CN110543992A - rainfall prediction method and device - Google Patents
rainfall prediction method and device Download PDFInfo
- Publication number
- CN110543992A CN110543992A CN201910842513.XA CN201910842513A CN110543992A CN 110543992 A CN110543992 A CN 110543992A CN 201910842513 A CN201910842513 A CN 201910842513A CN 110543992 A CN110543992 A CN 110543992A
- Authority
- CN
- China
- Prior art keywords
- rainfall
- target
- description information
- historical
- matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Optimization (AREA)
- Evolutionary Biology (AREA)
- Development Economics (AREA)
- Pure & Applied Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Environmental & Geological Engineering (AREA)
- Mathematical Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Marketing (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computational Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Databases & Information Systems (AREA)
- Quality & Reliability (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Educational Administration (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
Abstract
the invention discloses a rainfall prediction method and device, and belongs to the technical field of meteorological prediction. The method comprises the following steps: and acquiring target description information of the target rainfall process. Historical description information of one or more historical rainfall processes is obtained, and each historical rainfall process corresponds to one or more accompanying generation information. And selecting first history description information from the history description information, wherein the first history description information is the history description information with the highest matching degree with the target description information. And predicting the concomitant generation information corresponding to the target rainfall process according to the concomitant generation information of the historical rainfall process corresponding to the first historical description information. According to the method and the device, the historical rainfall process closest to the target rainfall process to be predicted is obtained, and the accompanying description information of the target rainfall process is predicted according to the accompanying description information of the historical rainfall process, so that the prediction accuracy is not affected by rainfall time, the prediction accuracy is high, and the prediction effect is good.
Description
Technical Field
The invention relates to the technical field of meteorological prediction, in particular to a rainfall prediction method and device.
background
With the development of weather forecasting technology, more and more predictable weather phenomena are generated, and the rainfall phenomenon is one of the weather forecasting technologies. Rainfall information such as a movement track, a development trend and the like of rainfall can be obtained by predicting the rainfall so as to take appropriate measures aiming at the rainfall information, thereby avoiding the rainfall from threatening personal safety or causing economic loss.
The related art provides a rainfall prediction method that transmits radio waves to a rainfall cloud by a radar and receives radar echoes reflected or scattered by the rainfall cloud. And then, calculating the optimal spatial correlation coefficient of the radar echo received at the adjacent time, and determining the movement characteristic of the radar echo according to the optimal spatial correlation coefficient. Finally, the position and shape of the future radar echo are estimated from the movement characteristics, and the rainfall is predicted based on the estimation result.
however, this method is only suitable for rainfall events of short duration (e.g., within two hours). If the rainfall lasts for a long time, the accuracy of the position and the shape of the presumed radar echo at the future time is low, so that the rainfall prediction accuracy is reduced, and the prediction effect is poor.
Disclosure of Invention
The embodiment of the application provides a rainfall prediction method and device, and aims to solve the problems of low prediction accuracy and poor prediction effect of the related technology. The technical scheme is as follows:
In one aspect, a method of rainfall prediction is provided, the method comprising:
Acquiring target description information of a target rainfall process;
acquiring historical description information of one or more historical rainfall processes, wherein each historical rainfall process corresponds to one or more accompanying generation information;
Selecting first history description information from the history description information, wherein the first history description information is the history description information with the highest matching degree with the target description information;
And predicting the accompanying generation information corresponding to the target rainfall process according to the accompanying generation information of the historical rainfall process corresponding to the first historical description information.
optionally, the obtaining target description information of the target rainfall process includes:
Acquiring first rainfall corresponding to more than two reference places in the target rainfall process;
and determining a target matrix according to the first rainfall, wherein one element in the target matrix is used for representing the first rainfall of one reference place at one moment.
optionally, before obtaining the historical description information of one or more historical rainfall processes, the method further includes:
Acquiring rainfall data of each reference place in past time;
Dividing the past time into the one or more historical rainfall processes according to a reference condition based on the rainfall data;
Wherein the reference condition includes: and taking a time point when the rainfall data of the reference place is not less than the reference value as a starting point of the historical rainfall process, and taking a time point when the rainfall data of the reference place is zero and is kept as zero reference time as an end point of the historical rainfall process.
