El Niño-Southern Oscillation prediction method and device

Abstract

Disclosed are an El Niño-Southern Oscillation prediction method and device. The method comprises: inputting a meteorological feature sequence of a target sea area in a target historical time period into an adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence, wherein the intermediate feature sequence includes a spatial dependency relationship between nodes of each hypergraph structure in a plurality of hypergraph structures, and each hypergraph structure is obtained based on a feature sequence of the meteorological feature sequence at a corresponding grid scale in a plurality of grid scales; obtaining an ocean Niño index for the target historical time period based on the intermediate feature sequence; inputting the ocean Niño index for the target historical time period and a target future time period into a frequency domain multi-scale time encoding and decoding module to obtain an ocean Niño index for the target sea area in the target future time period, wherein the ocean Niño index for the target future time period represents the extent to which the El Niño-Southern Oscillation phenomenon occurs in the target future time period.

Description

Technical Field

[0001] The present disclosure generally relates to the field of meteorological forecasting technology, and more specifically, to an El Niño-Southern Oscillation prediction method and apparatus. Background Art

[0002] The El Niño–Southern Oscillation (ENSO) cycle, characterized by fluctuations between unusually warm (El Niño) and cold (La Niña) conditions in the tropical Pacific Ocean, is Earth's most prominent interannual climate change. The term El Niño refers to the warming of the tropical Pacific Ocean that occurs every two to seven years, while the opposite cold phase is known as La Niña. The unusual warming or cooling conditions are associated with a large-scale, east-west seesaw in sea level pressure, known as the Southern Oscillation, which represents the atmospheric manifestation of the coupled ENSO phenomenon. ENSO originates in the tropical Pacific Ocean, but its impacts extend beyond regional climates and can trigger large-scale weather and climate anomalies worldwide. For example, ENSO can trigger droughts in Southeast Asia, heavy rains along the west coast of South America, and frequent extreme weather events in the Northern Hemisphere, directly impacting agricultural yields, energy demand, ecosystem balance, and the sustainable development of human society. ENSO predictions offer humanity an opportunity to prevent climate anomalies, potentially reducing the societal and economic impacts of this natural phenomenon and assisting in the management of natural resources and the environment.

[0003] Current ENSO prediction methods generally perform poorly in long-term prediction scenarios, with prediction accuracy decreasing as the prediction duration increases. In addition, the "Spring Prediction Barrier (SPB)" problem causes the prediction accuracy in spring months to be generally lower than that in other months. Summary of the Invention

[0004] The embodiments of the present disclosure provide an El Niño-Southern Oscillation prediction method and apparatus, which can effectively solve the problem in the prior art that the prediction accuracy of ENSO prediction decreases with increasing prediction time.

[0005] In a general aspect, a method for predicting the El Niño-Southern Oscillation is provided, comprising: inputting a meteorological feature sequence of a target sea area in a target historical time period into an adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence, wherein the target historical time period is a time period of a first predetermined length before a current moment, and the intermediate feature sequence comprises a spatial dependency relationship between nodes of each hypergraph structure in a plurality of hypergraph structures, each hypergraph structure being obtained based on a feature sequence of the meteorological feature sequence at a corresponding grid scale in a plurality of grid scales; obtaining an ocean Niño index for the target historical time period based on the intermediate feature sequence; inputting the ocean Niño index for the target historical time period and a target future time period into a frequency domain multi-scale time encoding and decoding module to obtain an ocean Niño index for the target sea area in a target future time period, wherein the target future time period is a time period of a second predetermined length after the current moment, and the ocean Niño index for the target future time period characterizes the extent of the occurrence of the El Niño-Southern Oscillation phenomenon in the target future time period.

[0006] Optionally, the adaptive multi-scale spatial hypergraph module includes a multi-scale data processing unit and multiple parallel network blocks, and the multiple grid scales correspond one-to-one to the multiple parallel network blocks, wherein the meteorological feature sequence of the target sea area in the target historical time period is input into the adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence, including: inputting the meteorological feature sequence of the target historical time period into the multi-scale data processing unit to obtain a feature sequence of the meteorological feature sequence at multiple grid scales; for the feature sequence at each grid scale in the multiple grid scales, inputting the current feature sequence into the network block corresponding to the current feature sequence to obtain the hidden features of the current feature sequence, wherein the hidden features include the spatial dependency between nodes in the hypergraph structure of the current feature sequence; and determining the hidden features of the feature sequences at multiple grid scales as the intermediate feature sequence.

[0007] Optionally, the network block includes a spatial graph construction unit and a hypergraph information aggregation unit, wherein the current feature sequence is input into the network block corresponding to the current feature sequence to obtain the hidden features of the current feature sequence, including: inputting the current feature sequence into the spatial graph construction unit to obtain the hypergraph structure of the current feature sequence, wherein the hypergraph structure of the current feature sequence includes multiple hypergraph structures of preset durations in the target historical time period; inputting the hypergraph structure into the hypergraph information aggregation unit to obtain the hidden features of the current feature sequence.

[0008] Optionally, the frequency domain multi-scale time coding and decoding module includes a frequency domain embedding unit, a frequency domain coding unit, a frequency domain decoding unit and a frequency domain fusion unit, wherein the ocean Niño index of the target historical time period and the target future time period are input into the frequency domain multi-scale time coding and decoding module to obtain the ocean Niño index of the target sea area in the target future time period, including: inputting the ocean Niño index of the target historical time period into the frequency domain embedding unit to obtain the time characteristics of different frequency segments; inputting the time characteristics of different frequency segments into the frequency domain coding unit to obtain coding characteristics; inputting the coding characteristics and the target future time period into the frequency domain decoding unit to obtain the predicted characteristics of the future frequency segment; inputting the predicted characteristics of the future frequency segment into the frequency domain fusion unit to obtain the ocean Niño index of the target sea area in the target future time period.

[0009] Optionally, the adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale time encoding and decoding module are trained in the following manner: based on the meteorological feature sequence and the ocean Niño index of each meteorological feature among the multiple meteorological features of the global sea surface, a training sample set is obtained, wherein each training sample includes a meteorological feature sequence of a first predetermined time length and a real ocean Niño index of a second predetermined time length after the first predetermined time length; for each training sample, the following processing is performed: the meteorological feature sequence of the first predetermined time length in the current training sample is input into the adaptive multi-scale spatial hypergraph module to obtain an estimated intermediate feature sequence; based on An estimated intermediate feature sequence is used to obtain a first estimated ocean Niño index, wherein the first estimated ocean Niño index is an estimated ocean Niño index for a first predetermined time length; the first estimated ocean Niño index and a second predetermined time length after the first predetermined time length are input into a frequency domain multi-scale time encoding and decoding module to obtain a second estimated ocean Niño index, wherein the second estimated ocean Niño index is an estimated ocean Niño index for a second predetermined time length after the first predetermined time length; based on the second estimated ocean Niño index and the true ocean Niño index, parameters of the adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale time encoding and decoding module are adjusted.

[0010] Optionally, a training sample set is obtained based on the meteorological feature sequence and ocean Niño index of each meteorological feature among multiple meteorological features of the global sea surface, including: obtaining multiple meteorological features of the global sea surface; preprocessing the multiple meteorological features of the global sea surface to obtain the meteorological feature sequence and ocean Niño index of each meteorological feature; performing sequential sliding sampling on the meteorological feature sequence and ocean Niño index of each meteorological feature; and obtaining a training sample set based on the sampling results.

[0011] In another general aspect, an El Niño-Southern Oscillation prediction device is provided, comprising: an extraction unit configured to input a meteorological feature sequence of a target sea area in a target historical time period into an adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence, wherein the target historical time period is a time period of a first predetermined length before a current moment, the intermediate feature sequence comprises a spatial dependency relationship between nodes of each hypergraph structure in a plurality of hypergraph structures, and each hypergraph structure is obtained based on a feature sequence of the meteorological feature sequence at a corresponding grid scale in a plurality of grid scales; an acquisition unit configured to obtain an ocean Niño index of the target historical time period based on the intermediate feature sequence; and a prediction unit configured to input the ocean Niño index of the target historical time period and a target future time period into a frequency domain multi-scale time encoding and decoding module to obtain an ocean Niño index of the target sea area in a target future time period, wherein the target future time period is a time period of a second predetermined length after the current moment, and the ocean Niño index of the target future time period characterizes the extent of the El Niño-Southern Oscillation phenomenon occurring in the target future time period.

