A feature extraction method and system for short-term photovoltaic power prediction

Abstract

The present invention relates to the field of photovoltaic power prediction technology, and discloses a feature extraction method and system for short-term photovoltaic power prediction, the method comprising: obtaining a data source, including historical measured irradiance data and historical meteorological data of a station, and performing data preprocessing on the data source; performing meteorological marking on the historical measured irradiance data through fast Fourier transform to obtain meteorological marking classification results; constructing a convolutional neural network and a long short-term memory network meteorological classification model to predict the meteorological classification of each day, and inputting the classification results as features into a photovoltaic power prediction model to output short-term prediction results of photovoltaic power generation. The present invention extracts meteorological features through fast Fourier transform, and classifies and predicts meteorological data in combination with a CNN+LSTM model, thereby achieving more accurate photovoltaic power generation prediction and improving the accuracy and reliability of the prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power prediction, and in particular to a feature extraction method and system for short-term photovoltaic power prediction. Background Art

[0002] With the rapid development of renewable energy, photovoltaic power generation has been widely used around the world due to its advantages such as cleanliness and renewableness. However, photovoltaic power generation is affected by factors such as weather conditions and irradiance, and the power output has large fluctuations. Under different weather types, the characteristics of photovoltaic power generation are quite different. In order to improve the accuracy of photovoltaic power prediction, it is particularly important to extract weather type characteristics.

[0003] Fast Fourier transform (FFT), as a signal processing method, can convert time domain signals into frequency domain signals, thereby revealing the spectral characteristics of the signal. By performing Fourier transform on the measured irradiance data, its frequency components can be analyzed, and then the features reflecting the weather conditions can be extracted. These features can be used as input features of the photovoltaic power prediction model, increasing the model's ability to resolve different weather scenarios and improving the accuracy of the prediction. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a feature extraction method for short-term photovoltaic power prediction to solve the problem of how to improve the accuracy of photovoltaic power prediction.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a feature extraction method for short-term photovoltaic power prediction, comprising: obtaining a data source, including historical measured irradiance data and historical meteorological data of a site, and performing data preprocessing on the data source;

[0008] Performing meteorological marking on the historical measured irradiance data by fast Fourier transform to obtain a meteorological marking classification result;

[0009] Convolutional neural network and long short-term memory network meteorological classification models are constructed to predict the daily meteorological classification conditions. The classification results are input as features into the photovoltaic power prediction model to output the short-term prediction results of photovoltaic power generation.

[0010] As a preferred solution of the feature extraction method for short-term photovoltaic power prediction described in the present invention, historical meteorological data is obtained through meteorological information, and the historical meteorological data includes: direct solar radiation in the entire sky, total cloud cover, low cloud cover, medium cloud cover, high cloud cover, cloud base height, large-scale precipitation and total precipitation.

[0011] As a preferred solution of the feature extraction method for short-term photovoltaic power prediction described in the present invention, the data preprocessing includes:

[0012] Remove abnormal values, including power limit values ​​and over-limit values;

[0013] Removing the power limit value includes: screening according to the power limit record of the station, removing the power data affected by the power limit, checking the power limit information in the data record, and removing part of the data affected by the power limit;

[0014] Removing the over-limit value includes: setting a meteorological quantity threshold, screening out data points exceeding the meteorological quantity threshold, and deleting all abnormal power data less than 0.

[0015] As a preferred solution of the feature extraction method for short-term photovoltaic power prediction described in the present invention, the meteorological label classification result obtained includes:

[0016] Prepare input data, and organize the daily irradiance data in the historical measured irradiance data into a time series form with equal time intervals;

[0017] The time series signal is converted into a frequency domain signal through the fast Fourier transform function, which is expressed as:

[0018] ,

[0019] in, represents the frequency domain signal, represents a time series signal, Indicates the signal length, Indicates frequency, is an imaginary unit;

[0020] According to the Fourier transform result, the positive frequency part is selected and the amplitude of each frequency component in the spectrum is calculated;

[0021] Sort the amplitudes, obtain the first three frequency components with the largest amplitudes, calculate the total amplitude of the frequency components, and calculate the proportion of each frequency component in the total amplitude;

[0022] Set the amplitude threshold and mark the weather according to the total amplitude ratio of the first three frequency components.

