CN116931129A - Short-term precipitation prediction method, device, equipment and medium based on multi-mode set - Google Patents
Short-term precipitation prediction method, device, equipment and medium based on multi-mode set Download PDFInfo
- Publication number
- CN116931129A CN116931129A CN202310924818.1A CN202310924818A CN116931129A CN 116931129 A CN116931129 A CN 116931129A CN 202310924818 A CN202310924818 A CN 202310924818A CN 116931129 A CN116931129 A CN 116931129A
- Authority
- CN
- China
- Prior art keywords
- sequence
- radar echo
- short
- prediction
- precipitation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000001556 precipitation Methods 0.000 title claims description 91
- 238000010586 diagram Methods 0.000 claims abstract description 57
- 238000007781 pre-processing Methods 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims description 26
- 238000012549 training Methods 0.000 claims description 25
- 238000013213 extrapolation Methods 0.000 claims description 23
- 238000003860 storage Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 10
- 230000007246 mechanism Effects 0.000 claims description 10
- 238000002310 reflectometry Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000002123 temporal effect Effects 0.000 claims description 3
- 238000013277 forecasting method Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/95—Radar or analogous systems specially adapted for specific applications for meteorological use
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/411—Identification of targets based on measurements of radar reflectivity
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/417—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section involving the use of neural networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Health & Medical Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Environmental Sciences (AREA)
- Electromagnetism (AREA)
- Ecology (AREA)
- Biodiversity & Conservation Biology (AREA)
- Atmospheric Sciences (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a short-term rainfall forecasting method, device, equipment and medium based on a multi-mode set, wherein the method comprises the steps of acquiring a radar echo diagram sequence of a radar acquired target area in a current period, and preprocessing; acquiring a radar echo diagram sequence of a short future period according to the preprocessed radar echo diagram sequence of the current period; according to the radar echo diagram sequence of the short future period, acquiring rainfall intensity sequence prediction of the target area in the short future period; obtaining rainfall intensity sequence predictions of target areas provided by various third parties in a long future period; according to the rainfall intensity sequence prediction of each long future period, obtaining the final prediction of the rainfall intensity sequence of the target area in the long future period; splicing the rainfall intensity sequence prediction of the target area in the short future period and the final rainfall intensity sequence prediction of the target area in the long future period to obtain short-term rainfall intensity sequence prediction and forecast; the rainfall forecast provided by the invention has higher accuracy.
Description
Technical Field
The invention relates to a short-term precipitation prediction method, device, equipment and medium based on a multi-mode set, and belongs to the technical field of precipitation prediction.
Background
Precipitation prediction is an important research field in meteorology, and aims to predict precipitation amount, precipitation space-time distribution and other information in a future period of time. The accuracy and precision of rainfall forecast have important significance and application value in agriculture, water conservancy, transportation, energy source, city planning and other aspects.
In precipitation forecasting, common methods include physical models, statistical methods, artificial neural networks, machine learning, and the like. The physical model is a mathematical model constructed based on physical principles and aerology theory, and can predict future rainfall condition by calculating the parameter change of the atmosphere environment. The statistical method is based on historical meteorological data and statistical principles for analysis and prediction, and is suitable for rainfall prediction in a certain period. The artificial neural network and the machine learning are realized by learning and training a large amount of data to establish a prediction model. However, the accuracy of the above-mentioned conventional methods still cannot meet the increasing use demands.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and provides a short-term precipitation prediction method, device, equipment and medium based on a multi-mode set, which solve the technical problem that the accuracy of precipitation prediction in the prior art still needs to be improved.
