Convection boundary layer height identification method and device, electronic equipment and storage medium

Abstract

The invention discloses a convection boundary layer height identification method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining high-altitude detection data; the high-altitude detection data is preprocessed; embedding time information into the preprocessed high-altitude detection data through position coding, and obtaining coded high-altitude detection data through a long-short term memory network; constructing a feature information capturing module based on a self-attention mechanism, and capturing meteorological information of the encoded high-altitude detection data by using the feature information capturing module; and converting the meteorological information into the convection boundary layer height through average pooling and linear mapping operation. The identification precision of the method is reduced by about 60-90m compared with the average absolute error of the existing optimal method in different seasons, and the performance is still kept stable when the effect test is carried out in the extreme arid region.

Description

Technical Field

[0001] The present application relates to the field of boundary layer height identification, and specifically to a convective boundary layer height identification method and device, electronic equipment, and storage medium. Background Art

[0002] The atmospheric boundary layer (ABL) is the lowest layer of direct interaction between Earth's atmosphere and the Earth's surface. Its height is a crucial parameter for describing its structure and dynamics. During the day, the ABL manifests as a convective boundary layer, and its height significantly influences important weather mechanisms such as dust storms, dry-wet climate variations, and monsoon precipitation. Traditional identification methods, often based on sounding data, deduce the BBL height using physical parameters such as temperature, humidity, and wind speed profiles. These methods include the relative humidity gradient method, the Liu-Liang method, the Richardson method, the potential temperature gradient method, and the dry adiabatic curve method. However, these methods often suffer from limited accuracy due to surface heterogeneity. With the advancement of machine learning, data-driven methods for identifying the BBL height have gradually emerged. For example, methods based on the XGBoost model have achieved good results in calculating the BBL height in the Beijing area. Deep forest algorithms, which fuse diverse remote sensing data, also achieve significantly higher accuracy in identifying the BBL height than traditional methods. However, these methods have limitations in processing time series data and capturing complex spatiotemporal features, making it difficult to effectively extract deep features. Summary of the Invention

[0003] In order to overcome the above technical problems, the present application implements and provides a convective boundary layer height identification method and device, electronic equipment, and storage medium to ensure the accuracy of identification.

[0004] According to a first aspect of an embodiment of the present application, a method for identifying a convective boundary layer height is provided, comprising: Acquire high-altitude detection data; Preprocessing the high-altitude detection data; The time information is embedded into the pre-processed high-altitude detection data through position coding, and the encoded high-altitude detection data is obtained through the long short-term memory network; Building a feature information capture module based on the self-attention mechanism, and using the feature information capture module to capture meteorological information of the encoded high-altitude sounding data; The meteorological information is converted into convective boundary layer height through average pooling and linear mapping operations.

[0005] According to a second aspect of an embodiment of the present application, a device for identifying a convective boundary layer height is provided, comprising: Data acquisition module, used to obtain high-altitude detection data; A preprocessing module, used for preprocessing the high-altitude detection data; The data encoding module is used to embed time information into the pre-processed high-altitude detection data through position encoding, and obtain the encoded high-altitude detection data through the long short-term memory network; An information capture module is used to construct a feature information capture module based on a self-attention mechanism, and use the feature information capture module to capture meteorological information of the encoded high-altitude sounding data; The identification module is used to convert the meteorological information into the convective boundary layer height through average pooling and linear mapping operations.

[0006] According to a third aspect of the embodiments of the present application, there is provided an electronic device, including: one or more sensors; one or more processors; a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in the first aspect.

[0007] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which computer instructions are stored. When the instructions are executed by a processor, the steps of the method described in the first aspect are implemented.

[0008] The technical solutions provided by the embodiments of the present application may have the following beneficial effects: This paper designs a method for identifying convective boundary layer height. It preprocesses high-altitude sounding data to eliminate interference from wind speed zero crossings. It uses positional encoding to capture temporal variations in convective boundary layer height, which is then encoded using a long-short-term memory network to effectively extract relevant information. A self-attention mechanism efficiently captures the characteristic information of the high-altitude sounding data, and transforms the captured features into convective boundary layer height through average pooling and linear mapping, effectively enhancing the model's applicability and accuracy. This method achieves a reduction in mean absolute error of approximately 60-90 meters across different seasons compared to the best existing methods, and maintains stable performance when tested in extremely arid regions.

