CN-121995546-A - Method, device and equipment for autonomous prediction of el Nino power

CN121995546ACN 121995546 ACN121995546 ACN 121995546ACN-121995546-A

Abstract

The application is suitable for the field of weather prediction, and provides an automatic prediction method, device and equipment for the dynamic of el Nino, wherein the method comprises the steps of acquiring historical analysis data and target surface sea temperature observation data; the method comprises the steps of performing space-time matching on target surface sea temperature observation data and historical analysis data to construct a target matching data set, pre-training a first neural network model to establish an initial mapping relation between surface sea temperature and deep sea temperature, performing migration learning on the pre-trained first neural network model to obtain a second neural network model, inputting the target surface sea temperature observation data into the second neural network model to reconstruct target three-dimensional sea temperature data, generating a target initial field based on an assimilation module of an el Nino power autonomous prediction model, driving the el Nino power autonomous prediction model to perform prediction integration by using the target initial field, and outputting an el Nino prediction result. By the method, the automation of the el Nino prediction is realized.

Inventors

SONG CHUNYANG
JIANG HUA
HE YUE
CHEN XINGRONG
TAN JING
HUANG YONGYONG

Assignees

国家海洋环境预报中心

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. An el nino power autonomy prediction method, comprising the steps of: Acquiring historical analysis data and target surface sea temperature observation data, wherein the target surface sea temperature observation data are acquired by domestic satellites; Performing space-time matching on the target surface sea temperature observation data and the historical analysis data to construct a target matching data set; Pre-training a first neural network model according to the historical analysis data, and establishing an initial mapping relation between the surface sea temperature and the deep sea temperature; Performing migration learning on the pre-trained first neural network model according to the target matching data set to obtain a second neural network model adapting to the target surface sea temperature observation data; inputting the target surface sea temperature observation data into the second neural network model to obtain corresponding target deep sea Wen Fanyan data, and reconstructing target three-dimensional sea temperature data; Assimilating the target three-dimensional sea temperature data into a sea mode based on an assimilation module of the el Nino power autonomous prediction model to generate a target initial field; and driving the el Nino power autonomous prediction model to perform prediction integration by using the target initial field, and outputting el Nino prediction result.
2. The method for autonomous prediction of el nino power according to claim 1, wherein the acquiring of the historical analysis data and the target skin sea temperature observation data comprises the steps of: GODAS analysis data of a first historical time period are obtained, and surface sea temperature observation data of a second historical time period are obtained; preprocessing the surface sea temperature observation data in the second historical time period, wherein the preprocessing comprises, but is not limited to, format unification and abnormal rejection; and interpolating and complementing the pretreated surface sea temperature observation data according to the GODAS re-analysis data to obtain the target surface sea temperature observation data.
3. The method for autonomous prediction of el nino power according to claim 1, wherein the pre-training the first neural network model according to the historical analysis data, and establishing an initial mapping relationship between the surface sea temperature and the deep sea temperature, comprises the steps of: Initializing a first neural network model based on a U-Net network architecture; and pre-training a first neural network model based on the U-Net network architecture according to the historical analysis data, and establishing an initial mapping relation between the surface sea temperature and the deep sea temperature, wherein the loss function is mean square error, and the optimizer is Adam.
4. The method for el nino power autonomy prediction according to any one of claims 1 to 3, wherein the assimilation module based on el nino power autonomy prediction model assimilates the target three-dimensional sea temperature data into a sea mode, and further comprises the steps of, before generating a target initial field: taking an initial field of a 20 th century climate state as a driving condition, and controlling the el Nino power autonomous prediction model to perform free integral operation for 50 years; Monitoring energy exchange flux between an atmospheric component and an upper ocean component in the el Nino power autonomous prediction model; When the fluctuation standard deviation of the energy exchange flux is converged in a preset threshold range, judging that the inside of the el Nino power autonomous prediction model reaches a physical balance state, activating the assimilation module, and inputting the target three-dimensional sea temperature data into the el Nino power autonomous prediction model.
5. The method for el nino power autonomy prediction according to any one of claims 1 to 3, wherein the assimilation module based on el nino power autonomy prediction model assimilates the target three-dimensional sea temperature data into a sea mode to generate a target initial field, comprising the steps of: and an assimilation module based on an el Nino power autonomous prediction model takes the three-dimensional sea temperature data of the target as an observation increment, assimilates the observation increment into a background field of the ocean mode in a Newton relaxation approximation mode, dynamically adjusts a relaxation coefficient according to an observation error covariance and a background error covariance, and generates the initial field of the target.
6. The method for el nino power autonomy prediction according to any one of claims 1 to 3, wherein the driving the el nino power autonomy prediction model with the target initial field performs prediction integration, and outputs el nino prediction results, comprising the steps of: And driving the el Nino power autonomous prediction model to perform prediction integration by taking the target initial field as an initial condition, and outputting a multi-element space-time evolution sequence to obtain an el Nino prediction result, wherein the integration period covers 12 months in the future.
7. The method for el nino power autonomy prediction according to any one of claims 1 to 3, further comprising the step of: acquiring historical el nino observation data; Performing space-time matching on the el Nino prediction result and historical el Nino observation data, and calculating a prediction error; And evaluating the prediction effect according to the prediction error.
8. An el nino power autonomy prediction device, comprising: The system comprises an acquisition unit, a control unit and a control unit, wherein the acquisition unit is used for acquiring historical analysis data and target surface sea temperature observation data, wherein the target surface sea temperature observation data are acquired by domestic satellites; the first processing unit is used for performing space-time matching on the target surface sea temperature observation data and the historical analysis data to construct a target matching data set; the second processing unit is used for pre-training the first neural network model according to the historical analysis data and establishing an initial mapping relation between the surface sea temperature and the deep sea temperature; The third processing unit is used for performing migration learning on the pre-trained first neural network model according to the target matching data set to obtain a second neural network model adapting to the target surface sea temperature observation data; The fourth processing unit is used for inputting the target surface sea temperature observation data into the second neural network model to obtain corresponding target deep sea Wen Fanyan data and reconstructing target three-dimensional sea temperature data; The fifth processing unit is used for assimilating the target three-dimensional sea temperature data into a sea mode based on an assimilation module of the el Nino power autonomous prediction model to generate a target initial field; And the sixth processing unit is used for driving the el Nino power autonomous prediction model to perform prediction integration by the target initial field and outputting el Nino prediction result.
9. An el nino power autonomy prediction device comprising a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the steps of the method according to any one of claims 1 to 7 are carried out by the processor when the computer program is executed.
10. A computer readable storage medium storing a computer program, which when executed by a processor implements the method of any one of claims 1 to 7.

