A method and system for time-space integrated processing of meteorological data

Through deep learning neural network, the integrated processing of meteorological data in space-time processing is solved, and the problems of low reliability and low efficiency of prediction results in traditional meteorological data processing are achieved, and higher accuracy and accurate meteorological data prediction are achieved to adapt to different regions and meteorological conditions. demand.

CN118349795BActive Publication Date: 2024-11-12FUJIAN PROVINCIAL METEOROLOGICAL INFORMATION CENT (FUJIAN PROVINCIAL METEOROLOGICAL ARCHIVES) +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410417183.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-08
Publication Date
2024-11-12
Estimated Expiration
2044-04-08

AI Technical Summary

Technical Problem

Traditional meteorological data processing relies on static numerical models and cannot fully consider the complexity and uncertainty of the meteorological system, resulting in low reliability of prediction results and low data cleaning and analysis efficiency.

Method used

Using the integrated processing method of meteorological data space-time processing based on deep learning neural network, we generate meteorological data in the area to be predicted by obtaining the meteorological data of the area to be predicted, and performing single-heat encoding, full-time-domain dimension encoding, local time-domain dimension enhancement encoding and sequence fusion to generate meteorological data time domain Autocorrelation fusion semantic features are used to intelligently predict meteorological data types.

Benefits of technology

It improves the accuracy and accuracy of meteorological data prediction, can better cope with meteorological changes and forecasting needs of meteorological conditions in different regions, and reduces the time and labor costs of data processing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118349795B_ABST
    Figure CN118349795B_ABST
Patent Text Reader

Abstract

The present application discloses a method and system for the integrated spatiotemporal processing of meteorological data, which obtains meteorological data of the area to be predicted, and uses data processing and analysis algorithms based on deep learning neural networks to perform semantic understanding and full-time domain fusion of the meteorological data of the area, so as to perform intelligent prediction of the type of meteorological data in the short term in the future according to the temporal changes of the meteorological data of the area. In this way, the characteristics of meteorological data on different time scales can be comprehensively considered to improve the precision and accuracy of meteorological data prediction, so as to better cope with the challenges brought by meteorological changes and adapt to the prediction needs of different regions and meteorological conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of intelligent processing, and more specifically, to a method and system for integrating time and space processing of meteorological data. Background Art

[0002] Meteorological data processing is used to manage and manipulate meteorological data for scientific research or prediction. These data usually include observations or model outputs of various meteorological elements such as temperature, humidity, wind speed, wind direction, air pressure, precipitation, etc. Meteorological data processing provides support for meteorological forecasting, climate research, disaster warning, etc.

[0003] Traditional meteorological data usually relies on static numerical meteorological models, which make predictions by numerically simulating the state and movement of the atmosphere. However, the meteorological system is a complex nonlinear system that is affected by many factors. This static prediction method cannot fully consider this uncertainty, resulting in low credibility of the prediction results. In addition, traditional meteorological data processing methods may require a lot of time and manpower for data cleaning, sorting and analysis, which is inefficient.

[0004] Therefore, a spatial and temporal integrated processing solution for meteorological data is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a method and system for the spatiotemporal integrated processing of meteorological data, which obtains the meteorological data of the area to be predicted, and uses a data processing and analysis algorithm based on a deep learning neural network to perform semantic understanding and full-time domain fusion of the meteorological data in the area, so as to make an intelligent prediction of the type of meteorological data in the short term in the future according to the temporal changes of the meteorological data in the area. In this way, the characteristics of meteorological data on different time scales can be comprehensively considered to improve the precision and accuracy of meteorological data prediction, so as to better cope with the challenges brought by meteorological changes and adapt to the prediction needs of different regions and meteorological conditions.

[0006] According to one aspect of the present application, a method for spatiotemporal integrated processing of meteorological data is provided, which comprises:

[0007] Obtain the time series of meteorological data for the area to be predicted;

[0008] One-hot encoding each meteorological data in the time series of the meteorological data to obtain a sequence of one-hot encoded vectors of the meteorological data;

[0009] Performing full-time-domain dimension encoding on the sequence of one-hot encoded meteorological data vectors to obtain a sequence of semantically encoded feature vectors of contextual meteorological data;

[0010] Passing the sequence of contextual meteorological data semantic encoding feature vectors through a local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local enhanced meteorological data semantic encoding feature vectors;

[0011] Sequentially fuse the sequence of the semantic encoding feature vectors of the local enhanced meteorological data to obtain the time-domain autocorrelation fusion semantic features of the meteorological data;

[0012] Based on the temporal autocorrelation fusion semantic features of the meteorological data, a prediction result is obtained.

