Training method and device of substation micrometeorological prediction model and computer equipment

Abstract

The invention relates to a training method and device for a substation micrometeorological prediction model and computer equipment. The method comprises the following steps: inputting historical meteorological sample data into an initial dynamic weighting module to obtain weighted meteorological sample data; inputting the weighted meteorological sample data into an initial time sequence feature extraction module to obtain meteorological time sequence feature data; inputting the meteorological time sequence characteristic data into the initial full connection layer to obtain meteorological prediction sample data; iteratively training according to the meteorological prediction sample data and the actual meteorological data to obtain a trained target dynamic weighting module, a target time sequence feature extraction module and a candidate full connection layer; iteratively training the candidate full connection layer by adopting a crisscross optimization model to obtain a trained target full connection layer; and according to the target dynamic weighting module, the target time sequence feature extraction module and the target full connection layer, obtaining a trained substation micrometeorological prediction model. The method can improve the accuracy of the meteorological prediction model.

Description

Technical Field

[0001] The present application relates to the field of meteorological forecasting technology, and in particular to a training method, apparatus, computer equipment, computer-readable storage medium, and computer program product for a substation micro-meteorological forecasting model. Background Art

[0002] In the power system, the operation of substations is closely related to environmental meteorological conditions. Especially in extreme weather conditions, accurate weather forecasts are of great significance to ensuring the stable operation of power equipment.

[0003] Currently, traditional weather forecasting methods are limited by the limitations of data sources and shallow machine learning prediction models. When the input sample dimension is large and the number of neurons is increased, the number of training parameters is too large and overfitting is prone to occur. It is impossible to better explore the time series patterns, which ultimately leads to low accuracy of the prediction model. Summary of the Invention

[0004] Based on this, it is necessary to provide a training method, device, computer equipment, computer-readable storage medium and computer program product for a substation micrometeorological prediction model that can improve the accuracy of the meteorological prediction model in response to the above technical problems.

[0005] In a first aspect, the present application provides a training method for a substation micrometeorological prediction model, comprising:

[0006] Obtain historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into an initial dynamic weighting module to obtain weighted meteorological sample data;

[0007] Inputting the weighted meteorological sample data into an initial time series feature extraction module to obtain meteorological time series feature data;

[0008] Input the meteorological time series feature data into the initial fully connected layer to obtain meteorological forecast sample data;

[0009] Acquire actual meteorological data, and iteratively train the initial dynamic weighting module, the initial time series feature extraction module, and the initial fully connected layer based on the meteorological forecast sample data and the actual meteorological data to obtain a trained target dynamic weighting module, a target time series feature extraction module, and a candidate fully connected layer;

[0010] Adopting a vertical and horizontal cross optimization model, based on the historical meteorological sample data, the actual meteorological data, the target dynamic weighting module and the target time series feature extraction module, iteratively training the candidate fully connected layer to obtain a trained target fully connected layer;

[0011] According to the target dynamic weighting module, the target time series feature extraction module and the target fully connected layer, a trained substation micrometeorological prediction model is obtained.

[0012] In one embodiment, inputting the historical meteorological sample data into an initial dynamic weighting module to obtain weighted meteorological sample data includes:

[0013] The historical meteorological sample data is subjected to channel feature weighting processing in the initial dynamic weighting module to obtain channel-weighted sample data;

[0014] The channel-weighted sample data is subjected to spatial feature weighting processing in the initial dynamic weighting module to obtain spatially weighted sample data as the weighted meteorological sample data.

[0015] In one embodiment, the step of subjecting the historical meteorological sample data to channel feature weighting in the initial dynamic weighting module to obtain channel-weighted sample data includes:

[0016] Inputting the historical meteorological sample data into the first maximum pooling layer and the first average pooling layer respectively, and obtaining two first local feature information accordingly;

[0017] Inputting the two first local feature information into a multilayer perceptron respectively to obtain two corresponding intermediate feature information;

[0018] Adding the two intermediate feature information to obtain added feature information, and processing the added feature information using an activation function to obtain channel weight information;

[0019] The channel-weighted sample data is obtained according to the channel weight information and the historical meteorological sample data.

[0020] In one embodiment, the channel-weighted sample data is subjected to spatial feature weighting processing in the initial dynamic weighting module to obtain spatially weighted sample data, including:

[0021] The weighted sample data of the channels are respectively input into the second maximum pooling layer and the second average pooling layer to obtain two second local feature information;

[0022] Merging the two second local feature information to obtain merged feature information;

[0023] Performing convolution processing and activation function processing on the combined feature information in sequence to obtain spatial weight information;

[0024] The spatially weighted sample data is obtained according to the spatial weight information and the channel-weighted sample data.

[0025] In one embodiment, inputting the weighted meteorological sample data into an initial time series feature extraction module to obtain meteorological time series feature data includes:

[0026] The weighted meteorological sample data is subjected to feature extraction processing to obtain candidate feature data;

[0027] The weighted meteorological sample data is subjected to one-dimensional convolution processing to obtain residual data;

[0028] The candidate feature data and the residual data are added together to obtain the meteorological time series feature data.

[0029] In one embodiment, the weighted meteorological sample data is subjected to feature extraction processing to obtain candidate feature data, including:

[0030] The weighted meteorological sample data is subjected to dilated causal convolution processing to obtain dilated causal convolution data;

[0031] The dilated causal convolution data is sequentially subjected to weight normalization processing, activation function processing, and random deactivation processing to obtain the candidate feature data.