[0012] Optionally, the adaptive multi-scale spatial hypergraph module includes a multi-scale data processing unit and multiple parallel network blocks, and the multiple grid scales correspond one-to-one to the multiple parallel network blocks, wherein the extraction unit is further configured to input the meteorological feature sequence of the target historical time period into the multi-scale data processing unit to obtain the feature sequence of the meteorological feature sequence at multiple grid scales; for the feature sequence at each grid scale in the multiple grid scales, the current feature sequence is input into the network block corresponding to the current feature sequence to obtain the hidden features of the current feature sequence, wherein the hidden features include the spatial dependency relationship between nodes in the hypergraph structure of the current feature sequence; the hidden features of the feature sequences at multiple grid scales are determined as intermediate feature sequences.

[0013] Optionally, the network block includes a spatial graph construction unit and a hypergraph information aggregation unit, wherein the extraction unit is further configured to input the current feature sequence into the spatial graph construction unit to obtain the hypergraph structure of the current feature sequence, wherein the hypergraph structure of the current feature sequence includes multiple hypergraph structures of preset durations in the target historical time period; and input the hypergraph structure into the hypergraph information aggregation unit to obtain the hidden features of the current feature sequence.

[0014] Optionally, the frequency domain multi-scale time coding and decoding module includes a frequency domain embedding unit, a frequency domain encoding unit, a frequency domain decoding unit and a frequency domain fusion unit, wherein the prediction unit is further configured to input the ocean Niño index of the target historical time period into the frequency domain embedding unit to obtain the time characteristics of different frequency segments; input the time characteristics of different frequency segments into the frequency domain encoding unit to obtain coding characteristics; input the coding characteristics and the target future time period into the frequency domain decoding unit to obtain the predicted characteristics of the future frequency segment; input the predicted characteristics of the future frequency segment into the frequency domain fusion unit to obtain the ocean Niño index of the target sea area in the target future time period.

[0015] Optionally, the training unit is configured to train the adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale time encoding and decoding module in the following manner: based on the meteorological feature sequence and the ocean Niño index of each meteorological feature in the global sea surface multiple meteorological features, a training sample set is obtained, wherein each training sample includes a meteorological feature sequence of a first predetermined time length and a real ocean Niño index of a second predetermined time length after the first predetermined time length; for each training sample, the following processing is performed: the meteorological feature sequence of the first predetermined time length in the current training sample is input into the adaptive multi-scale spatial hypergraph module to obtain an estimated intermediate feature sequence. sequence; based on the estimated intermediate feature sequence, obtaining a first estimated ocean Niño index, wherein the first estimated ocean Niño index is an estimated ocean Niño index for a first predetermined time length; inputting the first estimated ocean Niño index and a second predetermined time length after the first predetermined time length into the frequency domain multi-scale time encoding and decoding module to obtain a second estimated ocean Niño index, wherein the second estimated ocean Niño index is an estimated ocean Niño index for a second predetermined time length after the first predetermined time length; based on the second estimated ocean Niño index and the true ocean Niño index, adjusting the parameters of the adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale time encoding and decoding module.

[0016] Optionally, the training unit is also configured to obtain multiple meteorological characteristics of the global sea surface; pre-process the multiple meteorological characteristics of the global sea surface to obtain a meteorological characteristic sequence and an ocean Niño index for each meteorological characteristic; perform sequential sliding sampling on the meteorological characteristic sequence and the ocean Niño index for each meteorological characteristic; and obtain a training sample set based on the sampling results.

[0017] In another general aspect, a computer-readable storage medium storing instructions is provided, wherein when the instructions are executed by at least one computing device, the at least one computing device is prompted to perform any of the El Niño-Southern Oscillation prediction methods described above.

[0018] In another general aspect, a system is provided comprising at least one computing device and at least one storage device storing instructions, wherein the instructions, when executed by the at least one computing device, cause the at least one computing device to perform any of the El Niño-Southern Oscillation prediction methods described above.

[0019] In another general aspect, a computer program product is provided, comprising computer instructions, which, when executed by a processor, perform the El Niño-Southern Oscillation prediction method as described above.

[0020] According to the El Niño-Southern Oscillation prediction method and apparatus of the embodiments of the present disclosure, an adaptive multi-scale spatial hypergraph module is used to obtain an intermediate feature sequence containing spatial information for the target sea area within the target historical time period based on the meteorological feature sequence of the target sea area within the target historical time period. That is, the adaptive multi-scale spatial hypergraph module dynamically constructs a hypergraph structure of meteorological feature sequences at different grid scales, which can accurately depict the spatiotemporal evolution pattern of meteorological variables in the target sea area in the time dimension, achieve efficient aggregation of spatial information, and fully extract the multi-grid-scale spatial features of the ENSO evolution process. Moreover, a frequency-domain multi-scale time encoding and decoding module is used to obtain the ocean Niño index for the target sea area within the target future time period based on the intermediate feature sequence containing spatial information. That is, the frequency-domain multi-scale time encoding and decoding module is used to perform frequency-domain modeling on the intermediate feature sequence, effectively capturing the long-term evolution law of ENSO, and ultimately generating the ocean Niño index for the target future time period. This allows the extent of the El Niño-Southern Oscillation phenomenon in the target future time period to be known, thereby improving the efficiency, accuracy, and stability of predicting the ENSO evolution trend. Therefore, the present disclosure can effectively solve the problem in the prior art that the prediction accuracy of ENSO prediction decreases as the prediction time increases.

[0021] Additional aspects and / or advantages of the present general inventive concept will be set forth in part in the following description and in part will be apparent from the description, or may be learned through practice of the present general inventive concept. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and other objects and features of the embodiments of the present disclosure will become more apparent through the following description in conjunction with the accompanying drawings showing the embodiments, in which:

[0023] Figure 1 is a flow chart illustrating an El Niño-Southern Oscillation prediction method according to an embodiment of the present disclosure;

[0024] Figure 2 is a system flow chart illustrating a training method of an adaptive multi-scale spatiotemporal ENSO prediction model and an ENSO prediction method according to an embodiment of the present disclosure;

[0025] Figure 3 1 is a schematic diagram illustrating the structure of an adaptive multi-scale spatiotemporal ENSO prediction model according to an embodiment of the present disclosure;

[0026] Figure 4 is a schematic diagram showing the structure of an adaptive multi-scale spatial hypergraph module according to an embodiment of the present disclosure;

[0027] Figure 5 is a schematic diagram showing the structure of a frequency domain multi-scale time encoding and decoding module according to an embodiment of the present disclosure;

[0028] Figure 6 1 is a schematic diagram showing the overall structure of an adaptive multi-scale spatiotemporal ENSO prediction model according to an embodiment of the present disclosure;

[0029] Figure 7 is a block diagram illustrating an El Niño-Southern Oscillation prediction apparatus according to an embodiment of the present disclosure;

[0030] Figure 8 FIG. 1 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0031] The following detailed description is provided to help the reader gain a comprehensive understanding of the methods, devices and / or systems described herein. However, various changes, modifications and equivalents of the methods, devices and / or systems described herein will be clear after understanding the disclosure of the present application. For example, the order of operations described herein is merely an example and is not limited to those orders set forth herein, but can be changed as will be clear after understanding the disclosure of the present application, except for operations that must occur in a specific order. In addition, for greater clarity and conciseness, descriptions of features known in the art may be omitted.

[0032] The features described herein can be implemented in different forms and should not be construed as limited to the examples described herein. Rather, the examples described herein are provided to illustrate only some of the many possible ways to implement the methods, devices, and / or systems described herein, which will become clear after understanding the disclosure of this application.

[0033] As used herein, the term "and / or" includes any one of the associated listed items and any combination of any two or more.

[0034] Although terms such as "first," "second," and "third" may be used herein to describe various members, components, regions, layers, or portions, these members, components, regions, layers, or portions should not be limited by these terms. Instead, these terms are used solely to distinguish one member, component, region, layer, or portion from another member, component, region, layer, or portion. Thus, what is referred to as a first member, first component, first region, first layer, or first portion in the examples described herein may also be referred to as a second member, second component, second region, second layer, or second portion without departing from the teachings of the examples.