[0013] According to another aspect of the present application, a meteorological data spatiotemporal integrated processing system is provided, comprising:

[0014] A meteorological data acquisition module for the area to be predicted is used to obtain the time series of meteorological data for the area to be predicted;

[0015] A meteorological data one-hot encoding module, used for one-hot encoding each meteorological data in the time series of the meteorological data to obtain a sequence of one-hot encoding vectors of the meteorological data;

[0016] A full-time-domain dimension encoding module, used for performing full-time-domain dimension encoding on the sequence of meteorological data one-hot encoding vectors to obtain a sequence of contextual meteorological data semantic encoding feature vectors;

[0017] A local time domain dimension enhancement module, used for passing the sequence of contextual meteorological data semantic encoding feature vectors through a local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local enhanced meteorological data semantic encoding feature vectors;

[0018] A sequence fusion module, used for performing sequence fusion on the sequence of the semantic encoding feature vectors of the local enhanced meteorological data to obtain a time-domain autocorrelation fusion semantic feature of the meteorological data;

[0019] The prediction result generation module is used to obtain the prediction result based on the temporal autocorrelation fusion semantic features of the meteorological data.

[0020] Compared with the prior art, the present application provides a method and system for the spatiotemporal integration of meteorological data, which obtains meteorological data of the area to be predicted, and uses data processing and analysis algorithms based on deep learning neural networks to perform semantic understanding and full-time domain fusion of the meteorological data in the area, so as to make intelligent predictions of the type of meteorological data in the short term in the future based on the temporal changes of the meteorological data in the area. In this way, the characteristics of meteorological data on different time scales can be comprehensively considered to improve the precision and accuracy of meteorological data prediction, so as to better cope with the challenges brought by meteorological changes and adapt to the prediction needs of different regions and meteorological conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 Flow chart of a method for spatiotemporal integrated processing of meteorological data according to an embodiment of the present application.

[0023] Figure 2 Schematic diagram of the architecture of the method for spatiotemporal integrated processing of meteorological data according to an embodiment of the present application.

[0024] Figure 3 This is a flowchart for training the full time domain dimension encoder based on the converter structure, the local time domain dimension enhancement encoder based on the one-dimensional extended convolutional neural network, the sequence fuser based on the autocorrelation attention network, and the classifier-based predictor in the spatiotemporal integrated processing method of meteorological data according to an embodiment of the present application.

[0025] Figure 4 4 is a block diagram of a system for integrating time and space for processing meteorological data according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0027] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0028] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0029] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0030] Meteorological data processing is used to manage and manipulate meteorological data for scientific research or prediction. These data usually include observations or model outputs of various meteorological elements such as temperature, humidity, wind speed, wind direction, air pressure, precipitation, etc. Meteorological data processing provides support for meteorological forecasting, climate research, disaster warning, etc.

[0031] Traditional meteorological data usually relies on static numerical meteorological models, which make predictions by numerically simulating the state and movement of the atmosphere. However, the meteorological system is a complex nonlinear system that is affected by many factors. This static prediction method cannot fully consider this uncertainty, resulting in low credibility of the prediction results. In addition, traditional meteorological data processing methods may require a lot of time and manpower for data cleaning, sorting and analysis, which is inefficient.

[0032] Therefore, in response to the above technical problems, the technical concept of this application is to obtain the meteorological data of the area to be predicted, and use the data processing and analysis algorithm based on deep learning neural network to perform semantic understanding and full time domain fusion of the meteorological data in the area, so as to make intelligent predictions on the type of meteorological data in the short term in the future according to the temporal changes of the meteorological data in the area. In this way, the characteristics of meteorological data on different time scales can be comprehensively considered to improve the precision and accuracy of meteorological data prediction, so as to better cope with the challenges brought by meteorological changes and adapt to the prediction needs of different regions and meteorological conditions.

[0033] Figure 1 Flow chart of a method for spatiotemporal integrated processing of meteorological data according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of the method for integrating time and space processing of meteorological data according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the method for time-space integrated processing of meteorological data includes: S110, obtaining the time series of meteorological data of the area to be predicted; S120, one-hot encoding each meteorological data in the time series of the meteorological data to obtain a sequence of one-hot encoded vectors of meteorological data; S130, encoding the sequence of one-hot encoded vectors of meteorological data in the entire time domain dimension to obtain a sequence of semantic encoded feature vectors of contextual meteorological data; S140, passing the sequence of semantic encoded feature vectors of contextual meteorological data through a local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local enhanced semantic encoded feature vectors of meteorological data; S150, performing sequence fusion on the sequence of local enhanced semantic encoded feature vectors of meteorological data to obtain time domain autocorrelation fusion semantic features of meteorological data; and, S160, obtaining a prediction result based on the time domain autocorrelation fusion semantic features of meteorological data.