[0032] In a second aspect, the present application also provides a training device for a substation micrometeorological prediction model, comprising:

[0033] The sample weighting module is used to obtain historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into the initial dynamic weighting module to obtain weighted meteorological sample data;

[0034] A feature extraction module is used to input the weighted meteorological sample data into an initial time series feature extraction module to obtain meteorological time series feature data;

[0035] A fully connected layer module is used to input the meteorological time series feature data into the initial fully connected layer to obtain meteorological forecast sample data;

[0036] An initial training module is used to obtain actual meteorological data and iteratively train the initial dynamic weighting module, the initial time series feature extraction module, and the initial fully connected layer based on the meteorological forecast sample data and the actual meteorological data to obtain a trained target dynamic weighting module, a target time series feature extraction module, and a candidate fully connected layer;

[0037] A secondary training module is configured to employ a vertical and horizontal cross optimization model to iteratively train the candidate fully connected layer based on the historical meteorological sample data, the actual meteorological data, the target dynamic weighting module, and the target time series feature extraction module to obtain a trained target fully connected layer;

[0038] The model determination module is used to obtain a trained substation micrometeorological prediction model based on the target dynamic weighting module, the target time series feature extraction module and the target fully connected layer.

[0039] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0040] Obtain historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into an initial dynamic weighting module to obtain weighted meteorological sample data;

[0041] Inputting the weighted meteorological sample data into an initial time series feature extraction module to obtain meteorological time series feature data;

[0042] Input the meteorological time series feature data into the initial fully connected layer to obtain meteorological forecast sample data;

[0043] Acquire actual meteorological data, and iteratively train the initial dynamic weighting module, the initial time series feature extraction module, and the initial fully connected layer based on the meteorological forecast sample data and the actual meteorological data to obtain a trained target dynamic weighting module, a target time series feature extraction module, and a candidate fully connected layer;

[0044] Adopting a vertical and horizontal cross optimization model, based on the historical meteorological sample data, the actual meteorological data, the target dynamic weighting module and the target time series feature extraction module, iteratively training the candidate fully connected layer to obtain a trained target fully connected layer;

[0045] According to the target dynamic weighting module, the target time series feature extraction module and the target fully connected layer, a trained substation micrometeorological prediction model is obtained.

[0046] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0047] Obtain historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into an initial dynamic weighting module to obtain weighted meteorological sample data;

[0048] Inputting the weighted meteorological sample data into an initial time series feature extraction module to obtain meteorological time series feature data;

[0049] Input the meteorological time series feature data into the initial fully connected layer to obtain meteorological forecast sample data;

[0050] Acquire actual meteorological data, and iteratively train the initial dynamic weighting module, the initial time series feature extraction module, and the initial fully connected layer based on the meteorological forecast sample data and the actual meteorological data to obtain a trained target dynamic weighting module, a target time series feature extraction module, and a candidate fully connected layer;

[0051] Adopting a vertical and horizontal cross optimization model, based on the historical meteorological sample data, the actual meteorological data, the target dynamic weighting module and the target time series feature extraction module, iteratively training the candidate fully connected layer to obtain a trained target fully connected layer;

[0052] According to the target dynamic weighting module, the target time series feature extraction module and the target fully connected layer, a trained substation micrometeorological prediction model is obtained.

[0053] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the following steps:

[0054] Obtain historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into an initial dynamic weighting module to obtain weighted meteorological sample data;

[0055] Inputting the weighted meteorological sample data into an initial time series feature extraction module to obtain meteorological time series feature data;

[0056] Input the meteorological time series feature data into the initial fully connected layer to obtain meteorological forecast sample data;

[0057] Acquire actual meteorological data, and iteratively train the initial dynamic weighting module, the initial time series feature extraction module, and the initial fully connected layer based on the meteorological forecast sample data and the actual meteorological data to obtain a trained target dynamic weighting module, a target time series feature extraction module, and a candidate fully connected layer;

[0058] Adopting a vertical and horizontal cross optimization model, based on the historical meteorological sample data, the actual meteorological data, the target dynamic weighting module and the target time series feature extraction module, iteratively training the candidate fully connected layer to obtain a trained target fully connected layer;

[0059] According to the target dynamic weighting module, the target time series feature extraction module and the target fully connected layer, a trained substation micrometeorological prediction model is obtained.

[0060] The training method, device, computer equipment, computer-readable storage medium and computer program product of the above-mentioned substation micro-meteorological prediction model first obtain the historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into the initial dynamic weighting module to obtain weighted meteorological sample data. Through the dynamic weighting module, different weights can be given to different data elements, highlighting key features, reducing the influence of redundant features, improving the model's attention to important features, and enhancing the model's input feature representation ability; then, the weighted meteorological sample data is input into the initial time series feature extraction module to obtain meteorological time series feature data. The time series feature extraction module can capture the time dependency and time series regularity in the historical meteorological data, thereby more accurately extracting time series features related to future meteorological conditions, solving the problem of limited receptive field of traditional models when processing long time series; then, the meteorological time series feature data is input into the initial fully connected layer to obtain meteorological forecast sample data. Through further processing of the fully connected layer, the time series feature data is mapped to the output result of the future meteorological forecast; then, actual meteorological data is obtained, and iterative training is performed based on the meteorological forecast sample data and the actual meteorological data. The initial dynamic weighting module, initial time series feature extraction module and initial fully connected layer are trained to obtain the trained target dynamic weighting module, target time series feature extraction module and candidate fully connected layer. By comparing the feedback of actual meteorological data with meteorological forecast sample data, the initial dynamic weighting module, time series feature extraction module and fully connected layer are gradually optimized to make them better adapt to the actual situation, reduce errors and improve prediction accuracy. Then, a vertical and horizontal cross optimization model is used to iteratively train the candidate fully connected layer based on historical meteorological sample data, actual meteorological data, target dynamic weighting module and target time series feature extraction module to obtain the trained target fully connected layer. The vertical and horizontal cross optimization model is used to perform secondary optimization on the fully connected layer, which can effectively prevent the model parameters from falling into the local optimum, further improve the prediction ability and generalization ability of the fully connected layer, and enhance the overall performance of the model. Finally, based on the target dynamic weighting module, target time series feature extraction module and target fully connected layer, the trained substation micrometeorological prediction model is obtained. The target dynamic weighting module, target time series feature extraction module and target fully connected layer are integrated to generate a highly accurate micrometeorological prediction model, providing accurate meteorological forecasts for substations. This method, by introducing a dynamic weighting module, a time series feature extraction module, and a fully connected layer, combined with iterative training of actual meteorological data and a cross-optimization model, forms a highly efficient micrometeorological forecasting model. This model overcomes the limitations of traditional forecasting methods, enhances focus on key meteorological factors, and accurately captures the long- and short-term dependencies of time series, ultimately generating more accurate weather forecasts. The fully connected layer, after secondary optimization, improves the model's generalization capabilities, giving it stronger predictive capabilities under varying meteorological conditions, ensuring the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0062] Figure 1 1 is a flow chart of a method for training a substation micrometeorological prediction model in one embodiment;