Optionally, the selecting the first history description information from the history description information includes:
Calculating the characteristic distance between each historical description information and the target matrix;
and taking the history description information with the minimum characteristic distance with the target matrix as the first history description information.
optionally, before the selecting the first history description information from the history description information, the method further includes:
and reducing the dimension of the target matrix to obtain a reduced-dimension target matrix, wherein one element in the reduced-dimension target matrix is used for representing the first rainfall of a reference place at one moment.
Optionally, the predicting the accompanying generation information corresponding to the target rainfall process according to the accompanying generation information of the historical rainfall process corresponding to the first historical description information includes:
And taking the accompanying generation information of the historical rainfall process corresponding to the first historical description information as the accompanying generation information corresponding to the target rainfall process.
in another aspect, there is provided a rainfall prediction apparatus, comprising:
the first acquisition module is used for acquiring target description information of a target rainfall process;
The second acquisition module is used for acquiring historical description information of one or more historical rainfall processes, and each historical rainfall process corresponds to one or more accompanying generation information;
the selection module is used for selecting first history description information from the history description information, wherein the first history description information is the history description information with the highest matching degree with the target description information;
And the prediction module is used for predicting the concomitant generation information corresponding to the target rainfall process according to the concomitant generation information of the historical rainfall process corresponding to the first historical description information.
Optionally, the first obtaining module is configured to obtain first rainfall amounts corresponding to more than two reference locations in the target rainfall process; and determining a target matrix according to the first rainfall, wherein one element in the target matrix is used for representing the first rainfall of one reference place at one moment.
optionally, the apparatus further comprises:
the third acquisition module is used for acquiring first rainfall corresponding to more than two reference places in the target rainfall process; and determining a target matrix according to the first rainfall, wherein one element in the target matrix is used for representing the first rainfall of one reference place at one moment.
Optionally, the selecting module is configured to calculate a characteristic distance between each piece of historical description information and the target matrix; and taking the history description information with the minimum characteristic distance with the target matrix as the first history description information.
Optionally, the apparatus further comprises:
And the dimension reduction module is used for reducing the dimension of the target matrix to obtain a dimension-reduced target matrix, and one element in the dimension-reduced target matrix is used for representing the first rainfall of a reference place at one moment.
optionally, the predicting module is configured to use the accompanying generation information of the historical rainfall process corresponding to the first historical description information as the accompanying generation information corresponding to the target rainfall process.
The beneficial effects brought by the technical scheme provided by the embodiment of the application at least comprise:
according to the method and the device, the historical rainfall process which is most similar to the target rainfall process to be predicted is obtained based on the target description information and the historical description information, and the accompanying generation information of the target rainfall process is predicted according to the accompanying generation information of the historical rainfall process. Therefore, the forecasting process is not influenced by the rainfall duration of the target rainfall process, and even if the rainfall duration is longer (for example, more than two hours), the forecasting method and the forecasting device can still accurately forecast the target rainfall process, and are high in forecasting accuracy and good in forecasting effect.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flow chart of a method of rainfall prediction provided by an embodiment of the present application;
FIG. 2 is a schematic diagram of object description information provided by an embodiment of the present application;
FIG. 3 is a schematic diagram of history description information provided by an embodiment of the present application;
Fig. 4 is a schematic structural diagram of a rainfall prediction device according to an embodiment of the present application.
Detailed Description
to make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.
The application provides a rainfall prediction method, as shown in fig. 1, the method includes:
step 101, obtaining target description information of a target rainfall process.
In the present embodiment, it is considered that the target description information is used to describe the target rainfall process, and thus the target description information may be determined based on the information related to the target rainfall process. Therefore, in an alternative embodiment, the manner of obtaining the target description information includes: the method comprises the steps of obtaining first rainfall corresponding to more than two reference places in the target rainfall process, and determining a target matrix according to the first rainfall.
The reference location may include all weather monitoring stations in the reference area. Referring to fig. 2, the X-axis and the Y-axis in fig. 2 are used to indicate space, and the T-axis is used to indicate time. Therefore, each graph represents a plurality of weather monitoring stations in a reference area at a certain time, and the graphs are superposed according to the T axis to represent the first rainfall of more than two reference points in the target rainfall process (at multiple times). The number of the reference points is two or more, for example, two, three, four or more, which is not limited in this embodiment. It should be noted that, for different types of target rainfall processes, the manner of obtaining the corresponding first rainfall amount of each reference location is also different, including but not limited to the following three:
The first acquisition mode is as follows: the target rainfall process is a rainfall process which passes through the weather forecast but does not occur, and the first rainfall amount corresponding to each reference location can be obtained according to the weather forecast information. The information accompanying the target rainfall process predicted subsequently may be information such as a rainfall development tendency of the target rainfall process.