[0023] As a preferred solution of the feature extraction method for short-term photovoltaic power prediction described in the present invention, the short-term prediction result of the output photovoltaic power generation includes:

[0024] Using the historical meteorological data as model input data;

[0025] Serialize the input data, divide the daily observation data into time series data sets, and group the data sets by date;

[0026] Perform characterization on the serialized data, calculate the difference value, and perform difference calculation on the cumulative value;

[0027] Standardize and normalize the data and scale the data to the range of [0,1];

[0028] Construct a network model and perform model training, input the historical meteorological data, and output the meteorological classification results for each day;

[0029] The meteorological classification results and other meteorological data are input as features into a photovoltaic power prediction model to perform photovoltaic power prediction and output a short-term prediction result of photovoltaic power generation.

[0030] As a preferred solution of the feature extraction method for short-term photovoltaic power prediction described in the present invention, the construction of a meteorological classification model includes:

[0031] Feature extraction is performed through two one-dimensional convolutional layers. The first layer contains 16 convolution kernels and the second layer contains 32 convolution kernels. The activation function is ReLU and causal padding is used.

[0032] The third layer is a bidirectional long short-term memory network layer with 32 neurons, which processes the sequence data and returns the sequence, capturing the important features in the sequence through the attention mechanism layer;

[0033] The fourth layer passes through a fully connected layer and outputs the result of a single neuron.

[0034] As a preferred solution of the feature extraction method for short-term photovoltaic power prediction described in the present invention, the meteorological classification model training includes:

[0035] Optimize through the adam optimization algorithm, the loss function is the mean square error, and the model parameters are optimized by minimizing the loss function;

[0036] When training the model, set the number of epochs to 30. At the end of each epoch, use the validation set data to validate the model. When the validation set performance no longer improves, stop training early to obtain a trained network model.

[0037] In a second aspect, the present invention provides a feature extraction system for short-term photovoltaic power prediction, comprising:

[0038] A data collection and cleaning module is used to obtain data sources, including historical measured irradiance data and historical meteorological data of the station, and perform data preprocessing on the data sources;

[0039] A meteorological marking module is used to perform meteorological marking on the historical measured irradiance data through fast Fourier transform to obtain a meteorological marking classification result;

[0040] The prediction module is used to predict the daily meteorological classification through convolutional neural networks and long short-term memory network models, and input the classification results as features into the photovoltaic power prediction model to output the short-term prediction results of photovoltaic power generation.

[0041] In a third aspect, the present invention provides an electronic device, comprising:

[0042] Memory and processor;

[0043] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the feature extraction method for short-term photovoltaic power prediction are implemented.

[0044] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the feature extraction method for short-term photovoltaic power prediction.

[0045] Compared with the prior art, the present invention has the following beneficial effects: the present invention extracts meteorological features through fast Fourier transform, and classifies and predicts meteorological data in combination with the CNN+LSTM model, thereby achieving more accurate photovoltaic power generation prediction and improving the accuracy and reliability of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0047] Figure 1 A schematic diagram of the overall process of a feature extraction method for short-term photovoltaic power prediction according to an embodiment of the present invention;

[0048] Figure 2A comparative schematic diagram of a feature extraction method for short-term photovoltaic power prediction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0050] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0052] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0053] Example 1, reference Figure 1 , which is an embodiment of the present invention, provides a feature extraction method for short-term photovoltaic power prediction, comprising:

[0054] S101, obtaining data sources, including historical measured irradiance data and historical meteorological data of the station, and performing data preprocessing on the data sources.

[0055] Preferably, the historical meteorological data includes:

[0056] Historical meteorological data are obtained through meteorological information and stored in the database, including total sky direct solar radiation (fdir), total cloud cover (tcc), low cloud cover (cc), middle cloud cover (mcc), high cloud cover (hcc), cloud base height (cbh), large-scale precipitation (Isp), and total precipitation (tp). The above data sources are obtained through the database, and further, the database query function is used to extract the required meteorological data.

[0057] Data preprocessing includes:

[0058] To remove outliers, the following two types of outliers need to be removed, including power limit values ​​and over-limit values, to ensure data quality and model accuracy;

[0059] Removing power restriction values ​​includes: screening according to the power restriction records of the site, removing power data affected by power restriction, checking the power restriction information in the data records, and removing part of the data affected by power restriction.