In order to achieve the above purpose, the invention is realized by adopting the following technical scheme:
in a first aspect, the present invention provides a short-term precipitation prediction method based on a multimode set, comprising:
acquiring a radar echo diagram sequence of a target area acquired by a radar in a current period, and preprocessing;
inputting the preprocessed radar echo graph sequence of the current time period into a trained radar echo extrapolation model to obtain a radar echo graph sequence of a short future time period;
inputting the radar echo diagram sequence of the short future period into a fitted rainfall intensity estimation model, and obtaining rainfall intensity sequence prediction of a target area in the short future period;
obtaining rainfall intensity sequence predictions of target areas provided by various third parties in a long future period;
inputting the rainfall intensity sequence predictions of each long future period into a trained RNN stacking model, and obtaining the final prediction of the rainfall intensity sequence of the target area in the long future period;
and splicing the rainfall intensity sequence prediction of the target area in the short future period and the final rainfall intensity sequence prediction of the target area in the long future period to obtain the short-term rainfall intensity sequence prediction of the target area and forecast.
Optionally, the radar echo extrapolation model comprises a convolution layer, a max pooling layer, a seq2seq model and a deconvolution layer; the seq2seq model includes an encoder and a decoder, each of the encoder and the decoder including a plurality of stacked rain net modules including a convls (tm) module and a CBAM convolution attention module; the CBAM convolution attention module includes a spatial attention module and a channel attention module.
Optionally, the training process of the radar echo extrapolation model includes:
acquiring radar echo diagram sequences of a target area in a plurality of historical time periods, and preprocessing;
extracting radar echo diagram sequences of short future time periods corresponding to the radar echo diagram sequences of each history time period from the preprocessed radar echo diagram sequences of a plurality of the history time periods;
combining the radar echo map sequences of each historical period and the radar echo map sequences of the corresponding short future period to generate a first training set;
and inputting the first training set into the radar echo extrapolation model for training, and updating model parameters of the radar echo extrapolation model through a back propagation algorithm and an optimizer to obtain a trained radar echo extrapolation model.
Optionally, the fitting process of the precipitation intensity estimation model includes:
constructing a relation equation of radar reflectivity and precipitation intensity:
Z＝aR b
wherein Z, R is radar reflectivity and precipitation intensity, and a and b are coefficients to be solved;
constructing a relation equation of radar echo intensity and radar reflectivity:
dBZ＝10lgZ
wherein dBZ is radar echo intensity recorded by a radar echo diagram;
fitting a precipitation intensity estimation model according to a relation equation:
wherein R is precipitation intensity;
acquiring radar echo diagram sequences and precipitation intensity sequences of a target area in a plurality of historical time periods, and preprocessing the radar echo diagram sequences of the historical time periods;
substituting a plurality of precipitation intensity sequences of the historical time periods and the preprocessed radar echo diagram sequences into the precipitation intensity estimation model to solve coefficients a and b to be solved, and determining a precipitation intensity estimation model;
the radar echo diagram is fused through the attention mechanism, the rainfall intensity is corrected, and a final rainfall intensity estimation model is determined:
Q＝W q *dBZ T-1
K＝W k *dBZ T-1
V＝W v *dBZ T
in dBZ T-1 、dBZ T For time steps T, T-1Radar echo diagram, W q 、W k 、W v A weight matrix for the attention mechanism, Q, K, V is an input feature for the attention mechanism;is the Hadamard product; y is T The corrected precipitation intensity is the time step T.
Optionally, the preprocessing includes:
and carrying out interpolation processing on the radar echo map sequence by using a kriging interpolation method, and carrying out filtering processing on the radar echo map sequence subjected to interpolation processing by using Gaussian smoothing to obtain a preprocessed radar echo map sequence.
Optionally, the training process of the RNN stacking model includes:
obtaining rainfall intensity sequence predictions of target areas provided by various third parties in a plurality of historical time periods, and processing the rainfall intensity sequences on a spatial scale and a time scale;
acquiring precipitation intensity sequences of a plurality of long future time periods corresponding to the historical time periods, and integrating the prediction of the precipitation intensity sequences of the plurality of the historical time periods after the processing of the precipitation intensity sequences to generate a second training set;
and inputting the second training set into the RNN stacking model for training, and updating model parameters of the RNN stacking model through a back propagation algorithm and an optimizer to obtain a trained RNN stacking model.
