CN111445087A - Flood prediction method based on extreme learning machine - Google Patents

Flood prediction method based on extreme learning machine Download PDF

Info

Publication number
CN111445087A
CN111445087A CN202010305624.XA CN202010305624A CN111445087A CN 111445087 A CN111445087 A CN 111445087A CN 202010305624 A CN202010305624 A CN 202010305624A CN 111445087 A CN111445087 A CN 111445087A
Authority
CN
China
Prior art keywords
extreme learning
learning machine
index
hidden layer
output
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
Application number
CN202010305624.XA
Other languages
Chinese (zh)
Inventor
刘扬
刘雪梅
王立虎
闫新庆
杨礼波
刘明堂
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.)
North China University of Water Resources and Electric Power
Original Assignee
North China University of Water Resources and Electric Power
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 North China University of Water Resources and Electric Power filed Critical North China University of Water Resources and Electric Power
Priority to CN202010305624.XA priority Critical patent/CN111445087A/en
Publication of CN111445087A publication Critical patent/CN111445087A/en
Pending legal-status Critical Current

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
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Computational Linguistics (AREA)
  • Development Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Game Theory and Decision Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (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 discloses a flood prediction method based on an extreme learning machine, relating to the technical field of disaster predictionnsThe root mean square error RMSE, the average absolute error MAE and the related error RE verify the efficiency and the precision advantages of the extreme learning machine relative to the artificial neural network, and experimental results show that the learning speed of the extreme learning machine is 32 times that of the artificial neural network, the noise processing capacity of the extreme learning machine is superior to that of the artificial neural network, compared with the artificial neural network, the extreme learning has great advantages in the aspects of prediction capacity and efficiency, and the extreme learning is a suitable choice for a flood forecasting model.