[0034] In step S110, the time series of meteorological data of the area to be predicted is obtained. It should be understood that, considering that the meteorological data of the area to be predicted refers to the data collected and recorded the meteorological conditions of the area, these data usually include the measured values ​​of various meteorological elements, such as temperature, humidity, precipitation, wind speed and wind direction, etc. These meteorological data are of great significance for effective meteorological forecasting and responding to climate change. In particular, in a specific embodiment of the present application, the meteorological data of the area to be predicted can be collected by meteorological observation equipment such as meteorological stations, satellites, radars, etc. Based on this, in the technical solution of the present application, the time series of meteorological data of the area to be predicted is obtained, and the time series of the meteorological data is analyzed and correlated and fused in the entire time domain, so as to better make predictions of future short-term meteorological data types, so as to improve the accuracy and practicality of meteorological forecasts in the area.

[0035] In step S120, each meteorological data in the time series of the meteorological data is uniquely encoded to obtain a sequence of uniquely encoded vectors of meteorological data. Accordingly, considering that the time series of the meteorological data contains multiple meteorological data, some meteorological data such as weather conditions, precipitation types, etc. belong to categorical data, and machine learning algorithms usually need to input numerical data. Based on this, in order to better understand and analyze the semantic feature information contained in each meteorological data in the time series of the meteorological data in the future, in the technical solution of the present application, each meteorological data in the time series of the meteorological data is uniquely encoded to obtain a sequence of uniquely encoded vectors of meteorological data. It should be understood that one-hot encoding is a commonly used categorical data encoding method for converting categorical data into numerical data that can be processed by machine learning algorithms. In one-hot encoding, each categorical variable is converted into a vector with a length equal to the number of categories, in which only one element is 1 and the rest of the elements are 0. Specifically, if there are N different categories, each category will be encoded as a vector of length N, in which one element is 1, indicating the category, and the rest of the elements are 0. Therefore, converting each meteorological data in the time series of the meteorological data into a sequence of one-hot encoded vectors of numerical meteorological data that can be directly processed by a machine learning algorithm through one-hot encoding can help improve the accuracy and performance of the model.

[0036] In step S130, the sequence of meteorological data one-hot encoding vectors is encoded in the full time domain dimension to obtain a sequence of contextual meteorological data semantic encoding feature vectors. Specifically, in an embodiment of the present application, the sequence of meteorological data one-hot encoding vectors is encoded in the full time domain dimension to obtain a sequence of contextual meteorological data semantic encoding feature vectors, including: inputting the sequence of meteorological data one-hot encoding vectors into a full time domain dimension encoder based on a converter structure to obtain a sequence of contextual meteorological data semantic encoding feature vectors. It should be understood that, considering that there is a semantic association relationship between the contexts based on the time dimension between each meteorological data one-hot encoding vector in the sequence of meteorological data one-hot encoding vectors. Therefore, in order to better understand the semantic information between each meteorological data in the context of the time series in each meteorological data one-hot encoding vector, in the technical solution of the present application, the sequence of meteorological data one-hot encoding vectors is input into a full time domain dimension encoder based on a converter structure to obtain a sequence of contextual meteorological data semantic encoding feature vectors. It is worth mentioning that the converter structure can better capture the long-distance dependencies between different positions in the sequence. That is, the converter structure can capture the semantic association between each meteorological data at different time points, thereby obtaining a more meaningful semantic encoding feature representation of the contextual meteorological data.

[0037] In step S140, the sequence of the semantic coding feature vectors of the contextual meteorological data is passed through a local time domain dimension reinforcement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local reinforcement semantic coding feature vectors of meteorological data. Accordingly, considering that the sequences of the semantic coding feature vectors of the contextual meteorological data have certain local implicit feature information in the time dimension, that is, the semantic coding features of the contextual meteorological data at a certain moment are usually related to the semantic coding features of the contextual meteorological data at nearby moments, and there is also semantic association information about meteorological data between the local time series semantic coding feature vectors of the contextual meteorological data between different time lengths in the semantic coding feature vectors of the contextual meteorological data. However, the full time domain dimension encoder has a good perception and capture ability for the long-distance dependent and associated meteorological data time series features, but it is easy to ignore the semantic association between the local time series features of the meteorological data. Therefore, in order to focus more closely on the local temporal semantic coding features of contextual meteorological data under different time spans in the sequence of contextual meteorological data semantic coding feature vectors, so as to better characterize the sequence of contextual meteorological data semantic coding features, in the technical solution of the present application, the sequence of contextual meteorological data semantic coding feature vectors is passed through a local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local enhanced meteorological data semantic coding feature vectors. It should be understood that the one-dimensional extended convolutional neural network controls the number of null values ​​0 in the convolution kernel by introducing a hyperparameter called the expansion rate, so that the contextual semantic associations between the local time domains of meteorological data at different time lengths of the contextual meteorological data semantic coding feature vectors can be captured to better characterize the sequence of the local enhanced meteorological data semantic coding feature vectors.