[0009] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0011] Figure 1 This is a flow chart of a method for identifying the height of a convective boundary layer provided in an embodiment of the present application.

[0012] Figure 2 This is a block diagram of a convective boundary layer height identification device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0013] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0014] The convective boundary layer height identification method provided in this application can be applied to the convective boundary layer height identification in almost all regions. The following is a detailed explanation using the high-altitude detection data of the Korla station as an example.

[0015] Figure 1 This is a flow chart of a method for identifying the height of a convective boundary layer provided by the present application according to an exemplary implementation example. Figure 1 As shown, the method may include the following steps: S1: Acquire high-altitude detection data; Specifically, high-altitude detection sensors are used to collect meteorological elements such as temperature, air pressure, wind direction, and wind speed.

[0016] S2: Preprocessing the high-altitude detection data; this step may include the following sub-steps: S21: Acquire various meteorological element data collected by high-altitude detection sensors, and convert wind speed and direction into longitudinal wind and latitudinal wind; Specifically, for the two meteorological elements of wind direction and wind speed, they are converted into vector wind representation, that is, decomposed into longitudinal wind and latitudinal wind to eliminate the interference of wind direction crossing zero. The conversion formula is as follows: in, u For the meridian wind, v is the zonal wind, V is the wind speed, θ For wind direction.

[0017] S22: Based on the meteorological element data, perform data quality control, remove outliers, fill in missing values, and construct a high-altitude detection data matrix; Specifically, the abnormal data is removed, and the missing data and the missing data positions after removal are filled by linear interpolation. The filling formula is as follows: in, x 1, x2 is the height of the known position, y 1, y 2 is the corresponding observation value, such as temperature, wind speed, etc. x is the position to be interpolated, y is the observed value obtained after interpolation. n The high-altitude detection data matrix X of different meteorological elements is x 1, x 2,……, x n}, each meteorological element x i It also includes h values ​​of different heights, i.e. x i ={ x i 1 , x i 2 ,……, x i h}.

[0018] S23: normalizing the high-altitude detection data matrix using a maximum normalization method to obtain normalized high-altitude detection data; Specifically, for each meteorological element variable sequence x i , based on the observed values, the maximum and minimum values ​​of the variable are obtained, and through The high-altitude detection data after the maximum value normalization is calculated is X'={ x 1', x 2', ..., x n '}.

[0019] S3: embedding time information into the normalized high-altitude detection data through position encoding, and obtaining the encoded high-altitude detection data through a long short-term memory network; this step may include the following sub-steps: S31: Use position encoding to encode the temporal information of the upper-altitude sounding data into a vector and embed it into the normalized upper-altitude sounding data matrix to capture the temporal dimension of the convective boundary layer height. Specifically, the position encoding of the monthly and daily time elements of the high-altitude sounding data is first performed. This is because the convective boundary layer height has obvious seasonal variations. The encoding formula is as follows: in POS Indicates the position of the month or day, i represents the index of the encoding dimension, dmodel Represents the total dimension of the position encoding. After position encoding, the month encoding PE is obtained month , date code PE day , and perform splicing operation with the normalized high-altitude detection data to obtain the new high-altitude detection data X PE .

[0020] S32: Encode the high-altitude detection data matrix after embedding the time information through a long short-term memory network to obtain encoded high-altitude detection data; Specifically, the high-altitude detection data matrix X after embedding time information PE The data is input into the Long Short-Term Memory (LSTM) encoder for encoding to effectively capture the internal temporal correlation of the elements in the high-altitude detection data and obtain the encoded high-altitude detection data H. The encoding formula is as follows: .