Description

Flood prediction method based on extreme learning machine
Technical Field
The invention relates to the technical field of disaster prediction, in particular to a flood prediction method based on an extreme learning machine.
Background
Flood is one of the most destructive natural disasters, and the occurrence of flood causes great damage to lives and properties of residents. Therefore, the establishment of a flood model to achieve early warning of regional flood is urgent.
In recent years, machine learning algorithms such as artificial neural networks have been increasingly applied to various fields due to the development of information science. Looking up related documents to find that the current flood prediction method can be basically divided into two categories, namely, performing correlation calculation on regional rainfall and runoff based on a linear model to further analyze the runoff so as to achieve the flood prediction effect; the other method is based on a neural network, an artificial neural network is combined with a Geographic Information System (GIS), elevation data is used as input by analyzing relevant data of the GIS, and runoff is used as output to achieve the purpose of flood prediction.
However, the former method considers only rainfall and does not take other factors into consideration, resulting in inaccurate prediction results, and the latter method uses an artificial neural network, which has a problem of low efficiency.
Disclosure of Invention
The embodiment of the invention provides a flood prediction method based on an extreme learning machine, which integrates an extreme learning method and a multi-factor index into a prediction method and can solve the problems in the prior art.
The invention provides a flood prediction method based on an extreme learning machine, which comprises the following steps:
constructing a multi-factor index system, wherein indexes in the multi-factor index system comprise a maximum three-day rainfall M3PD, a digital elevation model DEM, a terrain humidity index TWI, a normalized vegetation index NDVI, a river power index SPI, a soil texture index ST, a river-to-river distance index DR, a gradient index S L and a land utilization mode L UP;
the method comprises the steps of inputting indexes in a multi-factor index system into an input layer of an extreme learning machine, calculating an output layer result, namely runoff according to data of the input layer, correcting output weights of nodes of a hidden layer in a reverse error propagation mode according to the runoff, finally obtaining a predicted runoff output result, and judging flood risk according to the runoff.
Preferably, the extreme learning machine is represented as:
Figure BDA0002455694230000021
where m is the number of hidden layer nodes, n is the number of input layer nodes, βiIs the output weight of the i-th hidden layer node, g (-) is the activation function, wiIs the input weight of the i-th hidden layer node, biIs the offset of the ith hidden layer node, xjInput data representing the jth input level node, OjRepresenting the results of the output layer;
the learning target of the extreme learning machine is the minimum error of the output result, and the expression
Figure BDA0002455694230000022
I.e. find wi、xjAnd biSo that
Figure BDA0002455694230000023
The target is represented by a matrix H β ═ T, where H is the output matrix of the hidden layer node, β is the output weight matrix of the hidden layer node, T represents the desired output matrix, TjIs an element in T;
at the input weight wiAnd bias of hidden layer node biAfter being determined, the output matrix H of the hidden layer node is uniquely determined, and the target for training the extreme learning machine is converted into the linear formula H β ═ T, according to which the evaluation value of the output weight of the hidden layer node is uniquely determined as
Figure BDA0002455694230000024
Wherein H+Representing the plus-sign generalized inverse of matrix H.
Preferably, the output matrix H of the hidden layer node is solved by using an orthogonal decomposition method.
Preferably, the activation function is a Sigmoid function.
Preferably, precipitation data in an early stage of flood formation is selected as the terrain humidity index.
The flood prediction method based on the extreme learning machine establishes a flood prediction model according to various flood causes, the extreme learning machine and a Geographic Information System (GIS), and determines a coefficient r, a Willemite index WI and a Nash efficiency index EnsThe root mean square error RMSE, the average absolute error MAE and the related error RE verify the efficiency and the precision advantages of the extreme learning machine relative to the artificial neural network, and experimental results show that the learning speed of the extreme learning machine is 32 times that of the artificial neural network, the noise processing capacity of the extreme learning machine is superior to that of the artificial neural network, compared with the artificial neural network, the extreme learning has great advantages in the aspects of prediction capacity and efficiency, and the extreme learning is a suitable choice for a flood forecasting model.
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 topological model of an extreme learning machine used in the extreme learning machine-based flood prediction method of the present invention;
FIG. 2 is a schematic view of a broken line of the observed values and the radial flow rates predicted by the artificial neural network and the extreme learning machine of the present invention;
FIG. 3 is a schematic view of a scatter plot of a radial flow and an observed value predicted by an artificial neural network and an extreme learning machine of the present invention;
FIG. 4 is a schematic diagram of the results of the artificial neural network and the runoff predicted by the extreme learning machine of the present invention under different evaluation indexes;
fig. 5 is a schematic diagram of an evaluation result of the runoff quantity predicted by the artificial neural network and the extreme learning machine of the present invention under the RE index.
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.
The invention provides a flood prediction method based on an extreme learning machine, which adopts the extreme learning machine, wherein the topological structure of the extreme learning machine is shown in figure 1 and comprises an input layer, a hidden layer and an output layer, and the flood prediction method based on the extreme learning machine comprises the following steps:
step 1, constructing a multi-factor index system, wherein the multi-factor adopted in the embodiment comprises nine factors including a maximum three-day rainfall M3PD, a digital elevation model DEM, a terrain humidity index TWI, a normalized vegetation index NDVI, a river power index SPI, a soil texture index ST, a river-to-river distance index DR, a slope index S L and a land utilization mode L UP, and the factors are input into an extreme learning machine as data of an input layer in a time sequence mode.
(1) Maximum three day precipitation
Precipitation is the important input parameter of flood accurate prediction, and an accurate flood forecast system is basically according to a large amount of hydrologic data, and this embodiment is according to data fitting degree, compares the precipitation data of one day to six days before the flood takes place and is regarded as the prediction training data, obtains the data fitting degree of three days before the highest, consequently chooses the biggest three days precipitation as the index of input layer.
(2) Digital elevation model