[0063] Figure 2 1 is a flow chart of steps for obtaining weighted meteorological sample data in one embodiment;

[0064] Figure 3 is a schematic structural diagram of a dynamic weighting module in one embodiment;

[0065] Figure 4 FIG1 is a flow chart of a channel feature weighting process in a dynamic weighting module according to an embodiment;

[0066] Figure 5 1. A schematic diagram of a flow chart of spatial feature weighting processing in a dynamic weighting module in one embodiment;

[0067] Figure 6 Schematic diagram of the structure of causal convolution in one embodiment;

[0068] Figure 7 Schematic diagram of the structure of dilated convolution in one embodiment;

[0069] Figure 8 Schematic diagram of the structure of a temporal feature extraction module in one embodiment;

[0070] Figure 9 A schematic diagram of a flow chart of the steps of secondary training of a fully connected layer in one embodiment;

[0071] Figure 10 1 is a structural block diagram of a training device for a substation micrometeorological prediction model in one embodiment;

[0072] Figure 11 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0074] In one embodiment, Figure 1 As shown, a training method for a substation micro-meteorological prediction model is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. Among them, the terminal can be, but is not limited to, various personal computers, laptops, smart phones, and tablet computers. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. In this embodiment, the method includes the following steps:

[0075] Step S101: Obtain historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into an initial dynamic weighting module to obtain weighted meteorological sample data.

[0076] Among them, historical meteorological sample data refers to the meteorological element data collected by the meteorological monitoring system in the substation over the past period of time, including but not limited to temperature, humidity, rainfall, wind speed, wind direction, air pressure and sunshine.

[0077] Among them, a sample data includes multiple meteorological element data at multiple consecutive moments, that is, meteorological forecasting is performed using multiple meteorological element data at multiple consecutive moments before the forecast time point.

[0078] Among them, the initial dynamic weighting module is a module used to automatically adjust the weights between meteorological elements. This module analyzes the mutual influence between different meteorological elements and assigns appropriate weights to different features to ensure that important features receive higher priority in subsequent processing stages.

[0079] For example, the terminal first obtains historical meteorological sample data from the substation to be predicted and performs preprocessing on the data, including missing value filling, outlier detection, and data normalization. The preprocessed meteorological data is then input into the initial dynamic weighting module. Within this module, the system uses an algorithm to calculate the relative importance of different meteorological factors during a specific time period and assigns dynamic weights to each factor based on their importance. This weighting method ensures that important meteorological features have a greater influence in subsequent analysis, while the influence of less important features is appropriately weakened.

[0080] Step S102: input the weighted meteorological sample data into an initial time series feature extraction module to obtain meteorological time series feature data.

[0081] Meteorological time series feature data refers to the changing trends and characteristic patterns of meteorological sample data over time after being processed by the time series feature extraction module. This data contains the time series characteristics of meteorological elements and can reflect the patterns of meteorological elements changing over time.

[0082] The initial time series feature extraction module extracts time series information from weighted meteorological sample data, capturing the temporal correlations and short- and long-term dependencies of meteorological data. This module identifies temporal correlations and trend changes between meteorological elements, providing a basis for subsequent forecasting.

[0083] For example, the terminal first inputs the weighted meteorological sample data into the initial time series feature extraction module. In this module, the terminal uses an algorithm to analyze the time series characteristics of the meteorological data and extract the temporal characteristics of each meteorological element. This module can capture the changing patterns of meteorological elements over different time periods and identify potential temporal dependencies between meteorological elements. Through this time series feature extraction, the terminal can effectively construct the time series characteristics of the meteorological samples, providing more accurate input data for subsequent model training and prediction.

[0084] Step S103: input the meteorological time series feature data into the initial fully connected layer to obtain meteorological forecast sample data.

[0085] Among them, the meteorological time series feature data is the data representing the time change characteristics of meteorological elements generated after being processed by the time series feature extraction module. It contains the time dependence of meteorological variables and reflects the dynamic characteristics of meteorological elements.

[0086] Among them, the initial fully connected layer is a neural network layer, which is used to nonlinearly combine and map the input time series feature data to generate the final weather forecast output.

[0087] For example, the terminal inputs the meteorological time series feature data into the initial fully connected layer. The fully connected layer performs a linear transformation on the input data through its weight matrix to obtain a meteorological forecast result for the input meteorological time series feature data, i.e., meteorological forecast sample data.

[0088] Step S104, obtain actual meteorological data, and iteratively train the initial dynamic weighting module, initial time series feature extraction module and initial fully connected layer based on the meteorological forecast sample data and the actual meteorological data to obtain the trained target dynamic weighting module, target time series feature extraction module and candidate fully connected layer.

[0089] Actual weather data refers to current weather conditions collected in real time from the substation's weather monitoring system. It covers forecast-related meteorological elements such as temperature, humidity, wind speed, and wind direction. Actual weather data is compared with forecast sample data to assess the accuracy of model predictions.

[0090] For example, the terminal first obtains actual meteorological data from the substation and compares it with the forecast sample data to calculate the forecast error. Using this error, the terminal uses a backpropagation algorithm to iteratively train each module of the model, gradually optimizing the parameters in the initial dynamic weighting module, the initial time series feature extraction module, and the initial fully connected layer. During this iterative training process, the model continuously updates weights, gradually improving its ability to extract meteorological time series features and enhancing the accuracy of weighted feature recognition.