[0038] Specifically, in an embodiment of the present application, the sequence of the contextual meteorological data semantic encoding feature vectors is passed through a local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local enhanced meteorological data semantic encoding feature vectors, including: the sequence of the contextual meteorological data semantic encoding feature vectors is processed through the local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network with the following extended convolution formula to obtain the sequence of the local enhanced meteorological data semantic encoding feature vectors;

[0039] Wherein, the extended convolution formula is:

[0040]

[0041] C i =f(ω·X i:i+j-l +b)

[0042] Among them, X1, X2, ..., Xn Respectively represent the respective contextual meteorological data semantic encoding feature vectors in the sequence of contextual meteorological data semantic encoding feature vectors, X 1:n a concatenated vector representing a sequence of semantically encoded feature vectors of the contextual meteorological data, Indicates the connection operation, X i:i+j-l X in the sequence representing the semantic encoding feature vector of the contextual meteorological data i , X i+1 , ..., X i+j-1 , ωω and b represent the weight matrix and bias vector respectively, f(·) represents the convolution operation, C i Represents the i-th local enhanced meteorological data semantic encoding feature vector in the sequence of the local enhanced meteorological data semantic encoding feature vectors.

[0043] In step S150, the sequence of the local enhanced meteorological data semantic coding feature vectors is sequence fused to obtain the meteorological data time-domain autocorrelation fusion semantic features. Specifically, in an embodiment of the present application, the sequence of the local enhanced meteorological data semantic coding feature vectors is sequence fused to obtain the meteorological data time-domain autocorrelation fusion semantic features, including: passing the sequence of the local enhanced meteorological data semantic coding feature vectors through a sequence fuser based on an autocorrelation attention network to obtain the meteorological data time-domain autocorrelation fusion semantic feature vector as the meteorological data time-domain autocorrelation fusion semantic feature. It should be understood that, considering the complex interrelated influences and dependencies between the sequences of the local enhanced meteorological data semantic coding feature vectors in the time dimension, for example, as the temperature rises over time, the humidity in the air may decrease, because warm air can accommodate more water vapor, resulting in a decrease in relative humidity. At the same time, there is also some redundant information accordingly. Therefore, in order to better extract and integrate the temporal semantic feature information between each meteorological data in the sequence of the local enhanced meteorological data semantic coding feature vector, in the technical solution of the present application, the sequence of the local enhanced meteorological data semantic coding feature vector is passed through a sequence fuser based on an autocorrelation attention network to obtain a meteorological data temporal autocorrelation fusion semantic feature vector. It is worth mentioning that the autocorrelation attention mechanism can dynamically learn the correlation between the semantic features of meteorological data at different time points, thereby assigning different weights to different features. This dynamic weight allocation enables the model to more flexibly process the temporal semantic information of different meteorological data, thereby obtaining a meteorological data temporal autocorrelation fusion semantic feature vector with more temporal autocorrelation fusion characteristics, providing a richer and more accurate feature representation for subsequent tasks.

[0044] More specifically, in an embodiment of the present application, the sequence of the semantic encoding feature vectors of the local enhanced meteorological data is passed through a sequence fusion device based on an autocorrelation attention network to obtain a meteorological data temporal autocorrelation fusion semantic feature vector as the meteorological data temporal autocorrelation fusion semantic feature, including: using the sequence fusion device based on the autocorrelation attention network to process the sequence of the semantic encoding feature vectors of the local enhanced meteorological data with the following autocorrelation extraction formula to obtain the meteorological data temporal autocorrelation fusion semantic feature vector;

[0045] Wherein, the autocorrelation extraction formula is:

[0046]

[0047]

[0048]

[0049] Among them, e i represents the attention scoring mechanism value of the i-th local enhanced meteorological data semantic encoding feature vector in the sequence of the local enhanced meteorological data semantic encoding feature vector, x i represents the i-th local enhanced meteorological data semantic coding feature vector in the sequence of the local enhanced meteorological data semantic coding feature vectors, and W i represents the weight coefficient vector and weight coefficient matrix, B i is the offset vector, tanh(·) represents the tanh function, a i represents the attention weight value of the i-th local enhanced meteorological data semantic encoding feature vector in the sequence of the local enhanced meteorological data semantic encoding feature vector, e k represents the attention scoring mechanism value of the kth local enhanced meteorological data semantic encoding feature vector, exp(·) represents an exponential function with the natural constant e as the base, and F represents the time domain autocorrelation fusion semantic feature vector of the meteorological data.

[0050] In step S160, a prediction result is obtained based on the time-domain autocorrelation fusion semantic features of the meteorological data. Specifically, in an embodiment of the present application, a prediction result is obtained based on the time-domain autocorrelation fusion semantic features of the meteorological data, including: passing the time-domain autocorrelation fusion semantic feature vector of the meteorological data through a classifier-based predictor to obtain a prediction result, and the prediction result is used to represent the type label of the short-term meteorological data. That is, the time-domain autocorrelation fusion semantic features of the meteorological data obtained by autocorrelation attention fusion of the sequence of the local enhanced meteorological data semantic encoding feature vectors are used for classification processing, so as to perform intelligent prediction of the type of meteorological data in the short term in the future according to the temporal changes of the meteorological data in the area. In this way, the characteristics of meteorological data on different time scales can be comprehensively considered, so as to improve the precision and accuracy of meteorological data prediction, so as to better cope with the challenges brought by meteorological changes and adapt to the prediction needs of different regions and meteorological conditions.