[0060] Removing out-of-limit values ​​includes: removing power data that exceeds the meteorological quantity threshold or is less than 0, setting the meteorological quantity threshold, filtering out data points that exceed the meteorological quantity threshold, and deleting all abnormal power data that is less than 0 to ensure the validity and accuracy of the input data.

[0061] S102, meteorologically marking the historical measured irradiance data through fast Fourier transform to obtain meteorological marking classification results.

[0062] Preferably, obtaining the meteorological mark classification result includes:

[0063] Prepare input data and organize the daily irradiance data in the historical measured irradiance data into a time series form with equal time intervals;

[0064] The time series signal is converted into a frequency domain signal through the fast Fourier transform (FFT) function, which is expressed as:

[0065] ,

[0066] in, represents the frequency domain signal, represents a time series signal, Indicates the signal length, Indicates frequency, is an imaginary unit;

[0067] According to the Fourier transform result, the positive frequency part is selected and the amplitude of each frequency component in the spectrum is calculated, that is, the modulus of the complex number;

[0068] It should be noted that the Fourier transform results contain positive and negative frequency parts. Since the negative frequency is the mirror image of the positive frequency, only the positive frequency part needs to be considered in actual analysis. The positive frequency part is selected and the amplitude of each frequency component is calculated. These amplitudes represent the intensity of the corresponding frequency component in the original signal.

[0069] Sort the amplitudes, obtain the first three frequency components with the largest amplitudes, calculate the total amplitude of the frequency components, and calculate the proportion of each frequency component in the total amplitude;

[0070] Set the amplitude threshold, mark the weather according to the total amplitude ratio of the first three frequency components, and mark the weather as sunny, cloudy or rainy according to the spectrum analysis results;

[0071] Specifically, the amplitude threshold is preferably [40, 70]. If the proportion is higher than the set amplitude threshold, that is, higher than 70%, it is marked as sunny; if it is between the amplitude threshold range, that is, between 40% and 70%, it is marked as cloudy; if it is lower than the amplitude threshold, that is, lower than 40%, it is marked as rainy.

[0072] It should be noted that through the fast Fourier transform (FFT), a periodic or non-periodic time signal is decomposed into a superposition of a series of sine waves or cosine waves, which have different frequencies, amplitudes and phases. On sunny days, the solar energy received by the photovoltaic system is relatively stable, and the power output is similar to a sine wave. After the Fourier transform, the spectrum is mainly concentrated on a few frequency components, and the amplitudes of these frequency components are large. On non-sunny days (such as cloudy or rainy days), the solar energy received by the photovoltaic system is unstable, and the power output fluctuates greatly. After the Fourier transform, the spectrum is relatively scattered, and the amplitudes of multiple frequency components are small.

[0073] The format of the input data is time series data, which records the daily irradiance values. The data is usually indexed by timestamps, and each time point corresponds to a measured irradiance value. The data required for the Fourier transform to be input is a time series with equal time intervals. The power data is Fourier transformed to obtain a frequency domain signal. The Fourier transform reveals the periodic components in the signal by converting the time domain signal to the frequency domain.

[0074] S103, constructing a convolutional neural network and a long short-term memory network meteorological classification model, predicting the meteorological classification of each day, and inputting the classification results as features into a photovoltaic power prediction model, and outputting a short-term prediction result of photovoltaic power generation.

[0075] Preferably, the short-term prediction results of the output photovoltaic power generation power include:

[0076] The historical meteorological data are used as the model input data, including the total sky direct solar radiation (fdir), total cloud cover (tcc), low cloud cover (cc), middle cloud cover (mcc), high cloud cover (hcc), cloud base height (cbh), large-scale precipitation (Isp), and total precipitation (tp) at the station.