[0091] In step S105 , a vertical and horizontal cross optimization model is used to iteratively train the candidate fully connected layer based on historical meteorological sample data, actual meteorological data, target dynamic weighting module and target time series feature extraction module to obtain a trained target fully connected layer.

[0092] The vertical and horizontal cross-optimization model is an optimization algorithm that combines horizontal and vertical cross-optimization mechanisms to effectively avoid falling into local optimality during fully connected layer parameter training, thereby improving the global optimization capabilities of the fully connected layer. This optimization model iteratively adjusts the weights and biases of the fully connected layer to better adapt it to the relationship between historical and actual meteorological data, thereby enhancing the model's generalization capabilities.

[0093] For example, the terminal first obtains historical meteorological sample data and actual meteorological data, and inputs this data into a candidate fully connected layer, combining it with the outputs of the optimized target dynamic weighting module and target time series feature extraction module. The terminal then optimizes the candidate fully connected layer using a vertical and horizontal cross-optimization model, gradually adjusting its weight parameters. This model alternates between horizontal cross-optimization (exploring the global optimal solution) and vertical cross-optimization (deepening local optimization) to ensure that the fully connected layer avoids falling into local optimality during training while fully exploiting the global characteristics of the meteorological data.

[0094] Step S106: A trained substation micrometeorological prediction model is obtained according to the target dynamic weighting module, the target time series feature extraction module and the target fully connected layer.

[0095] Exemplarily, the terminal integrates the trained target dynamic weighting module, target time series feature extraction module, and target fully connected layer to form the final substation micro-meteorological prediction model. Specifically, in actual applications, the terminal will first input the substation meteorological data collected in real time into the target dynamic weighting module. This module assigns weights based on the importance of different features and passes the weighted data to the target time series feature extraction module. The latter extracts time series features useful for current predictions by learning the time series patterns in historical data. The extracted time series features are then input into the target fully connected layer, which, through the deep learning capabilities of this layer, ultimately generates a forecast value for future meteorological conditions.

[0096] In the training method of the above-mentioned substation micro-meteorological prediction model, first, the historical meteorological sample data of the substation to be predicted is obtained, and the historical meteorological sample data is input into the initial dynamic weighting module to obtain weighted meteorological sample data. Through the dynamic weighting module, different weights can be given to different data elements, highlighting key features, reducing the influence of redundant features, improving the model's attention to important features, and enhancing the model's input feature representation ability; then, the weighted meteorological sample data is input into the initial time series feature extraction module to obtain meteorological time series feature data. The time series feature extraction module can capture the time dependency and time series regularity in the historical meteorological data, thereby more accurately extracting time series features related to future meteorological conditions, and solving the problem of limited receptive field of traditional models when processing long time series; then, the meteorological time series feature data is input into the initial fully connected layer to obtain meteorological forecast sample data. Through further processing of the fully connected layer, the time series feature data is mapped to the output result of the future meteorological forecast; then, the actual meteorological data is obtained, and the initial dynamic weighting module, initial time series feature extraction module and meteorological forecast sample data and actual meteorological data are iteratively trained. The trained target dynamic weighting module, target time series feature extraction module and candidate fully connected layer are obtained by combining the feedback of actual meteorological data with the meteorological forecast sample data, and gradually optimizing the initial dynamic weighting module, time series feature extraction module and fully connected layer to make them better adapt to the actual situation, reduce errors and improve prediction accuracy. Then, a vertical and horizontal cross optimization model is adopted to iteratively train the candidate fully connected layer based on historical meteorological sample data, actual meteorological data, target dynamic weighting module and target time series feature extraction module to obtain the trained target fully connected layer. The vertical and horizontal cross optimization model is used to perform secondary optimization on the fully connected layer, which can effectively prevent the model parameters from falling into the local optimum, further improve the prediction ability and generalization ability of the fully connected layer, and enhance the overall performance of the model. Finally, the trained substation micrometeorological prediction model is obtained based on the target dynamic weighting module, target time series feature extraction module and target fully connected layer. The target dynamic weighting module, target time series feature extraction module and target fully connected layer are integrated to generate a highly accurate micrometeorological prediction model, providing accurate meteorological forecasts for substations. This method, by introducing a dynamic weighting module, a time series feature extraction module, and a fully connected layer, combined with iterative training of actual meteorological data and a cross-optimization model, forms a highly efficient micrometeorological forecasting model. This model overcomes the limitations of traditional forecasting methods, enhances focus on key meteorological factors, and accurately captures the long- and short-term dependencies of time series, ultimately generating more accurate weather forecasts. The fully connected layer, after secondary optimization, improves the model's generalization capabilities, giving it stronger predictive capabilities under varying meteorological conditions, ensuring the safe and stable operation of the power system.

[0097] In an exemplary embodiment, Figure 2As shown, the above step S101 inputs the historical meteorological sample data into the initial dynamic weighting module to obtain the weighted meteorological sample data, which can also be achieved by the following steps:

[0098] Step S201: The historical meteorological sample data is subjected to channel feature weighting processing in an initial dynamic weighting module to obtain channel-weighted sample data;

[0099] Step S202 : The channel-weighted sample data is subjected to spatial feature weighting processing in an initial dynamic weighting module to obtain spatially weighted sample data as weighted meteorological sample data.

[0100] Exemplarily, the terminal inputs the historical meteorological sample data into the initial dynamic weighting module, and first performs channel weighting processing, specifically generating a feature map through maximum pooling and average pooling operations, and then performing nonlinear mapping on the feature map. Finally, the channel attention weight is calculated and multiplied with the sample data to generate weighted sample data. The terminal then inputs the channel-weighted sample data into the spatial weighting processing part of the initial dynamic weighting module. Two spatial feature maps are generated by performing maximum pooling and average pooling operations on the feature map. Then, the two feature maps are merged through a 1×1 convolution operation and weight calculation is performed to obtain a spatial attention matrix. Ultimately, the sample data obtained after combining channel and spatial weighting can more accurately reflect the spatiotemporal characteristics of meteorological data. As Figure 3 FIG. 1 is a schematic diagram of the structure of a dynamic weighting module in one embodiment.

