CN113779113A - Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation - Google Patents

Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation Download PDF

Info

Publication number
CN113779113A
CN113779113A CN202111336042.9A CN202111336042A CN113779113A CN 113779113 A CN113779113 A CN 113779113A CN 202111336042 A CN202111336042 A CN 202111336042A CN 113779113 A CN113779113 A CN 113779113A
Authority
CN
China
Prior art keywords
flood
time
rainfall
historical
similarity
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202111336042.9A
Other languages
Chinese (zh)
Other versions
CN113779113B (en
Inventor
云兆得
李伶杰
王银堂
胡庆芳
王磊之
刘勇
崔婷婷
盖永伟
李笑天
高轩
张野
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.)
Jiangsu Water Resources Service Center
Nanjing Hydraulic Research Institute of National Energy Administration Ministry of Transport Ministry of Water Resources
Original Assignee
Jiangsu Water Resources Service Center
Nanjing Hydraulic Research Institute of National Energy Administration Ministry of Transport Ministry of Water Resources
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 Jiangsu Water Resources Service Center, Nanjing Hydraulic Research Institute of National Energy Administration Ministry of Transport Ministry of Water Resources filed Critical Jiangsu Water Resources Service Center
Priority to CN202111336042.9A priority Critical patent/CN113779113B/en
Publication of CN113779113A publication Critical patent/CN113779113A/en
Application granted granted Critical
Publication of CN113779113B publication Critical patent/CN113779113B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2465—Query processing support for facilitating data mining operations in structured databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2474—Sequence data queries, e.g. querying versioned data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/29—Geographical information databases
    • 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
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G06Q50/265—Personal security, identity or safety
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2216/00—Indexing scheme relating to additional aspects of information retrieval not explicitly covered by G06F16/00 and subgroups
    • G06F2216/03—Data mining
    • 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)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Data Mining & Analysis (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Computational Linguistics (AREA)
  • Development Economics (AREA)
  • General Business, Economics & Management (AREA)
  • Software Systems (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Computer Security & Cryptography (AREA)
  • Educational Administration (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Remote Sensing (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Alarm Systems (AREA)

Abstract

The invention provides a flood dynamic estimation method and a system based on rainfall flood space-time process similarity excavation, wherein the method comprises the following steps: acquiring historical and real-time remote sensing inversion observation rainfall, numerical prediction rainfall and hydrological station historical site flood information in a research area, and establishing a historical rainfall flood database; splicing the early stage of the moment, the observed rainfall and the numerical forecast rainfall, and retrieving the rainfall similar to the time-space process and a corresponding historical flood process set 1 by adopting the structural similarity of the structural similarity index and the dynamic time warping DTW; retrieving a historical flood process set 2 with similar time sequence by adopting DTW; merging the sets 1 and 2 to estimate the subsequent flood by the average similar flood process; and updating information along with the time, and realizing dynamic flood estimation. The similarity analysis is expanded to a three-dimensional space-time process from a one-dimensional time sequence, and the accuracy of the estimated subsequent water inflow is higher than that of the traditional method.