The second acquisition mode is as follows: if the target rainfall process is a rainfall process in the rainfall process, the actual rainfall amount of rainfall and the rainfall amount provided in the weather forecast information can be used as the first rainfall amount of any reference location. The higher the proportion of the actual rainfall in the first rainfall is, the more accurate the subsequent predicted accompanying generation information is, and the accompanying generation information may include information such as waterlogging water in the target rainfall process.
the third acquisition mode is as follows: the target rainfall process is a finished rainfall process, and at this time, the actual rainfall capacity of the target rainfall process is used as the first rainfall capacity. The follow-up predicted accompanying generation information can be information such as the man (object) force scheduling condition of disaster relief engineering suitable for the target rainfall process.
no matter which acquisition mode is adopted, the first rainfall corresponding to each reference point can be acquired. It should be noted that the first rainfall is obtained once every reference time interval, and the reference time interval may be five minutes, one hour, and the like. After the first rainfall is obtained, the following target matrix P can be determined according to the first rainfall:
Where an element in the target matrix is used to represent a first rainfall of a reference location at a time, for example pmn is used to represent a rainfall of an mth reference location at an nth time, and each column is a first rainfall of a reference location corresponding to a plurality of times.
Of course, no matter how the target description information is obtained, the historical description information of the historical rainfall process needs to be obtained accordingly. In this embodiment, the order of acquiring the target description information and acquiring the history description information is not limited, and for example, the target description information may be acquired first and then the history description information may be acquired first and then the target description information may be acquired first, or the target description information and the history description information may be acquired at the same time. The manner in which the history description information is obtained can be seen in step 102 below.
step 102, obtaining historical description information of one or more historical rainfall processes, wherein each historical rainfall process corresponds to one or more accompanying information.
The accompanying information corresponding to the historical rainfall process includes but is not limited to: the moving track of the rainfall cloud cluster, the development trend of rainfall, the river water regime, the waterlogging and water accumulation condition, the man (object) force scheduling condition in the disaster relief project and other information in the historical rainfall process.
historical descriptive information for one or more historical rainfall events can be seen in fig. 3. As can be seen from the description in step 101, the object description information may be in a matrix form, and thus, the history description information may also be in a matrix form. Optionally, obtaining historical description information of one or more historical rainfall processes, including: and acquiring second rainfall corresponding to the reference location in one or more historical rainfall processes, and determining a historical matrix according to the second rainfall, wherein one element in the historical matrix is used for representing the second rainfall of one reference location at one historical moment.
The reference location in step 102 may be the same as or different from the reference location in step 101. In addition, in the prediction process of the target rainfall process, the history matrix of each historical rainfall process can be acquired before the prediction process, and the acquired history matrix is stored. And when the prediction is carried out subsequently, the stored history matrix is directly called. Alternatively, the history matrix may be obtained in real time during the prediction process.
In this embodiment, the historical rainfall process is a rainfall process that has ended in the past time, so the second rainfall amount corresponding to each reference location is the actual rainfall amount of the reference location, and the second rainfall amount may also include rainfall amounts at one or more times. Therefore, the history matrix determined according to the second rainfall is as follows:
wherein i is used for distinguishing different history matrixes, one column in each Qi is a second rainfall corresponding to one reference location, and qmn is used for representing the rainfall at the nth time of the mth reference location. In addition, as can be seen from the above equation, Qi is a matrix of m rows and n columns, and expressed as Qi ∈ Rm × n. All history matrices can form a sample set Ω, and Ω ═ Q1, Q2, … Qi, … QN }, and it can be seen that there are N history matrices in the sample set, i.e. i has any positive integer in the range of [1, N ].