[0021] S4: Constructing a feature information capture module based on the self-attention mechanism, and using the feature information capture module to capture meteorological information from the encoded high-altitude sounding data; this step may include the following sub-steps: S41: Construct a multi-layer self-attention module based on the multi-head self-attention mechanism and residual normalization structure; Specifically, a multi-layered "multi-head self-attention layer + residual connection and normalization" structure was designed, combining the sequence modeling capabilities of long short-term memory networks to further enhance the model's ability to focus on key features. Each layer of multi-head self-attention captures information from different subspaces of the input features by computing multiple attention distributions in parallel. To enhance model stability and prevent gradient vanishing, residual connections and layer normalization are added after each layer of multi-head self-attention.

[0022] S42: Using a multi-layer self-attention module to extract features from the encoded high-altitude detection data to obtain meteorological data; Specifically, the multi-head self-attention calculation formula is as follows: in, , W Q , W K , W V They represent the learnable parameter matrices, W O is the output projection matrix, Q , K , VThey are query, key, and value vectors respectively. d k is the dimension of the key vector, which is used to scale the dot product to stabilize the gradient. Through this module, the vector matrix Z is obtained.

[0023] S5: Converting the meteorological information into convective boundary layer height through average pooling and linear mapping operations; this step may include the following sub-steps: S51: Using an average pooling layer to reduce the dimension of the meteorological information to extract effective compressed sequence information and retain key features; Specifically, the input feature sequence is reduced in dimension through the average pooling layer to extract global features. The formula is as follows: in, Z t is the t-th time step feature output by the attention mechanism, that is, the data feature of the t-th sounding altitude layer, T is the length of the high-altitude detection data sequence, Z pool is the key feature of the output.

[0024] S52: Map the pooled features to a one-dimensional output through a linear connection layer to obtain the final convective boundary layer height; Specifically, the pooled features are mapped to a one-dimensional output through a linear connection layer to represent the convective boundary layer height, as follows: in W fc and b fc are the weights and biases of the linear layer, both of which are learnable parameters, and we finally get This is the convective boundary layer height.

[0025] This application uses high-altitude detection data to identify the convective boundary layer height through a deep learning method, effectively improving the recognition accuracy. Through data processing, the wind is decomposed into vector wind, effectively eliminating the interference problem caused by wind speed crossing zero in the model, and improving the performance of the deep learning model; using position encoding to embed time information into the normalized high-altitude detection data to capture the temporal characteristics of the convective boundary layer height change, and using LSTM encoding to capture the intrinsic characteristics of the high-altitude detection data; based on the multi-head self-attention mechanism, the characteristics of each element of the high-altitude detection data are effectively extracted, and the key features are effectively retained through average pooling to improve the performance of the model.

[0026] The convective boundary layer height identification method of the present invention is implemented on an Ubuntu 18.04 system running on an Intel core 2.80 GHz server with 12GiB (GPU) and 192GB (RAM). The accuracy and efficiency of the system are tested seasonally under the daily high-altitude sounding dataset of Korla from 2014 to 2023. The original data contains 10 elements, including altitude, time, temperature, air pressure, humidity, dew point temperature, temperature-dew difference, virtual temperature, wind direction, and wind speed. For the measurement of system accuracy, we use the absolute mean error (MAE) and the root mean square error (RMSE). The calculation formula of the absolute mean error is: , the formula for calculating the root mean square error is The smaller the values ​​of these two indicators, the higher the recognition accuracy of the system.

[0027] Experimental testing analyzed the performance of the proposed convective boundary layer height identification method (i.e., LSTM-Attention), with the results shown in Table 1. It can be seen that LSTM-Attention reduces the mean absolute error by an average of over 70 meters compared to existing convective boundary layer height identification systems (relative humidity gradient, LiuLiang, Richardson, potential temperature gradient, dry adiabatic curve, AdaBoost, RandomForest, and SVM). This is because the proposed method fully learns the inherent correlations in high-altitude sounding data, overcoming the regional errors associated with previous physical calculation methods. Furthermore, wind vector decomposition provides data support for the model's accurate predictions, and temporal information encoding enables the model to better learn the temporal variations in convective boundary layer height.