Description

Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation
Technical Field
The invention relates to G06F: the field of electric digital data processing, in particular to a flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation.
Background
The rain flood process is the result of coupling of factors such as weather system evolution and river basin underlying surface conditions, and although the factors cannot be completely repeated, the weather system dominating one region is not completely and invisibly found, and certain regularity still exists, so that rain flood events occurring in different periods may have certain similarity. Along with the accumulation of the number of meteorological hydrological events and the deepening of cause recognition, a historical rainfall flood database with a certain scale is constructed, on the basis of similarity mining of rainfall flood events, the estimation of the development situation of the flood by utilizing the historical similar rainfall flood process is another important idea for improving the prediction capability, and the method has important significance for safe and efficient utilization of the rainfall flood.
At present, research about rainfall flood similarity excavation is mainly developed based on rainfall flood real-time observation information and historical data, wherein the related rainfall time sequence is mostly the surface average rainfall calculated based on the observation data of a ground rainfall station network, the consideration of rainfall spatial distribution and rainfall landing area positions is lacked, and particularly, the drainage basin with a large water collection area is provided. In addition, precipitation similarity analysis in the existing research only utilizes real-time observation precipitation information, and does not consider the development situation of the late stage of the rain condition.
Disclosure of Invention
The purpose of the invention is as follows: the invention provides a flood dynamic estimation method based on rainfall flood time-space process similarity mining, which comprises four core links of historical rainfall flood database construction, historical flood process retrieval based on rainfall time-space process similarity, historical flood process retrieval based on flood time sequence process similarity, and dynamic estimation based on historical similar flood. The method utilizes observation and forecast information in a coupling mode, corresponding historical similar floods are retrieved according to similarity of time-space processes and similarity of time sequence processes aiming at precipitation and flood respectively, and the historical similar floods considering occurrence and development processes of rainfall flood more comprehensively can be obtained by integrating retrieval results of the similarity of precipitation and the similarity of flood, so that the future development situation of the flood is estimated, and the accuracy of estimation of the subsequent incoming water amount in the flood process is improved.
The technical scheme is as follows: on the first hand, a flood dynamic estimation method based on rainfall flood space-time process similarity excavation is provided, and the method comprises the following steps:
step 1, obtaining historical remote sensing inversion rainfall information, real-time remote sensing inversion rainfall information, numerical forecast rainfall information and historical stage flood information of a preset number of hydrologic stations in a research area, matching time scales of historical rainfall and flood information, and constructing a hydrologic station historical rainfall flood database;
step 2, aiming at the flood which is happening in the hydrological station, matching and splicing the space-time scales of early-stage remote sensing inversion precipitation before the flood rising time, observed remote sensing inversion precipitation from the flood rising time to the confrontation time and numerical forecast precipitation information after the confrontation time, calculating a surface average precipitation time sequence, and searching a historical precipitation process with similar surface average precipitation time sequence in the historical rainfall database established in the step 1 by adopting a time sequence similarity analysis method; screening historical rainfall processes similar to the spatio-temporal process by adopting a spatio-temporal information similarity analysis method, and extracting a corresponding similarity historical flood set 1;
step 3, searching a similarity historical flood set 2 with similar time sequence process in the historical rainfall flood database established in the step 1 by adopting a time sequence similarity analysis method according to flood information in an effective forecast period forecasted by the constantly observed flood and flood forecast model at the splicing face;
step 4, merging the similarity history flood sets 1 and 2 retrieved in the steps 2 and 3 to obtain a similarity history flood set 3, reasonably determining the splicing time point of each field of similar flood, calculating the average similar flood process after the splicing time point by adopting a weighted average method, and estimating the subsequent flood process through translation and splicing operation;
step 5, judging whether the follow-up flood process in the step 4 contains a flood peak or not, and stopping dynamic estimation if the follow-up flood process does not contain the flood peak; otherwise, updating the rainfall flood observation and forecast information along with the time, repeating the steps 2-4, and dynamically estimating the flood development situation.
In a further embodiment of the first aspect, the step 1 is further:
step 1-1, comprehensively performing remote sensing inversion on precipitation information and historical field flood information, reasonably determining time points and step lengths of field flood information at equal intervals, and acquiring flood data of sampling time points by adopting a cubic spline interpolation method;
step 1-2, judging whether the time step length of the historical remote sensing inversion precipitation information is consistent with the flood data or not; if the difference is not consistent, distributing the precipitation amount to a time period smaller than the time step of the flood information by an arithmetic mean method, and obtaining historical remote sensing inversion precipitation information matched with the time point and the time step of the flood data by accumulation;
step 1-3, intercepting corresponding historical rainfall spatio-temporal information from the historical remote sensing inversion rainfall information obtained in the step 1-2 according to the rising and ending time of each flood process, and recording the historical rainfall spatio-temporal information as the rainfall information in the flood period; before flood rising A1The early-stage rainfall information of each period is spliced with the rainfall information of the flood period to form complete rainfall information corresponding to the flood process; therefore, the rainfall and flood information of the field is stored by taking the flood occurrence date as a serial number, and a historical rainfall flood database is constructed.
In a further embodiment of the first aspect, the step 2 is further:
step 2-1, aiming at the occurring flood process, obtaining A consistent with the time point and the time step of the historical remote sensing inversion precipitation information by adopting the method of the step 1-21Early-stage remote sensing inversion rainfall information I of each period1、A2Remote sensing inversion rainfall information I observed in each time period2；
2-2, aiming at rainfall data forecasted in a nearest numerical mode before the moment, adjusting the spatial resolution of the forecast rainfall information to be consistent with the early-stage rainfall information and the observed rainfall information by methods such as bilinear interpolation, correcting errors by methods such as joint probability distribution, and processing by the method in the step 1-3 to obtain numerical forecast rainfall information with the same time step as that in the step 2-1; extracting A after face time3Individual time interval numerical forecast rainfall information I3(ii) a Splicing I1、I2And I3Obtaining the information P of observing and forecasting the rainfall at the momenttarget_stThe total number of time segments is A1+ A2+ A3；
Step 2-3, observing and forecasting rainfall information P of the facing moment with three-dimensional space-time attribute in the step 2-2target_stConversion into a time sequence P of mean precipitation of the surfacetarget_t(ii) a Calculating Ptarget_tAverage precipitation sequence P corresponding to each rain flood process in historical rain flood databasedatabase_tThe dynamic time of the method is used for regulating and sequencing DTW indexes, the smaller the DTW indexes are, the highest similarity is obtained, and the highest similarity is obtained in the screening time sequence process1Field dewatering;
Figure 797586DEST_PATH_IMAGE002
Figure 520691DEST_PATH_IMAGE004
in the formula: i and j are respectively Ptarget_tAnd Pdatabase_tIndex of a certain period in the information, d (P)target_t,i, Pdatabase_t,j) Is Ptarget_t,iAnd Pdatabase_t,jThe DTW index is substantially the minimum cumulative Euclidean distance from (1, 1) to (i, j),
Figure DEST_PATH_IMAGE005
is the minimum cumulative euclidean distance traversed before point (i, j);
step 2-4, observing and forecasting rainfall information P at facing moment with three-dimensional space-time attributetarget_stTaking each grid in the plane space of the research area as an analysis unit, aiming at the grid (u, v) and at Ptarget_stCorresponding time range ST1Within, with tiIn time interval, the grid precipitation is centered at N × N × N time windows3The precipitation of each grid is used as a calculation monomer
Figure 600643DEST_PATH_IMAGE006
(N may be 3, 5, 7 …); for X1Remote sensing inversion of precipitation P in field precipitation processdatabase_stAt Pdatabase_stCorresponding time range ST2In the interior, the grid (u, v) and the t-th grid are extractedjComputing unit for same space-time range with time interval precipitation as center
Figure DEST_PATH_IMAGE007
(ii) a Calculating the structural similarity index of two monomers
Figure 751001DEST_PATH_IMAGE008
The index takes into account the integrated similarity of the mean, variance and covariance,
Figure 112975DEST_PATH_IMAGE008
the larger the value is, the higher the similarity of the space-time structures of the two is; transforming the index sequence number of the precipitation period and calculating the position of the grid (u, v)Comprehensive similarity index of space-time variation of precipitation in adjacent range
Figure DEST_PATH_IMAGE009
；
Figure DEST_PATH_IMAGE011
Wherein:
Figure DEST_PATH_IMAGE013
Figure DEST_PATH_IMAGE015
Figure DEST_PATH_IMAGE017
in the formula: SSIM monomers
Figure 596695DEST_PATH_IMAGE006
And
Figure 593470DEST_PATH_IMAGE007
regarding the images as two spatio-temporal images, the spatio-temporal images include L, C, S three parts, which respectively represent brightness, contrast and structure between the images, and α, β and γ are indexes of the three parts, and usually 1 is taken;
Figure 586834DEST_PATH_IMAGE018
、
Figure DEST_PATH_IMAGE019
are respectively as
Figure 137901DEST_PATH_IMAGE006
Mean and standard deviation of precipitation in the monomer;
Figure 969853DEST_PATH_IMAGE020
is a single body
Figure 962080DEST_PATH_IMAGE006
And
Figure 985400DEST_PATH_IMAGE007
covariance of internal precipitation; c. C1、c2、c3Is constant, avoids instability caused when the denominator is close to 0,
Figure DEST_PATH_IMAGE021