Furthermore, as the rainfall data recorded in the past time is more, the number of the historical rainfall processes can be divided according to the rainfall data in the past time, so that one or more historical rainfall processes can be obtained. That is to say, optionally, before acquiring one or more pieces of history description information, the method provided in this embodiment further includes: acquiring rainfall data of a reference place in past time; based on the rainfall data, past time is divided into one or more historical rainfall processes according to reference conditions.
the reference conditions include: and taking a time point when the rainfall data of one of the reference places is not less than the reference value as a starting point of the historical rainfall process. The reference value is not limited in this embodiment, and for example, the reference value may be 0.1mm/h (unit: mm/h), or may be other values set according to needs or experience. Taking the above example that the number of the reference sites is 9 and the reference value is 0.1mm/h, as long as the rainfall data of any one reference site in the 9 reference sites is not less than 0.1mm/h, the time point at which the rainfall data of the reference site is not less than 0.1mm/h, which is called the starting point of the historical rainfall process, is considered to be that one historical rainfall process has started.
in addition, the reference conditions further include: and taking the time point when the rainfall data of all the reference places return to zero and the reference time length is kept unchanged after the rainfall data of all the reference places return to zero as the end point of the historical rainfall process. The reference time period may also be set according to needs or experience, for example, the reference time period may be two hours in this embodiment. Taking the number of the reference sites as 9 and the reference time as two hours as an example, after a historical rainfall process starts, the rainfall of one or more of the 9 reference sites changes from zero to a positive value successively. Then, after the rainfall data of the 9 reference locations are all zeroed by positive values and the rainfall data is kept zero for two hours, the time point two hours after the rainfall data is kept zero can be used as the end point of the historical rainfall process, that is, the historical rainfall process is considered to be ended.
it should be noted that, only after the rainfall data is maintained at the zero reference time length, the historical rainfall is considered to be finished, and the starting point of the next historical rainfall process in the past time can be further confirmed. That is, taking the reference time length as two hours as an example, if all the rainfall data of all the reference places are zeroed by positive values, but five minutes after the zeroed rainfall data of one or more reference places become positive values again, the historical rainfall process is considered to be not finished. The end of the historical rainfall event can only be considered when the rainfall data for all reference locations is re-zeroed and held at zero for two hours. By the reference conditions, the acquired historical rainfall process is more practical, that is, the actual rainfall process of rainfall may be restarted after the rainfall is temporarily interrupted, so that the prediction effect of predicting the target rainfall process according to the historical rainfall process in the following process is better, and the prediction process can be referred to in step 103.
Step 103, selecting first history description information from the history description information.
the first history description information refers to one history description information with the highest matching degree with the target description information in all history description information. Optionally, the selecting means includes: and calculating the characteristic distance between each piece of history description information and the target matrix, and taking the history description information with the minimum characteristic distance to the target matrix as the first history description information.
under the condition that the target description information and the historical description information are both in a matrix form, the selection mode is as follows: and calculating the characteristic distance between each history matrix and the target matrix, and taking the history matrix with the minimum characteristic distance to the target matrix as a first history matrix.
The calculated characteristic distance may be an euclidean distance, and then the characteristic distance D between any history matrix and the target matrix in the sample set may be calculated according to the following formula:
D＝‖Q-P‖
it should be noted that both the history matrix and the target matrix can be regarded as vectors in a multidimensional space, and the characteristic distance is used for measuring the difference between the two vectors in each dimension. That is, for the target matrix and any history matrix, the smaller the characteristic distance D, the smaller the difference between the target matrix and the history matrix in all dimensions is considered, that is, the higher the matching degree between the target matrix and the history matrix is. Therefore, the history matrix having the smallest characteristic distance from the target matrix may be used as the first history matrix.
further, the more the second rainfall included in the history matrix, the larger the dimension of the history matrix, and the better the history matrix can describe the history rainfall process. Similarly, the larger the dimension of the target matrix, the better the target matrix can describe the target rainfall process. However, the larger the dimension of the target matrix and the history matrix is, the larger the calculation amount of selecting the history matrix with the highest matching degree with the target matrix from the history matrix is, and the calculation difficulty is high, which may affect the selection effect and further affect the prediction accuracy.
Based on the above consideration, optionally, before selecting the history matrix with the highest matching degree with the target matrix from the history matrices, the method further includes: and reducing the dimension of the target matrix to obtain the reduced target matrix, wherein one element in the reduced target matrix is still used for representing the first rainfall of a reference place at one moment. Correspondingly, for any historical matrix, performing dimensionality reduction on the historical matrix to obtain a dimensionality-reduced historical matrix, wherein one element in the dimensionality-reduced historical matrix is still used for representing the second rainfall of one reference place at a past moment.