[0101] In this embodiment, the combination of channel weighting and spatial weighting effectively improves the efficiency and accuracy of extracting spatiotemporal features from meteorological sample data. Channel weighting highlights key features, while spatial weighting enhances the model's sensitivity to location information. The combined effects of these two greatly enhance the accuracy and robustness of the micrometeorological forecast model, making predictions more reliable, especially in complex meteorological environments.

[0102] In an exemplary embodiment, the above-mentioned step S201 subjects the historical meteorological sample data to channel feature weighting processing in the initial dynamic weighting module to obtain channel-weighted sample data, and also includes: inputting the historical meteorological sample data into the first maximum pooling layer and the first average pooling layer respectively, and correspondingly obtaining two first local feature information; inputting the two first local feature information into the multi-layer perceptron respectively, and correspondingly obtaining two intermediate feature information; adding the two intermediate feature information to obtain the added feature information, and using the activation function to process the added feature information to obtain channel weight information; obtaining channel-weighted sample data based on the channel weight information and the historical meteorological sample data.

[0103] Among them, the first maximum pooling layer, the first average pooling layer, the multi-layer perceptron and the activation function are all units required for the channel feature weighting processing in the initial dynamic weighting module.

[0104] Among them, the maximum pooling layer highlights the important features in the data by retaining the maximum value in each feature channel, while the average pooling layer smoothes the noise in the data by calculating the average value of each feature channel, thereby obtaining two different first local feature information.

[0105] Among them, the multi-layer perceptron is a nonlinear model structure used to enhance the representation ability of features.

[0106] Exemplarily, the terminal inputs the historical meteorological sample data into the first maximum pooling layer and the first average pooling layer, and compresses them respectively to obtain two first local feature information with the number of retained feature channels (one channel is a meteorological factor in this embodiment). These two feature information have different feature focus methods, one focuses on highlighting local extreme value features, and the other emphasizes the overall smoothness of the features. Next, the terminal inputs these two first local feature information into the multi-layer perceptron respectively, and generates two intermediate feature information through the nonlinear mapping capability of the perceptron. Then, the terminal uses an activation function to process the added feature information. By using activation functions such as Sigmoid, ReLU, etc., nonlinear processing further enhances the expression ability of the model and generates the final channel weight information. The channel weight information represents the importance of each channel. According to the weight information, the terminal weights the historical meteorological sample data to obtain the final channel-weighted sample data. As Figure 4 FIG. 1 is a flow chart of channel feature weighting processing in a dynamic weighting module in one embodiment.

[0107] In this embodiment, channel-weighted processing effectively improves the ability to distinguish and represent features in historical meteorological sample data through the synergistic effects of max pooling, average pooling, a multi-layer perceptron, and activation functions. This channel-weighted sample data more accurately reflects the contribution of meteorological factors across different channels, helping to improve the accuracy and robustness of subsequent forecast models.

[0108] In an exemplary embodiment, the above-mentioned step S202 subjects the channel-weighted sample data to spatial feature weighting processing in the initial dynamic weighting module to obtain spatially weighted sample data, and also includes: inputting the channel-weighted sample data into the second maximum pooling layer and the second average pooling layer respectively, and correspondingly obtaining two second local feature information; merging the two second local feature information to obtain merged feature information; performing convolution processing and activation function processing on the merged feature information in turn to obtain spatial weight information; and obtaining spatially weighted sample data based on the spatial weight information and the channel-weighted sample data.

[0109] Among them, the second maximum pooling layer, the second average pooling layer and the activation function are all units required for the spatial feature weighting processing in the initial dynamic weighting module.

[0110] For example, the terminal first inputs the channel-weighted sample data into the second max pooling layer and the second average pooling layer, extracting two different second local feature information channels from the input feature data. The terminal then merges these two local feature information channels, fusing the feature information from the max pooling layer and the average pooling layer to form a more comprehensive merged feature information. This merging operation ensures that key spatial features (in this embodiment, the time dimension is analogous to "space") are preserved while eliminating redundant information. The merged feature information is further processed through a convolution operation. Convolution aims to capture local dependencies between features in the spatial dimension (in this embodiment, the temporal dimension), particularly the mutual influence between different locations. The convolution kernel traverses the merged feature information using a sliding window, extracting deep features of the spatial distribution and enhancing the model's understanding of spatial structure. The terminal then applies an activation function to the convolved feature information, typically using nonlinear activation functions such as sigmoid and ReLU, to further enhance the model's nonlinear expression capabilities and ultimately generate spatial weight information. Finally, the terminal weights the channel-weighted sample data according to the calculated spatial weight information to generate the final spatially weighted sample data, ensuring that the model can not only adjust the weight according to the importance of the channel features, but also optimize the processing according to the changing characteristics of the samples in the spatial dimension, thereby obtaining more accurate weighted sample data. Figure 5 FIG. 1 is a flow chart of spatial feature weighting processing in a dynamic weighting module in one embodiment.

[0111] In this embodiment, weighted spatial feature processing fully accounts for the distribution of meteorological features across different spatial locations. By combining max pooling, average pooling, convolution, and activation functions, the model's feature extraction capabilities in the spatial dimension are enhanced, effectively increasing the influence of spatial features on prediction results and further improving the model's overall prediction performance.

[0112] In an exemplary embodiment, the above-mentioned step S102 inputs the weighted meteorological sample data into the initial time series feature extraction module to obtain meteorological time series feature data, and also includes: subjecting the weighted meteorological sample data to feature extraction processing to obtain candidate feature data; subjecting the weighted meteorological sample data to one-dimensional convolution processing to obtain residual data; and adding the candidate feature data and the residual data to obtain meteorological time series feature data.