Optionally, the splicing the prediction of the precipitation intensity sequence of the target area in the short future period and the final prediction of the precipitation intensity sequence in the long future period includes:
performing spatial interpolation processing on the rainfall intensity sequence prediction of the short future period by using an inverse distance weight method to obtain the rainfall intensity sequence prediction of the short future period on a target spatial scale;
performing spatial and temporal interpolation processing on the final prediction of the precipitation intensity sequence of the long future period by using an inverse distance weight method to obtain a final prediction of the precipitation intensity sequence of the long future period on a target spatial scale and a target time scale; the target time scale is a time scale predicted by a precipitation intensity sequence of the short future period;
and finally predicting the precipitation intensity sequence of the short future time period in the long future time period is replaced by the precipitation intensity sequence prediction of the short future time period on the target space scale and the target time scale to finish splicing.
In a second aspect, the present invention provides a short-term precipitation prediction device based on a multimodal set, the device comprising:
the echo acquisition module is used for acquiring a radar echo diagram sequence of a target area acquired by the radar in the current period and preprocessing the radar echo diagram sequence;
the echo prediction module is used for inputting the preprocessed radar echo graph sequence of the current time period into a trained radar echo extrapolation model to obtain a radar echo graph sequence of a short future time period;
the rainfall estimation module is used for inputting the radar echo diagram sequence of the short future period into a fitted rainfall intensity estimation model, and obtaining rainfall intensity sequence prediction of the target area in the short future period;
the multi-mode acquisition module is used for acquiring rainfall intensity sequence predictions of the target area provided by each third party in a long future period;
the rainfall prediction module is used for inputting the rainfall intensity sequence predictions of each long future period into a trained RNN stacking model, and obtaining the final prediction of the rainfall intensity sequence of the target area in the long future period;
and the precipitation prediction module is used for splicing the precipitation intensity sequence prediction of the target area in the short future period with the final prediction of the precipitation intensity sequence of the long future period to obtain the short-term precipitation intensity sequence prediction of the target area and predict the short-term precipitation intensity sequence.
In a third aspect, the present invention provides an electronic device, including a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is operative according to the instructions to perform steps according to the method described above.
In a fourth aspect, the present invention provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the above method.
Compared with the prior art, the invention has the beneficial effects that:
the method acquires a radar echo diagram through a radar, and acquires rainfall intensity prediction in a short future period based on a radar echo extrapolation model and a rainfall intensity estimation model; obtaining a rainfall intensity prediction of a long future period through a third party, and obtaining a final rainfall intensity prediction of the long future period based on an RNN stacking model; splicing the rainfall intensity sequence prediction of the target area in the short future period and the final prediction of the rainfall intensity sequence of the target area in the long future period to obtain the short-term rainfall intensity sequence prediction of the target area and forecast; different technical means are adopted in a long future period and a short future period, so that accurate prediction data of the corresponding period are obtained, more accurate prediction data are obtained after splicing, and the prediction accuracy is greatly improved compared with the prior art; the device, the equipment and the medium can realize the same technical effect by adopting the method.
Drawings
FIG. 1 is a flow chart of a short-term precipitation prediction method based on a multi-mode set provided by an embodiment of the invention;
FIG. 2 is a schematic illustration of correcting precipitation intensity by an attention mechanism provided by an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a radar echo extrapolation model according to an embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for more clearly illustrating the technical aspects of the present invention, and are not intended to limit the scope of the present invention.
Embodiment one:
as shown in fig. 1, the invention provides a short-term precipitation prediction method based on a multi-mode set, which comprises the following steps:
s1, acquiring a radar echo diagram sequence of a target area acquired by a radar in a current period, and preprocessing;
in this embodiment, the preprocessing includes: and carrying out interpolation processing on the radar echo map sequence by using a kriging interpolation method, and carrying out filtering processing on the radar echo map sequence subjected to interpolation processing by using Gaussian smoothing to obtain a preprocessed radar echo map sequence.