Next, a dimension reduction method will be described. It should be noted that the dimension reduction method can be applied to the dimension reduction of the history matrix, and can also be applied to the dimension reduction of the target matrix, and the history matrix Qi is taken as an example in the following description:
The dimension reduction process includes performing dimension reduction from both row and column directions through the projection matrix, for example, performing dimension reduction from the column direction first and then from the row direction, or performing dimension reduction from the row direction first and then from the column direction. Taking the dimension reduction from the column direction as an example, optionally, the dimension reduction method includes the following steps a 1-a 4:
and A1, acquiring a first projection matrix based on the historical matrix, wherein the number of columns of the first projection matrix is less than that of the historical matrix.
assuming that the projection matrix is X (X is a column vector matrix), the history matrix Qi is projected by X to obtain the following projection result Yi:
Y＝QX
It should be noted that the projection matrix X should make the dispersion of the covariance matrix of the projection result Y larger, so as to ensure that the projection result Yi can keep the information (i.e. the second rainfall) in the history matrix Qi as much as possible, thereby reducing the information loss degree caused by dimension reduction. The covariance matrix S of the projection result Yi can be calculated according to the following formula:
S＝E[(Y-E(Y))(Y-E(Y))]
＝E[(QX-E(QX))(QX-E(QX))]
＝E[((Q-E(Q))X)((Q-E(Q))X)X
＝XE[(Q-E(Q))(Q-E(Q))]X
Wherein E () represents a mean value, e.g., E (Yi), which represents the mean value of the projection result Yi; t represents the transpose of the matrix, and E [ (Qi-E (Qi)) T (Qi-E (Qi)) ] X in the above equation just represents the covariance matrix of the history matrix Qi, so the first projection matrix can be obtained based on the history matrix.
In this embodiment, when E [ (Qi-E (Qi)) T (Qi-E (Qi)) ] X is denoted as Gt, the covariance matrix S of the projection result Y becomes XTGtX. Since the projection matrix X is to be solved, the dispersion of S can be maximized by maximizing the dispersion of Gt. Qi is located in the sample set omega, and N history matrixes are in total in the sample set, so the covariance matrix G of Qi can be further expressed according to the following formula:
The average value of all history matrixes Qi is known according to the matrix operation rule, namely Gt belongs to Rn x n, because Qi belongs to Rm x n. And then, calculating one or more characteristic vectors of the Gt, wherein each characteristic vector corresponds to one characteristic value, and each vector is a column vector. And arranging the eigenvectors according to the sequence of the eigenvalues from big to small to obtain an eigenvector sequence. The first k (k < n) eigenvectors in the eigenvector sequence are selected to form a first projection matrix U ═ U1, U2 … uk ∈ Rn × k. It should be noted that, as for the determination method of the k numerical value, the determination may be performed according to a rule that the ratio of the sum of the feature values corresponding to the k feature vectors to the sum of the feature values corresponding to all the feature vectors is a reference value. The reference value is not limited in this embodiment, and for example, the reference value may be any one of 0.9 to 0.99.
And A2, multiplying the history matrix and the first projection matrix to obtain a projection matrix in the column direction.
after the first projection matrix U is obtained, a product calculation may be performed on the first projection matrix U and the history matrix Qi, and according to a matrix operation rule, a projection matrix Fi in the column direction may be obtained as Qi · U ∈ Rm × k, where the number of projection matrices in the column direction is the same as the number of history matrices, and is N. Since k is smaller than n, Fi is reduced compared to the column number of Qi, the number of rows is unchanged, i.e. dimensionality reduction in the column direction is accomplished. Then, the projection matrix Fi in the column direction may be further reduced in dimension in the row direction.
And A3, acquiring a second projection matrix based on the projection matrix in the column direction, wherein the row number of the second projection matrix is less than that of the history matrix.
when the dimension reduction in the row direction is carried out, a covariance matrix Gr of Fi is obtained according to the following formula:
The average value of the projection matrices Fi in all column directions is Fi ∈ Rm × k, so that Gr ∈ Rm × m can be obtained according to a matrix operation rule. Then, d (d < m) eigenvectors of Gr are selected and obtained according to the method in the above description, and a second projection matrix V ═ V1, V2 … Vd ∈ Rm × d is formed.
and A4, taking the product of the transpose matrix of the second projection matrix and the projection matrix in the column direction as the history matrix after dimension reduction.