[0113] Feature extraction involves extracting candidate features from weighted meteorological sample data using specific algorithms (such as convolutional neural networks or other time series feature extraction techniques). These candidate features represent significant patterns and trends in the time series data. This process aims to enhance relevant information in the time series data and reduce unnecessary redundant information, making subsequent convolution processing more efficient.

[0114] If the number of network layers is too large, it can easily cause gradient explosion or vanishing defects, causing the network to crash and reduce the network's learning ability. The introduction of residual data effectively solves the problem of network degradation caused by the deepening of network layers.

[0115] Exemplarily, the terminal inputs the weighted meteorological sample data into the time series feature extraction module, and obtains candidate feature data through a specific feature extraction algorithm. These candidate feature data contain representative time series features extracted from the input samples, aiming to capture key changes and trends in meteorological data. Next, the terminal inputs the weighted meteorological sample data into the one-dimensional convolution processing module to calculate the residual data. The residual data captures the feature information that is not fully expressed during the convolution processing process and retains the key information in the input sample. This process can effectively alleviate the information loss problem that may occur as the network depth increases, and ensure the integrity of the feature information. The terminal then adds the candidate feature data and the residual data to fuse the residual and candidate features obtained from the convolution process, retaining the time series features captured by the convolution network while enhancing the stability and generalization ability of the network. The added feature data contains the refined meteorological time series feature information, providing a more reliable input for the subsequent prediction layer.

[0116] In this example, feature extraction and residual addition processing are used to effectively improve the accuracy of meteorological time series feature extraction and ensure the model's sensitivity to meteorological changes in the time series dimension. This approach allows the model to better understand the complex time series patterns in meteorological data, improving the accuracy and robustness of forecasts.

[0117] In an exemplary embodiment, the above-mentioned weighted meteorological sample data is subjected to feature extraction processing to obtain candidate feature data, and also includes: subjecting the weighted meteorological sample data to dilated causal convolution processing to obtain dilated causal convolution data; and performing weight normalization processing, activation function processing and random deactivation processing on the dilated causal convolution data in sequence to obtain candidate feature data.

[0118] Dilated causal convolution involves performing a convolution operation on weighted sample data using dilated causal convolution. This method maintains the causal relationships within the time series while also expanding the receptive field of the convolution kernel by introducing holes (dilation) during the convolution operation, thereby better capturing long-term dependent feature information. This convolution process can extract important temporal patterns within time series data over longer timeframes, avoiding the loss of critical historical information.

[0119] The causal convolution structure is composed of a series of one-dimensional convolution stacks. The structure of causal convolution is as follows: Figure 6 As shown in the figure, it can be clearly seen that the output at the current moment is closely related to the historical input. For example, the output at time t is only convolved with the two inputs at the current moment and the previous moment. t and historical information {x t-3 , x t-2 , x t-1 , x t However, when processing long time series, causal convolution has to increase the number of hidden layers or the size of the convolution kernel, which will make the network structure more complicated. Causal convolution also has the defect of too small receptive field.

[0120] In order to overcome the problem of small receptive field of causal convolution, the temporal convolutional neural network adds dilated convolution to the network structure. The advantage of this structure is that when the number of hidden layers and the size of the convolution kernel are the same, the dilated convolution can obtain a larger receptive field, which is conducive to solving the problem of long-term series prediction of micro-meteorological elements. Figure 7 shown.

[0121] For example, the terminal inputs weighted meteorological sample data into the dilated causal convolution module. Dilated causal convolution processes time series data by increasing the span of the convolution kernel, ultimately generating dilated causal convolution data. This process more effectively captures important information in long time series while avoiding the problem of overlooking long-range dependencies often seen in convolutions of short time series. The terminal then performs a series of processing on the dilated causal convolution data to enhance feature representation and model robustness: Weight normalization: Weight normalization standardizes the convolution results, constrains the range of network weights, ensures consistent feature distribution, and prevents model instability caused by large weight differences. This step helps accelerate model convergence and prevents problems such as exploding or vanishing gradients. Activation function processing: After weight normalization, the terminal applies an activation function to the data, typically using a nonlinear activation function such as ReLU (rectified linear unit). Activation functions introduce nonlinear characteristics, enabling the model to learn and represent complex temporal relationships. The ReLU function effectively filters out irrelevant or negative eigenvalues while retaining positively activated feature information. Dropout: Finally, the terminal performs Dropout on the activated data. Dropout is a regularization technique used to prevent the model from overfitting during training. By randomly discarding the connections of some neurons, the network can avoid over-reliance on certain specific features and improve the generalization ability of the model. After these processing steps, the terminal finally obtains candidate feature data. These candidate feature data combine the time series features extracted by dilated causal convolution, the standardized features after weight normalization, the nonlinear features processed by the activation function, and the generalization ability enhanced by random dropout. Through these steps, the model can better capture and express complex time series features and provide high-quality input for the subsequent prediction stage. Figure 8 FIG. 1 is a schematic diagram of the structure of a temporal feature extraction module in an embodiment.

[0122] In an exemplary embodiment, Figure 9 As shown, the above step S105 adopts a vertical and horizontal cross optimization model to iteratively train the candidate fully connected layer based on historical meteorological sample data, actual meteorological data, target dynamic weighting module and target time series feature extraction module to obtain the trained target fully connected layer. It can also be achieved by the following steps:

[0123] Step S901, initialize multiple groups of fully connected layer parameters as the initial population, obtain basic parameters of the vertical and horizontal cross optimization model and obtain the fitness function.

[0124] Step S902: Calculate the fitness of all individuals in the population.

[0125] Step S903: Record the optimal parameters of each individual and the global optimal parameters.

[0126] Step S904, determine whether the number of iterations reaches a preset value, if not, execute step S905, if yes, execute step S908.