The digital elevation model DEM is a model for expressing the ground elevation by utilizing a group of ordered numerical values, and has wide application in the aspects of hydrology, meteorology, geology, soil and the like.
(3) Topographic moisture index
The topography humidity divide into absolute humidity and relative humidity, because the topography humidity is not convenient for measure, consequently chooses for use the rainfall data of flood formation earlier stage as topography humidity index, when soil completely reached the saturation state, the maximum value of soil absorption moisture this moment, later, the increment of rainfall directly equals the flood volume.
(4) Normalized vegetation index
The normalized vegetation index NDVI reflects the vegetation growth state and the vegetation coverage, the normalized vegetation index can reflect the ground moisture degree, snow, dead leaves and the like, and the calculation formula is as follows: NDVI ═ (NIR-R)/(NIR + R), where NIR is the reflectance in the near infrared band and R is the reflectance in the red band.
(5) Power index of river
According to the theory of Florinsky,2011, the river power index describes the potential water erosion degree of a certain river section, and as the gradient of a water collection area increases, the amount of water and the water flow speed contributed by an uphill region increase, so that the river power index increases, and therefore the index should be used as an evaluation index of flood risk together with a digital elevation model.
(6) Index of soil texture
In the expression Florinsky,2011, the soil texture index is the soil type in the reaction research area, and describes the content and relative proportion of different particles, so as to detect the infiltration capacity of the soil, the soil infiltration rate refers to the infiltration water amount of the surface area soil in unit time, the infiltration rate is larger in the initial infiltration period, then along with the increase of rainfall, the infiltration rate is changed from big to small so as to keep a stable value, and the initial infiltration rate and the final stable value are respectively called as an initial infiltration rate and a stable infiltration rate.
(7) Distance from river index
In a Geographic Information System (GIS), river data is extracted by using a digital elevation model, the value of the river data is set to be zero, and then the river data is continuously increased according to the distance from a river.
(8) Grade index
The slope index reflects the change degree of the terrain, the slope indexes of a mountain area and a plain are greatly different, the flow rate of flood can be increased by a severe slope of the mountain area, and then water collection is prevented, and flood can be well collected in a low-lying and gentle area.
(9) Land use mode
Land utilization is all activities of purposefully developing and utilizing land resources for human beings, and analysis of land utilization change is to judge a terrain change rule according to land use conditions. Under different land utilization modes, the runoff quantity has large difference, so that the runoff coefficient and the land utilization mode can be used for establishing a correlation.
And 2, calculating an output layer result, namely runoff according to the data of the input layer, correcting the output weight of the hidden layer node in a reverse error propagation mode according to the runoff, and improving the accuracy of the runoff output result, wherein the runoff is used as a basis for judging flood occurrence risk.
The number of input layer nodes in the extreme learning machine is n, and is defined as Q ═ Q1Q2... Qn]TThe number of hidden layer nodes is m, so an extreme learning machine with n input layer nodes and m hidden layer nodes is represented as:
Figure BDA0002455694230000061
β thereiniIs the output weight of the i-th hidden layer node, g (-) is the activation function, in this embodiment, the activation function is Sigmoid function, wiIs the input weight of the i-th hidden layer node, biIs the offset of the ith hidden layer node, xjInput data representing the jth input level node, wixjDenotes wiAnd xjInner product of, OjThe results of the output layer are shown.
The learning object of the extreme learning machine described above is to minimize the error of the output result, and the expression can be expressed as
Figure BDA0002455694230000062
I.e. the learning objective is to find wi、xjAnd biSo that
Figure BDA0002455694230000063
The target can be expressed by a matrix H β ═ T, where H is the output matrix of the hidden layer node, β is the output weight matrix of the hidden layer node, T represents the desired output matrix, and the expressions are:
Figure BDA0002455694230000064
Figure BDA0002455694230000065
to train the extreme learning machine, w needs to be obtainedi、biAnd βiEvaluation value of
Figure BDA0002455694230000066
And
Figure BDA0002455694230000067
so that
Figure BDA0002455694230000068
This equation can be equated with a minimum loss function
Figure BDA0002455694230000071
The conventional gradient descent algorithm may solve for hidden layer node output weights β in the minimization loss functioniBasic gradient learning based algorithms require adjustment of parameters in an iterative process, but in E L M (extreme learning machine), once the weights w are inputiAnd bias of hidden layer node biRandomly determining the output matrix H of the hidden layer node, uniquely determining the output matrix H of the hidden layer node, converting the training extreme learning machine into a linear formula H β -T, and uniquely determining the output weight of the hidden layer node
Figure BDA0002455694230000072
WhereinH+The generalized inverse of the plus sign (Moore-Penrose) of matrix H is shown and the found norm β is the smallest and β unique.
For the output matrix H of the hidden layer node, in the past, the matrix H is solved by using singular value decomposition in the process of least square solution solving, and then the plus sign generalized inverse is solved, so that the most core part of the extreme learning machine is the matrix H+And (4) solving. In the embodiment, the matrix H is solved by using orthogonal decomposition, the method has excellent performance when solving a linear problem, and the calculation process is simple and effective. In order to improve the calculation performance, the output weight is solved by adopting orthogonal decomposition, and compared with singular value decomposition, the orthogonal decomposition method has the advantages of high efficiency, greatly simplified calculation process and stable performance. The orthogonal decomposition presents efficient performance advantages regardless of the matrix H size.
Assuming that H is an M x N matrix, the singular value decomposition of H into H-U-Sigma V can be obtained from the theory of matrix theoryTWherein
Figure BDA0002455694230000073
HHIs the conjugate transpose of H, sigma is a diagonal matrix with the singular value λ of H as the diagonal element, VTIs HHThe feature vector corresponding to the orthonormal basis of (1). The complete orthogonal decomposition method is adopted to divide H into S R, namely H is S R, S is an orthogonal matrix obtained by solving a standard orthogonal base of a column vector of the matrix H, R is an upper triangular matrix obtained by calculation according to S and H, and according to the property of the orthogonal matrix, when the inverse matrix of S is solved, the transposition of the inverse matrix can be directly solved, so that the matrix inversion can be solved without specially carrying out inversion on the decomposed matrix.
The results obtained with the method of the invention are verified below.