According to the operation rule of the matrix, after the second projection matrix V is transposed to obtain the transposed matrix VT, the transposed matrix and the first projection matrix are multiplied to obtain the projection result Yi as follows:
Y＝V.F＝V.Q·U∈R
It can be seen that, in the embodiment, the projection matrix X is divided into the first projection matrix U and the second projection matrix V, and the projection result Yi of d rows and k columns can be calculated from the history matrix Qi of m rows and n columns through the two projection matrices. Since d is less than m and k is less than n, the projection result is the history matrix after dimension reduction.
in addition to the above-mentioned dimension reduction of the target matrix and the history matrix from both the row direction and the column direction, the dimension reduction of the target matrix and the history matrix from one of the row direction and the column direction may be performed. Since the dimension reduction method from one direction of row or column is included in the above description, it is not repeated here.
After the dimensionality reduced target matrix and the dimensionality reduced history matrix are obtained, the dimensionality reduced history matrix with the highest matching degree with the dimensionality reduced target matrix can be further selected from the dimensionality reduced history matrices to serve as a first history matrix, so that the information accompanying the target rainfall process can be conveniently predicted in the subsequent process, and detailed step 104 is shown.
and step 104, predicting the concomitant generation information corresponding to the target rainfall process according to the concomitant generation information of the historical rainfall process corresponding to the first historical description information.
In this embodiment, different prediction modes may be selected according to different matching degrees between the first history description information and the target description information. For example, in the present embodiment, a reference threshold may be set empirically, and if the matching degree is higher than the reference threshold, it indicates that the similarity between the selected history matrix and the target matrix is higher, and further, it may be considered that the similarity between the history rainfall process indicated by the history matrix and the target rainfall process is higher. Thus, in an alternative embodiment, the information associated with the historical rainfall event may be used directly as information associated with the target rainfall event to complete the forecast.
Correspondingly, if the matching degree is lower than the reference threshold, the similarity between the historical rainfall process indicated by the historical matrix and the target rainfall process is low, and the associated production information of the historical rainfall process can be analyzed and updated by combining the relevant rainfall data of the target rainfall process to obtain the updated associated production information. And then, the updated accompanying generation information is used as the accompanying generation information of the target rainfall process to finish the prediction. By the mode, the accuracy of prediction can be improved, and the prediction effect of the embodiment is guaranteed.
In summary, in the embodiment of the present application, based on the target description information and the historical description information, the historical rainfall process most similar to the target rainfall process to be predicted is obtained, and then the accompanying generation information of the target rainfall process is predicted according to the accompanying generation information of the historical rainfall process. Therefore, the forecasting process is not influenced by the rainfall duration of the target rainfall process, and even if the rainfall duration is longer (for example, more than two hours), the forecasting method and the forecasting device can still accurately forecast the target rainfall process, and are high in forecasting accuracy and good in forecasting effect.
Furthermore, the embodiment also performs dimension reduction on the target description information and the historical description information, so that the calculation amount required by prediction is reduced, the calculation time is shortened, and the efficiency of predicting the target rainfall process is improved.
Based on the same concept, the present embodiment further provides a rainfall prediction apparatus, referring to fig. 4, including:
a first obtaining module 401, configured to obtain target description information of a target rainfall process;
A second obtaining module 402, configured to obtain historical description information of one or more historical rainfall processes, where each historical rainfall process corresponds to one or more pieces of accompanying generation information;
a selecting module 403, configured to select first history description information from the history description information, where the first history description information is history description information with a highest matching degree with the target description information;
the predicting module 404 is configured to predict the accompanying generation information corresponding to the target rainfall process according to the accompanying generation information of the historical rainfall process corresponding to the first historical description information.
Optionally, the first obtaining module 401 is configured to obtain first rainfall corresponding to more than two reference points in the target rainfall process; and determining a target matrix according to the first rainfall, wherein one element in the target matrix is used for representing the first rainfall of one reference place at one moment.
Optionally, the apparatus further comprises: the third acquisition module is used for acquiring first rainfall corresponding to more than two reference places in the target rainfall process; and determining a target matrix according to the first rainfall, wherein one element in the target matrix is used for representing the first rainfall of one reference place at one moment.
optionally, the selecting module 403 is configured to calculate a characteristic distance between each piece of history description information and the target matrix; and taking the history description information with the minimum characteristic distance with the target matrix as the first history description information.
Optionally, the apparatus further comprises: and the dimension reduction module is used for reducing the dimension of the target matrix to obtain the target matrix after dimension reduction, and one element in the target matrix after dimension reduction is used for representing the first rainfall of one reference place at one moment.
Optionally, the predicting module 404 is configured to use the accompanying generation information of the historical rainfall process corresponding to the first historical description information as the accompanying generation information corresponding to the target rainfall process.