[0035] In the specification, when an element (such as a layer, region, or substrate) is described as being “on,” “connected to,” or “coupled to” another element, the element may be directly “on,” “connected to,” or “coupled to” the other element, or one or more other elements may be present therebetween. Conversely, when an element is described as being “directly on,” “directly connected to,” or “directly coupled to” another element, there may be no other elements present therebetween.

[0036] The terms used herein are only used to describe various examples and are not intended to limit the disclosure. Unless the context clearly indicates otherwise, the singular is intended to include the plural. The terms "comprise," "include," and "have" indicate the presence of the recited features, quantities, operations, components, elements, and / or combinations thereof, but do not preclude the presence or addition of one or more other features, quantities, operations, components, elements, and / or combinations thereof.

[0037] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present disclosure pertains after understanding the present disclosure. Unless expressly defined as such herein, terms (such as those defined in commonly used dictionaries) should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and should not be interpreted in an idealized or overly formal manner.

[0038] Furthermore, in describing the examples, when it is deemed that a detailed description of well-known related structures or functions would cause ambiguous interpretation of the present disclosure, such detailed description will be omitted.

[0039] Current ENSO prediction models use past time series meteorological characteristics (such as global sea surface temperature anomalies and heat capacity anomalies) to predict the future Oceanic Nino Index (ONI). The problem of predicting the El Niño-Southern Oscillation (ENSO) is essentially based on past temporal and spatial correlations of sea surface temperature, heat capacity, and other factors. This can be formally expressed as: , where t represents the current time and T represents the number of past time points. Represents the number of future moments. Traditional ENSO prediction models only consider historical spatiotemporal correlation sequences at the original grid scale, establishing a fixed spatial topology at the original grid scale and extracting temporal features in the time domain. Consequently, they are unable to capture both local and global information simultaneously, resulting in a one-sided and simplistic capture of spatiotemporal evolution patterns. Furthermore, as the prediction time increases, computational complexity and error accumulation gradually increase, reducing long-term prediction accuracy and leading to low prediction efficiency, poor accuracy, and poor stability.

[0040] This paper starts from the difficulties in capturing spatiotemporal information. First, in the spatial dimension, an adaptive multi-scale spatial hypergraph module is introduced to dynamically construct hypergraph structures of different grid scales, realize fine modeling of spatial dependencies and information aggregation, and fully extract the multi-scale spatial characteristics of the ENSO evolution process; secondly, in the temporal dimension, a frequency domain multi-scale time encoding and decoding module is used to map the characteristic sequence from the time domain to the frequency domain space, and the periodic characteristics of different frequency bands in the frequency domain are used for modeling, which effectively alleviates the problem of long-term prediction error accumulation in traditional time domain modeling, and can restore the time signal through inverse Fourier transform to achieve accurate prediction of ONI in future time periods, so as to know the extent of the El Niño-Southern Oscillation phenomenon in the target future time period, thereby improving the efficiency, accuracy and stability of predicting the evolution trend of ENSO.

[0041] The El Niño-Southern Oscillation prediction method and apparatus disclosed herein are described in detail below with reference to the accompanying drawings.

[0042] This paper proposes a method for predicting El Niño-Southern Oscillation. Figure 1 is a flow chart illustrating an El Niño-Southern Oscillation prediction method according to an embodiment of the present disclosure. Figure 1 , the El Nino-Southern Oscillation prediction method comprises the following steps:

[0043] In step S101, the meteorological feature sequence of the target sea area in the target historical time period is input into the adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence, wherein the target historical time period is a time period of the first predetermined length before the current moment, and the intermediate feature sequence includes a spatial dependency relationship between nodes of each hypergraph structure in a plurality of hypergraph structures, and each hypergraph structure is obtained based on the feature sequence of the meteorological feature sequence at a corresponding grid scale in a plurality of grid scales.

[0044] In step S102, the ocean Niño index for the target historical time period is obtained based on the intermediate feature sequence.

[0045] In step S103, the ocean Niño index of the target historical time period and the target future time period are input into the frequency domain multi-scale time encoding and decoding module to obtain the ocean Niño index of the target sea area in the target future time period, wherein the target future time period is the time period of the second predetermined length after the current moment, and the ocean Niño index of the target future time period represents the degree of El Niño-Southern Oscillation phenomenon in the target future time period.

[0046] As an example, the El Niño-Southern Oscillation prediction method of this embodiment can be applied to an ENSO prediction model. Specifically, the ENSO prediction model includes an input adaptive multiscale spatial hypergraph module, a conversion module, and a frequency-domain multiscale temporal encoding and decoding module. The conversion module generates an ocean Niño index for a target historical time period based on an intermediate feature sequence. The conversion module can be a multilayer perceptron (MLP), which is not limited in this disclosure.

[0047] As an example, the ENSO prediction model can be trained based on the meteorological characteristic sequences and corresponding ONIs of the target sea area over a historical period. For example, the historical meteorological characteristic sequences of the target sea area can be used as training samples, and the ONIs of the target sea area in the future period can be used as the prediction result labels of the training samples to train the ENSO prediction model. For another example, if it is necessary to predict the ONI for the next 12 months based on the meteorological characteristic sequences of the past 12 months, when training the ENSO prediction model, both the historical and future time periods are set to 12 months, the meteorological characteristic sequences of the target sea area from 1 to 12 months are used as training samples, and the ONIs of the target sea area from 12 to 24 months are used as the prediction result labels.

[0048] It should be noted that the specific training process will be explained in detail later and will not be explained here.

[0049] As an example, the meteorological characteristic sequence of the target sea area during the target historical time period can be obtained by preprocessing the global sea surface multi-feature sample database. For example, the meteorological characteristic sequence of the target sea area during the target historical time period can be the sea surface temperature, heat capacity, etc. of the target sea area from a certain time in the past to the current time (e.g., the past 12 months). It should be noted that the global sea surface multi-feature data sample database can be obtained from, but is not limited to, the following: the National Oceanic and Atmospheric Administration of the United States, the Climate Observing Center of the National Weather Service of the United States, the meteorological satellite data download website provided by NASA, the climate data storage and query platform of the World Meteorological Organization, etc.

[0050] Based on the characteristics and requirements of the ocean El Niño-Southern Oscillation prediction problem, the adaptive multi-scale spatial hypergraph module is designed to construct feature sequences at multiple grid scales, from fine to coarse granularity, to enhance the model's ability to comprehensively extract local information and global spatial features. By introducing a hypergraph structure, multiple connections are established between distant and nearby nodes in physical space, enabling each spatial node to aggregate information from more relevant regions. Furthermore, the historical time period can be based on monthly time steps, with each time step having its own hypergraph results at multiple grid scales. This allows the module to adaptively adjust the hypergraph structure at each grid scale at each time step, combining the dynamic evolution of the time dimension to more accurately simulate the spatial dynamic evolution of the ENSO phenomenon in real physical time.

[0051] The frequency-domain multi-scale time encoding and decoding module proposes a frequency-domain-based time series processing strategy based on the characteristics and requirements of the ocean El Niño-Southern Oscillation prediction problem. This module maps the input historical time series into the frequency domain and decomposes it into time features of multiple frequency bands, thereby achieving decoupled modeling of short-term fluctuations and long-term trends. In the frequency domain, an attention mechanism is introduced to the time features of each frequency band for feature enhancement and encoding, and then restored to the time domain representation through the frequency-domain decoding unit and the frequency-domain fusion unit.

[0052] The adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale temporal encoding and decoding module are introduced below.

[0053] According to an embodiment of the present disclosure, an adaptive multi-scale spatial hypergraph module may include a multi-scale data processing unit and multiple parallel network blocks, and multiple grid scales correspond one-to-one to the multiple parallel network blocks. In step S101, the meteorological feature sequence of the target sea area in the target historical time period is input into the adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence. This can be achieved in the following manner: the meteorological feature sequence of the target historical time period is input into the multi-scale data processing unit to obtain a feature sequence of the meteorological feature sequence at multiple grid scales; for the feature sequence at each grid scale in the multiple grid scales, the current feature sequence is input into the network block corresponding to the current feature sequence to obtain the hidden features of the current feature sequence, wherein the hidden features include the spatial dependency between nodes in the hypergraph structure of the current feature sequence; and the hidden features of the feature sequences at multiple grid scales are determined as intermediate feature sequences.