wherein R is the difference between the maximum value and the minimum value of the precipitation in the monomer.
Figure DEST_PATH_IMAGE023
In the formula:
Figure 492604DEST_PATH_IMAGE024
is composed of
Figure 626782DEST_PATH_IMAGE006
And
Figure 739095DEST_PATH_IMAGE007
index of structural similarity
Figure 808682DEST_PATH_IMAGE008
A function of, i.e.
Figure DEST_PATH_IMAGE025
；
Figure 98456DEST_PATH_IMAGE026
Is composed of
Figure DEST_PATH_IMAGE027
Minimum accumulation of past dots
Figure 239587DEST_PATH_IMAGE024
；
Step 2-5, calculating the grid position in sequence
Figure 331040DEST_PATH_IMAGE009
Drawing a DTW (SSIM) grid distribution diagram in the range of the research area; aiming at X searched in step 2-3 in sequence1In the field rain flood process, sequentially calculating to obtain a corresponding DTW (SSIM) grid distribution map; calculation face moment observation and forecast rainfall information Ptarget_stThe accumulated precipitation of each grid is sorted from big to small, the grid with the B% of the highest rank is selected, and the weight w of each selected grid is given according to the accumulated precipitationgridWeighted calculation DTW (SSIM) distribution diagram, and marked as DTW (SSIM) -SCORE; the smaller the DTW (SSIM) -SCORE is, the highest degree of temporal and spatial similarity of precipitation is, for X1Sorting DTW (SSIM) -SCORE in the field precipitation process from small to large, and selecting the top X2Taking the field precipitation as historical precipitation information with higher degree of similarity in the space-time process, and taking the corresponding flood process as a historical similarity flood set 1 based on precipitation space-time similarity retrieval;
Figure DEST_PATH_IMAGE029
in the formula: q is the number of grids ranked B% before the cumulative precipitation,
Figure 899424DEST_PATH_IMAGE030
in order to be a grid position,
Figure DEST_PATH_IMAGE031
is a grid
Figure 679424DEST_PATH_IMAGE030
By weight of (2), in a grid
Figure 30771DEST_PATH_IMAGE030
The proportion of the precipitation quantity of the grid to the total precipitation quantity of the first q grids is calculated,
Figure 242309DEST_PATH_IMAGE032
。
in a further embodiment of the first aspect, the step 3 is further:
step 3-1, acupunctureFor historical rainfall flood information of the hydrological station, A is the time before the flood rises1Time period and flood rise to face time A2A between the average precipitation and flood rise of the research area of each period2Actual measurement flow (including facing time) at +1 time is used as BP neural network input data to forecast flood flow at v times in the future; taking the minimum of flood average (1-NSE) of the field times as an objective function, utilizing 70% of field times of rainfall flood information in a historical rainfall flood database, and calibrating the weight and the threshold of the BP neural network through a genetic algorithm, wherein the rest 30% of field times of rainfall flood information is used as test data of the precision of the BP neural network model; by gradually increasing the quantity v of the output flow of the BP neural network model, the average NSE of the test data is not lower than 0.70 (the second-level standard of the hydrological information forecasting specification (GB/T22482-2008)), so that the maximum v is determined, namely the effective forecasting period of flood forecasting;
Figure 653699DEST_PATH_IMAGE034
in the formula:
Figure DEST_PATH_IMAGE035
and
Figure 888371DEST_PATH_IMAGE036
are respectively the v thiThe current flow is measured according to the measured flow;
Figure DEST_PATH_IMAGE037
the measured flow mean value is obtained; v is the total time of forecasting future flow.
Step 3-2, aiming at the flood process occurring in the hydrological station to face the moment A1+A2Area average observed precipitation sum A for each time period2Inputting the observed flow at +1 moments into the calibrated BP neural network model, and forecasting the flow Q of the hydrological station at v moments in the futureBPFlood forecasting is realized;
step 3-3, will face A before time2The observed flow at +1 time and the forecast flow at v times are spliced to generate a packetFlood time series f containing observed information and forecast informationtargetSequentially calculating each flood time sequence f in the historical rainfall flood databasedatabaseAnd ftargetThe DTW indexes of (1) are sorted from small to large, and the first X is taken3And (3) taking the field flood as a similarity history flood set 2 based on the similarity retrieval of the flood time sequence process.
Figure DEST_PATH_IMAGE039
In the formula: f. ofi,fjAre respectively ftargetAnd fdatabaseAn index of a time in the time series;
Figure 27815DEST_PATH_IMAGE040
is composed of
Figure DEST_PATH_IMAGE041
And
Figure 562701DEST_PATH_IMAGE042
the Euclidean distance between them;
Figure DEST_PATH_IMAGE043
is (f)i，fj) The minimum cumulative euclidean distance traveled before the point.
In a further embodiment of the first aspect, the step 4 is further:
step 4-1, merging the similarity historical flood set 1 based on rainfall space-time similarity retrieval and the similarity historical flood set 2 based on flood time sequence process similarity retrieval to obtain X2+X3The historical flood process that the field is similar to the flood is recorded as a similarity historical flood set 3;
step 4-2, recording ftargetLength of (L), sliding computation of different starting points and times (I) of a flood process in the similarity historical flood set 30To the following item I0Flood sequence f between + L-1 momentsdatabase_partAnd ftargetTaking the sum of squared deviations of
Figure 738468DEST_PATH_IMAGE044
Minimum corresponding item I0,bestThe + L-1 moments are used as splicing positions of the similar flood and the flood in the field; respectively determining a splicing position aiming at each flood in the similarity historical flood set 3, and after the splicing positions are weighted and calculated by adopting an equal weight method, X2+X3Average similar flood process f of field similar flood processessimilarity，fsimilarityLength and X of2+X3The lengths of the shortest flood time sequences after the splicing positions in the field similar flood process are consistent and are recorded as Lsimilarity；
Figure 132540DEST_PATH_IMAGE046
In the formula:
Figure DEST_PATH_IMAGE047
representing the flow of the Kth field similar flood sequence in the similarity historical flood set 3 at the ith moment after the splicing position;
step 4-3, beyond the effective forecasting period of flood forecasting, the flood forecasting cannot provide reliable forecasting flow, and the average similar flood process fsimilarityMove up and down
Figure 779684DEST_PATH_IMAGE048
And splicing the prediction flow with the forecast flow at v moments after the confrontation moment to obtain a complete flood prediction result after the confrontation moment, namely:
Figure 575602DEST_PATH_IMAGE050
in the formula: t is t0Dynamically estimating the confrontation moment for the flood; flood process estimation result based on similar excavation of rainfall flood space-time process
Figure DEST_PATH_IMAGE051
；
Figure 656690DEST_PATH_IMAGE052
The estimation results are for the complete flood process after the moment is encountered.
In a further embodiment of the first aspect, the process of step 5 further comprises:
step 5-1, judging flood process estimation results based on precipitation and flood similarity
Figure 928272DEST_PATH_IMAGE051
Whether flood peak information is contained or not is judged, if yes, the dynamic estimation stopping condition is not met, otherwise, if no flood peak information is contained, namely, all the flood peak information is in the process of water withdrawal, the dynamic estimation is stopped;
step 5-2, if the stopping condition is not met, moving to the next facing moment along with time, updating rainfall observation and forecast information and flood observation and forecast information in the effective forecast period, repeating the step 2-4, and realizing the flood process forecast result based on rainfall and flood similarity
Figure 159533DEST_PATH_IMAGE051
Is dynamically updated.
In a further embodiment of the first aspect, the steps 1 to 5 are repeated to obtain a dynamic estimation result of the flood process of the hydrological station in the research area.
In a second aspect, a flood dynamic estimation system based on similarity of rain flood space-time processes is provided, the system comprising:
the system comprises a first module, a second module and a third module, wherein the first module is used for acquiring historical remote sensing inversion rainfall information, real-time remote sensing inversion rainfall information, numerical forecast rainfall information and historical site flood information of a preset number of hydrologic stations in a research area, matching the time scales of historical rainfall and flood information and constructing a hydrologic station historical rainfall flood database;
the second module is used for matching and splicing the space-time scales of the remote sensing inversion precipitation and numerical forecast precipitation information facing the earlier stage of time and observed aiming at the flood which is happening in the hydrological station, and searching historical precipitation processes with similar surface average precipitation time sequences in a historical rainfall flood database established by the first module by adopting a time sequence similarity analysis method; screening historical rainfall processes similar to the spatio-temporal process by adopting a spatio-temporal information similarity analysis method, and extracting a corresponding similarity historical flood set 1;
the third module is used for splicing flood information in an effective forecast period of constantly observed flood and flood forecast model forecast, and searching a historical flood process set 2 with similar time sequence processes in a historical rainfall flood database by adopting a time sequence similarity analysis method;
the fourth module is used for merging the historical flood process sets 1 and 2 to obtain a similar historical flood process set 3, reasonably determining the splicing time point of similar floods in each field, calculating the average similar flood process after the splicing time point by adopting a weighted average method, and estimating the subsequent flood process through translation and splicing operation;
the fifth module judges whether the follow-up flood process estimated by the fourth module contains a flood peak or not, and stops dynamic estimation if the follow-up flood process does not contain the flood peak; otherwise, updating the rainfall flood observation and forecast information along with the time, repeatedly executing the second module to the fourth module, and dynamically estimating the development situation of the flood.