In summary, in the embodiment of the present application, based on the target description information and the historical description information, the historical rainfall process most similar to the target rainfall process to be predicted is obtained, and then the accompanying generation information of the target rainfall process is predicted according to the accompanying generation information of the historical rainfall process. Therefore, the forecasting process is not influenced by the rainfall duration of the target rainfall process, and even if the rainfall duration is longer (for example, more than two hours), the forecasting method and the forecasting device can still accurately forecast the target rainfall process, and are high in forecasting accuracy and good in forecasting effect.
Furthermore, the embodiment also performs dimension reduction on the target description information and the historical description information, so that the calculation amount required by prediction is reduced, the calculation time is shortened, and the rainfall prediction efficiency is improved.
all the above optional technical solutions may be combined arbitrarily to form the optional embodiments of the present disclosure, and are not described herein again.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The above description is only exemplary of the present application and should not be taken as limiting the present application, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (10)
1. A method of rainfall prediction, the method comprising:
Acquiring target description information of a target rainfall process;
acquiring historical description information of one or more historical rainfall processes, wherein each historical rainfall process corresponds to one or more accompanying generation information;
selecting first history description information from the history description information, wherein the first history description information is the history description information with the highest matching degree with the target description information;
And predicting the accompanying generation information corresponding to the target rainfall process according to the accompanying generation information of the historical rainfall process corresponding to the first historical description information.
2. The method of claim 1, wherein obtaining target description information of a target rainfall process comprises:
acquiring first rainfall corresponding to more than two reference places in the target rainfall process;
and determining a target matrix according to the first rainfall, wherein one element in the target matrix is used for representing the first rainfall of one reference place at one moment.
3. The method of claim 2, wherein prior to obtaining historical descriptive information for one or more historical rainfall events, the method further comprises:
Acquiring rainfall data of each reference place in past time;
Dividing the past time into the one or more historical rainfall processes according to a reference condition based on the rainfall data;
wherein the reference condition includes: and taking a time point when the rainfall data of the reference place is not less than the reference value as a starting point of the historical rainfall process, and taking a time point when the rainfall data of the reference place is zero and is kept as zero reference time as an end point of the historical rainfall process.
4. the method according to claim 2 or 3, wherein the selecting the first history description information from the history description information comprises:
calculating the characteristic distance between each historical description information and the target matrix;
And taking the history description information with the minimum characteristic distance with the target matrix as the first history description information.
5. The method of claim 4, wherein prior to said selecting the first history description information from the history description information, the method further comprises:
and reducing the dimension of the target matrix to obtain a reduced-dimension target matrix, wherein one element in the reduced-dimension target matrix is used for representing the first rainfall of a reference place at one moment.
6. The method according to claim 2 or 3, wherein the predicting the co-production information corresponding to the target rainfall process according to the co-production information corresponding to the historical rainfall process corresponding to the first historical description information comprises:
and taking the accompanying generation information of the historical rainfall process corresponding to the first historical description information as the accompanying generation information corresponding to the target rainfall process.
7. an apparatus for rainfall prediction, the apparatus comprising:
The first acquisition module is used for acquiring target description information of a target rainfall process;
The second acquisition module is used for acquiring historical description information of one or more historical rainfall processes, and each historical rainfall process corresponds to one or more accompanying generation information;
the selection module is used for selecting first history description information from the history description information, wherein the first history description information is the history description information with the highest matching degree with the target description information;
and the prediction module is used for predicting the concomitant generation information corresponding to the target rainfall process according to the concomitant generation information of the historical rainfall process corresponding to the first historical description information.
8. The apparatus of claim 7, further comprising:
the third acquisition module is used for acquiring first rainfall corresponding to more than two reference places in the target rainfall process; and determining a target matrix according to the first rainfall, wherein one element in the target matrix is used for representing the first rainfall of one reference place at one moment.
9. The apparatus of claim 8, further comprising:
And the dimension reduction module is used for reducing the dimension of the target matrix to obtain a dimension-reduced target matrix, and one element in the dimension-reduced target matrix is used for representing the first rainfall of a reference place at one moment.