The embodiment adopts the determination coefficients r, the Wilminder index WI and the Nash efficiency index EnsThe root mean square error RMSE, the average absolute error MAE and the related error RE are used as verification indexes so as to verify the advantages and the characteristics of the invention. The expression of the verification index is as follows:
Figure BDA0002455694230000081
Figure BDA0002455694230000082
Figure BDA0002455694230000083
Figure BDA0002455694230000084
Figure BDA0002455694230000085
Figure BDA0002455694230000086
wherein Zo、Zf、
Figure BDA0002455694230000087
And
Figure BDA0002455694230000088
RMSE compares to MAE, RMSE equals L2 norm, MAE equals L1 norm, RMSE describes the deviation between a set of data measurements and truth, applicable to the case where the overall distribution is normal, and MAE is applicable to the case where the overall distribution is uniformNumbers are also often applied in hydrological models, which can provide a complementary information between the correlation coefficients r, RMSE and MAE. However, Willmott found E in 1981nsThe performance of the prediction model cannot be well reflected because of EnsIs the magnitude of the weight that establishes the difference between the observed and predicted values. The Willmott coefficient is therefore employed in embodiments to represent the mean and absolute values (including both observed and absolute values) between observed and predicted values.
In the following, a flood prediction experiment is carried out according to the probability of future flood occurrence predicted by the past time sequence data, and an artificial neural network ANN is compared with an extreme learning machine E L M, so that the maximum three-day rainfall is found to have great correlation with the flood occurrence in the experiment, the influence coefficient of the soil texture and the distance from the river on the flood is relatively small, in order to fully explain the advantages of the method, the input data is divided into five parts, four parts of data are trained and verified, the four parts of training data are respectively named as a model I to a model IV, and the experiment verification is carried out by using a cross verification mode.
As can be seen from fig. 2, the extreme learning machine shows a satisfactory degree of fitting in time series data prediction, the prediction accuracy of the extreme learning machine is higher than that of the artificial neural network in the models one to four, and the extreme learning machine is particularly obvious in the models two and four. Errors of the initial data segment and the final data segment of the model II are larger, but the extreme learning machine does not show fluctuation at the moment, which shows that the extreme learning machine shows stronger stability in the time sequence data prediction process. In the fourth model, the artificial neural network with coordinates ten to twenty-five has larger instability, because the artificial neural network is influenced by some data with larger fluctuation in the training process, which is called noise, and if the problem is to be avoided, the data needs to be regularized. In the aspect of learning efficiency, the artificial neural network is more advantageous than extreme learning, and the learning efficiency of the extreme learning machine is 32 times that of the artificial neural network under the condition that the learning rate of the artificial neural network is set to be lower 0.02.
In order to better evaluate the accuracy of the method of the present invention, a scatter plot is plotted as shown in fig. 3, wherein the x-axis represents the observed value and the y-axis represents the predicted value, and for better observed data, a function of y ═ x is also given, from which it can be understood that the points of the plot around y ═ x are the more accurate points to be predicted. The artificial neural network in the model two and the model three only has poor performance in the early stage, the performance in the later stage is basically equal to that of the extreme learning machine, and when the artificial neural network converges to the allowable error, the prediction effect is basically acceptable. Compared with an artificial neural network, the extreme learning machine has better performance from the model I and the model II, the artificial neural network has poorer performance in the early stage and the later stage of prediction because the convergence speed of the artificial neural network is lower in the training process, and the problem of larger error fluctuation also occurs in the later stage of prediction because the artificial neural network has poorer performance on the generalization capability, and some data far away from the center have larger influence on the model, so that the convergence speed of the extreme learning machine is higher compared with the artificial neural network. After the data are input into the extreme learning machine, Moore-Penrose is directly utilized to solve the optimal solution of the weights, and the optimal solution is unique. It can also be seen from fig. 3 that the extreme learning machine is relatively considerable in data prediction performance compared to the artificial neural network due to its excellent convergence rate in the early stage of prediction.
TABLE 1 validation index results
Figure BDA0002455694230000101
By combining table 1 and fig. 4, it can be found that, in the four models, only the determination coefficient r of the model four-pole limit learning machine is smaller than the artificial neural network, and the determination coefficient is the reliability degree of the response random variable, while the stability degree of the limit learning machine is larger than the artificial neural network according to the performance of the model. Index of Nash efficiency EnsIn hydrological models, E, which are commonly used to evaluate the quality of the modelnsValue of minus infinity to 1, EnsClose to 1, denotes the modeThe formula quality is good, and the model reliability is high; ensThe simulation result is close to 0, which means that the simulation result is close to the average value level of the observed value, namely the overall result is credible, but the process simulation error is large; ensFar less than 0, the model is not trusted. The Willemmer's index WI corresponding to the Nash efficiency index is used for proving the reliability of the model, and unlike the Nash efficiency index, the Willemer's index is the absolute value of the prediction and observation used, and the comprehensive Nash efficiency index EnsIn comparison with the Willemt index WI, the extreme learning machines in the four models are all higher than the artificial neural network, which indicates that the credibility is also higher than the artificial neural network. The RMSE and the MAE are complementary information in the verification index, so that the error level of the artificial neural network and the error level of extreme learning are basically not in an order of magnitude, and the error level of the artificial neural network and the model prediction accuracy are further explained to be smaller than the extreme learning.
Fig. 5 reflects the artificial neural network and the extreme learning RE index level, and the curve at the bottom of the graph is the RE index level of the extreme learning machine, from which it can be seen that the convergence degree of the artificial neural network and the extreme learning is faster in the training process of the model, which indicates that the artificial neural network will have a larger error in some data predictions in the previous period, although some errors may exist in the extreme learning, the accuracy of the extreme learning and the neural network will increase with the increase of the sample, but the convergence degree of the extreme learning is faster than that of the neural network.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (5)