[0051] It is worth mentioning that those skilled in the art should be aware that before applying a deep neural network model for inference, the deep neural network model must be trained so that the deep neural network can implement specific functional capabilities.

[0052] Specifically, in an embodiment of the present application, it also includes a method for training the full time domain dimension encoder based on the converter structure, the local time domain dimension enhancement encoder based on the one-dimensional extended convolutional neural network, the sequence fuser based on the autocorrelation attention network, and the classifier-based predictor.

[0053] Figure 3 The present invention is a flowchart for training the full-time-domain dimension encoder based on the converter structure, the local-time-domain dimension enhancement encoder based on the one-dimensional extended convolutional neural network, the sequence fusion device based on the autocorrelation attention network, and the predictor based on the classifier in the spatiotemporal integrated processing method of meteorological data according to an embodiment of the present application. Figure 3As shown, the training step includes: S210, obtaining training data, the training data including the time series of training meteorological data of the area to be predicted, and the true value of the type label of the short-term meteorological data; S220, performing one-hot encoding on each training meteorological data in the time series of the training meteorological data to obtain a sequence of one-hot encoded vectors of the training meteorological data; S230, inputting the sequence of one-hot encoded vectors of the training meteorological data into the full time domain dimension encoder based on the converter structure to obtain a sequence of semantic encoding feature vectors of training context meteorological data; S240, passing the sequence of semantic encoding feature vectors of the training context meteorological data through the local time domain dimension reinforcement encoder based on the one-dimensional extended convolutional neural network to obtain a sequence of semantic encoding feature vectors of training local reinforcement meteorological data; S250, passing the sequence of semantic encoding feature vectors of the training local reinforcement meteorological data through the sequence fusion device based on the autocorrelation attention network to obtain to the time-domain autocorrelation fusion semantic feature vector of training meteorological data; S260, feature optimization is performed on the time-domain autocorrelation fusion semantic feature vector of training meteorological data to obtain the optimized time-domain autocorrelation fusion semantic feature vector of training meteorological data; S270, the optimized time-domain autocorrelation fusion semantic feature vector of training meteorological data is passed through the classifier-based predictor to obtain the classification loss function value; and, S280, based on the classification loss function value and through the back propagation of gradient descent, the full time-domain dimension encoder based on the converter structure, the local time-domain dimension reinforcement encoder based on the one-dimensional extended convolutional neural network, the sequence fuser based on the autocorrelation attention network and the classifier-based predictor are trained, wherein each time the time-domain autocorrelation fusion semantic feature vector of training meteorological data is iteratively optimized when it passes through the classifier-based predictor for classification regression iteration.

[0054] Specifically, in step S280, the full-time-domain dimension encoder based on the transformer structure, the local-time-domain dimension reinforcement encoder based on the one-dimensional extended convolutional neural network, the sequence fusion based on the autocorrelation attention network, and the predictor based on the classifier are trained based on the classification loss function value and through the back propagation of gradient descent, wherein each time the training meteorological data time-domain autocorrelation fusion semantic feature vector is iteratively optimized when the classification regression is iterated by the predictor based on the classifier. It should be understood that in the technical solution of the present application, each training meteorological data unique hot encoding vector in the sequence of the training meteorological data unique hot encoding vectors respectively represents the unique hot encoding vectorization representation of the meteorological elements at each predetermined time point. When the sequence of the training meteorological data unique hot encoding vectors is input into the full-time-domain dimension encoder based on the transformer structure, its essence is to use the Transformer mechanism of the full-time-domain dimension encoder based on the transformer structure to perform full-time-domain meteorological element context semantic association encoding based on the multi-head attention mechanism on the sequence of the training meteorological data unique hot encoding vectors to capture the context semantic association information of the meteorological elements in the time domain space. The sequence of semantic coding feature vectors of training context meteorological data is passed through a local time domain dimension enhanced encoder based on a one-dimensional extended convolutional neural network. Its purpose is to capture the high-dimensional semantic features of meteorological elements at each predetermined time point through one-dimensional convolutional coding and the semantic association features of meteorological elements in the local time domain determined by the one-dimensional convolution kernel to obtain a sequence of semantic coding feature vectors of training local enhanced meteorological data.