Description

Method, device and equipment for autonomous prediction of el Nino power Technical Field The application belongs to the technical field of weather prediction, and particularly relates to an automatic prediction method, device and equipment for el Nino power. Background Since the eighties of the last century, the study of the el nino/lanina phenomenon has been the focus of the field of marine and atmospheric science. The existing satellite remote sensing technology has limitation in acquiring the temperature data from the subsurface layer to the deep layer of the ocean, so that the accuracy of the traditional estimation method is insufficient. This problem makes it difficult to achieve high accuracy in the monitoring and prediction of the el nino phenomenon. The early-stage el nino monitoring and predicting system mainly depends on an analysis data set released abroad and lacks an independent and reliable data source. The dependence makes China face the risk of data acquisition when dealing with international relation changes, and influences the autonomy and safety of prediction. The existing prediction model has insufficient inversion capability on deep sea temperature, and can not fully utilize target surface sea temperature observation data acquired by domestic satellites, so that the adaptability and accuracy of the model are reduced. The data assimilation module of the current el nino prediction model has limited processing capability on new data, and is difficult to effectively integrate data from different sources, especially domestic satellite data, so that the construction of an initial field and the accuracy of a prediction result are affected. Disclosure of Invention The embodiment of the application provides an el Nino power autonomous prediction method, an el Nino power autonomous prediction device and el Nino power autonomous prediction equipment, which can solve the problems. In a first aspect, an embodiment of the present application provides a method for autonomous prediction of el nino power, including: Acquiring historical analysis data and target surface sea temperature observation data, wherein the target surface sea temperature observation data are acquired by domestic satellites; Performing space-time matching on the target surface sea temperature observation data and the historical analysis data to construct a target matching data set; Pre-training a first neural network model according to the historical analysis data, and establishing an initial mapping relation between the surface sea temperature and the deep sea temperature; Performing migration learning on the pre-trained first neural network model according to the target matching data set to obtain a second neural network model adapting to the target surface sea temperature observation data; inputting the target surface sea temperature observation data into the second neural network model to obtain corresponding target deep sea Wen Fanyan data, and reconstructing target three-dimensional sea temperature data; Assimilating the target three-dimensional sea temperature data into a sea mode based on an assimilation module of the el Nino power autonomous prediction model to generate a target initial field; and driving the el Nino power autonomous prediction model to perform prediction integration by using the target initial field, and outputting el Nino prediction result. In a second aspect, an embodiment of the present application provides an el nino power autonomy prediction device, including: The system comprises an acquisition unit, a control unit and a control unit, wherein the acquisition unit is used for acquiring historical analysis data and target surface sea temperature observation data, wherein the target surface sea temperature observation data are acquired by domestic satellites; the first processing unit is used for performing space-time matching on the target surface sea temperature observation data and the historical analysis data to construct a target matching data set; the second processing unit is used for pre-training the first neural network model according to the historical analysis data and establishing an initial mapping relation between the surface sea temperature and the deep sea temperature; The third processing unit is used for performing migration learning on the pre-trained first neural network model according to the target matching data set to obtain a second neural network model adapting to the target surface sea temperature observation data; The fourth processing unit is used for inputting the target surface sea temperature observation data into the second neural network model to obtain corresponding target deep sea Wen Fanyan data and reconstructing target three-dimensional sea temperature data; The fifth processing unit is used for assimilating the target three-dimensional sea temperature data into a sea mode based on an assimilation module of the el Nino power autonomous prediction model to generate