[0127] Step S905: Execute horizontal crossover and competition operator.

[0128] Step S906: Execute vertical crossover and competition operator.

[0129] Step S907: obtain a new vertical and horizontal cross population, and return to step S902.

[0130] Step S908: Obtain the optimal fully connected layer parameters.

[0131] For example, during the optimization process, it is necessary to determine the fitness function. The mean square error and regularization term are selected as the fitness function. The mean square error ensures that the prediction result is close to the actual value. The introduction of the regularization term prevents the model from overfitting and improves the generalization ability of the model. The fitness function can be expressed as follows:

[0132]

[0133] Where N represents the total number of training set samples; represents the parameters (threshold or bias) of the fully connected layer; M is the total number of parameters of the fully connected layer used for training; and Indicates the actual and predicted values of micrometeorological elements; represents the coefficient of the regularization term, which is set to 0.005.

[0134] The salient feature of the vertical-crossover algorithm is the dual crossover mechanism that combines horizontal crossover (hc) and vertical crossover (vc). During the optimization process, the two crossover operators are performed alternately and iterated continuously until the number of iterations reaches the upper limit or the loss meets the requirements.

[0135] Crossover is the exchange of information between individuals in a population. Let S represent the population, the number of individuals in the population is M, the dimension of each individual is D, and S(i) and S(j) represent two different individuals in the population. S(i, d) and S(j, d) represent the d-th dimension information of individuals i and j respectively. Then the formula for crossover information exchange between these two individuals is:

[0136]

[0137] in, 、 For the parent and New individuals generated after horizontal crossover; , represents the probability that two groups of parents exchange information on the same dimension; ,Increased emphasis on the difference between the two parent features can explore a larger search space beyond the parent features, introducing additional randomness and diversity, which is very important for the global search ability of the algorithm.

[0138] The object of vertical crossover is the offspring after horizontal crossover, which is to cross-replace the information of different dimensions of a single individual in the population. In this way, the diversity of the population can be increased, and at the same time, the individual can escape from the local optimum under certain conditions and enhance the optimization ability. The formula of vertical crossover is:

[0139]

[0140] in, The offspring produced by vertical cross; , used to measure the degree of information exchange between individuals in different dimensions; 、 Respectively represent the d1 and d2 dimension information of the parent individual S.

[0141] In this embodiment, by adopting a vertical and horizontal cross optimization model to iteratively optimize the fully connected layer, the training efficiency and global search capability of the model can be effectively improved, and ultimately a target fully connected layer with better performance can be obtained. By initializing multiple fully connected layer parameters as the initial population, and combining the double cross mechanism of horizontal cross and vertical cross, sufficient information exchange can be carried out between individuals and within individuals, thereby enhancing the model's ability to explore the global optimal solution. At the same time, the introduction of the competition operator ensures that individuals with lower fitness are gradually eliminated, thereby promoting the overall evolution of the population in a better direction. The setting of the fitness function ensures the model's ability to fit the meteorological data during training, and gradually optimizes the parameters of the fully connected layer during the iterative update process.

[0142] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0143] Based on the same inventive concept, an embodiment of the present application also provides a training device for a substation micrometeorological prediction model for implementing the aforementioned training method for a substation micrometeorological prediction model. The solution to the problem provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of the embodiments of the training device for one or more substation micrometeorological prediction models provided below can be found in the limitations of the training method for a substation micrometeorological prediction model described above and will not be repeated here.

[0144] In an exemplary embodiment, Figure 10 As shown, a training device for a substation micrometeorological prediction model is provided, comprising: a sample weighting module 1001, a feature extraction module 1002, a fully connected layer module 1003, an initial training module 1004, a secondary training module 1005, and a model determination module 1006, wherein:

[0145] The sample weighting module 1001 is used to obtain historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into the initial dynamic weighting module to obtain weighted meteorological sample data;

[0146] The feature extraction module 1002 is used to input the weighted meteorological sample data into the initial time series feature extraction module to obtain meteorological time series feature data;

[0147] The fully connected layer module 1003 is used to input the meteorological time series feature data into the initial fully connected layer to obtain meteorological forecast sample data;

[0148] Initial training module 1004 is used to obtain actual meteorological data and iteratively train the initial dynamic weighting module, initial time series feature extraction module, and initial fully connected layer based on the meteorological forecast sample data and actual meteorological data to obtain the trained target dynamic weighting module, target time series feature extraction module, and candidate fully connected layer;

[0149] The secondary training module 1005 is used to iteratively train the candidate fully connected layer using a vertical and horizontal cross optimization model based on historical meteorological sample data, actual meteorological data, a target dynamic weighting module, and a target time series feature extraction module to obtain a trained target fully connected layer;

[0150] The model determination module 1006 is used to obtain a trained substation micrometeorological prediction model based on the target dynamic weighting module, the target time series feature extraction module and the target fully connected layer.

[0151] In one embodiment, the above-mentioned sample weighting module 1001 is also used to subject the historical meteorological sample data to channel feature weighting processing in the initial dynamic weighting module to obtain channel-weighted sample data; and subject the channel-weighted sample data to spatial feature weighting processing in the initial dynamic weighting module to obtain spatially weighted sample data as weighted meteorological sample data.

[0152] In one embodiment, the above-mentioned sample weighting module 1001 is also used to input the historical meteorological sample data into the first maximum pooling layer and the first average pooling layer respectively, and obtain two first local feature information accordingly; input the two first local feature information into the multi-layer perceptron respectively, and obtain two intermediate feature information accordingly; add the two intermediate feature information to obtain the added feature information, and use the activation function to process the added feature information to obtain channel weight information; obtain channel-weighted sample data based on the channel weight information and the historical meteorological sample data.