[0054] Through this embodiment, the meteorological characteristic sequence of the target historical time is divided into characteristic sequences at multiple network scales, which can accurately characterize the evolution of the meteorological variables of the target sea area in the time dimension, realize efficient aggregation of spatial information, and use multiple network blocks to process the characteristic sequence at each network scale in parallel, so as to quickly obtain the hidden features of the characteristic sequence at each network scale, and then quickly obtain the intermediate characteristic sequence.

[0055] As an example, the multi-scale data processing unit creates feature sequences at multiple grid scales with different granularities from the meteorological feature sequence of the target historical time period. Specifically, the multi-scale data processing unit divides the meteorological feature sequence of the target historical time period into feature sequences at multiple grid scales from fine granularity to coarse granularity based on the preset granularity window size; for example, the original grid scale of the meteorological feature sequence of the target historical time period is latitude. ,longitude Total 24 The global real physical space range of 48 grid points, the preset granularity window size is 24 and 48, and then, 4 can be obtained through two-dimensional convolution. 12 and 2 6 The feature sequences at two different grid scales represent the feature sequence representations at different grid scales from local to global.

[0056] As an example, the parameters of the above-mentioned multiple parallel network blocks can be shared, which is not limited in this disclosure. Each network block can adopt an adaptive hypergraph learning neural network, which is not limited in this disclosure.

[0057] Specifically, each network block is used to output the hidden features of the feature sequence at the corresponding grid scale based on the feature sequence at different grid scales. In other words, the input of each network block is the feature sequence of the target sea area at a grid scale within the target historical time period. The feature sequence at the corresponding grid scale is processed independently to obtain the feature time series with spatial information of the target sea area at the corresponding network scale within the target historical time period, that is, the above-mentioned hidden features. Multiple network blocks can derive hidden features at multiple grid scales in parallel, and combine multiple hidden features as the output of the adaptive multi-scale spatial hypergraph module.

[0058] According to an embodiment of the present disclosure, the above-mentioned network block may include a spatial graph construction unit and a hypergraph information aggregation unit, wherein the current feature sequence is input into the network block corresponding to the current feature sequence to obtain the hidden features of the current feature sequence, which can be achieved in the following manner: the current feature sequence is input into the spatial graph construction unit to obtain the hypergraph structure of the current feature sequence, wherein the hypergraph structure of the current feature sequence includes multiple hypergraph structures of preset durations in the target historical time period; the hypergraph structure is input into the hypergraph information aggregation unit to obtain the hidden features of the current feature sequence.

[0059] Through this embodiment, the spatial graph construction unit can construct a hypergraph structure of the feature sequence of each preset time length at the corresponding network scale, so that combined with the dynamic evolution characteristics of the time dimension, the hypergraph structure at each grid scale of each preset time length can be adaptively adjusted, thereby more accurately simulating the spatial dynamic evolution process of the ENSO phenomenon in real physical time, and on the basis of this hypergraph structure, through the hypergraph information aggregation unit, the hyperedge weight is adaptively adjusted according to the changes in the feature sequence in each preset time length, thereby realizing effective modeling of the dynamic spatial evolution structure, and thus relatively accurate hidden features can be obtained.

[0060] As an example, the hypergraph information aggregation unit can adaptively adjust the hyperedge weights based on changes in the network state during each preset duration, aggregate the hypergraph results, and obtain corresponding accurate hidden features. The preset duration can be measured in months, which is not limited in this disclosure. It should be noted that the preset duration can also be referred to as a time step. The network block can include a spatial graph construction unit and a hypergraph information aggregation unit, and the two units have the same number of channels. It should be noted that changes in the network state are changes in the feature sequence for each preset duration (time step).

[0061] The spatial graph construction unit and the hypergraph information aggregation unit are described below:

[0062] 1) The above-mentioned spatial graph construction unit can construct a hypergraph structure based on the geographical distance between meteorological observation points and the similarity of the meteorological feature sequences of the meteorological observation points. Specifically, based on the geographical coordinate information of the meteorological observation points in the current feature sequence, the Euclidean distance between each observation point can be calculated, and the observation points with a Euclidean distance less than a first set threshold can be selected, and the selected observation points can be included in the same hyperedge; hyperedges can also be constructed based on the correlation between the meteorological feature sequences of the meteorological observation points in the current feature sequence. When the correlation of the feature sequences of two meteorological observation points in the historical time period exceeds the second set threshold, the two meteorological observation points are included in the same hyperedge; meteorological observation points that meet the above two conditions at the same time can also be included in the same hyperedge, and this disclosure does not limit this, as long as a set of nodes with high meteorological similarity can be formed. The hypergraph structure output by the spatial graph construction unit is used for subsequent hypergraph information aggregation operations to enhance spatial dependency modeling capabilities.

[0063] It should be noted that for s The hypergraph structure expression constructed based on the geographical distance between meteorological observation points or the similarity of meteorological characteristic sequences of meteorological observation points can be expressed as follows:

[0064]

[0065] Among them, when constructing a hypergraph structure based on the geographical distance between meteorological observation points, represents the first set threshold; when constructing a hypergraph structure based on the similarity of the meteorological feature sequences of meteorological observation points, represents the second set threshold; v represents the node position at the sth grid scale, e Indicates the s Hyperedges at the grid scale.

[0066] 2) The hypergraph information aggregation unit outputs the corresponding hidden features based on the hypergraph structure. Specifically, the hypergraph information aggregation unit can use a hypergraph convolution operation to perform weighted summation or attention-weighted fusion on the features of multiple nodes connected to each hyperedge in a node-hyperedge-node aggregation path at each preset time length to extract high-order spatial dependency information between nodes, i.e., the hidden features mentioned above. This allows each node to not only aggregate the information of its adjacent nodes, but also integrate the global dynamic features of multiple nodes in the same hyperedge, thereby enhancing the spatiotemporal modeling capabilities within each grid scale. This aggregation process can be combined with a dynamic adjustment mechanism to adaptively adjust the hyperedge weights according to the changes in the network state at each time step, thereby achieving effective modeling of dynamic spatial evolution structures.

[0067] The weights used in the above weighting method can be static weights, such as calculating fixed edge weights based on the geographical distance between nodes or the similarity of meteorological characteristics; the weights used in the above weighting method can also be dynamic learnable weights, such as by introducing an attention mechanism to adaptively allocate the aggregation weights of different nodes according to the feature context at the current moment.

[0068] For example, in a hyperedge connecting five meteorological observation points, different weights can be dynamically assigned to each adjacent node based on the similarity of their meteorological characteristics or geographic distance to the central node, ultimately generating a spatially aggregated representation of the node under the current hyperedge. Furthermore, the hypergraph information aggregation unit also fuses feature information from multiple hyperedges, further enhancing spatial perception capabilities through multi-hyperedge aggregation strategies (such as average aggregation, maximum aggregation, or weighted fusion), resulting in more discriminative hidden features.

[0069] According to an embodiment of the present disclosure, the frequency domain multi-scale time coding and decoding module includes a frequency domain embedding unit, a frequency domain encoding unit, a frequency domain decoding unit and a frequency domain fusion unit, wherein the ocean Niño index of the target historical time period and the target future time period are input into the frequency domain multi-scale time coding and decoding module to obtain the ocean Niño index of the target sea area in the target future time period, which can be achieved in the following manner: the ocean Niño index of the target historical time period is input into the frequency domain embedding unit to obtain the time characteristics of different frequency segments; the time characteristics of different frequency segments are input into the frequency domain encoding unit to obtain coding characteristics; the coding characteristics are input into the frequency domain decoding unit of the target future time period to obtain the predicted characteristics of the future frequency segment; the predicted characteristics of the future frequency segment are input into the frequency domain fusion unit to obtain the ocean Niño index of the target sea area in the target future time period.

[0070] Through this embodiment, the ocean Niño index of the input target historical time period is mapped to the frequency domain space and decomposed into time features of multiple different frequency segments, thereby realizing decoupling modeling of short-term fluctuations and long-term trends; and in the frequency domain, the time features of each frequency segment are respectively enhanced and encoded, and then restored to the time domain representation through the frequency domain decoding unit and the frequency domain fusion unit.