10. the apparatus according to claim 8, wherein the prediction module is configured to use the accompanying information of the historical rainfall process corresponding to the first historical description information as the accompanying information corresponding to the target rainfall process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910842513.XA CN110543992A (en) | 2019-09-06 | 2019-09-06 | rainfall prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910842513.XA CN110543992A (en) | 2019-09-06 | 2019-09-06 | rainfall prediction method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110543992A true CN110543992A (en) | 2019-12-06 |
Family
ID=68712834
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910842513.XA Pending CN110543992A (en) | 2019-09-06 | 2019-09-06 | rainfall prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110543992A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100274542A1 (en) * | 2009-04-23 | 2010-10-28 | Honeywell International Inc. | Enhanced prediction of atmospheric parameters |
| US20160370500A1 (en) * | 2015-06-18 | 2016-12-22 | J.F. Sabourin et associés Inc. | Method and server for providing alerts for rainfall return periods |
| CN108009596A (en) * | 2017-12-26 | 2018-05-08 | 中国水利水电科学研究院 | The method and apparatus for determining characteristics of rainfall |
| CN109376940A (en) * | 2018-11-02 | 2019-02-22 | 中国水利水电科学研究院 | The method and apparatus for obtaining the rainfall time space distribution in rainfall |
-
2019
- 2019-09-06 CN CN201910842513.XA patent/CN110543992A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100274542A1 (en) * | 2009-04-23 | 2010-10-28 | Honeywell International Inc. | Enhanced prediction of atmospheric parameters |
| US20160370500A1 (en) * | 2015-06-18 | 2016-12-22 | J.F. Sabourin et associés Inc. | Method and server for providing alerts for rainfall return periods |
| CN108009596A (en) * | 2017-12-26 | 2018-05-08 | 中国水利水电科学研究院 | The method and apparatus for determining characteristics of rainfall |
| CN109376940A (en) * | 2018-11-02 | 2019-02-22 | 中国水利水电科学研究院 | The method and apparatus for obtaining the rainfall time space distribution in rainfall |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109376940B (en) | Method and device for acquiring rainfall spatial-temporal distribution rule in rainfall process | |
| EP3161527A1 (en) | Solar power forecasting using mixture of probabilistic principal component analyzers | |
| CN109752710B (en) | Rapid target angle estimation method based on sparse Bayesian learning | |
| KR101984274B1 (en) | Apparatus for detecting target and identifying shape in surveillance scout and method thereof | |
| CN113111524B (en) | Antenna isolation degree prediction method based on far field test, storage medium and device | |
| CN110543992A (en) | rainfall prediction method and device | |
| Epaillard et al. | Hidden Markov models based on generalized Dirichlet mixtures for proportional data modeling | |
| Tanaka et al. | Method selection in different regions for short-term wind speed prediction in Japan | |
| JP2019138862A (en) | Radar signal processor | |
| CN111832623A (en) | Echo state network time sequence prediction algorithm based on phase space reconstruction | |
| US20210018319A1 (en) | Search support apparatus, search support method, and computer-readable recording medium | |
| CN114581781B (en) | Target detection method and device for high-resolution remote sensing image | |
| CN110133387A (en) | A kind of phased array antenna Electrical Analysis method estimated based on random field | |
| Choi et al. | Doppler coherent focusing doa method for efficient radar map generation | |
| CN115526402A (en) | Index prediction method, index prediction device, electronic equipment and storage medium | |
| Avendaño-Valencia et al. | Markov Chain Monte Carlo estimation of non-stationary GSC-TARMA models: application to random vibration modelling and analysis for an operating wind turbine | |
| Bedin et al. | The use of a genetic algorithm for parameter adjustment of the multi-hypothesis aircraft tracking algorithm | |
| CN113740802A (en) | Signal source positioning method and system for performing matrix completion by using adaptive noise estimation | |
| Bremnes et al. | Evaluation of forecasts by a global data-driven weather model with and without probabilistic post-processing at Norwegian stations | |
| Yang et al. | Probability hypothesis density filter with adaptive parameter estimation for tracking multiple maneuvering targets | |
| Li et al. | Shape-driven multiple extended target tracking and classification based on B-Spline and PHD filter | |
| CN108897072B (en) | Cloud quantity numerical forecasting method and forecasting system for business remote satellites | |
| Yamada et al. | Multi-dimensional multiple hypothesis tracking with a Gaussian mixture model to suppress grating lobes | |
| Subashini et al. | Imputation of missing data using Bayesian Principal Component Analysis on TEC ionospheric satellite dataset | |
| CN113176783B (en) | Positioning control method based on map matching, chip and robot |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191206 |
|
| RJ01 | Rejection of invention patent application after publication |