[0077] Serialize the input data, divide the daily observation data into time series data sets, and group the data sets by date. Specifically, the number of data points per day is 96;

[0078] Perform characterization on the serialized data, calculate the difference value to capture the change trend of the data, and perform difference calculation on the cumulative value to eliminate the cumulative effect;

[0079] According to the characteristics of each key meteorological data, the data is standardized and normalized, and the data is scaled to the range of [0,1]. Specifically, the Min-Max normalization method is used;

[0080] Build a meteorological classification model and train the model, that is, use the trained CNN+LSTM model, input historical meteorological data, and output daily meteorological classification results;

[0081] The meteorological classification results and other meteorological data are input as features into the photovoltaic power prediction model to perform photovoltaic power prediction and output short-term prediction results of photovoltaic power generation. The photovoltaic power prediction model can be an xgboost model, thereby improving the accuracy of photovoltaic power prediction.

[0082] Preferably, constructing a meteorological classification model includes:

[0083] Feature extraction is performed through two one-dimensional convolutional layers. The first layer contains 16 convolution kernels and the second layer contains 32 convolution kernels. The activation function is ReLU and causal padding is used.

[0084] The third layer is a bidirectional long short-term memory network layer with 32 neurons, which processes the sequence data and returns the sequence, capturing the important features in the sequence through the attention mechanism layer;

[0085] The fourth layer outputs the results of a single neuron through a fully connected layer.

[0086] Specifically, the model includes the following layers: the first layer is a one-dimensional convolution layer with 16 convolution kernels, using the ReLU activation function and causal padding, which is used to receive input data of time step multiplied by 2 plus 1, as well as multiple feature dimensions; the second layer is a one-dimensional convolution layer with 32 convolution kernels, using the ReLU activation function and causal padding, and feature extraction is performed through these two one-dimensional convolution layers; the third layer is a bidirectional long short-term memory network (BiLSTM) layer with 32 neurons, which returns a sequence, processes the sequence data, and captures the complex relationship in the time series; the fourth layer is an attention mechanism layer, which is used to capture important features in the sequence and strengthen attention to key time step information. Finally, a fully connected layer outputs the results of a single neuron.

[0087] Preferably, the meteorological classification model training includes:

[0088] Optimize through the adam optimization algorithm, the loss function is the mean square error (MSE), and the model parameters are optimized by minimizing the loss function;

[0089] When training the model, the number of epochs is set to 30. At the end of each epoch, the model is verified using the validation set data to monitor the performance of the model. When the performance of the validation set no longer improves, the training is stopped early to obtain a trained meteorological classification model.

[0090] In summary, the present invention extracts meteorological features through fast Fourier transform and classifies and predicts meteorological data in combination with a deep learning model, thereby achieving more accurate photovoltaic power generation prediction and improving the accuracy and reliability of the prediction.

[0091] The above is a schematic scheme of a feature extraction method for short-term photovoltaic power prediction in this embodiment. It should be noted that the technical scheme of the feature extraction system for short-term photovoltaic power prediction and the technical scheme of the feature extraction method for short-term photovoltaic power prediction described above belong to the same concept. For details not described in detail in the technical scheme of the feature extraction system for short-term photovoltaic power prediction in this embodiment, please refer to the description of the technical scheme of the feature extraction method for short-term photovoltaic power prediction described above.

[0092] The feature extraction system for short-term photovoltaic power prediction in this embodiment includes:

[0093] The data collection and cleaning module is used to obtain data sources, including the historical measured irradiance data and historical meteorological data of the station, and perform data preprocessing on the data sources;

[0094] The meteorological marking module is used to perform meteorological marking on the historical measured irradiance data through fast Fourier transform to obtain meteorological marking classification results;

[0095] The prediction module is used to predict the daily meteorological classification through convolutional neural networks and long short-term memory network models, and input the classification results as features into the photovoltaic power prediction model to output the short-term prediction results of photovoltaic power generation.

[0096] This embodiment further provides an electronic device, which is applicable to the case of feature extraction for short-term photovoltaic power prediction, including:

[0097] A memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the feature extraction method for short-term photovoltaic power prediction as proposed in the above embodiment.

[0098] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the feature extraction method for short-term photovoltaic power prediction proposed in the above embodiment is implemented.

[0099] The storage medium proposed in this embodiment and the feature extraction method for implementing short-term photovoltaic power prediction proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0100] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0101] Example 2, refer to Table 1 and Figure 2 , which is an embodiment of the present invention, provides a feature extraction method for short-term photovoltaic power prediction. In order to verify its beneficial effects, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0102] An actual case of the present invention in a wind farm in North China is as follows: the installed capacity Cap of the wind power station is 50.0MW, and the measured power data for 14 days from May 1 to May 14, 2024 and the instantaneous value of 15 minutes of short-term predicted power data at the corresponding time are collected as sample data.