[0071] As an example, the frequency domain multi-scale temporal encoding and decoding module can convert the intermediate feature sequence of the target sea area within the target historical time period into the frequency domain based on the self-attention mechanism architecture, and extract the ONI prediction results. It should be noted that the frequency domain multi-scale temporal encoding and decoding module can include multiple frequency domain operation units, which are introduced below:

[0072] 1) The frequency domain embedding unit performs a discrete Fourier transform (DFT) on the ocean Niño index for the target historical time period, mapping the time series of the ocean Niño index from the time domain to the frequency domain. This unit then extracts frequency band features that reflect variations at different time scales. This extracts temporal features across different frequency bands, which are used to capture the periodicity and oscillation patterns in the ENSO phenomenon. Specifically, the frequency domain embedding unit first performs a Fourier transform on the input ocean Niño index time series to obtain its amplitude or phase spectrum representation across multiple frequency components. These frequency components are then further divided into several predefined frequency bands using a frequency band segmentation module, such as low frequency (long-term variation trends), mid-frequency (interannual fluctuations), and high frequency (seasonal oscillations), thereby extracting frequency domain features representing variations at different time scales.

[0073] For example, when the input is the ocean Niño index time series of sea surface temperature, the frequency domain embedding unit can obtain its characteristics in frequency intervals with periods of 1 year, 6 months, 1 month, etc. after performing FFT. The characteristics in these frequency intervals help the subsequent frequency domain encoding unit to determine the ENSO-related oscillation period, thereby improving the perception of ONI trends.

[0074] It should be noted that in order to enhance the stability and distinguishability of frequency representation, the frequency domain embedding unit can also introduce processing operations such as amplitude normalization and spectrum enhancement, so that the model's dependence on key frequency bands is more significant, thereby improving the prediction accuracy.

[0075] 2) The frequency-domain encoding unit encodes the temporal features of the multiple frequency bands extracted by the frequency-domain embedding unit, and highlights the influence of key frequency bands by constructing a frequency-domain attention map. For example, a frequency-domain attention module based on the Transformer structure can be used to model the dependencies between multiple frequency bands. Specifically, the frequency-domain encoding unit constructs a frequency-wise self-attention network (Frequency-wise Transformer Encoder) in the frequency dimension, capturing the long-range dependencies and interactions between different frequencies through a multi-head attention mechanism, and then outputs the corresponding encoded features. The frequency-domain encoding unit can stack multiple frequency-domain Transformer layers, each of which contains a frequency-attention sublayer and a feedforward network sublayer. Residual connections and normalization operations are used to enable the network to more stably focus on key frequency bands that are strongly correlated with ENSO.

[0076] 3) The frequency domain decoding unit constructs a prediction structure for future frequency distribution based on the frequency representation output by the frequency domain encoding unit. Specifically, it predicts the Ocean Niño Index (ONI) for the target sea area within the target future time period. For example, the frequency domain decoding unit can employ a frequency time series modeling network based on a Transformer decoder structure. Specifically, this decoder structure is a frequency-aware Transformer decoder. During the training phase, it fuses historical frequency representations (i.e., the encoded features output by the frequency domain encoding unit) with target frequency position information (the actual ONI corresponding to the encoded features) to perform sequence prediction in the frequency domain. Each layer of the decoding module incorporates a frequency attention mechanism, which gradually predicts the evolutionary characteristics of each frequency band in the future and outputs the predicted spectrum results for subsequent fusion and restoration.

[0077] 4) The frequency domain fusion unit performs a weighted fusion of the prediction results from different frequency bands and converts the frequency domain results back to the time domain to form the Ocean Niño Index (ONI) for the target sea area within the target future time period. Specifically, the frequency domain fusion unit introduces a band weighting strategy, dynamically adjusting the fusion weights of each band based on indicators such as the importance and confidence of the band prediction results. Then, through an inverse Fourier transform (IDFT), the fused spectrum signal is restored to a time series to obtain the final ONI prediction result.

[0078] According to an embodiment of the present disclosure, the adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale time encoding and decoding module are trained in the following manner: based on the meteorological feature sequence and the ocean Niño index of each meteorological feature among the various meteorological features of the global sea surface, a training sample set is obtained, wherein each training sample includes a meteorological feature sequence of a first predetermined time length and a real ocean Niño index of a second predetermined time length after the first predetermined time length; for each training sample, the following processing is performed: the meteorological feature sequence of the first predetermined time length in the current training sample is input into the adaptive multi-scale spatial hypergraph module to obtain an estimated intermediate feature sequence ; Based on the estimated intermediate feature sequence, a first estimated ocean Niño index is obtained, wherein the first estimated ocean Niño index is the estimated ocean Niño index for the first predetermined time length; the first estimated ocean Niño index and the second predetermined time length after the first predetermined time length are input into the frequency domain multi-scale time encoding and decoding module to obtain a second estimated ocean Niño index, wherein the second estimated ocean Niño index is the estimated ocean Niño index for the second predetermined time length after the first predetermined time length; based on the second estimated ocean Niño index and the true ocean Niño index, the parameters of the adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale time encoding and decoding module are adjusted.

[0079] As an example, during the training process, the frequency domain multi-scale time encoding and decoding module can also input the real ocean Niño index of different time periods, such as the real ocean Niño index of the first predetermined time period, or the real ocean Niño index of the second half of the first predetermined time period. This is not limited in the present disclosure.

[0080] Through this embodiment, the training of the adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale temporal encoding and decoding module can be conveniently and accurately achieved.

[0081] According to an embodiment of the present disclosure, a training sample set is obtained based on the meteorological feature sequence and the ocean Niño index of each meteorological feature among multiple meteorological features of the global sea surface, including: obtaining multiple meteorological features of the global sea surface; preprocessing the multiple meteorological features of the global sea surface to obtain the meteorological feature sequence and the ocean Niño index of each meteorological feature; sequentially sliding sampling the meteorological feature sequence and the ocean Niño index of each meteorological feature; and obtaining a training sample set based on the sampling results.

[0082] Through this embodiment, the pre-processed meteorological characteristic sequence and the ocean Niño index are sampled, which can increase the sample size and alleviate the problem of insufficient effective data in the ocean field.

[0083] As an example, global sea surface multiple meteorological characteristic data can be obtained from the global sea surface multi-feature data sample library, and the global sea surface multi-feature data sample library can be obtained from the National Oceanic and Atmospheric Administration of the United States, the Climate Observation Center of the National Weather Service of the United States, the meteorological satellite data download website provided by NASA, the climate data storage and query platform of the World Meteorological Organization, etc., and this disclosure is not limited to this. The prediction data samples of the ocean El Niño / Southern Oscillation in the global sea surface multi-feature data sample library can be divided into two categories: historical observation data and model simulation data, where model simulation data is observation data simulated based on historical observation data. These two types of data are usually used in combination to improve the accuracy and reliability of the prediction. For example, historical observation data can be used to calibrate the model and serve as benchmark data for model input; model simulation data can help predict the possibility and trend of future ocean El Niño / Southern Oscillation time.

[0084] As an example, the meteorological characteristic sequence and ocean Niño index for each meteorological feature can be obtained by preprocessing a global sea surface multi-feature sample database. For example, the meteorological characteristic sequence for the target sea area over a historical time period can be the sea surface temperature, heat capacity, etc., from a certain point in the past to the current moment (e.g., the past 12 months), while the ONI over a historical time period can be the actual ONI for the target sea area from a certain point in the past to the current moment (e.g., the past 12 months). It should be noted that the actual ONI over a historical time period can be used to represent the extent of the ENSO phenomenon from a certain point in the past to the current moment (e.g., the past 12 months).

[0085] The above-mentioned preprocessing can be performed using data preprocessing software, python data processing packages, etc., and the present disclosure is not limited to this. The purpose of preprocessing the obtained global sea surface multi-feature sample database is to obtain a sample set in a specific format containing features related to the ocean El Niño / Southern Oscillation prediction. The preprocessed features are prediction factors closely related to the time of the ocean El Niño / Southern Oscillation, such as sea surface temperature, heat capacity, etc. For ease of understanding, the sea surface temperature and heat capacity are explained below. Data preprocessing can be performed using data preprocessing software, python data processing packages, etc., and the final data formats of the sea surface temperature and heat capacity are meteorological feature sequences and ONI time series data, where the meteorological feature sequences can be expressed as , ONI time series data can be expressed as ,in, n represents the maximum moment of the meteorological characteristic sequence, R H×W express X i Has spatial dimensions, i.e. X iIn addition to the time dimension, it also has longitude and latitude. H Indicates longitude, W Indicates latitude, R express Y i It has no spatial dimension, i.e. Y i It only has the time dimension.