S2, inputting the preprocessed radar echo map sequence in the current time period into a trained radar echo extrapolation model, and obtaining a radar echo map sequence in a short future time period;
the current time period and the short future time period can be flexibly set according to actual requirements, for example, the current time period is set to be 2 hours before the current time, and the short future time period is set to be 2 hours after the current time.
In this embodiment, the training process of the radar echo extrapolation model includes:
acquiring radar echo diagram sequences of a target area in a plurality of historical time periods, and preprocessing;
extracting radar echo diagram sequences of short future time periods corresponding to the radar echo diagram sequences of each history time period from the preprocessed radar echo diagram sequences of a plurality of history time periods; for example, the plurality of history periods are respectively continuous history periods 1 to 10, each history period is 2 hours, and the short future period is also 2 hours, then the history period 2 can be regarded as the short future period of the history period 1;
combining the radar echo map sequences of each historical period and the radar echo map sequences of the corresponding short future period to generate a first training set;
and inputting the first training set into a radar echo extrapolation model for training, and updating model parameters of the radar echo extrapolation model through a back propagation algorithm and an optimizer (e.g. an Adam optimizer) to obtain a trained radar echo extrapolation model.
S3, inputting the radar echo diagram sequence of the short future period into a fitted rainfall intensity estimation model, and obtaining rainfall intensity sequence prediction of the target area in the short future period;
in this embodiment, the fitting process of the precipitation intensity estimation model includes:
constructing a relation equation of radar reflectivity and precipitation intensity:
Z＝aR b
wherein Z, R is radar reflectivity and precipitation intensity, and a and b are coefficients to be solved;
constructing a relation equation of radar echo intensity and radar reflectivity:
dBZ＝10lgZ
wherein dBZ is radar echo intensity recorded by a radar echo diagram;
fitting a precipitation intensity estimation model according to a relation equation:
wherein R is precipitation intensity;
acquiring radar echo diagram sequences and precipitation intensity sequences of a target area in a plurality of historical time periods, and preprocessing the radar echo diagram sequences in the plurality of historical time periods;
substituting the rainfall intensity sequences of a plurality of historical time periods and the preprocessed radar echo diagram sequences into a rainfall intensity estimation model to solve coefficients a and b to be solved, and determining the rainfall intensity estimation model;
as shown in fig. 2, the radar echo diagram is fused through the attention mechanism, the rainfall intensity is corrected, and a final rainfall intensity estimation model is determined:
Q＝W q *dBZ T-1
K＝W k *dBZ T-1
V＝W v *dBZ T
in dBZ T-1 、dBZ T Radar echo diagram for time step T, T-1，W q 、W k 、W v A weight matrix for the attention mechanism, Q, K, V is an input feature for the attention mechanism;is the Hadamard product; y is T The corrected precipitation intensity is the time step T.
S4, obtaining rainfall intensity sequence predictions of target areas provided by all third parties in a long future period;
the long future period can be flexibly set according to actual requirements, for example, the long future period is 72 hours after the current moment.
The third party forecasting products are many, and the embodiment adopts EC European mode data and GFS mode data, wherein the EC European mode data is called as a European center numerical forecasting mode, the English is called as European Centre for Medium-Range Weather Forecasts (ECMWF), and the EC European mode data is developed and maintained by the European center numerical forecasting mode. ECMWF is one of the most representative numerical forecasting modes worldwide, and is widely applied to the fields of weather forecasting, climate research and the like. The GFS mode data, collectively referred to as "global forecast system mode", and the english, collectively referred to as "Global Forecast System", was developed and maintained by the united states national marine and atmospheric administration (NOAA). The GFS mode is one of the oldest and most widely used numerical forecasting modes worldwide, and is widely applied to the fields of weather forecasting, wind energy forecasting, environmental pollution forecasting and the like
S5, inputting the rainfall intensity sequence predictions of each long future period into a trained RNN stacking model, and obtaining the final prediction of the rainfall intensity sequence of the target area in the long future period;
in this embodiment, the training process of the RNN stacking model includes:
obtaining rainfall intensity sequence predictions of target areas provided by various third parties in a plurality of historical time periods, and processing the rainfall intensity sequences on a spatial scale and a time scale;
acquiring the rainfall intensity sequences of the long future time periods corresponding to the plurality of historical time periods (although the rainfall intensity sequences of the target areas provided by the third parties in the plurality of historical time periods are different in prediction, the rainfall intensity sequences in the plurality of historical time periods are practically unique), and generating a second training set by integrating the rainfall intensity sequences of the plurality of historical time periods after the rainfall intensity sequences are processed;
and inputting the second training set into the RNN stacking model for training, and updating model parameters of the RNN stacking model through a back propagation algorithm and an optimizer to obtain a trained RNN stacking model.