Has the advantages that: the method mainly comprises the steps of historical rainfall flood database construction, historical flood process retrieval based on rainfall spatio-temporal process similarity, historical flood process retrieval based on flood time sequence process similarity and dynamic prediction based on historical similar flood. In the historical rainfall flood database construction stage, long-series historical remote sensing is used for inverting precipitation information, and three-dimensional space-time precipitation processes corresponding to the historical flood processes are extracted, so that a historical rainfall flood event library reflecting the space-time evolution process is constructed. In a historical flood process retrieval stage based on rainfall space-time process similarity, rainfall information and numerical forecast rainfall information are inverted by coupling the early stage and observed remote sensing, on the basis of only considering the similarity of a surface average rainfall time sequence in the prior art, a coupling structure similarity index SSIM and a DTW (SSIM) index of dynamic time warping DTW are provided, the space-time similarity of the historical rainfall process and the observation and forecast rainfall process is measured, and therefore the influence of rainfall falling area distribution on the flood process is considered. In the historical flood process retrieval stage based on the similarity of the flood time sequence process, the observed flow information and the effective forecast period flow information forecasted by the flood model are utilized in a coupling mode, so that reliable information is added in the similarity analysis of the flood time sequence process, and the retrieval result of the historical flood process with the similarity is improved. In a dynamic estimation stage based on historical similar flood, the historical similar flood which considers the occurrence and development processes of the rainfall flood more comprehensively can be obtained by integrating retrieval results of rainfall spatio-temporal similarity and flood time sequence similarity, and the flood estimation precision is gradually improved along with the time lapse and the rainfall flood information accumulation and update. Generally, the method comprehensively utilizes historical, real-time observed and forecasted rainfall and flood information in the aspect of information, achieves the expansion from a one-dimensional time sequence to a three-dimensional space-time process in the aspect of similarity analysis dimensionality, provides a quantitative measurement method of space-time process similarity in the aspect of a mathematical method, and can powerfully promote the development of a flood dynamic prediction technology based on the rain flood process similarity mining.
Drawings
Fig. 1 is a flowchart of a first embodiment of the present invention.
Fig. 2 shows a reservoir control catchment area of a third watershed a according to an embodiment of the invention.
Detailed Description
In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art, that the present invention may be practiced without one or more of these specific details. In other instances, well-known features have not been described in order to avoid obscuring the invention.
The first embodiment is as follows:
a flood dynamic estimation method based on rainfall flood space-time process similarity excavation comprises the following steps:
acquiring historical remote sensing retrieval rainfall information, real-time remote sensing retrieval rainfall information, numerical forecast rainfall information and historical site flood information of a preset number of hydrological stations in a research area, matching time scales of historical rainfall and flood information, and constructing a hydrological station historical rainfall flood database; aiming at the flood which is happening in a hydrological station, matching and splicing the space-time scales of the early stage of the moment and the observed remote sensing inversion precipitation and numerical forecast precipitation information, and searching a historical precipitation process with similar surface average precipitation time sequence in a historical rainfall flood database by adopting a dynamic time warping DTW index; further screening historical precipitation processes similar to the time-space process by adopting a coupling Structure Similarity Index (SSIM) and a Dynamic Time Warping (DTW) index, and extracting a corresponding historical flood process set 1; the method comprises the steps that flood information in an effective forecast period forecasted by a flood forecast model and observed floods at the moment are spliced, and a historical flood process set 2 with similar time sequence processes is searched in a historical rainfall flood database by adopting a DTW index; merging the historical flood process sets 1 and 2 to obtain a similar historical flood process set 3, reasonably determining the splicing time point of each field of similar flood, calculating the average similar flood process after the splicing time point by adopting an equal weight weighted average method, and pre-estimating the subsequent flood process through translation and splicing operation; and judging whether the current estimated follow-up flood process at the moment contains a flood peak or not, if not, stopping dynamic estimation, otherwise, updating rainfall flood observation and forecast information along with the time, and dynamically estimating the flood development situation.
Example two:
on the basis of the first embodiment, the second embodiment further provides a refinement step of a flood dynamic estimation method based on the rainfall flood space-time process similarity mining, which comprises the following steps:
step 1, obtaining historical remote sensing inversion rainfall information, real-time remote sensing inversion rainfall information, numerical forecast rainfall information and historical stage flood information of a preset number of hydrologic stations in a research area, matching time scales of historical rainfall and flood information, and constructing a hydrologic station historical rainfall flood database;
step 1-1, acquiring historical and real-time remote sensing inversion observation rainfall information, numerical forecast rainfall information and historical field flood information of a research area;
step 1-2, comprehensively performing remote sensing inversion on precipitation information and historical field flood information, reasonably determining time points and step lengths of field flood information at equal intervals, and acquiring flood data of sampling time points by adopting a cubic spline interpolation method;
step 1-3, judging whether the time step length of the historical remote sensing inversion precipitation information is consistent with the flood data or not; if the difference is not consistent, distributing the precipitation amount to a time period smaller than the time step of the flood information by an arithmetic mean method, and obtaining historical remote sensing inversion precipitation information matched with the time point and the time step of the flood data by accumulation;
step 1-4, intercepting corresponding historical rainfall spatio-temporal information from the historical remote sensing inversion rainfall information obtained in the step 1-3 according to the rising and ending time of each flood process, and recording the historical rainfall spatio-temporal information as the rainfall information in the flood period; before flood rising A1The early-stage rainfall information of each period is spliced with the rainfall information of the flood period to form complete rainfall information corresponding to the flood process; therefore, the rainfall and flood information of the field is stored by taking the flood occurrence date as a serial number, and a historical rainfall flood database is constructed.
Step 2, aiming at the flood which is happening in the hydrological station, matching and splicing the space-time scales of early-stage remote sensing inversion precipitation before the flood rising time, observed remote sensing inversion precipitation from the flood rising time to the confrontation time and numerical forecast precipitation information after the confrontation time, calculating a surface average precipitation time sequence, and searching a historical precipitation process with similar surface average precipitation time sequence in the historical rainfall database established in the step 1 by adopting a time sequence similarity analysis method; screening historical rainfall processes similar to the spatio-temporal process by adopting a spatio-temporal information similarity analysis method, and extracting a corresponding similarity historical flood set 1;
step 2-1, aiming at the occurring flood process, obtaining A consistent with the time point and the time step of the historical remote sensing inversion precipitation information by adopting the method of the step 1-21Individual period early stage remote sensing inversion precipitation information I1、A2Remote sensing inversion rainfall information I observed in each time period2；
Step 2-2, aiming at rainfall data forecasted in a nearest numerical mode before the moment, adjusting the spatial resolution of the forecast rainfall information to be consistent with the early-stage rainfall information and the observed rainfall information by methods such as bilinear interpolation, correcting errors by methods such as joint probability distribution, processing by the method in step 1-3 to obtain numerical forecast rainfall with the same time step as that in step 2-1Water information; extracting A after face time3Individual time interval numerical forecast rainfall information I3(ii) a Splicing I1、I2And I3Obtaining the information P of observing and forecasting the rainfall at the momenttarget_stThe total number of time segments is A1+ A2+ A3；
Step 2-3, observing and forecasting rainfall information P of the facing moment with three-dimensional space-time attribute in the step 2-2target_stConversion into a time sequence P of mean precipitation of the surfacetarget_t(ii) a Calculating Ptarget_tAverage precipitation sequence P corresponding to each rain flood process in historical rain flood databasedatabase_tThe dynamic time of the method is used for regulating and sequencing DTW indexes, the smaller the DTW indexes are, the highest similarity is obtained, and the highest similarity is obtained in the screening time sequence process1Field dewatering;
Figure 934591DEST_PATH_IMAGE002
Figure 858684DEST_PATH_IMAGE004
in the formula: i and j are respectively Ptarget_tAnd Pdatabase_tIndex of a certain period in the information, d (P)target_t,i, Pdatabase_t,j) Is Ptarget_t,iAnd Pdatabase_t,jThe DTW index is substantially the minimum cumulative Euclidean distance from (1, 1) to (i, j),
Figure 492928DEST_PATH_IMAGE005
is the minimum cumulative euclidean distance traversed before point (i, j);
step 2-4, observing and forecasting rainfall information P at facing moment with three-dimensional space-time attributetarget_stTaking each grid in the plane space of the research area as an analysis unit, aiming at the grid (u, v) and at Ptarget_stCorresponding time range ST1Within, with tiIn time interval, the grid precipitation is centered at N × N × N time windows3The precipitation of each grid is used as a calculation monomer
Figure 151049DEST_PATH_IMAGE006
(N may be 3, 5, 7 …); for X1Remote sensing inversion of precipitation P in field precipitation processdatabase_stAt Pdatabase_stCorresponding time range ST2In the interior, the grid (u, v) and the t-th grid are extractedjComputing unit for same space-time range with time interval precipitation as center
Figure 655980DEST_PATH_IMAGE007
(ii) a Calculating the structural similarity index of two monomers
Figure 875609DEST_PATH_IMAGE008
The index considers the comprehensive similarity of the mean value, the variance and the covariance, and the larger the value is, the higher the similarity of the space-time structures of the mean value, the variance and the covariance is; transforming the index sequence number of the precipitation time interval, and calculating the comprehensive similarity index of the precipitation space-time change in the adjacent range at the grid (u, v) position
Figure 997148DEST_PATH_IMAGE009
；
Figure 694846DEST_PATH_IMAGE011
Wherein:
Figure DEST_PATH_IMAGE053
Figure 913338DEST_PATH_IMAGE054
Figure 913655DEST_PATH_IMAGE017
in the formula: SSIM monomers
Figure 617431DEST_PATH_IMAGE006
And
Figure 525344DEST_PATH_IMAGE007