[0028] Table 1: Comparison of recognition accuracy in different seasons (unit: m) Corresponding to the aforementioned embodiment of the method for identifying the height of a convective boundary layer, the present application also provides an embodiment of the method for identifying the height of a convective boundary layer.

[0029] Figure 2 This is a block diagram of the convective boundary layer height identification method device of the present application. Figure 2 , the device comprises: Data acquisition module 1, used to obtain high-altitude detection data; Preprocessing module 2, used for preprocessing the high-altitude detection data; Data encoding module 3, used to embed time information into the pre-processed high-altitude detection data through position coding, and obtain the encoded high-altitude detection data through the long short-term memory network; Information capture module 4, used to construct a feature information capture module based on the self-attention mechanism, and use the feature information capture module to capture the meteorological information of the encoded high-altitude sounding data; The identification module 5 is used to convert the meteorological information into the convective boundary layer height through average pooling and linear mapping operations.

[0030] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0031] Correspondingly, the present application also provides an electronic device, comprising: one or more sensors; one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in the first aspect.

[0032] Accordingly, the present application also provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-mentioned convective boundary layer height identification method.

[0033] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the contents disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered merely as exemplary, and the true scope and spirit of the present application are indicated by the claims.

[0034] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for identifying the height of a convective boundary layer, characterized in that: include: Acquire high-altitude detection data; Preprocessing the high-altitude detection data; The time information is embedded into the pre-processed high-altitude detection data through position coding, and the encoded high-altitude detection data is obtained through the long short-term memory network; Building a feature information capture module based on the self-attention mechanism, and using the feature information capture module to capture meteorological information of the encoded high-altitude sounding data; The meteorological information is converted into convective boundary layer height through average pooling and linear mapping operations.

2. The method according to claim 1, characterized in that Preprocessing the high-altitude detection data includes: Obtain various meteorological element data collected by high-altitude detection sensors, and convert wind speed and direction into longitudinal wind and latitudinal wind; Based on the meteorological element data, data quality control is performed, outliers are eliminated, missing values ​​are filled, and a high-altitude detection data matrix is ​​constructed; The high-altitude detection data matrix is ​​normalized using a maximum normalization method to obtain normalized high-altitude detection data.

3. The method according to claim 1, characterized in that The time information is embedded into the pre-processed high-altitude detection data through position encoding, and the encoded high-altitude detection data is obtained through the long short-term memory network, including: Using position encoding, the temporal information of the upper-altitude sounding data is encoded into a vector and embedded into the pre-processed upper-altitude sounding data to capture the temporal dimension changes in the convective boundary layer height. The high-altitude detection data matrix with embedded time information is encoded through the long short-term memory network to obtain the encoded high-altitude detection data.

4. The method according to claim 1, wherein A feature information capture module is constructed based on the self-attention mechanism, and the feature information capture module is used to capture meteorological information of the encoded high-altitude sounding data, including: Based on the multi-head self-attention mechanism and residual normalization structure, a multi-layer self-attention module is constructed; A multi-layer self-attention module is used to extract features from the encoded high-altitude detection data to obtain meteorological information.

5. The method according to claim 1, wherein The meteorological information is converted into convective boundary layer height through average pooling and linear mapping operations, including: The meteorological information is reduced in dimension using an average pooling layer to extract effective compressed sequence information and retain key features; The pooled features are mapped to a one-dimensional output through a linear connection layer to obtain the final convective boundary layer height.

6. A device for identifying the height of a convective boundary layer, characterized in that: include: Data acquisition module, used to obtain high-altitude detection data; A preprocessing module, used for preprocessing the high-altitude detection data; The data encoding module is used to embed time information into the pre-processed high-altitude detection data through position encoding, and obtain the encoded high-altitude detection data through the long short-term memory network; An information capture module is used to construct a feature information capture module based on a self-attention mechanism, and use the feature information capture module to capture meteorological information of the encoded high-altitude sounding data; The identification module is used to convert the meteorological information into the convective boundary layer height through average pooling and linear mapping operations.

7. An electronic device, characterized in that: include: one or more sensors; one or more processors; a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images