1. The flood prediction method based on the extreme learning machine is characterized by comprising the following steps of:
constructing a multi-factor index system, wherein indexes in the multi-factor index system comprise a maximum three-day rainfall M3PD, a digital elevation model DEM, a terrain humidity index TWI, a normalized vegetation index NDVI, a river power index SPI, a soil texture index ST, a river-to-river distance index DR, a gradient index S L and a land utilization mode L UP;
the method comprises the steps of inputting indexes in a multi-factor index system into an input layer of an extreme learning machine, calculating an output layer result, namely runoff according to data of the input layer, correcting output weights of nodes of a hidden layer in a reverse error propagation mode according to the runoff, finally obtaining a predicted runoff output result, and judging flood risk according to the runoff.
2. The extreme learning machine-based flood prediction method of claim 1, wherein the extreme learning machine is represented as:
Figure FDA0002455694220000011
where m is the number of hidden layer nodes, n is the number of input layer nodes, βiIs the output weight of the i-th hidden layer node, g (-) is the activation function, wiIs the input weight of the i-th hidden layer node, biIs the offset of the ith hidden layer node, xjInput data representing the jth input level node, OjRepresenting the results of the output layer;
the learning target of the extreme learning machine is the minimum error of the output result, and the expression
Figure FDA0002455694220000012
I.e. find wi、xjAnd biSo that
Figure FDA0002455694220000013
The target is represented by a matrix H β ═ T, where H is the output matrix of the hidden layer node, β is the output weight matrix of the hidden layer node, T represents the desired output matrix, TjIs an element in T;
at the input weight wiAnd bias of hidden layer node biAfter being determined, the output matrix H of the hidden layer node is uniquely determined, and the target for training the extreme learning machine is converted into the linear formula H β ═ T, according to which the evaluation value of the output weight of the hidden layer node is uniquely determined as
Figure FDA0002455694220000021
Wherein H+Representing the plus-sign generalized inverse of matrix H.
3. The extreme learning machine-based flood prediction method of claim 2, wherein the solving of the output matrix H of the hidden layer nodes is calculated using an orthogonal decomposition method.
4. The extreme learning machine-based flood prediction method of claim 2, wherein the activation function employs a Sigmoid function.
5. The extreme learning machine-based flood prediction method of claim 1, wherein precipitation data from an earlier stage of flood formation is selected as the terrain moisture index.
CN202010305624.XA 2020-04-17 2020-04-17 Flood prediction method based on extreme learning machine Pending CN111445087A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010305624.XA CN111445087A (en) 2020-04-17 2020-04-17 Flood prediction method based on extreme learning machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010305624.XA CN111445087A (en) 2020-04-17 2020-04-17 Flood prediction method based on extreme learning machine

