flow predicted value; where K ∈ {1, …, K }, wkRepresenting the weight Q of the kth reference flood sample in the K reference flood samples corresponding to the real-time water flow characteristic and the real-time rainfall characteristickAnd representing an optimal flow prediction model corresponding to the kth reference flood sample in the K reference flood samples, predicting flow values according to the real-time water flow characteristics and the real-time rainfall characteristics, and obtaining a flow value prediction result.
CN201710228116.4A 2017-04-10 2017-04-10 River flood prediction method based on similarity matching and extreme learning machine Active CN107145965B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710228116.4A CN107145965B (en) 2017-04-10 2017-04-10 River flood prediction method based on similarity matching and extreme learning machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710228116.4A CN107145965B (en) 2017-04-10 2017-04-10 River flood prediction method based on similarity matching and extreme learning machine

Publications (2)

Publication Number Publication Date
CN107145965A CN107145965A (en) 2017-09-08
CN107145965B true CN107145965B (en) 2020-01-24

Family

ID=59773697

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710228116.4A Active CN107145965B (en) 2017-04-10 2017-04-10 River flood prediction method based on similarity matching and extreme learning machine

Country Status (1)

Country Link
CN (1) CN107145965B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP7347918B2 (en) * 2017-11-20 2023-09-20 日本無線株式会社 Water level prediction method, water level prediction program, and water level prediction device
CN111323847B (en) * 2018-12-13 2023-05-23 北京金风慧能技术有限公司 Method and apparatus for determining weight ratios for analog integration algorithms
CN110688773A (en) * 2019-10-14 2020-01-14 中国电建集团成都勘测设计研究院有限公司 System and method for quickly positioning drainage basin blockage
CN111027763B (en) * 2019-12-06 2020-07-31 中国水利水电科学研究院 Basin flood response similarity analysis method based on machine learning
CN111445087A (en) * 2020-04-17 2020-07-24 华北水利水电大学 Flood prediction method based on extreme learning machine
CN112711917B (en) * 2021-03-26 2021-07-16 长江水利委员会水文局 Real-time flood forecasting intelligent method based on face recognition algorithm
CN113762618B (en) * 2021-09-07 2022-03-01 中国水利水电科学研究院 Lake water level forecasting method based on multi-factor similarity analysis
CN113779113B (en) * 2021-11-12 2022-03-08 水利部交通运输部国家能源局南京水利科学研究院 Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104408900A (en) * 2014-11-10 2015-03-11 柳州师范高等专科学校 Dynamic optimization based neural network flood warning device and method
CN105678067A (en) * 2015-12-31 2016-06-15 广东省水利水电科学研究院 Urban river surge flood control and drainage dynamic forecasting control method and system
CN105912770A (en) * 2016-04-08 2016-08-31 中山大学 Real-time hydrologic forecasting system
CN106249318A (en) * 2016-07-12 2016-12-21 清华大学 River basin flood Real-time Forecasting Method and predictor

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104408900A (en) * 2014-11-10 2015-03-11 柳州师范高等专科学校 Dynamic optimization based neural network flood warning device and method
CN105678067A (en) * 2015-12-31 2016-06-15 广东省水利水电科学研究院 Urban river surge flood control and drainage dynamic forecasting control method and system
CN105912770A (en) * 2016-04-08 2016-08-31 中山大学 Real-time hydrologic forecasting system
CN106249318A (en) * 2016-07-12 2016-12-21 清华大学 River basin flood Real-time Forecasting Method and predictor

Also Published As

Publication number Publication date
CN107145965A (en) 2017-09-08

Similar Documents

Publication Publication Date Title
CN107145965B (en) River flood prediction method based on similarity matching and extreme learning machine
CN108491970B (en) Atmospheric pollutant concentration prediction method based on RBF neural network
CN111639748B (en) Watershed pollutant flux prediction method based on LSTM-BP space-time combination model
CN108304668B (en) Flood prediction method combining hydrologic process data and historical prior data
CN114254561B (en) Waterlogging prediction method, system and storage medium
CN107730151B (en) Basin design flood calculation method based on conceptual hydrological model
CN108876021B (en) Medium-and-long-term runoff forecasting method and system
CN112288164B (en) Wind power combined prediction method considering spatial correlation and correcting numerical weather forecast
CN107292098A (en) Medium-and Long-Term Runoff Forecasting method based on early stage meteorological factor and data mining technology
CN108021773B (en) DSS database-based distributed hydrological model multi-field secondary flood parameter calibration method
CN110619432A (en) Feature extraction hydrological forecasting method based on deep learning
CN106971237B (en) A kind of Medium-and Long-Term Runoff Forecasting method for optimization algorithm of being looked for food based on bacterium
Chen et al. Groundwater level prediction using SOM-RBFN multisite model
CN101480143A (en) Method for predicating single yield of crops in irrigated area
Nie et al. Simulation and uncertainty analysis for groundwater levels using radial basis function neural network and support vector machine models
CN106228030A (en) A kind of fine granulation self study integrated prediction method based on wavelet transformation
CN116796799A (en) Method for creating small-river basin flood rainfall threshold model in area without hydrologic data
CN113033081A (en) Runoff simulation method and system based on SOM-BPNN model
CN114417740B (en) Deep sea breeding situation sensing method
CN109272144B (en) BPNN-based prediction method for NDVI (normalized difference of variance) in northern grassland area of China
CN111914488B (en) Data area hydrologic parameter calibration method based on antagonistic neural network
CN111914487B (en) Data-free regional hydrological parameter calibration method based on antagonistic neural network
CN117787081A (en) Hydrological model parameter uncertainty analysis method based on Morris and Sobol methods
CN106980764B (en) A kind of general hydrological simulation method directly evaporated using remote sensing
CN114169646A (en) Water bloom prediction method, device, electronic equipment and computer readable storage medium

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