[0103] Furthermore, the short-term forecast assessment accuracy of the power grid to which the station belongs is as follows:

[0104] Acc =1 - ∑ i=1 n [ ( P i - P i ' ) 2 | P i - P i ' | ∑ i=1 n | P i - P i ' | ] Cap ,

[0105] in, is the day-ahead prediction accuracy, The maximum operating capacity of the new energy power station on the assessment day. is the actual power of the new energy power station at time i, is the predicted power of the new energy power station at time i, and n is the predicted points for the assessment on that day.

[0106] Furthermore, in step S101, historical measured irradiance and historical forecast meteorological data are obtained, and power limit values ​​and over-limit abnormal data are removed to ensure data quality and model accuracy. In step S102, meteorological marking is performed based on fast Fourier transform (FFT) to identify sunny, cloudy and rainy days. In step S103, the CNN+LSTM model is used to perform meteorological classification and prediction, and feature extraction and standardization processing are performed on each meteorological factor. The trained CNN+LSTM model is used to output the daily meteorological classification results, and these results are input as features into the photovoltaic power prediction model (for example, xgboost). Finally, the short-term prediction results of photovoltaic power generation are obtained. Compared with the traditional model that does not use weather classification features, the results are shown in Table 1 and Figure 2 shown.

[0107] Table 1: Comparison of prediction results with and without using weather classification features

[0108]

[0109] Furthermore, Table 1 shows the prediction accuracy of the two methods for 14 days. From the data in Table 1, it can be found that in the 14-day instance data, the prediction results of the method using weather classification features are higher than those of the method not using weather classification features on 11 days, and Figure 2 The data in the comparison effect diagram show that the feature extraction method can effectively improve the prediction accuracy.

[0110] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A feature extraction method for short-term photovoltaic power prediction, characterized in that: include: Acquire data sources, including historical measured irradiance data and historical meteorological data of the station, and perform data preprocessing on the data sources; Performing meteorological marking on the historical measured irradiance data by fast Fourier transform to obtain a meteorological marking classification result; Construct a convolutional neural network and a long short-term memory network meteorological classification model to predict the daily meteorological classification, and input the classification results as features into the photovoltaic power prediction model to output the short-term prediction results of photovoltaic power generation; Acquire historical meteorological data through meteorological information, wherein the historical meteorological data includes: total sky solar direct radiation, total cloud cover, low cloud cover, medium cloud cover, high cloud cover, cloud base height, large-scale precipitation and total precipitation; The data preprocessing includes: Remove abnormal values, including power limit values ​​and over-limit values; Removing the power limit value includes: screening according to the power limit record of the station, removing the power data affected by the power limit, checking the power limit information in the data record, and removing part of the data affected by the power limit; Removing the over-limit value includes: setting a meteorological quantity threshold, filtering out data points exceeding the meteorological quantity threshold, and deleting all abnormal power data less than 0; The meteorological mark classification result obtained includes: Prepare input data, and organize the daily irradiance data in the historical measured irradiance data into a time series form with equal time intervals; The time series signal is converted into a frequency domain signal through the fast Fourier transform function, which is expressed as: Among them, X(f) represents the frequency domain signal, x(n) represents the time series signal, N represents the signal length, f represents the frequency, and j is the imaginary unit; According to the Fourier transform result, the positive frequency part is selected and the amplitude of each frequency component in the spectrum is calculated; Sorting the amplitudes, obtaining the first three frequency components with the largest amplitudes, calculating the total amplitude of the frequency components, and calculating the proportion of each frequency component in the total amplitude; Set the amplitude threshold, mark the weather according to the total amplitude ratio of the first three frequency components, and mark the weather as sunny, cloudy or rainy according to the spectrum analysis results; The short-term prediction results of the output photovoltaic power generation power include: Using the historical meteorological data as model input data; Serialize the input data, divide the daily observation data into time series data sets, and group the data sets by date; Perform characterization on the serialized data, calculate the difference value, and perform difference calculation on the cumulative value; Standardize and normalize the data and scale the data to the range of [0,1]; Construct a meteorological classification model and perform model training, input the historical meteorological data, and output the meteorological classification results for each day; The meteorological classification results and other meteorological data are input as features into a photovoltaic power prediction model, where the photovoltaic power prediction model is an xgboost model, to perform photovoltaic power prediction and output a short-term prediction result of photovoltaic power generation.