Publications (1)

Publication Number Publication Date
CN111445087A true CN111445087A (en) 2020-07-24

Family

ID=71653326

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010305624.XA Pending CN111445087A (en) 2020-04-17 2020-04-17 Flood prediction method based on extreme learning machine

Country Status (1)

Country Link
CN (1) CN111445087A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112016839A (en) * 2020-08-31 2020-12-01 华北水利水电大学 Flood disaster prediction and early warning method based on QR-BC-ELM
CN113378484A (en) * 2021-07-12 2021-09-10 中国水利水电科学研究院 Plain river network area flood process forecasting method based on machine learning
CN114970315A (en) * 2022-04-19 2022-08-30 河海大学 Urban accumulated water simulation and rapid prediction method based on spatial dynamic characteristic deep learning

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103218522A (en) * 2013-04-01 2013-07-24 民政部国家减灾中心 Method and device for grading flood risk
CN106202790A (en) * 2016-07-20 2016-12-07 中国水利水电科学研究院 A kind of novel distributed Hebei Model construction method and application thereof
CN107145965A (en) * 2017-04-10 2017-09-08 河海大学 A kind of river flood Forecasting Methodology based on similarity mode and extreme learning machine
CN107563567A (en) * 2017-09-18 2018-01-09 河海大学 Core extreme learning machine Flood Forecasting Method based on sparse own coding
CN108647454A (en) * 2018-05-16 2018-10-12 中国水利水电科学研究院 Basin synthetic disaster information processing method based on hydrological simulation and system
CN108876021A (en) * 2018-05-31 2018-11-23 华中科技大学 A kind of Medium-and Long-Term Runoff Forecasting method and system
CN110276477A (en) * 2019-05-29 2019-09-24 河海大学 A kind of Flood Forecasting Method based on layering Bayesian network and incremental learning
US10579749B1 (en) * 2015-06-26 2020-03-03 State Farm Mutual Automobile Insurance Company Systems and methods for augmented reality for disaster simulation

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103218522A (en) * 2013-04-01 2013-07-24 民政部国家减灾中心 Method and device for grading flood risk
US10579749B1 (en) * 2015-06-26 2020-03-03 State Farm Mutual Automobile Insurance Company Systems and methods for augmented reality for disaster simulation
CN106202790A (en) * 2016-07-20 2016-12-07 中国水利水电科学研究院 A kind of novel distributed Hebei Model construction method and application thereof
CN107145965A (en) * 2017-04-10 2017-09-08 河海大学 A kind of river flood Forecasting Methodology based on similarity mode and extreme learning machine
CN107563567A (en) * 2017-09-18 2018-01-09 河海大学 Core extreme learning machine Flood Forecasting Method based on sparse own coding
CN108647454A (en) * 2018-05-16 2018-10-12 中国水利水电科学研究院 Basin synthetic disaster information processing method based on hydrological simulation and system
CN108876021A (en) * 2018-05-31 2018-11-23 华中科技大学 A kind of Medium-and Long-Term Runoff Forecasting method and system
CN110276477A (en) * 2019-05-29 2019-09-24 河海大学 A kind of Flood Forecasting Method based on layering Bayesian network and incremental learning

Non-Patent Citations (9)