[0153] In one embodiment, the above-mentioned sample weighting module 1001 is also used to input the channel-weighted sample data into the second maximum pooling layer and the second average pooling layer respectively, and obtain two second local feature information accordingly; merge the two second local feature information to obtain merged feature information; perform convolution processing and activation function processing on the merged feature information in sequence to obtain spatial weight information; and obtain spatially weighted sample data based on the spatial weight information and the channel-weighted sample data.

[0154] In one embodiment, the feature extraction module 1002 is further used to subject the weighted meteorological sample data to feature extraction processing to obtain candidate feature data; subject the weighted meteorological sample data to one-dimensional convolution processing to obtain residual data; and add the candidate feature data and the residual data to obtain meteorological time series feature data.

[0155] In one embodiment, the feature extraction module 1002 is further used to subject the weighted meteorological sample data to dilated causal convolution processing to obtain dilated causal convolution data; the dilated causal convolution data is sequentially subjected to weight normalization processing, activation function processing and random inactivation processing to obtain candidate feature data.

[0156] Each module in the aforementioned substation micrometeorological prediction model training device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor within a computer device in hardware form, or stored in a computer device memory in software form, allowing the processor to call and execute the corresponding operations of each module.

[0157] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 11 As shown. The computer device includes a processor, memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication, and the wireless communication can be achieved via Wi-Fi, mobile cellular networks, near-field communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for training a substation micrometeorological prediction model. The display unit of the computer device is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse.

[0158] Those skilled in the art will understand that Figure 11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0159] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0160] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0161] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0162] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0163] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0164] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0165] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A training method for a substation micrometeorological prediction model, characterized in that: The method comprises: Obtain historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into an initial dynamic weighting module to obtain weighted meteorological sample data; Inputting the weighted meteorological sample data into an initial time series feature extraction module to obtain meteorological time series feature data; Input the meteorological time series feature data into the initial fully connected layer to obtain meteorological forecast sample data; Acquire actual meteorological data, and iteratively train the initial dynamic weighting module, the initial time series feature extraction module, and the initial fully connected layer based on the meteorological forecast sample data and the actual meteorological data to obtain a trained target dynamic weighting module, a target time series feature extraction module, and a candidate fully connected layer; Adopting a vertical and horizontal cross optimization model, based on the historical meteorological sample data, the actual meteorological data, the target dynamic weighting module and the target time series feature extraction module, iteratively training the candidate fully connected layer to obtain a trained target fully connected layer; According to the target dynamic weighting module, the target time series feature extraction module and the target fully connected layer, a trained substation micrometeorological prediction model is obtained.

2. The method according to claim 1, characterized in that The step of inputting the historical meteorological sample data into an initial dynamic weighting module to obtain weighted meteorological sample data includes: The historical meteorological sample data is subjected to channel feature weighting processing in the initial dynamic weighting module to obtain channel-weighted sample data; The channel-weighted sample data is subjected to spatial feature weighting processing in the initial dynamic weighting module to obtain spatially weighted sample data as the weighted meteorological sample data.

3. The method according to claim 2, characterized in that The step of subjecting the historical meteorological sample data to channel feature weighting processing in the initial dynamic weighting module to obtain channel-weighted sample data includes: Inputting the historical meteorological sample data into the first maximum pooling layer and the first average pooling layer respectively, and obtaining two first local feature information accordingly; Inputting the two first local feature information into a multilayer perceptron respectively to obtain two corresponding intermediate feature information; Adding the two intermediate feature information to obtain added feature information, and processing the added feature information using an activation function to obtain channel weight information; The channel-weighted sample data is obtained according to the channel weight information and the historical meteorological sample data.

4. The method according to claim 2, characterized in that The channel-weighted sample data is subjected to spatial feature weighting processing in the initial dynamic weighting module to obtain spatially weighted sample data, including: The weighted sample data of the channels are respectively input into the second maximum pooling layer and the second average pooling layer to obtain two pieces of second local feature information; Merging the two second local feature information to obtain merged feature information; Performing convolution processing and activation function processing on the combined feature information in sequence to obtain spatial weight information; The spatially weighted sample data is obtained according to the spatial weight information and the channel-weighted sample data.

5. The method according to claim 1, wherein The weighted meteorological sample data is input into the initial time series feature extraction module to obtain meteorological time series feature data, including: The weighted meteorological sample data is subjected to feature extraction processing to obtain candidate feature data; The weighted meteorological sample data is subjected to one-dimensional convolution processing to obtain residual data; The candidate feature data and the residual data are added together to obtain the meteorological time series feature data.

6. The method according to claim 5, characterized in that The weighted meteorological sample data is subjected to feature extraction processing to obtain candidate feature data, including: The weighted meteorological sample data is subjected to dilated causal convolution processing to obtain dilated causal convolution data; The dilated causal convolution data is sequentially subjected to weight normalization processing, activation function processing, and random deactivation processing to obtain the candidate feature data.

7. A training device for a substation micro-meteorological prediction model, characterized in that: The device comprises: The sample weighting module is used to obtain historical meteorological sample data of the substation to be predicted, and input the historical meteorological sample data into the initial dynamic weighting module to obtain weighted meteorological sample data; A feature extraction module is used to input the weighted meteorological sample data into an initial time series feature extraction module to obtain meteorological time series feature data; A fully connected layer module is used to input the meteorological time series feature data into the initial fully connected layer to obtain meteorological forecast sample data; An initial training module is used to obtain actual meteorological data and iteratively train the initial dynamic weighting module, the initial time series feature extraction module, and the initial fully connected layer based on the meteorological forecast sample data and the actual meteorological data to obtain a trained target dynamic weighting module, a target time series feature extraction module, and a candidate fully connected layer; A secondary training module is configured to employ a vertical and horizontal cross optimization model to iteratively train the candidate fully connected layer based on the historical meteorological sample data, the actual meteorological data, the target dynamic weighting module, and the target time series feature extraction module to obtain a trained target fully connected layer; The model determination module is used to obtain a trained substation micrometeorological prediction model based on the target dynamic weighting module, the target time series feature extraction module and the target fully connected layer.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images