2. The feature extraction method for short-term photovoltaic power prediction according to claim 1, characterized in that: The construction of the meteorological classification model comprises: Feature extraction is performed through two one-dimensional convolutional layers. The first layer contains 16 convolution kernels and the second layer contains 32 convolution kernels. The activation function is ReLU and causal padding is used. The third layer is a bidirectional long short-term memory network layer with 32 neurons, which processes the sequence data and returns the sequence, capturing the important features in the sequence through the attention mechanism layer; The fourth layer passes through a fully connected layer and outputs the result of a single neuron.

3. The feature extraction method for short-term photovoltaic power prediction according to claim 2, characterized in that: Weather classification model training includes: Optimize through the adam optimization algorithm, the loss function is the mean square error, and the model parameters are optimized by minimizing the loss function; When training the model, set the number of epochs to 30. At the end of each epoch, use the validation set data to validate the model. When the validation set performance no longer improves, stop training early to obtain a trained network model.

4. A feature extraction system for short-term photovoltaic power prediction, characterized in that: include, A data collection and cleaning module is used to obtain data sources, including historical measured irradiance data and historical meteorological data of the station, and perform data preprocessing on the data sources; A meteorological marking module is used to perform meteorological marking on the historical measured irradiance data through fast Fourier transform to obtain a meteorological marking classification result; The prediction module is used to predict the daily meteorological classification through convolutional neural networks and long short-term memory network models, and input the classification results as features into the photovoltaic power prediction model to output the short-term prediction results of photovoltaic power generation; The historical meteorological data are obtained through meteorological information, and the historical meteorological data include: direct solar radiation from all over the sky, total cloud cover, low cloud cover, medium cloud cover, high cloud cover, cloud base height, large-scale precipitation, and total precipitation; The data preprocessing includes: Remove abnormal values, including power limit values ​​and over-limit values; Removing the power limit value includes: screening according to the power limit record of the station, removing the power data affected by the power limit, checking the power limit information in the data record, and removing part of the data affected by the power limit; Removing the over-limit value includes: setting a meteorological quantity threshold, filtering out data points exceeding the meteorological quantity threshold, and deleting all abnormal power data less than 0; The meteorological mark classification result obtained includes: Prepare input data, and organize the daily irradiance data in the historical measured irradiance data into a time series form with equal time intervals; The time series signal is converted into a frequency domain signal through the fast Fourier transform function, which is expressed as: Among them, X(f) represents the frequency domain signal, x(n) represents the time series signal, N represents the signal length, f represents the frequency, and j is the imaginary unit; According to the Fourier transform result, the positive frequency part is selected and the amplitude of each frequency component in the spectrum is calculated; Sorting the amplitudes, obtaining the first three frequency components with the largest amplitudes, calculating the total amplitude of the frequency components, and calculating the proportion of each frequency component in the total amplitude; Set the amplitude threshold, mark the weather according to the total amplitude ratio of the first three frequency components, and mark the weather as sunny, cloudy or rainy according to the spectrum analysis results; The short-term prediction results of the output photovoltaic power generation power include: Using the historical meteorological data as model input data; Serialize the input data, divide the daily observation data into time series data sets, and group the data sets by date; Perform characterization on the serialized data, calculate the difference value, and perform difference calculation on the cumulative value; Standardize and normalize the data and scale the data to the range of [0,1]; Construct a meteorological classification model and perform model training, input the historical meteorological data, and output the meteorological classification results for each day; The meteorological classification results and other meteorological data are input as features into a photovoltaic power prediction model, where the photovoltaic power prediction model is an xgboost model, to perform photovoltaic power prediction and output a short-term prediction result of photovoltaic power generation.

5. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the feature extraction method for short-term photovoltaic power prediction described in any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of the feature extraction method for short-term photovoltaic power prediction as described in any one of claims 1 to 3.

Images