S6, splicing the rainfall intensity sequence prediction of the target area in the short future period and the rainfall intensity sequence final prediction of the target area in the long future period to obtain the short-term rainfall intensity sequence prediction of the target area and forecast;
in this embodiment, the splicing process includes:
performing spatial interpolation processing on the rainfall intensity sequence prediction of the short future time period by using an inverse distance weight method to obtain the rainfall intensity sequence prediction of the short future time period on a target spatial scale;
performing spatial and temporal interpolation processing on the final prediction of the precipitation intensity sequence of the long future period by using an inverse distance weight method to obtain the final prediction of the precipitation intensity sequence of the long future period on a target spatial scale and a target time scale; the target time scale is the time scale of rainfall intensity sequence prediction of a short future period;
and finally predicting the precipitation intensity sequence of the short future time period in the long future time period is replaced by predicting the precipitation intensity sequence of the short future time period on the target space scale and the target time scale to finish splicing.
For example, the final prediction of the precipitation intensity sequence 2 hours after the current time in the long future period (72 hours after the current time) is replaced by the prediction of the precipitation intensity sequence in the short future period (2 hours after the current time) to complete the splice.
As shown in fig. 3, in the present embodiment, the radar echo extrapolation model includes a convolution layer, a max-pooling layer, a seq2seq model, and a deconvolution layer; the seq2seq model comprises an encoder and a decoder, wherein the encoder and the decoder comprise a plurality of stacked RainNet modules, and the RainNet modules comprise ConvLSTM modules and CBAM convolution attention modules; the CBAM convolution attention module includes a spatial attention module and a channel attention module.
The input image is extracted and processed by the convolution layer and the maximum pooling layer to obtain characteristic data suitable for the input form of the ConvLSTM module, the ConvLSTM module extracts and calculates the input characteristic, the space attention module (Spatial Attention Module) and the channel attention module (Channel Attention Module) are combined, the CBAM convolution attention module is a lightweight module, the additional expenditure generated by adding the module into a network is very small, good effects are obtained on a plurality of data sets and a plurality of tasks, and the ConvLSTM module has good applicability. The output characteristics obtained by the ConvLSTM module are processed by the CBAM convolution attention module, the channel attention module in the CBAM firstly carries out weight distribution on the data input by the ConvLSTM module to obtain characteristic data suitable for processing by the space attention module, and the space attention module carries out calculation processing on the input characteristic data and outputs the characteristic data. Since the RainNet model is finally required to generate prediction data, a prediction image consistent with the original image in size needs to be generated, and operations such as rolling and pooling tend to reduce the image size, so that an output image is restored to the same size as the original image through an upsampling operation, the size of the output data is restored through a specific operation by the output of the spatial attention module, the function can be just realized by the deconvolution operation, and finally the feature data meeting the requirements is output.
The method comprises the steps of dividing the method into two parts of a coding network and a prediction network through an encoder-decoder (seq 2 seq), wherein the initial state and the unit output of the prediction network are copied from the final state of the coding network; the coding network compresses the whole sequence input into the network into a hidden state, and the prediction network expands the hidden state and gives out the final prediction result.