[0055] Considering that each training context meteorological data semantic coding feature vector in the sequence of the training context meteorological data semantic coding feature vector has a difference in feature expression in the semantic feature space, the sequence of the training local enhanced meteorological data semantic coding feature vector also has obvious feature distribution inconsistency. When the sequence of the training local enhanced meteorological data semantic coding feature vector passes through the sequence fusion device based on the autocorrelation attention network, the autocorrelation attention enhancement can be performed based on the respective feature distribution of each training local enhanced meteorological data semantic coding feature vector in the sequence of the training local enhanced meteorological data semantic coding feature vector, thereby improving the significance and importance of the obtained training meteorological data time domain autocorrelation fusion semantic feature vector in the time dimension. However, considering that the sequence of the training local enhanced meteorological data semantic coding feature vector also has obvious feature distribution inconsistency, the autocorrelation attention enhancement will also lead to the inconsistency of the local feature distribution of the training meteorological data time domain autocorrelation fusion semantic feature vector based on the numerical representation of the feature value, affecting the convergence effect of the training meteorological data time domain autocorrelation fusion semantic feature vector through the class regression iteration of the classifier-based predictor. Based on this, in the technical solution of the present application, each time the time-domain autocorrelation fusion semantic feature vector of the training meteorological data is iteratively optimized when the classification regression is iterated through the classifier-based predictor.

[0056] More specifically, in an embodiment of the present application, each time the time-domain autocorrelation fusion semantic feature vector of the training meteorological data is iteratively optimized when the classification regression is iterated through the classifier-based predictor, including: each time the time-domain autocorrelation fusion semantic feature vector of the training meteorological data is iteratively optimized through the classifier-based predictor, the time-domain autocorrelation fusion semantic feature vector of the training meteorological data is iteratively optimized using the following optimization formula to obtain an optimized time-domain autocorrelation fusion semantic feature vector of the training meteorological data; wherein the optimization formula is:

[0057]

[0058] Among them, v i is the i-th eigenvalue of the temporal autocorrelation fusion semantic feature vector of the training meteorological data, φ(v i ) represents a probabilistic function of the eigenvalue, for example, a probabilistic function based on mapping the maximum eigenvalue to the interval [0, 1], and p is a class probability value obtained by a predictor based on a classifier, and L is the time domain autocorrelation fusion semantic feature vector v of the training meteorological data i length, α is the weight hyperparameter, exp(·) represents the exponential function with the natural constant e as the base, vi It is the i-th eigenvalue of the time-domain autocorrelation fusion semantic feature vector of the optimized training meteorological data.

[0059] In particular, here, the feature value v of the temporal autocorrelation fusion semantic feature vector of the training meteorological data is carried out by taking the Bernoulli probability likelihood distribution as the authoritative meta-entity i Based on the high-level class decision of Bernoulli probability distribution (p, 1-p), and introducing the probability density space mapping allocation relative to the whole feature set, the eigenvalue v of the time domain autocorrelation fusion semantic feature vector of the training meteorological data is determined. i The key insight relative to the class probability mapping is to achieve the eigenvalue v of the semantic feature vector fused with temporal autocorrelation of the training meteorological data. i Based on the correctness of the matching function of class probability distribution (p, 1-p) and class convergence, the convergence performance is superior to promote sample iteration convergence and improve the classification training speed. In this way, the characteristics of meteorological data on different time scales can be comprehensively considered to improve the precision and accuracy of meteorological data prediction, so as to better cope with the challenges brought by meteorological changes and adapt to the prediction needs of different regions and meteorological conditions.

[0060] In summary, the spatiotemporal integrated processing method of meteorological data based on the embodiment of the present application is explained, which obtains the meteorological data of the area to be predicted, and uses the data processing and analysis algorithm based on deep learning neural network to perform semantic understanding and full time domain fusion of the meteorological data in the area, so as to make intelligent predictions on the type of meteorological data in the short term in the future according to the temporal changes of the meteorological data in the area. In this way, the characteristics of meteorological data on different time scales can be comprehensively considered to improve the precision and accuracy of meteorological data prediction, so as to better cope with the challenges brought by meteorological changes and adapt to the prediction needs of different regions and meteorological conditions.

[0061] Figure 4 FIG. 1 is a block diagram of a system for integrating time and space processing of meteorological data according to an embodiment of the present application. Figure 4As shown, according to the meteorological data spatiotemporal integrated processing system 100 of the embodiment of the present application, it includes: a meteorological data acquisition module 110 for the area to be predicted, which is used to obtain the time series of meteorological data in the area to be predicted; a meteorological data one-hot encoding module 120, which is used to one-hot encode each meteorological data in the time series of the meteorological data to obtain a sequence of meteorological data one-hot encoding vectors; a full-time domain dimension encoding module 130, which is used to encode the sequence of meteorological data one-hot encoding vectors in the full-time domain dimension to obtain a sequence of contextual meteorological data semantic encoding feature vectors; a local time domain dimension enhancement module 140, which is used to pass the sequence of contextual meteorological data semantic encoding feature vectors through a local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local enhanced meteorological data semantic encoding feature vectors; a sequence fusion module 150, which is used to sequence fuse the sequence of local enhanced meteorological data semantic encoding feature vectors to obtain meteorological data time domain autocorrelation fusion semantic features; and a prediction result generation module 160, which is used to obtain a prediction result based on the meteorological data time domain autocorrelation fusion semantic features.