regarding the images as two spatio-temporal images, the spatio-temporal images include L, C, S three parts, which respectively represent brightness, contrast and structure between the images, and α, β and γ are indexes of the three parts, and usually 1 is taken;
Figure 129501DEST_PATH_IMAGE018
、
Figure 566298DEST_PATH_IMAGE019
are respectively as
Figure 255905DEST_PATH_IMAGE006
Mean and standard deviation of precipitation in the monomer;
Figure 436351DEST_PATH_IMAGE020
is a single body
Figure 35960DEST_PATH_IMAGE006
And
Figure 768292DEST_PATH_IMAGE007
covariance of internal precipitation; c. C1、c2、c3Is constant, avoids instability caused when the denominator is close to 0,
Figure 86141DEST_PATH_IMAGE021
wherein R is the difference between the maximum value and the minimum value of the precipitation in the monomer.
Figure DEST_PATH_IMAGE055
In the formula:
Figure 636989DEST_PATH_IMAGE024
is composed of
Figure 215738DEST_PATH_IMAGE006
And
Figure 994338DEST_PATH_IMAGE007
index of structural similarity
Figure 924117DEST_PATH_IMAGE008
A function of, i.e.
Figure 446365DEST_PATH_IMAGE025
；
Figure 879620DEST_PATH_IMAGE026
Is composed of
Figure 94701DEST_PATH_IMAGE027
Minimum accumulation of past dots
Figure 855984DEST_PATH_IMAGE024
；
Step 2-5, calculating the grid position in sequence
Figure 73601DEST_PATH_IMAGE009
Drawing a DTW (SSIM) grid distribution diagram in the range of the research area; aiming at X searched in step 2-3 in sequence1In the field rain flood process, sequentially calculating to obtain a corresponding DTW (SSIM) grid distribution map; calculation face moment observation and forecast rainfall information Ptarget_stThe accumulated precipitation of each grid is sorted from big to small, the grid with the B% of the highest rank is selected, and the weight w of each selected grid is given according to the accumulated precipitationgridWeighted calculation DTW (SSIM) distribution diagram, and marked as DTW (SSIM) -SCORE; the smaller the DTW (SSIM) -SCORE is, the highest degree of temporal and spatial similarity of precipitation is, for X1Sorting DTW (SSIM) -SCORE in the field precipitation process from small to large, and selecting the top X2Taking the field precipitation as historical precipitation information with higher degree of similarity in the space-time process, and taking the corresponding flood process as a historical similarity flood set 1 based on precipitation space-time similarity retrieval;
Figure 236729DEST_PATH_IMAGE029
in the formula: q is the number of grids ranked B% before the cumulative precipitation,
Figure 481765DEST_PATH_IMAGE030
in order to be a grid position,
Figure 995923DEST_PATH_IMAGE031
is a grid
Figure 125553DEST_PATH_IMAGE030
By weight of (2), in a grid
Figure 267822DEST_PATH_IMAGE030
The proportion of the precipitation quantity of the grid to the total precipitation quantity of the first q grids is calculated,
Figure 90284DEST_PATH_IMAGE032
。
step 3, searching a similarity historical flood set 2 with similar time sequence process in the historical rainfall flood database established in the step 1 by adopting a time sequence similarity analysis method according to flood information in an effective forecast period forecasted by the constantly observed flood and flood forecast model at the splicing face;
step 3-1, aiming at historical rainfall flood information of the hydrological station, raising flood to a height A1Time period and flood rise to face time A2A between the average precipitation and flood rise of the research area of each period2Actual measurement flow (including facing time) at +1 time is used as BP neural network input data to forecast flood flow at v times in the future; taking the minimum of flood average (1-NSE) of the field times as an objective function, utilizing 70% of field times of rainfall flood information in a historical rainfall flood database, and calibrating the weight and the threshold of the BP neural network through a genetic algorithm, wherein the rest 30% of field times of rainfall flood information is used as test data of the precision of the BP neural network model; by gradually increasing the quantity v of the output flow of the BP neural network model, the average NSE of the test data is not lower than 0.70 (the second-level standard of the hydrological information forecasting specification (GB/T22482-2008)), so that the maximum v is determined, namely the effective forecasting period of flood forecasting;
Figure 950793DEST_PATH_IMAGE056
in the formula: and
Figure 87376DEST_PATH_IMAGE036
are respectively the v thiThe current flow is measured according to the measured flow;
Figure 848265DEST_PATH_IMAGE037
the measured flow mean value is obtained; v is the total time of forecasting future flow.
Step 3-2, aiming at the flood process occurring in the hydrological station to face the moment A1+A2Area average observed precipitation sum A for each time period2Inputting the observed flow at +1 moments into the calibrated BP neural network model, and forecasting the flow Q of the hydrological station at v moments in the futureBPFlood forecasting is realized;
step 3-3, will face A before time2Splicing the observed flow at +1 moments and the forecast flow at v moments to generate a flood time sequence f containing observed information and forecast informationtargetSequentially calculating each flood time sequence f in the historical rainfall flood databasedatabaseAnd ftargetThe DTW indexes of (1) are sorted from small to large, and the first X is taken3And (3) taking the field flood as a similarity history flood set 2 based on the similarity retrieval of the flood time sequence process.
Figure DEST_PATH_IMAGE057
In the formula: f. ofi,fjAre respectively ftargetAnd fdatabaseAn index of a time in the time series;
Figure 169525DEST_PATH_IMAGE040
is composed of
Figure 127117DEST_PATH_IMAGE041
And
Figure 723183DEST_PATH_IMAGE042
the Euclidean distance between them;
Figure 980989DEST_PATH_IMAGE043
is (f)i，fj) The minimum cumulative euclidean distance traveled before the point.
Step 4, merging the similarity history flood sets 1 and 2 retrieved in the steps 2 and 3 to obtain a similarity history flood set 3, reasonably determining the splicing time point of each field of similar flood, calculating the average similar flood process after the splicing time point by adopting a weighted average method, and estimating the subsequent flood process through translation and splicing operation;
step 4-1, merging the similarity historical flood set 1 based on rainfall space-time similarity retrieval and the similarity historical flood set 2 based on flood time sequence process similarity retrieval to obtain X2+X3The historical flood process that the field is similar to the flood is recorded as a similarity historical flood set 3;
step 4-2, recording ftargetLength of (L), sliding computation of different starting points and times (I) of a flood process in the similarity historical flood set 30To the following item I0Flood sequence f between + L-1 momentsdatabase_partAnd ftargetSum of squared deviations of
Figure 738730DEST_PATH_IMAGE044
Get it
Figure 449197DEST_PATH_IMAGE044
Minimum corresponding item I0,bestThe + L-1 moments are used as splicing positions of the similar flood and the flood in the field; respectively determining a splicing position aiming at each flood in the similarity historical flood set 3, and after the splicing positions are weighted and calculated by adopting an equal weight method, X2+X3Average similar flood process f of field similar flood processessimilarity，fsimilarityLength and X of2+X3The lengths of the shortest flood time sequences after the splicing positions in the field similar flood process are consistent and are recorded as Lsimilarity；
Figure 84840DEST_PATH_IMAGE046
In the formula:
Figure 197152DEST_PATH_IMAGE047
representing the flow of the Kth field similar flood sequence in the similarity historical flood set 3 at the ith moment after the splicing position;
step 4-3, beyond the effective forecasting period of flood forecasting, the flood forecasting cannot provide reliable forecasting flow, and the average similar flood process fsimilarityMove up and down
Figure 266739DEST_PATH_IMAGE048
And splicing the prediction flow with the forecast flow at v moments after the confrontation moment to obtain a complete flood prediction result after the confrontation moment, namely:
Figure 323557DEST_PATH_IMAGE058
in the formula: t is t0Dynamically estimating the confrontation moment for the flood; estimating a flood process estimation result based on similar excavation of the rainfall flood space-time process;
Figure 995847DEST_PATH_IMAGE052
the estimation results are for the complete flood process after the moment is encountered.
Step 5, judging whether the follow-up flood process in the step 4 contains a flood peak or not, and stopping dynamic estimation if the follow-up flood process does not contain the flood peak; otherwise, updating the rainfall flood observation and forecast information along with the time, repeating the steps 2-4, and dynamically estimating the flood development situation.
Step 5-1, judging flood process estimation results based on precipitation and flood similarity
Figure 228245DEST_PATH_IMAGE051
If the flood peak information is contained, the dynamic estimation stopping condition is not met, otherwise, if the flood peak information is not contained, all the steps are the water returning process,stopping dynamic estimation;
step 5-2, if the stopping condition is not met, moving to the next facing moment along with time, updating rainfall observation and forecast information and flood observation and forecast information in the effective forecast period, repeating the step 2-4, and realizing the flood process forecast result based on rainfall and flood similarity
Figure 203154DEST_PATH_IMAGE051
Is dynamically updated.
Example three:
in this embodiment, a reservoir in a certain watershed a is taken as a research object, and fig. 2 is a water collection area for controlling the reservoir in the certain watershed a according to the embodiment of the invention. A flood dynamic estimation method based on rainfall flood space-time process similarity excavation is provided. The main stream flows through 2 provinces in Shaanxi and Hubei of Changjiang river, and converges to Yangtze river in Wuhan, with the total length of 1577 km. The range of the drainage basin is between 106 DEG 15 'E and 114 DEG 20' E and 30 DEG 10 'N and 34 DEG 20' N, and the total area is about 15.9 km2. The drainage basin is complex in terrain, is high in the west and low in the east, belongs to a monsoon climate area in the east subtropical zone, has average rainfall of about 900mm for many years, is uneven in rainfall spatial-temporal distribution, has a distribution rule that annual rainfall shows that a south bank is larger than a north bank, the upstream and the downstream are large, and the midstream is small, and 70-80% of rainfall in the whole year is concentrated in a flood season (5-10 months).
In the embodiment, Multi-Source Weighted-Ensemble Precipitation version 2.1 (MSWEP V2.1) long-series remote sensing inversion Precipitation data (time Range: 1979-2016, time resolution: 3h, UTC time; spatial resolution: 0.1 degree) and control forecast Precipitation information of ECMWF (European Centre for Medium-Range Weather forms) mode (time Range: 2008-2016, time resolution: 6h, forecast period: 10 days, UTC time, UTC00:00 update every day, spatial resolution: 0.5 degree), and 58-field flood data of A reservoir between 1979 and 2010 (time resolution 6h, North time 08:00, 14:00, 20:00, 02: 00) are collected. Forecast precipitation data only include two floods, namely 20100716 floods and 20100722 floods, between 2008 and 2010, in the embodiment, a historical rainfall flood database is constructed by the rest 56 floods and corresponding MSWEP V2.1 precipitation information, two warehouse-in floods in 2010 are selected as estimation samples, and a dynamic flood estimation method based on similarity mining of a rainfall flood space-time process is applied, and the method mainly comprises the following steps:
s1): constructing a historical rainfall flood database: and (3) integrating the time and the spatial resolution of the MSWEP V2.1 inversion precipitation, ECMWF forecast precipitation data and flood data, and determining that the time of the rainfall flood is uniformly adjusted to be Beijing time, the time points are 02:00, 08:00, 14:00 and 20:00 respectively, the time step length is 6h, and the spatial resolution of the rainfall is 0.5 degrees multiplied by 0.5 degrees. In the aspect of time, the MSWEP V2.1 data are accumulated to obtain precipitation data with the step length of 6h, the time mark is increased by 8h (the difference between the Beijing time and the UTC time is 8 h) to obtain the MSWEP V2.1 precipitation data of four time periods of 02: 00-08: 00, 08: 00-14: 00, 14: 00-20: 00 and 20: 00-02: 00 every day at the Beijing time, and the time point of the flood data is consistent with the target time; spatially, a bilinear interpolation method is used to scale up a 0.1 ° × 0.1 ° MSWEP V2.1 to 0.5 ° × 0.5 °. The method comprises the steps of extracting rainfall corresponding to historical flood from long-series MSWEP V2.1 data, intercepting MSWEP V2.1 within 3 days (12 time periods) before the flood rises and within the time range from the flood rise to the flood end in terms of time, performing mask processing on global MSWEP V2.1 data by adopting an A reservoir control water collection area boundary file in terms of space, and extracting MSWEP V2.1 rainfall space-time data of a research area range corresponding to each historical flood. And matching the field MSWEP V2.1 precipitation and flood data, and storing the rainfall flood data by taking the flood rising date as a serial number, thereby constructing a historical rainfall flood database of the reservoir A.
S2): searching historical flood processes based on precipitation space-time process similarity: aiming at the occurring flood process, because the ECMWF forecast precipitation information is updated at the time 08:00 of Beijing every day (UTC 00: 00), the facing moment of flood dynamic estimation is taken as the time 08:00 of Beijing every day. At the moment of facing, extracting MSWEP V2.1 data which are within the range of the research area, have the same time point and time step length as the historical database, are 3 days before the flood rises and rise to the moment of facing by adopting the same method as the step S1), and recording the data as early stage and observed precipitation MP; for the ECMWF control forecast rainfall data which is most adjacent to the moment, increasing the time mark by 8h, converting the data into Beijing time, correcting the ECMWF control forecast rainfall by adopting a joint probability distribution method, intercepting forecast rainfall data 8 time intervals (2 days) after the moment and recording the data as forecast rainfall PP; and splicing the MP and the PP according to time to form the MPP _ ST which has three-dimensional space-time attributes and faces the earlier stage of time, is observed and forecasts precipitation. And (3) converting the MPP into a surface average precipitation time sequence MPP _ T by adopting an equal weight weighting method, sequentially evaluating the time sequence similarity degree of the MPP _ T and each precipitation process in a historical rainfall flood database by adopting a DTW index, and selecting the precipitation process with the 6 surface average time sequence process with the minimum DTW index. On the basis, the DTW (SSIM) distribution diagram of the 6 similar precipitation processes is calculated in a sliding mode by utilizing a 3X 3 space-time window. Calculating the accumulated precipitation amount of each grid of MPP _ ST data, selecting q grids of the first 30% from large to small as main precipitation falling areas, weighting and calculating the DTW (SSIM) -SCORE of a DTW (SSIM) distribution diagram of 6-field similar precipitation processes by taking the proportion of the accumulated precipitation amount of each grid in the total precipitation amount of q grids as weight, taking the first 3 fields as space-time similar precipitation processes after sorting from small to large, and taking the corresponding flood process as a similarity historical flood set 1 based on precipitation space-time similar retrieval.
S3): and (3) searching historical flood processes based on similarity of flood time sequence processes: a BP neural network is adopted as a flood forecasting model, the average precipitation of a 3-day (12 periods) face in the early period of the facing moment and the corresponding facing moment flow (1 moment) are used as the input of the BP neural network, the effective forecasting period is tested by gradually increasing the number of output flow data, and finally the average NSE is determined to be not lower than 0.70 when the model outputs flood information at 2 moments in the future. And splicing the Flood rising to the Flood facing moment observation information and the Flood information predicted by the BP model at the future 2 moments, and recording as the Flood facing moment observation and Flood forecasting information Flood _ T. And sequentially evaluating the time sequence similarity degree of the Flood _ T and each Flood process in the historical rainfall Flood database by adopting the DTW indexes, and selecting the Flood process with the 3 fields with the minimum DTW indexes and similar time sequence process as a similarity historical Flood set 2 based on Flood time sequence similarity retrieval.
S4): estimating the flood development situation based on historical similar flood: and merging the similarity history flood sets 1 and 2 to form a similarity history flood set 3, determining the splicing position of each similarity history flood process by adopting a mode of sliding calculation of least sum of squared deviations, calculating an average similar flood process after the splicing position by adopting an equal weight weighting method, and splicing the average similar flood process with the flow predicted by the BP neural network model by vertically translating the average similar flood process to realize the prediction of the flood development situation.
S5): flood dynamic estimation and termination condition judgment: if the average similar flood process in the S4) contains flood peak information, the stopping condition is not met, and the steps S2) -S5) are repeated; otherwise, if only the flood water-dropping process is involved, the dynamic estimation is stopped.
S6): estimating the accuracy of the subsequent water inflow of the flood: the flood estimation precision is measured by RBIAS facing the relative error between the subsequent incoming water volume of the estimated flood and the actually measured flood at any moment, and the calculation formula is as follows:
Figure 12847DEST_PATH_IMAGE060
s7): in order to analyze the effectiveness of the method, the third embodiment compares the method with two traditional methods of flood dynamic estimation, namely the traditional method considering that the average observed precipitation of the surface is similar to the forecast precipitation. It should be noted that the difference between the two traditional methods and the method of the present invention is only in the precipitation similarity retrieval part, and the flood similarity retrieval part is consistent.
Table 1 shows the prediction accuracy of the BP neural network model for the flow at the 1 st and 2 nd time points in the future. According to the method, 70% of rainfall flood information of a field in a historical rainfall flood database is used as training data, a Nash efficiency coefficient NSE average value is used as a target function, a genetic algorithm is adopted to calibrate model parameters, and 30% of rainfall flood information of the field is used as a verification set. The average flood forecast NSE of the training set and the validation set at the time 2 is higher than 0.70. The BP neural network model has higher forecasting precision on No. 20100716 flood and No. 20100722 flood.
TABLE 1 prediction accuracy of BP neural network model for future 1 st and 2 nd moment flows
Forecast time period Training set
Figure DEST_PATH_IMAGE061
 Verification set 20100716NSE 20100722 NSE
1st 90% 85% 90% 89%
2nd 87% 82% 90% 87%
Table 2 shows the flood dynamic estimation method based on the similarity mining of the rainfall flood space-time process and the traditional method for estimating the subsequent water inflow amount of the flood and the relative error thereof. The estimation precision of the two traditional methods for the subsequent inflow amount of the two-field flood is superior to that of the method only in the 1 st dynamic estimation; although the post-estimation water volume of 20100716 flood is closer to the actual measurement flood in the 4 th dynamic estimation, the actual result is due to the overestimation of the water withdrawal process, and the control of the flood process, the flood peak and the flood volume is not good. Generally, in a dynamic estimation test, rainfall and flood observation information gradually accumulates along with the lapse of the faced moment, the relative error of the estimated subsequent water inflow tends to be reduced by the method, and the traditional method does not show the trend of increasing estimation precision and has unstable performance. Therefore, the method effectively exerts the advantages of observation and forecast information coupling utilization and similarity analysis dimension expansion, and except for the 1 st rolling forecast and the 4 th rolling forecast, the relative error of the estimated subsequent water inflow is relatively reduced by 9.7-38.5% compared with that of the traditional method.
TABLE 2 flood dynamic estimation method based on similarity excavation of rainfall flood time-space process and traditional method for estimating subsequent water inflow amount of flood and relative error thereof
Figure DEST_PATH_IMAGE063
By adopting the technical scheme disclosed by the invention, the following beneficial effects are obtained: the flood dynamic estimation method based on the rainfall and flood space-time process similarity excavation coupling utilizes observation and forecast rainfall and flood information, measures rainfall and flood space-time process similarity by adopting a coupling structure similarity index SSIM and a DTW (SSIM) index of dynamic time warping DTW, realizes the expansion of rainfall and flood process similarity analysis from a one-dimensional time sequence to a three-dimensional space-time process, more comprehensively considers the occurrence and evolution processes of flood, improves the retrieval effect of similar flood, and estimates the accuracy of the subsequent water volume of flood to be higher than that of the traditional method utilizing the surface average rainfall information.
In the aspect of information utilization, the embodiment mainly relates to remote sensing precipitation MSWEP V2.1 and numerical forecast precipitation ECMWF. In the aspect of using a flood forecasting model, the method mainly relates to a BP neural network model. It should be noted that the method has strong expansibility, and other remote sensing inversion, atmosphere reanalysis, numerical forecast precipitation information and flood forecast models can be used for the method.
As noted above, while the present invention has been shown and described with reference to certain preferred embodiments, it is not to be construed as limited thereto. Various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (8)