[0086] As an example, after obtaining the meteorological characteristic sequence and ONI of the target sea area in the historical period, the sampling interval Sliding sampling is performed on meteorological characteristic series and ONI. Sampling interval It can be set for the year or for the time resolution of the sequence data (usually monthly). The advantage of the latter is that it can greatly increase the sample size and alleviate the problem of insufficient effective data in the marine field. The disadvantage is that the repeated use of data may cause the model to overfit. When set to 12, it is equivalent to the sampling interval for the year Set to 1. Therefore, considering the insufficient amount of data and the risk of model overfitting, the sampling interval for the month The value range can be .

[0087] As an example, a loss can be determined based on the second estimated ocean Niño index and the true ocean Niño index, and the parameters of the adaptive multi-scale spatial hypergraph module and the frequency-domain multi-scale temporal encoding and decoding module can be adjusted based on this loss. The loss function used to calculate the above loss can be the mean square error loss between the second estimated ocean Niño index and the true ocean Niño index, which is not limited in this disclosure. It should be noted that in terms of training, relevant training parameters, methods, and functions can be selected, and meteorological feature sequence samples can be input into the adaptive multi-scale spatiotemporal ENSO prediction model. The model is trained to obtain a trained adaptive multi-scale spatiotemporal ENSO prediction model.

[0088] In order to better understand the present disclosure, Figure 2 Provide a description of the system.

[0089] Figure 2 The training method of the adaptive multi-scale spatiotemporal ENSO prediction model and the systematic process of the ENSO prediction method are demonstrated. Figure 2 As shown, the following steps may be included:

[0090] 1) Obtain a global ocean surface multi-feature data sample library;

[0091] 2) Preprocess the data in the global sea surface multi-feature data sample library to extract meteorological characteristic series data and ONI time series data containing ocean El Niño-Southern Oscillation prediction-related features.

[0092] 3) Based on the spatiotemporal characteristics and evolution trends of the ocean El Niño-Southern Oscillation, an adaptive multi-scale spatiotemporal ENSO prediction model is constructed and trained using meteorological characteristic series data such as sea surface temperature and heat capacity and ONI time series data.

[0093] 4) Based on the trained adaptive multi-scale spatiotemporal ENSO prediction model, ONI time series data for future time periods are predicted.

[0094] The typical problems addressed by this adaptive multi-scale spatiotemporal ENSO prediction model are spatiotemporal sequence feature extraction and time series prediction. Specifically, this model extracts an intermediate feature sequence containing spatial feature information from past meteorological feature sequence data. This intermediate feature sequence is then converted into past ONI time series data. This ONI time series data is then used to predict future ONI time series data. The key here is capturing both the spatial feature information and the temporal evolution of the meteorological feature sequence data.

[0095] Figure 3 The structure of the adaptive multi-scale spatiotemporal ENSO prediction model is presented, e.g. Figure 3 As shown in the figure, the input data of the adaptive multi-scale spatiotemporal ENSO prediction model is the meteorological characteristic sequence data of the target sea area in the past T moments. , and also input the future τ time points, Figure 3 Not shown in the figure, the meteorological feature sequence data is extracted by the adaptive multi-scale spatial hypergraph module to obtain the intermediate feature sequence containing spatial feature information, and then the intermediate feature sequence is converted into ONI time series data for the past T moments. , and then, time series data The ONI prediction results of the target sea area at the next τ time are obtained through the frequency domain multi-scale time encoding and decoding module .

[0096] Figure 4 The structure of the adaptive multi-scale spatial hypergraph module is presented, e.g. Figure 4 As shown, the adaptive multi-scale spatial hypergraph module includes a multi-scale data processing unit, multiple parallel network blocks, and each network block includes a spatial graph construction unit and a hypergraph information aggregation unit.

[0097] The adaptive multi-scale spatial hypergraph module targets the meteorological characteristic sequence data of the past time period. Its purpose is to extract the spatial feature information from the meteorological characteristic sequence data of the past time period, obtain the intermediate feature sequence containing the spatial feature information, and then convert it into the ONI time series data of the corresponding target sea area in the past time period. Figure 4 The example shows the case where the feature dimension (i.e., the number of channels) at each moment is 1. In practice, the feature dimension at each moment is greater than 1, but the principle remains the same. The functions of each unit in the adaptive multi-scale spatial hypergraph module are described as follows:

[0098] The multi-scale data processing unit can perform scale processing on meteorological characteristic sequence data of past time periods, such as forming a coarse-grained spatial representation based on two-dimensional convolution. Figure 4 The processing of time t-T+1, ..., t is shown. For example, two-dimensional convolution can be used to perform convolution processing on the spatial grid to obtain the spatiotemporal correlation grid sequence representation (i.e., the above sequence features) of spatial grid points of multiple different grid scales, such as Figure 4 The S shown here represents the number of different grid scales. Each of the spatiotemporal grid sequences output by the multi-scale data processing unit is input into a joint unit (i.e., a network block) consisting of a corresponding spatial graph construction unit and a hypergraph information aggregation module for subsequent spatial structure modeling and aggregation.

[0099] The spatial graph construction unit constructs a spatial graph connection structure (i.e., a hypergraph structure) at each grid scale based on the geographic distances or meteorological similarities between ocean physical variables. For example, at each time step t in the past (e.g., a month is a time step, which is the preset duration in the above embodiment), all grid points within a grid scale at the current time step are considered as graph nodes. The Euclidean distance and meteorological characteristic similarity between each graph node are calculated to form a weighted adjacency matrix, thereby establishing a physically meaningful spatial graph. Furthermore, to capture the potential dynamic correlations between distant regions, a hypergraph structure is introduced, aggregating multiple nodes with similar change patterns within each hyperedge, thereby constructing a hypergraph connection relationship that reflects the spatial coupling pattern.

[0100] The hypergraph information aggregation unit uses hypergraph convolution or attention aggregation mechanisms to propagate information within the constructed hypergraph structure. At each time step, the module performs feature fusion along a node-hyperedge-node aggregation path, enabling each node to aggregate not only information from its neighboring nodes but also the global dynamic features of multiple nodes within the same hyperedge, thereby enhancing spatiotemporal modeling capabilities within each grid scale. This aggregation process can be combined with a dynamic adjustment mechanism to adaptively adjust hyperedge weights based on changes in network state at each time step, effectively modeling dynamic spatial evolution structures.

[0101] It's important to note that in a hypergraph, the node-hyperedge-node aggregation path is a mechanism for information transfer and feature updates. This aggregation path allows nodes and hyperedges to reinforce each other, better representing complex relationships within the graph structure. This facilitates modeling and analysis of hypergraph data, and has broad applications in fields such as image segmentation, genetic medicine, and natural language processing. For example, in image segmentation, nodes can be pixels, and hyperedges represent relationships between pixels. This aggregation path allows for better utilization of contextual information in images for pixel classification. In genetic medicine, nodes represent genes, and hyperedges correspond to gene sets or genetic pathways, facilitating analysis of gene interactions and functional relationships.

[0102] Figure 5 The structure of the frequency domain multi-scale time encoding and decoding module is shown, such as Figure 5 As shown, the frequency domain multi-scale time coding and decoding module includes a frequency domain embedding unit, a frequency domain encoding unit, a frequency domain decoding unit and a frequency domain fusion unit.

[0103] The frequency domain multi-scale time encoding and decoding module targets the ONI time series data of the target sea area in the past time period. Its purpose is to capture the temporal evolution of the ONI time series data in the past time period and predict the ONI time series data in the future time period. The functions of each unit of the frequency domain multi-scale time encoding and decoding module are described as follows:

[0104] The frequency-domain embedding unit transforms the input ONI time series data from the past time period from the time domain to the frequency domain to more effectively extract periodic and trend features, that is, to extract time features of different frequency segments. Specifically, the discrete Fourier transform (DFT) is used to perform window segmentation and spectrum extraction on the ONI time series data, obtaining frequency component information at different time locations and decoupling the long sequence signal into multiple sub-features in different frequency ranges.