[0062] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned meteorological data spatiotemporal integrated processing system have been described in the above reference. Figures 1 to 3 The description of the spatiotemporal integrated processing method of meteorological data has been introduced in detail, and therefore, its repeated description will be omitted.

[0063] As described above, the meteorological data spatiotemporal integrated processing system 100 according to the embodiment of the present disclosure can be implemented in various wireless terminals, such as a server with a meteorological data spatiotemporal integrated processing algorithm. In a possible implementation, the meteorological data spatiotemporal integrated processing system 100 according to the embodiment of the present disclosure can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the meteorological data spatiotemporal integrated processing system 100 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the meteorological data spatiotemporal integrated processing system 100 can also be one of the many hardware modules of the wireless terminal.

[0064] Alternatively, in another example, the meteorological data spatiotemporal integrated processing system 100 and the wireless terminal may also be separate devices, and the meteorological data spatiotemporal integrated processing system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0065] The foregoing is merely an example of the principles of the present disclosure, and various modifications may be made by those skilled in the art without departing from the scope of the present disclosure. The above embodiments are presented for the purpose of illustration and not limitation. The present disclosure may also take many forms other than those explicitly described herein. Therefore, it is emphasized that the present disclosure is not limited to the methods, systems and devices explicitly disclosed, but is intended to include variations and modifications within the spirit and scope of the appended claims.

Claims

1. A method for integrating time and space in meteorological data processing, characterized in that: include: Obtain the time series of meteorological data for the area to be predicted; One-hot encoding each meteorological data in the time series of the meteorological data to obtain a sequence of one-hot encoded vectors of the meteorological data; Performing full-time-domain dimension encoding on the sequence of one-hot encoded meteorological data vectors to obtain a sequence of semantically encoded feature vectors of contextual meteorological data; Passing the sequence of contextual meteorological data semantic encoding feature vectors through a local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local enhanced meteorological data semantic encoding feature vectors; Sequentially fuse the sequence of the semantic encoding feature vectors of the local enhanced meteorological data to obtain the time-domain autocorrelation fusion semantic features of the meteorological data; Obtaining a prediction result based on the temporal autocorrelation fusion semantic features of the meteorological data; Performing full-time-domain dimension encoding on the sequence of meteorological data one-hot encoded vectors to obtain a sequence of contextual meteorological data semantic encoding feature vectors, including: inputting the sequence of meteorological data one-hot encoded vectors into a full-time-domain dimension encoder based on a converter structure to obtain a sequence of contextual meteorological data semantic encoding feature vectors; The sequence of the local enhanced meteorological data semantic coding feature vectors is sequence-fused to obtain the meteorological data time-domain autocorrelation fusion semantic features, including: passing the sequence of the local enhanced meteorological data semantic coding feature vectors through a sequence fuser based on an autocorrelation attention network to obtain the meteorological data time-domain autocorrelation fusion semantic feature vector as the meteorological data time-domain autocorrelation fusion semantic features.

2. The method for time-space integrated processing of meteorological data according to claim 1, characterized in that: The sequence of the contextual meteorological data semantic coding feature vectors is passed through a local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local enhanced meteorological data semantic coding feature vectors, including: the sequence of the contextual meteorological data semantic coding feature vectors is processed through the local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network using the following extended convolution formula to obtain the sequence of the local enhanced meteorological data semantic coding feature vectors; Wherein, the extended convolution formula is: C i =f(ω·X i:i+j-l +b) Among them, X1, X2, ..., X n Respectively represent the respective contextual meteorological data semantic encoding feature vectors in the sequence of contextual meteorological data semantic encoding feature vectors, X 1:n a concatenated vector representing a sequence of semantically encoded feature vectors of the contextual meteorological data, Indicates the connection operation, X i:i+j-l X in the sequence representing the semantic encoding feature vector of the contextual meteorological data i , X i+1 , ..., X i+j-1 , ω and b represent the weight matrix and bias vector respectively, f(·) represents the convolution operation, C i Represents the i-th local enhanced meteorological data semantic encoding feature vector in the sequence of the local enhanced meteorological data semantic encoding feature vectors.

3. The method for time-space integrated processing of meteorological data according to claim 1, characterized in that: The sequence of the semantic encoding feature vectors of the local enhanced meteorological data is passed through a sequence fusion device based on an autocorrelation attention network to obtain a meteorological data temporal autocorrelation fusion semantic feature vector as the meteorological data temporal autocorrelation fusion semantic feature, including: using the sequence fusion device based on the autocorrelation attention network to process the sequence of the semantic encoding feature vectors of the local enhanced meteorological data with the following autocorrelation extraction formula to obtain the meteorological data temporal autocorrelation fusion semantic feature vector; Wherein, the autocorrelation extraction formula is: Among them, e i represents the attention scoring mechanism value of the lth local enhanced meteorological data semantic encoding feature vector in the sequence of the local enhanced meteorological data semantic encoding feature vector, x i represents the i-th local enhanced meteorological data semantic encoding feature vector in the sequence of the local enhanced meteorological data semantic encoding feature vectors, and W i represents the weight coefficient vector and weight coefficient matrix, B i is the offset vector, tanh(·) represents the tanh function, a i represents the attention weight value of the lth local enhanced meteorological data semantic encoding feature vector in the sequence of the local enhanced meteorological data semantic encoding feature vector, e k represents the attention scoring mechanism value of the kth local enhanced meteorological data semantic encoding feature vector, exp(·) represents an exponential function with the natural constant e as the base, and F represents the time domain autocorrelation fusion semantic feature vector of the meteorological data.