Embodiment two:
the embodiment of the invention provides a short-term precipitation forecasting device based on a multi-mode set, which comprises the following components:
the echo acquisition module is used for acquiring a radar echo diagram sequence of a target area acquired by the radar in the current period and preprocessing the radar echo diagram sequence;
the echo prediction module is used for inputting the preprocessed radar echo graph sequence in the current time period into a trained radar echo extrapolation model to obtain a radar echo graph sequence in a short future time period;
the rainfall estimation module is used for inputting the radar echo diagram sequence of the short future period into a fitted rainfall intensity estimation model, and obtaining rainfall intensity sequence prediction of the target area in the short future period;
the multi-mode acquisition module is used for acquiring rainfall intensity sequence predictions of the target area provided by each third party in a long future period;
the rainfall prediction module is used for inputting the rainfall intensity sequence predictions of each long future period into a trained RNN stacking model, and obtaining the final prediction of the rainfall intensity sequence of the target area in the long future period;
and the precipitation prediction module is used for splicing the precipitation intensity sequence prediction of the target area in the short future period with the final prediction of the precipitation intensity sequence of the target area in the long future period to obtain the short-term precipitation intensity sequence prediction of the target area and predict the short-term precipitation intensity sequence.
Embodiment III:
based on the first embodiment, the embodiment of the invention provides an electronic device, which is characterized by comprising a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is operative according to the instructions to perform steps according to the method described above.
Embodiment four:
based on the first embodiment, the embodiment of the present invention provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the steps of the above method.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing is merely a preferred embodiment of the present invention, and it should be noted that modifications and variations could be made by those skilled in the art without departing from the technical principles of the present invention, and such modifications and variations should also be regarded as being within the scope of the invention.
Claims (10)
1. A short-term precipitation prediction method based on a multimode set, comprising:
acquiring a radar echo diagram sequence of a target area acquired by a radar in a current period, and preprocessing;
inputting the preprocessed radar echo graph sequence of the current time period into a trained radar echo extrapolation model to obtain a radar echo graph sequence of a short future time period;
inputting the radar echo diagram sequence of the short future period into a fitted rainfall intensity estimation model, and obtaining rainfall intensity sequence prediction of a target area in the short future period;
obtaining rainfall intensity sequence predictions of target areas provided by various third parties in a long future period;
inputting the rainfall intensity sequence predictions of each long future period into a trained RNN stacking model, and obtaining the final prediction of the rainfall intensity sequence of the target area in the long future period;
and splicing the rainfall intensity sequence prediction of the target area in the short future period and the final rainfall intensity sequence prediction of the target area in the long future period to obtain the short-term rainfall intensity sequence prediction of the target area and forecast.
2. The short-term precipitation prediction method based on a multimode set according to claim 1, wherein the radar echo extrapolation model comprises a convolution layer, a max pooling layer, a seq2seq model and an inverse convolution layer; the seq2seq model includes an encoder and a decoder, each of the encoder and the decoder including a plurality of stacked rain net modules including a convls (tm) module and a CBAM convolution attention module; the CBAM convolution attention module includes a spatial attention module and a channel attention module.
3. The short-term precipitation prediction method based on a multi-mode set according to claim 1, wherein the training process of the radar echo extrapolation model comprises:
acquiring radar echo diagram sequences of a target area in a plurality of historical time periods, and preprocessing;
extracting radar echo diagram sequences of short future time periods corresponding to the radar echo diagram sequences of each history time period from the preprocessed radar echo diagram sequences of a plurality of the history time periods;
combining the radar echo map sequences of each historical period and the radar echo map sequences of the corresponding short future period to generate a first training set;
and inputting the first training set into the radar echo extrapolation model for training, and updating model parameters of the radar echo extrapolation model through a back propagation algorithm and an optimizer to obtain a trained radar echo extrapolation model.