[0105] The above-mentioned frequency-domain encoding unit (FRE) deeply encodes the temporal features of different frequency bands. That is, it can construct an attention map in the frequency dimension through the frequency attention mechanism to identify key frequency bands that have a significant influence on the evolution of ENSO.

[0106] The frequency-domain decoding unit performs feature conversion on the temporal features of the multiple frequency segments output by the frequency-domain encoding unit to generate prediction results for each frequency segment. This effectively converts the temporal features of the multiple frequency segments output by the frequency-domain encoding unit into a unified feature space representation that facilitates fusion. This represents the predicted features of future frequency segments, which are then used for fusion of subsequent prediction outputs.

[0107] The frequency-domain fusion unit integrates the prediction results of each frequency band to predict the final ONI. Specifically, it applies an inverse Fourier transform (IDFT) to restore the fused frequency-domain features to a time-domain signal and predict the ONI time series data for the target area in the future.

[0108] Figure 6 The overall structure of the adaptive multi-scale spatiotemporal ENSO prediction model is shown. Figure 6 As shown in the figure, the result of data preprocessing of the meteorological characteristic sequence of the target sea area in the target historical time period is [ ] Input the adaptive multi-scale spatial hypergraph module to obtain the multi-scale spatial enhanced meteorological feature time series, that is, the intermediate feature sequence, and aggregate the information of multiple scales of the intermediate feature sequence through the multi-layer perceptron (MLP) to obtain the ONI time series data of the target sea area in the target historical time period. ], then, the ONI time series data [ ] Input the frequency domain multi-scale time encoding and decoding module to predict the final ONI result of the target sea area in the target future time period[ ].

[0109] In summary, this paper introduces an adaptive multi-scale spatial hypergraph module to dynamically model the spatiotemporal correlation structure in the meteorological feature sequence, accurately characterizes the time-varying spatial evolution pattern of the target sea area meteorological variables in the time dimension, realizes efficient aggregation of spatial information, and constructs the ONI time feature sequence of the historical time period with spatial dependence characteristics; then, the frequency domain multi-scale time encoding and decoding module is used to perform frequency domain modeling on the ONI time series data of the above historical time period, thereby effectively capturing the long-term temporal evolution law of ENSO-related variables, and finally predicting the ONI time series data of the future time period, which significantly improves the prediction efficiency, accuracy and stability of the evolution trend of ENSO events.

[0110] Figure 7 is a block diagram illustrating an El Nino-Southern Oscillation prediction apparatus according to an embodiment of the present disclosure, Figure 7 As shown, the device includes an extraction unit 70 , an acquisition unit 72 and a prediction unit 74 .

[0111] The extraction unit 70 is configured to input the meteorological feature sequence of the target sea area in the target historical time period into the adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence, wherein the target historical time period is the time period of the first predetermined length before the current moment, and the intermediate feature sequence includes the spatial dependency relationship between the nodes of each hypergraph structure in multiple hypergraph structures, and each hypergraph structure is obtained based on the feature sequence of the meteorological feature sequence at the corresponding grid scale in multiple grid scales; the acquisition unit 72 is configured to obtain the ocean Niño index of the target historical time period based on the intermediate feature sequence; the prediction unit 74 is configured to input the ocean Niño index of the target historical time period and the target future time period into the frequency domain multi-scale time encoding and decoding module to obtain the ocean Niño index of the target sea area in the target future time period, wherein the target future time period is the time period of the second predetermined length after the current moment, and the ocean Niño index of the target future time period represents the degree of El Niño-Southern Oscillation phenomenon in the target future time period.

[0112] According to an embodiment of the present disclosure, the adaptive multi-scale spatial hypergraph module includes a multi-scale data processing unit and multiple parallel network blocks, and the multiple grid scales correspond one-to-one to the multiple parallel network blocks, wherein the extraction unit 70 is further configured to input the meteorological feature sequence of the target historical time period into the multi-scale data processing unit to obtain the feature sequence of the meteorological feature sequence at multiple grid scales; for the feature sequence at each grid scale in the multiple grid scales, the current feature sequence is input into the network block corresponding to the current feature sequence to obtain the hidden features of the current feature sequence, wherein the hidden features include the spatial dependency relationship between the nodes in the hypergraph structure of the current feature sequence; the hidden features of the feature sequences at multiple grid scales are determined as intermediate feature sequences.

[0113] According to an embodiment of the present disclosure, the network block includes a spatial graph construction unit and a hypergraph information aggregation unit, wherein the extraction unit 70 is further configured to input the current feature sequence into the spatial graph construction unit to obtain the hypergraph structure of the current feature sequence, wherein the hypergraph structure of the current feature sequence includes multiple hypergraph structures of preset durations in the target historical time period; and input the hypergraph structure into the hypergraph information aggregation unit to obtain the hidden features of the current feature sequence.

[0114] According to an embodiment of the present disclosure, the frequency domain multi-scale time coding and decoding module includes a frequency domain embedding unit, a frequency domain encoding unit, a frequency domain decoding unit and a frequency domain fusion unit, wherein the prediction unit 74 is further configured to input the ocean Niño index of the target historical time period into the frequency domain embedding unit to obtain the time characteristics of different frequency segments; input the time characteristics of different frequency segments into the frequency domain encoding unit to obtain coding characteristics; input the coding characteristics and the target future time period into the frequency domain decoding unit to obtain the predicted characteristics of the future frequency segment; input the predicted characteristics of the future frequency segment into the frequency domain fusion unit to obtain the ocean Niño index of the target sea area in the target future time period.

[0115] According to an embodiment of the present disclosure, the training unit is configured to train the adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale time encoding and decoding module in the following manner: based on the meteorological feature sequence and the ocean Niño index of each meteorological feature in a variety of global meteorological features, a training sample set is obtained, wherein each training sample includes a meteorological feature sequence of a first predetermined time length and a real ocean Niño index of a second predetermined time length after the first predetermined time length; for each training sample, the following processing is performed: the meteorological feature sequence of the first predetermined time length in the current training sample is input into the adaptive multi-scale spatial hypergraph module, and an estimated intermediate characteristic sequence; based on the estimated intermediate characteristic sequence, obtaining a first estimated ocean Niño index, wherein the first estimated ocean Niño index is an estimated ocean Niño index for a first predetermined time length; inputting the first estimated ocean Niño index and a second predetermined time length after the first predetermined time length into a frequency domain multi-scale time encoding and decoding module to obtain a second estimated ocean Niño index, wherein the second estimated ocean Niño index is an estimated ocean Niño index for a second predetermined time length after the first predetermined time length; based on the second estimated ocean Niño index and the true ocean Niño index, adjusting parameters of the adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale time encoding and decoding module.

[0116] According to an embodiment of the present disclosure, the training unit is also configured to obtain multiple meteorological features of the global sea surface; pre-process the multiple meteorological features of the global sea surface to obtain a meteorological feature sequence and an ocean Niño index for each meteorological feature; perform sequential sliding sampling on the meteorological feature sequence and the ocean Niño index for each meteorological feature; and obtain a training sample set based on the sampling results.

[0117] Figure 8 The schematic diagram of the structure of the electronic device provided by the present disclosure is shown as follows: Figure 8As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840. The processor 810, the communications interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 may invoke logic instructions in the memory 830 to execute any of the ENSO prediction methods provided in the above embodiments.

[0118] Furthermore, the logic instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0119] It should be noted here that the electronic device disclosed in the present invention can implement all the method steps implemented in the above method embodiment and can achieve the same technical effects. The parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0120] According to an embodiment of the present disclosure, a computer-readable storage medium storing instructions is provided, wherein, when the instructions are executed by at least one computing device, the at least one computing device is prompted to perform the El Niño-Southern Oscillation prediction method as described in any of the above embodiments.

[0121] It should be noted here that the non-transitory computer-readable storage medium provided by the present disclosure can implement all the method steps implemented in the above-mentioned method embodiment and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0122] According to an embodiment of the present disclosure, a system is provided, comprising at least one computing device and at least one storage device storing instructions, wherein the instructions, when executed by the at least one computing device, prompt the at least one computing device to execute the El Niño-Southern Oscillation prediction method as described in any of the above embodiments.

[0123] According to an embodiment of the present disclosure, a computer program product is provided, including computer instructions. When the computer instructions are executed by a processor, the El Niño-Southern Oscillation prediction method described above is performed.