1. A flood dynamic estimation method based on similarity excavation of a rainfall flood space-time process is characterized by comprising the following steps:
step 1, obtaining historical remote sensing inversion rainfall information, real-time remote sensing inversion rainfall information, numerical forecast rainfall information and historical stage flood information of a preset number of hydrologic stations in a research area, matching time scales of historical rainfall and flood information, and constructing a hydrologic station historical rainfall flood database;
step 2, aiming at the flood which is happening in the hydrological station, matching and splicing the space-time scales of early-stage remote sensing inversion precipitation before the flood rising time, observed remote sensing inversion precipitation from the flood rising time to the confrontation time and numerical forecast precipitation information after the confrontation time, calculating a surface average precipitation time sequence, and searching a historical precipitation process with similar surface average precipitation time sequence in the historical rainfall database established in the step 1 by adopting a time sequence similarity analysis method; screening historical rainfall processes similar to the spatio-temporal process by adopting a spatio-temporal information similarity analysis method, and extracting a corresponding similarity historical flood set 1;
step 3, searching a similarity historical flood set 2 with similar time sequence process in the historical rainfall flood database established in the step 1 by adopting a time sequence similarity analysis method according to flood information in an effective forecast period forecasted by the constantly observed flood and flood forecast model at the splicing face;
step 4, merging the similarity historical flood set 1 and the similarity historical flood set 2 retrieved in the steps 2 and 3 to obtain a similarity historical flood set 3, determining the splicing time point of each field of similar flood, calculating the average similar flood process after the splicing time point by adopting a weighted average method, and pre-estimating the subsequent flood process through translation and splicing operation;
step 5, judging whether the follow-up flood process in the step 4 contains a flood peak or not, and stopping dynamic estimation if the follow-up flood process does not contain the flood peak; otherwise, updating the rainfall flood observation and forecast information along with the time, repeating the steps 2-4, and dynamically estimating the flood development situation.
2. The flood dynamic estimation method based on the rainfall flood space-time process similarity mining according to claim 1, wherein the step 1 further comprises:
step 1-1, comprehensively performing remote sensing inversion on precipitation information and historical field flood information, determining time points and step lengths of field flood information at equal intervals, and acquiring flood data of sampling time points by adopting a cubic spline interpolation method;
step 1-2, judging whether the time step of the historical remote sensing inversion precipitation information is consistent with the flood data or not; if the difference is not consistent, distributing the precipitation amount to a time period smaller than the time step of the flood information by an arithmetic mean method, and obtaining historical remote sensing inversion precipitation information matched with the time point and the time step of the flood data by accumulation;
step 1-3, intercepting corresponding historical rainfall spatio-temporal information from the historical remote sensing inversion rainfall information obtained in the step 1-2 according to the rising and ending time of each flood process, and recording the historical rainfall spatio-temporal information as the rainfall information in the flood period; before flood rising A1The early-stage rainfall information of each period is spliced with the rainfall information of the flood period to form complete rainfall information corresponding to the flood process; therefore, the rainfall and flood information of the field is stored by taking the flood occurrence date as a serial number, and a historical rainfall flood database is constructed.
3. The flood dynamic estimation method based on the rainfall flood space-time process similarity mining according to claim 1, wherein the step 2 further comprises:
step 2-1, aiming at the occurring flood process, obtaining A consistent with the rainfall time point and the time step of the historical remote sensing inversion by adopting the method of the step 1-21Individual period early stage remote sensing inversion precipitation information I1、A2Remote sensing inversion rainfall information I observed in each time period2；
Step 2-2, acquiring precipitation data of the nearest numerical model forecast before the confrontation moment, matching the spatio-temporal scale and correcting errors, and intercepting A after the confrontation moment3Number of time periodForecast precipitation information I3(ii) a Splicing I1、I2And I3Obtaining the information P of observing and forecasting the rainfall at the momenttarget_st；
Step 2-3, adding Ptarget_stConversion into a time sequence P of mean precipitation of the surfacetarget_t(ii) a Calculating Ptarget_tAverage precipitation sequence P corresponding to each rain flood process in historical rain flood databasedatabase_tThe DTW indexes are normalized and sorted according to the dynamic time, and the front X with the highest similarity degree in the time sequence process is screened1Field dewatering;
step 2-4 for Ptarget_stAnd X1Remote sensing inversion of precipitation P in field precipitation processdatabase_stFor spatial positions (u, v), respectively, by ti、tjTwo time periods of grid precipitation are taken as centers, and the surrounding N multiplied by N grid precipitation are used for constructing the mobile units
Figure DEST_PATH_IMAGE002
And
Figure DEST_PATH_IMAGE004
calculating the structural similarity index of the two
Figure DEST_PATH_IMAGE006
(ii) a Moving along the time axis, will
Figure 670300DEST_PATH_IMAGE006
Embedding DTW formula, calculating position
Figure DEST_PATH_IMAGE008
Space-time similarity index of
Figure DEST_PATH_IMAGE010
(ii) a Thus, for X in step 2-31In the field precipitation process, SSIM grid distribution maps in the range of the research area are drawn one by one;
wherein, the position
Figure 712074DEST_PATH_IMAGE008
Space-time similarity index of
Figure 960653DEST_PATH_IMAGE010
The calculation process is as follows:
Figure DEST_PATH_IMAGE012
in the formula:
Figure DEST_PATH_IMAGE014
is a single body
Figure 838085DEST_PATH_IMAGE002
And
Figure 121299DEST_PATH_IMAGE004
index of structural similarity
Figure 537237DEST_PATH_IMAGE006
A function of, i.e.
Figure DEST_PATH_IMAGE016
；
Figure DEST_PATH_IMAGE018
Is composed of
Figure DEST_PATH_IMAGE020
Minimum accumulation of past dots
Figure 227107DEST_PATH_IMAGE014
；
Step 2-5, calculating Ptarget_stCumulative precipitation at each grid location and weighting w according to precipitation sizegridWeighting and calculating the average score of the SSIM distribution diagram, and selecting the front X with the minimum average score2Precipitation on the ground, corresponding to the flood process as a search based on the spatio-temporal similarity of precipitationSimilarity historical flood set 1.
4. The flood dynamic estimation method based on the rainfall flood space-time process similarity mining according to claim 1, wherein the step 3 further comprises:
step 3-1, aiming at historical rainfall flood information of the hydrological station, raising flood to a height A1Time period and flood rise to face time A2A between the average precipitation and flood rise of the research area of each period2Actual measurement flow at +1 moments is used as BP neural network input data to forecast flood flow at v moments in the future; taking the field flood average 1-NSE minimum as an objective function, and calibrating the weight and the threshold of the BP neural network through a genetic algorithm; the average NSE of the test data is not lower than a preset standard by gradually increasing the quantity v of the output flow of the BP neural network model, so that the maximum v is determined, namely the effective forecast period of the flood forecast;
step 3-2, aiming at the flood process occurring in the hydrological station to face the moment A1+A2Area average observed precipitation sum A for each time period2Inputting the observed flow at +1 moments into the calibrated BP neural network model, and forecasting the flow Q of the hydrological station at v moments in the futureBPFlood forecasting is realized;
step 3-3, will face A before time2Splicing the observed flow at +1 moments and the forecast flow at v moments to generate a flood time sequence f containing observed information and forecast informationtargetSequentially calculating each flood time sequence f in the historical rainfall flood databasedatabaseAnd ftargetThe DTW indexes of (1) are sorted from small to large, and the first X is taken3And (3) taking the field flood as a similarity history flood set 2 based on the similarity retrieval of the flood time sequence process.
5. The flood dynamic estimation method based on the rainfall flood space-time process similarity mining according to claim 1, wherein the step 4 further comprises:
step 4-1, merging similarity calendars based on rainfall spatio-temporal similarity retrievalObtaining the similarity between the historical flood set 1 and the historical flood set 2 based on flood time sequence process similarity retrieval to obtain X2+X3The historical flood process that the field is similar to the flood is recorded as a similarity historical flood set 3;
step 4-2, recording ftargetLength of (L), sliding computation of different starting points and times (I) of a flood process in the similarity historical flood set 30To the following item I0Flood sequence f between + L-1 momentsdatabase_partAnd ftargetSum of squared deviations of
Figure DEST_PATH_IMAGE022
Get it
Figure 488324DEST_PATH_IMAGE022
Minimum corresponding item I0,bestThe + L-1 moments are used as splicing positions of the similar flood and the flood in the field; respectively determining splicing positions aiming at each flood in the similarity historical flood set 3, and after the splicing positions, adopting an equal weight method to perform weighted calculation X2+X3Average similar flood process sequence f of field similar flood processessimilarity，fsimilarityLength and X of2+X3The lengths of the shortest flood time sequences after the splicing positions in the field similar flood process are consistent and are recorded as Lsimilarity；
Figure DEST_PATH_IMAGE024
In the formula:
Figure DEST_PATH_IMAGE026
representing the flow of the Kth field similar flood sequence in the similarity historical flood set 3 at the ith moment after the splicing position;
step 4-3, beyond the effective forecasting period of flood forecasting, the flood forecasting cannot provide reliable forecasting flow, and the average similar flood process fsimilarityMove up and down
Figure DEST_PATH_IMAGE028
And splicing the prediction flow with the forecast flow at v moments after the confrontation moment to obtain a complete flood prediction result after the confrontation moment, namely:
Figure DEST_PATH_IMAGE030
in the formula: t is t0Dynamically estimating the confrontation moment for the flood; flood process estimation result based on similar excavation of rainfall flood space-time process
Figure DEST_PATH_IMAGE032
；
Figure DEST_PATH_IMAGE034
The estimation results are for the complete flood process after the moment is encountered.
6. The flood dynamic estimation method based on the rainfall flood space-time process similarity mining according to claim 1, wherein the step 5 further comprises:
step 5-1, judging flood process estimation results based on similar excavation of rainfall flood space-time process
Figure 525244DEST_PATH_IMAGE032
Whether flood peak information is contained or not is judged, if yes, the dynamic estimation stopping condition is not met, otherwise, if no flood peak information is contained, namely, all the flood peak information is in the process of water withdrawal, the dynamic estimation is stopped;
step 5-2, if the stopping condition is not met, moving to the next facing moment along with time, updating rainfall observation and forecast information, flood observation and forecast information in the effective forecast period, and repeating the step 2 to the step 4 to realize the flood process estimated result based on similar excavation of the rainfall flood space-time process
Figure 987450DEST_PATH_IMAGE032
Is dynamically updated.
7. The flood dynamic estimation method based on rainfall flood space-time process similarity excavation according to claim 1, wherein the steps 1-5 are repeated to obtain dynamic estimation results of the flood process of the hydrological station in the research area.
8. Flood dynamic estimation system, its characterized in that includes:
the system comprises a first module, a second module and a third module, wherein the first module is used for acquiring historical remote sensing inversion rainfall information, real-time remote sensing inversion rainfall information, numerical forecast rainfall information and historical site flood information of a preset number of hydrologic stations in a research area, matching the time scales of historical rainfall and flood information and constructing a hydrologic station historical rainfall flood database;
the second module is used for matching and splicing the space-time scales of the remote sensing inversion precipitation and numerical forecast precipitation information facing the earlier stage of time and observed aiming at the flood which is happening in the hydrological station, and searching historical precipitation processes with similar surface average precipitation time sequences in a historical rainfall flood database established by the first module by adopting a time sequence similarity analysis method; screening historical rainfall processes similar to the spatio-temporal process by adopting a spatio-temporal information similarity analysis method, and extracting a corresponding similarity historical flood set 1;
the third module is used for splicing flood information in an effective forecast period of constantly observed flood and flood forecast model forecast, and searching a historical similarity flood set 2 with similar time sequence process in a historical rainfall flood database by adopting a time sequence similarity analysis method;
the fourth module is used for merging the similarity historical flood set 1 and the similarity historical flood set 2 to obtain a similarity historical flood set 3, reasonably determining the splicing time point of each similar flood, calculating the average similar flood process after the splicing time point by adopting a weighted average method, and estimating the subsequent flood process through translation and splicing operation;
the fifth module judges whether the follow-up flood process estimated by the fourth module contains a flood peak or not, and stops dynamic estimation if the follow-up flood process does not contain the flood peak; otherwise, updating the rainfall flood observation and forecast information along with the time, repeatedly executing the second module to the fourth module, and dynamically estimating the development situation of the flood.
CN202111336042.9A 2021-11-12 2021-11-12 Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation Active CN113779113B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111336042.9A CN113779113B (en) 2021-11-12 2021-11-12 Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111336042.9A CN113779113B (en) 2021-11-12 2021-11-12 Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation

Publications (2)

Publication Number Publication Date
CN113779113A true CN113779113A (en) 2021-12-10
CN113779113B CN113779113B (en) 2022-03-08

Family

ID=78957062

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111336042.9A Active CN113779113B (en) 2021-11-12 2021-11-12 Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation

Country Status (1)

Country Link
CN (1) CN113779113B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114997534A (en) * 2022-07-29 2022-09-02 长江水利委员会水文局 Similar rainfall forecasting method and equipment based on visual features
CN116011879A (en) * 2023-02-01 2023-04-25 长江水利委员会长江科学院 Ecological system stability assessment method and device, electronic equipment and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101864750A (en) * 2010-06-29 2010-10-20 西安理工大学 Multi-model meta-synthesis flood forecasting system and forecasting method thereof
CN103488871A (en) * 2013-08-27 2014-01-01 国家电网公司 Flood forecasting method for area without runoff data
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
WO2019173202A1 (en) * 2018-03-07 2019-09-12 General Electric Company Systems and methods for decision tree ensembles for selecting actions
CN110728409A (en) * 2019-10-16 2020-01-24 中国科学院地理科学与资源研究所 Flood process type similarity mining and rapid prediction method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101864750A (en) * 2010-06-29 2010-10-20 西安理工大学 Multi-model meta-synthesis flood forecasting system and forecasting method thereof
CN103488871A (en) * 2013-08-27 2014-01-01 国家电网公司 Flood forecasting method for area without runoff data
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
WO2019173202A1 (en) * 2018-03-07 2019-09-12 General Electric Company Systems and methods for decision tree ensembles for selecting actions
CN110728409A (en) * 2019-10-16 2020-01-24 中国科学院地理科学与资源研究所 Flood process type similarity mining and rapid prediction method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114997534A (en) * 2022-07-29 2022-09-02 长江水利委员会水文局 Similar rainfall forecasting method and equipment based on visual features
CN116011879A (en) * 2023-02-01 2023-04-25 长江水利委员会长江科学院 Ecological system stability assessment method and device, electronic equipment and storage medium

Also Published As

Publication number Publication date
CN113779113B (en) 2022-03-08

Similar Documents

Publication Publication Date Title
US11333796B2 (en) Spatial autocorrelation machine learning-based downscaling method and system of satellite precipitation data
CN108304668B (en) Flood prediction method combining hydrologic process data and historical prior data
Boer et al. Kriging and thin plate splines for mapping climate variables
Mwangi et al. Forecasting droughts in East Africa
CN112800634B (en) Rainfall estimation method and system coupling dry-wet state identification and multi-source information fusion
CN101354757B (en) Method for predicting dynamic risk and vulnerability under fine dimension
CN113779113B (en) Flood dynamic estimation method and system based on rainfall flood space-time process similarity excavation
CN111612055B (en) Weather situation typing method, air pollution condition prediction method and device
CN115271255B (en) Rainfall flood similarity analysis method and system based on knowledge graph and machine learning
CN110648014A (en) Regional wind power prediction method and system based on space-time quantile regression
CN112418346B (en) Numerical weather forecast total radiation system error classification calculation method
Chen et al. Probabilistic forecasting of drought: a hidden Markov model aggregated with the RCP 8.5 precipitation projection
CN107133686A (en) City-level PM2.5 concentration prediction methods based on Spatio-Temporal Data Model for Spatial
CN112579922B (en) Interest point recommendation method based on user track
CN110276477B (en) Flood forecasting method based on hierarchical Bayesian network and incremental learning
CN105974495A (en) Method for pre-judging future average cloud amount of target area by using classification fitting method
CN113836808A (en) PM2.5 deep learning prediction method based on heavy pollution feature constraint
CN117332909B (en) Multi-scale urban waterlogging road traffic exposure prediction method based on intelligent agent
CN114254802B (en) Prediction method for vegetation coverage space-time change under climate change drive
Zhao et al. Comparison and correction of IDW based wind speed interpolation methods in urbanized Shenzhen
CN114357737A (en) Agent optimization calibration method for time-varying parameters of large-scale hydrological model
CN111860974B (en) Drought multistage prediction method based on state space and joint distribution
CN110046756B (en) Short-term weather forecasting method based on wavelet denoising and Catboost
CN116776581A (en) Knowledge and data fusion driven flood forecast model parameter online optimization method
Mishra et al. Discovering flood rising pattern in hydrological time series data mining during the pre monsoon period

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