4. The short-term precipitation prediction method based on a multi-mode set according to claim 1, wherein the fitting process of the precipitation intensity estimation model comprises:
constructing a relation equation of radar reflectivity and precipitation intensity:
Z＝aR b
wherein Z, R is radar reflectivity and precipitation intensity, and a and b are coefficients to be solved;
constructing a relation equation of radar echo intensity and radar reflectivity:
dBZ＝10lgZ
wherein dBZ is radar echo intensity recorded by a radar echo diagram;
fitting a precipitation intensity estimation model according to a relation equation:
wherein R is precipitation intensity;
acquiring radar echo diagram sequences and precipitation intensity sequences of a target area in a plurality of historical time periods, and preprocessing the radar echo diagram sequences of the historical time periods;
substituting a plurality of precipitation intensity sequences of the historical time periods and the preprocessed radar echo diagram sequences into the precipitation intensity estimation model to solve coefficients a and b to be solved, and determining a precipitation intensity estimation model;
the radar echo diagram is fused through the attention mechanism, the rainfall intensity is corrected, and a final rainfall intensity estimation model is determined:
Q＝W q *dBZ T-1
K＝W k *dBZ T-1
V＝W v *dBZ T
in dBZ T-1 、dBZ T Radar echo diagram for time steps T, T-1, W q 、W k 、W v A weight matrix for the attention mechanism, Q, K, V is an input feature for the attention mechanism; DEG is Hadamard product; y is T The corrected precipitation intensity is the time step T.
5. The short-term precipitation prediction method based on a multi-mode set according to claim 1, wherein the preprocessing comprises:
and carrying out interpolation processing on the radar echo map sequence by using a kriging interpolation method, and carrying out filtering processing on the radar echo map sequence subjected to interpolation processing by using Gaussian smoothing to obtain a preprocessed radar echo map sequence.
6. The short-term precipitation prediction method based on a multimodal set according to claim 1, wherein the training process of the RNN stack model comprises:
obtaining rainfall intensity sequence predictions of target areas provided by various third parties in a plurality of historical time periods, and processing the rainfall intensity sequences on a spatial scale and a time scale;
acquiring precipitation intensity sequences of a plurality of long future time periods corresponding to the historical time periods, and integrating the prediction of the precipitation intensity sequences of the plurality of the historical time periods after the processing of the precipitation intensity sequences to generate a second training set;
and inputting the second training set into the RNN stacking model for training, and updating model parameters of the RNN stacking model through a back propagation algorithm and an optimizer to obtain a trained RNN stacking model.
7. The method of claim 1, wherein the concatenating the target zone's prediction of the sequence of precipitation intensities for the short future period and the final prediction of the sequence of precipitation intensities for the long future period comprises:
performing spatial interpolation processing on the rainfall intensity sequence prediction of the short future period by using an inverse distance weight method to obtain the rainfall intensity sequence prediction of the short future period on a target spatial scale;
performing spatial and temporal interpolation processing on the final prediction of the precipitation intensity sequence of the long future period by using an inverse distance weight method to obtain a final prediction of the precipitation intensity sequence of the long future period on a target spatial scale and a target time scale; the target time scale is a time scale predicted by a precipitation intensity sequence of the short future period;
and finally predicting the precipitation intensity sequence of the short future time period in the long future time period is replaced by the precipitation intensity sequence prediction of the short future time period on the target space scale and the target time scale to finish splicing.
8. A short-term precipitation forecasting device based on a multi-mode set, the device comprising:
the echo acquisition module is used for acquiring a radar echo diagram sequence of a target area acquired by the radar in the current period and preprocessing the radar echo diagram sequence;
the echo prediction module is used for inputting the preprocessed radar echo graph sequence of the current time period into a trained radar echo extrapolation model to obtain a radar echo graph sequence of a short future time period;
the rainfall estimation module is used for inputting the radar echo diagram sequence of the short future period into a fitted rainfall intensity estimation model, and obtaining rainfall intensity sequence prediction of the target area in the short future period;
the multi-mode acquisition module is used for acquiring rainfall intensity sequence predictions of the target area provided by each third party in a long future period;
the rainfall prediction module is used for inputting the rainfall intensity sequence predictions of each long future period into a trained RNN stacking model, and obtaining the final prediction of the rainfall intensity sequence of the target area in the long future period;
and the precipitation prediction module is used for splicing the precipitation intensity sequence prediction of the target area in the short future period with the final prediction of the precipitation intensity sequence of the long future period to obtain the short-term precipitation intensity sequence prediction of the target area and predict the short-term precipitation intensity sequence.