* Cited by examiner, † Cited by third party
Title
PUNYAPHOL HORATAN等: "Robust extreme learning machine", NEUROCOMPUTING *
WEI GUO等: "An integrated chaotic time series prediction model based on efficient extreme learning machine and differential evolution", NATURAL COMPUTING APPLICATIONS *
刘静: "并行极限学习机在洪水预测中的应用研究", 《西北大学学报(自然科学版)》, no. 04, pages 1 - 2 *
孙娜等: "基于正则极限学习机的非平稳径流组合预测", 《水力发电学报》, no. 08 *
徐涛;郭威;吕宗磊;: "基于快速极限学习机和差分进化的机场噪声预测模型", 电子与信息学报, no. 06 *
楚永贺;闫德勤;徐丽丽;高晴;: "基于微分同胚的鲁棒激活函数的极端学习机", 计算机应用研究, no. 09 *
王强;吕政;王霖青;王伟;: "基于深度去噪核映射的长期预测模型", 控制与决策, no. 05 *
苏辉东等: "机器学习在径流预测中的应用研究", 中国农村水利水电 *
马凯凯等: "数据驱动的中小河流智能洪水预报方法对比研究", 《中国科学技术大学学报》, no. 09 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112016839A (en) * 2020-08-31 2020-12-01 华北水利水电大学 Flood disaster prediction and early warning method based on QR-BC-ELM
CN113378484A (en) * 2021-07-12 2021-09-10 中国水利水电科学研究院 Plain river network area flood process forecasting method based on machine learning
CN114970315A (en) * 2022-04-19 2022-08-30 河海大学 Urban accumulated water simulation and rapid prediction method based on spatial dynamic characteristic deep learning
CN114970315B (en) * 2022-04-19 2023-08-04 河海大学 Urban ponding simulation and rapid prediction method based on space dynamic characteristic deep learning

Similar Documents

Publication Publication Date Title
Kisi et al. Precipitation forecasting using wavelet-genetic programming and wavelet-neuro-fuzzy conjunction models
CN111582755B (en) Mountain torrent disaster comprehensive risk dynamic assessment method based on multi-dimensional set information
Yu et al. Comparison of random forests and support vector machine for real-time radar-derived rainfall forecasting
CN109272146B (en) Flood prediction method based on deep learning model and BP neural network correction
Quej et al. ANFIS, SVM and ANN soft-computing techniques to estimate daily global solar radiation in a warm sub-humid environment
Nourani et al. Multi-step ahead modeling of reference evapotranspiration using a multi-model approach
Ramirez et al. Artificial neural network technique for rainfall forecasting applied to the Sao Paulo region
Luk et al. A study of optimal model lag and spatial inputs to artificial neural network for rainfall forecasting
Kisi et al. River suspended sediment estimation by climatic variables implication: Comparative study among soft computing techniques
CN109299812B (en) Flood prediction method based on deep learning model and KNN real-time correction
CN111445087A (en) Flood prediction method based on extreme learning machine
Kar et al. Rain gauge network design for flood forecasting using multi-criteria decision analysis and clustering techniques in lower Mahanadi river basin, India
Shiri et al. Estimation of daily suspended sediment load by using wavelet conjunction models
JP2019094640A (en) Water level prediction method, water level prediction program and water level prediction device
Wang et al. Flood simulation using parallel genetic algorithm integrated wavelet neural networks
Tabatabaei et al. Suspended sediment load prediction using non-dominated sorting genetic algorithm II
CN112287294A (en) Time-space bidirectional soil water content interpolation method based on deep learning
CN111968019A (en) Landslide hazard susceptibility prediction method based on CNN and RNN integration
CN113837352A (en) Rainfall-runoff space-time relation simulation method based on long-short term memory neural network
Zhao et al. Risk assessment of flood disaster and forewarning model at different spatial-temporal scales
Zhao et al. Optimally integrating multi-source products for improving long series precipitation precision by using machine learning methods
Schmidinger et al. Validation of uncertainty predictions in digital soil mapping
Sojitra et al. Comparative study of daily rainfall forecasting models using adaptive-neuro fuzzy inference system (ANFIS)
Santra et al. Delineation of hydrologically similar units in a watershed based on fuzzy classification of soil hydraulic properties
CN107608938A (en) The factor screening method towards two-value classification of tree algorithm is returned based on enhancing

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
CB03 Change of inventor or designer information
CB03 Change of inventor or designer information

Inventor after: Liu Yang

Inventor after: Wang Lihu

Inventor after: Liu Xuemei

Inventor after: Wu Huixin

Inventor after: Yang Libo

Inventor before: Liu Yang

Inventor before: Liu Xuemei

Inventor before: Wang Lihu

Inventor before: Yan Xinqing

Inventor before: Yang Libo

Inventor before: Liu Mingtang