[0124] It should be noted here that the computer program product provided by the present disclosure can implement all the method steps implemented by the above-mentioned method embodiment and can achieve the same technical effects. The parts and beneficial effects of this embodiment that are the same as those of the method embodiment will not be described in detail here.

[0125] While some embodiments of the present disclosure have been shown and described, it will be appreciated by those skilled in the art that changes may be made to these embodiments without departing from the principles and spirit of the disclosure, the scope of which is defined by the claims and their equivalents.

[0126] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0127] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

Claims

1. A method for predicting El Niño-Southern Oscillation, characterized in that: include: Inputting the meteorological feature sequence of the target sea area in the target historical time period into the adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence, wherein the target historical time period is a time period of a first predetermined length before the current moment, and the intermediate feature sequence includes a spatial dependency relationship between nodes of each hypergraph structure in a plurality of hypergraph structures, each hypergraph structure being obtained based on a feature sequence of the meteorological feature sequence at a corresponding grid scale in a plurality of grid scales; Based on the intermediate characteristic sequence, obtaining the ocean Niño index for the target historical time period; Inputting the ocean Niño index of the target historical time period and the target future time period into a frequency domain multi-scale time encoding and decoding module to obtain the ocean Niño index of the target sea area in the target future time period, wherein the target future time period is a time period of a second predetermined length after the current moment, and the ocean Niño index of the target future time period represents the degree of El Niño-Southern Oscillation phenomenon in the target future time period; The adaptive multi-scale spatial hypergraph module includes a multi-scale data processing unit and multiple parallel network blocks. The multiple grid scales correspond one-to-one to the multiple parallel network blocks. The meteorological feature sequence of the target sea area in the target historical time period is input into the adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence, including: Inputting the meteorological characteristic sequence of the target historical time period into a multi-scale data processing unit to obtain characteristic sequences of the meteorological characteristic sequence at multiple grid scales; For a feature sequence at each grid scale in the plurality of grid scales, inputting a current feature sequence into a network block corresponding to the current feature sequence to obtain a hidden feature of the current feature sequence, wherein the hidden feature includes a spatial dependency relationship between nodes in a hypergraph structure of the current feature sequence; Determining the hidden features of the feature sequences at the multiple grid scales as the intermediate feature sequences; The frequency domain multi-scale time coding and decoding module includes a frequency domain embedding unit, a frequency domain coding unit, a frequency domain decoding unit, and a frequency domain fusion unit. Inputting the ocean Niño index of the target historical time period and the target future time period into the frequency domain multi-scale time coding and decoding module to obtain the ocean Niño index of the target sea area in the target future time period includes: Inputting the ocean Niño index of the target historical time period into the frequency domain embedding unit to obtain time characteristics of different frequency segments; Inputting the time features of the different frequency segments into a frequency domain coding unit to obtain coding features; Inputting the coding feature and the target future time period into a frequency domain decoding unit to obtain a predicted feature of the future frequency segment; The predicted features of the future frequency segment are input into a frequency domain fusion unit to obtain the ocean Niño index of the target sea area in the target future time period.

2. The method according to claim 1, wherein in, The network block includes a spatial graph construction unit and a hypergraph information aggregation unit. The step of inputting the current feature sequence into a network block corresponding to the current feature sequence to obtain hidden features of the current feature sequence includes: Inputting the current feature sequence into a spatial graph construction unit to obtain a hypergraph structure of the current feature sequence, wherein the hypergraph structure of the current feature sequence includes hypergraph structures of multiple preset time lengths in the target historical time period; The hypergraph structure is input into a hypergraph information aggregation unit to obtain hidden features of the current feature sequence.

3. The method according to claim 1, wherein The adaptive multi-scale spatial hypergraph module and the frequency domain multi-scale temporal codec module are trained in the following manner: Acquire a training sample set based on a meteorological feature sequence and an ocean Niño index for each of a plurality of meteorological features of the global sea surface, wherein each training sample includes a meteorological feature sequence for the first predetermined time length and a true ocean Niño index for the second predetermined time length after the first predetermined time length; For each training sample, the following processing is performed: Inputting the meteorological feature sequence of the first predetermined time length in the current training sample into the adaptive multi-scale spatial hypergraph module to obtain an estimated intermediate feature sequence; Obtaining a first estimated ocean Niño index based on the estimated intermediate characteristic sequence, wherein the first estimated ocean Niño index is the estimated ocean Niño index for the first predetermined time period; Inputting the first estimated ocean Niño index and the second predetermined time length after the first predetermined time length into a frequency domain multi-scale time encoding and decoding module to obtain a second estimated ocean Niño index, wherein the second estimated ocean Niño index is the estimated ocean Niño index for the second predetermined time length after the first predetermined time length; Based on the second estimated ocean Niño index and the actual ocean Niño index, parameters of the adaptive multi-scale spatial hypergraph module and the frequency-domain multi-scale temporal encoding and decoding module are adjusted.

4. The method according to claim 3, wherein The training sample set is obtained based on the meteorological feature sequence of each meteorological feature among the multiple meteorological features of the global sea surface and the ocean Niño index, including: Obtain various meteorological characteristics of the global sea surface; Preprocessing of various global sea surface meteorological features to obtain meteorological characteristic sequences and ocean Niño index for each meteorological feature; Sequential sliding sampling of meteorological characteristic series and ocean Niño index for each meteorological feature; Based on the sampling results, a training sample set is obtained.

5. An El Niño-Southern Oscillation prediction device, characterized in that: include: an extraction unit configured to input a meteorological feature sequence of a target sea area in a target historical time period into an adaptive multi-scale spatial hypergraph module to obtain an intermediate feature sequence, wherein the target historical time period is a time period of a first predetermined length before a current moment, and the intermediate feature sequence includes a spatial dependency relationship between nodes of each hypergraph structure in a plurality of hypergraph structures, each hypergraph structure being obtained based on a feature sequence of the meteorological feature sequence at a corresponding grid scale among a plurality of grid scales; an acquisition unit configured to obtain the ocean Niño index of the target historical time period based on the intermediate characteristic sequence; a prediction unit configured to input the ocean Niño index of the target historical time period and the target future time period into a frequency domain multi-scale time encoding and decoding module to obtain the ocean Niño index of the target sea area in the target future time period, wherein the target future time period is a time period of a second predetermined length after the current moment, and the ocean Niño index of the target future time period represents the degree of El Niño-Southern Oscillation phenomenon in the target future time period; Wherein, the adaptive multi-scale spatial hypergraph module includes a multi-scale data processing unit and a plurality of parallel network blocks, the plurality of grid scales corresponding to the plurality of parallel network blocks one-to-one, the extraction unit is further configured to input the meteorological feature sequence of the target historical time period into the multi-scale data processing unit to obtain feature sequences of the meteorological feature sequence at a plurality of grid scales; for the feature sequence at each grid scale in the plurality of grid scales, input the current feature sequence into the network block corresponding to the current feature sequence to obtain hidden features of the current feature sequence, wherein the hidden features include the spatial dependency relationship between nodes in the hypergraph structure of the current feature sequence; and determine the hidden features of the feature sequences at the plurality of grid scales as the intermediate feature sequence; Among them, the frequency domain multi-scale time coding and decoding module includes a frequency domain embedding unit, a frequency domain encoding unit, a frequency domain decoding unit and a frequency domain fusion unit. The prediction unit is also configured to input the ocean Niño index of the target historical time period into the frequency domain embedding unit to obtain the time characteristics of different frequency segments; input the time characteristics of the different frequency segments into the frequency domain encoding unit to obtain coding characteristics; input the coding characteristics and the target future time period into the frequency domain decoding unit to obtain the predicted characteristics of the future frequency segment; input the predicted characteristics of the future frequency segment into the frequency domain fusion unit to obtain the ocean Niño index of the target sea area in the target future time period.

6. A computer-readable storage medium storing instructions, characterized in that: When the instructions are executed by at least one computing device, the at least one computing device is prompted to perform the El Niño-Southern Oscillation prediction method according to any one of claims 1 to 4.

7. A system comprising at least one computing device and at least one storage device storing instructions, characterized in that: When the instructions are executed by the at least one computing device, the instructions prompt the at least one computing device to perform the El Niño-Southern Oscillation prediction method according to any one of claims 1 to 4.

8. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the El Niño-Southern Oscillation prediction method according to any one of claims 1 to 4 is implemented.

Images