4. The method for time-space integrated processing of meteorological data according to claim 3, characterized in that: Based on the time-domain autocorrelation fusion semantic features of the meteorological data, a prediction result is obtained, including: passing the time-domain autocorrelation fusion semantic feature vector of the meteorological data through a classifier-based predictor to obtain a prediction result, and the prediction result is used to represent the type label of the short-term meteorological data.

5. The method for time-space integrated processing of meteorological data according to claim 4, characterized in that: It also includes a training step: for training the full time domain dimension encoder based on the converter structure, the local time domain dimension enhancement encoder based on the one-dimensional extended convolutional neural network, the sequence fuser based on the autocorrelation attention network and the classifier-based predictor.

6. The method for time-space integrated processing of meteorological data according to claim 5, characterized in that: The training step comprises: Acquire training data, wherein the training data includes a time series of training meteorological data of a region to be predicted and a true value of a type label of the short-term meteorological data; One-hot encoding each training meteorological data in the time series of the training meteorological data to obtain a sequence of one-hot encoded vectors of the training meteorological data; Inputting the sequence of one-hot encoded vectors of the training meteorological data into the full-time-domain dimensional encoder based on the transformer structure to obtain a sequence of semantically encoded feature vectors of the training context meteorological data; Passing the sequence of semantically encoded feature vectors of the training context meteorological data through the local time domain dimension enhancement encoder based on the one-dimensional extended convolutional neural network to obtain a sequence of semantically encoded feature vectors of the training local enhanced meteorological data; Passing the sequence of the semantic encoding feature vectors of the training local enhanced meteorological data through the sequence fusion device based on the autocorrelation attention network to obtain the time domain autocorrelation fusion semantic feature vector of the training meteorological data; Performing feature optimization on the time-domain autocorrelation fusion semantic feature vector of the training meteorological data to obtain an optimized time-domain autocorrelation fusion semantic feature vector of the training meteorological data; Passing the optimized training meteorological data time domain autocorrelation fusion semantic feature vector through the classifier-based predictor to obtain a classification loss function value; The full time domain dimension encoder based on the converter structure, the local time domain dimension reinforcement encoder based on the one-dimensional extended convolutional neural network, the sequence fuser based on the autocorrelation attention network and the classifier-based predictor are trained based on the classification loss function value and through back propagation of gradient descent, wherein the time domain autocorrelation fusion semantic feature vector of the training meteorological data is iteratively optimized each time the time domain autocorrelation fusion semantic feature vector of the training meteorological data is iteratively classified and regressed through the classifier-based predictor.

7. A meteorological data spatiotemporal integrated processing system, characterized in that: include: A meteorological data acquisition module for the area to be predicted is used to obtain the time series of meteorological data for the area to be predicted; A meteorological data one-hot encoding module, used for one-hot encoding each meteorological data in the time series of the meteorological data to obtain a sequence of one-hot encoding vectors of the meteorological data; A full-time-domain dimension encoding module, used for performing full-time-domain dimension encoding on the sequence of meteorological data one-hot encoding vectors to obtain a sequence of contextual meteorological data semantic encoding feature vectors; A local time domain dimension enhancement module, used for passing the sequence of contextual meteorological data semantic encoding feature vectors through a local time domain dimension enhancement encoder based on a one-dimensional extended convolutional neural network to obtain a sequence of local enhanced meteorological data semantic encoding feature vectors; A sequence fusion module, used for performing sequence fusion on the sequence of the semantic encoding feature vectors of the local enhanced meteorological data to obtain a time-domain autocorrelation fusion semantic feature of the meteorological data; A prediction result generation module is used to obtain a prediction result based on the temporal autocorrelation fusion semantic features of the meteorological data. The full-time-domain dimensional encoding module is used to: input the sequence of meteorological data one-hot encoding vectors into a full-time-domain dimensional encoder based on a converter structure to obtain a sequence of contextual meteorological data semantic encoding feature vectors; The sequence fusion module is used to: pass the sequence of the local enhanced meteorological data semantic encoding feature vector through a sequence fuser based on an autocorrelation attention network to obtain a meteorological data time domain autocorrelation fusion semantic feature vector as the meteorological data time domain autocorrelation fusion semantic feature.