9. An electronic device, comprising a processor and a storage medium;
the storage medium is used for storing instructions;
the processor being operative according to the instructions to perform the steps of the method according to any one of claims 1-7.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310924818.1A CN116931129A (en) | 2023-07-25 | 2023-07-25 | Short-term precipitation prediction method, device, equipment and medium based on multi-mode set |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310924818.1A CN116931129A (en) | 2023-07-25 | 2023-07-25 | Short-term precipitation prediction method, device, equipment and medium based on multi-mode set |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116931129A true CN116931129A (en) | 2023-10-24 |
Family
ID=88393916
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310924818.1A Pending CN116931129A (en) | 2023-07-25 | 2023-07-25 | Short-term precipitation prediction method, device, equipment and medium based on multi-mode set |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116931129A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117950088A (en) * | 2024-03-26 | 2024-04-30 | 南京满星数据科技有限公司 | Multi-mode-based precipitation prediction data fusion correction method |
-
2023
- 2023-07-25 CN CN202310924818.1A patent/CN116931129A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117950088A (en) * | 2024-03-26 | 2024-04-30 | 南京满星数据科技有限公司 | Multi-mode-based precipitation prediction data fusion correction method |
| CN117950088B (en) * | 2024-03-26 | 2024-06-04 | 南京满星数据科技有限公司 | Multi-mode-based precipitation prediction data fusion correction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114254561B (en) | Waterlogging prediction method, system and storage medium | |
| CN110363327A (en) | Short based on ConvLSTM and 3D-CNN faces Prediction of Precipitation method | |
| CN111737640B (en) | Water level prediction method, device and computer readable storage medium | |
| CN113255972B (en) | Short-term rainfall prediction method based on Attention mechanism | |
| CN116931129A (en) | Short-term precipitation prediction method, device, equipment and medium based on multi-mode set | |
| CN114493052B (en) | Multi-model fusion self-adaptive new energy power prediction method and system | |
| CN116449462B (en) | Method, system, storage medium and equipment for predicting effective wave height space-time sequence of sea wave | |
| CN116720156A (en) | Weather element forecasting method based on graph neural network multi-mode weather data fusion | |
| CN110909447A (en) | High-precision short-term prediction method for ionization layer region | |
| CN117096875A (en) | Short-term load prediction method and system based on ST-transducer model | |
| CN117008217A (en) | Short-time strong precipitation prediction method integrating self-attention module and Unet model | |
| Wang et al. | Customized deep learning for precipitation bias correction and downscaling | |
| CN116152206A (en) | Photovoltaic output power prediction method, terminal equipment and storage medium | |
| CN115358060A (en) | Efficient algorithm framework supporting high-precision prediction of new energy power generation power | |
| CN113984198B (en) | Shortwave radiation prediction method and system based on convolutional neural network | |
| CN117391139A (en) | Weather phenomenon prediction correction method based on improved UNet neural network | |
| CN117313990A (en) | Air conditioner load prediction method and system based on meta-integrated learning and electronic equipment | |
| CN116911178A (en) | Method and system for predicting capacity of small and medium-sized reservoirs based on weather forecast | |
| CN112488413A (en) | AWA-DRCN-based population spatialization method | |
| CN116596169A (en) | Power system prediction method, device and storage medium | |
| CN114169646B (en) | Water bloom prediction method, device, electronic equipment and computer readable storage medium | |
| Li et al. | Deepphysinet: Bridging deep learning and atmospheric physics for accurate and continuous weather modeling | |
| CN113962432A (en) | Wind power prediction method and system integrating three-dimensional convolution and light-weight convolution threshold unit | |
| CN117421601B (en) | Sea surface evaporation waveguide near-future rapid forecasting method | |
| JP3245238U (en) | Prediction correction system equipped with a learning device and a prediction correction device |
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 |