A photovoltaic power prediction method, device, equipment and readable storage medium

Abstract

The application provides a photovoltaic power prediction method, device and equipment and a readable storage medium, relates to the photovoltaic power prediction field, and comprises the following steps: acquiring prediction day and a plurality of candidate day data; determining the meteorological factor with the highest correlation degree with photovoltaic power by using the Pearson correlation coefficient method according to meteorological factor data and photovoltaic power data; performing wavelet packet decomposition on the meteorological factor data with the highest correlation degree, and selecting a plurality of similar days from a plurality of candidate days according to the wavelet packet decomposition result and the remaining meteorological factor data; performing wavelet packet decomposition on the photovoltaic power data of all similar days to obtain a plurality of power subsequences, and constructing a plurality of similar day data sets by using the plurality of power subsequences and the meteorological factor data of similar days; and training a gated recurrent unit network by using all the similar day data sets to predict the photovoltaic power of the prediction day. The application solves the technical problem that a single method cannot accurately complete photovoltaic power prediction in the prior art.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power prediction, specifically to a photovoltaic power prediction method, apparatus, device, and readable storage medium. Background Technology

[0002] New energy power generation is receiving increasing attention, especially clean and renewable energy sources like solar power. However, the intermittency and uncertainty of photovoltaic (PV) power generation can significantly impact the reliability and stability of power systems, and PV power prediction is an effective solution to these problems. Currently, the most commonly used technique for PV power prediction is statistical methods, which typically utilize machine learning or regression models to describe the relationship between historical power generation data and weather variables for prediction. Machine learning in statistical methods often employs shallow neural networks as the prediction network, leveraging their unique ability to express arbitrary nonlinear mappings to achieve better prediction results; however, as the amount and dimensionality of data increase, the efficiency of shallow neural networks decreases when solving prediction problems, making it difficult to uncover the deep features contained within large datasets. Summary of the Invention

[0003] The purpose of this invention is to provide a photovoltaic power prediction method, apparatus, device, and readable storage medium to improve the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows:

[0004] In a first aspect, this application provides a photovoltaic power prediction method, including:

[0005] Acquire data for the predicted date and multiple candidate dates, wherein the multiple candidate date data includes at least two meteorological impact data and photovoltaic power data;

[0006] Based on meteorological factor data and photovoltaic power data, the meteorological factor with the highest correlation to photovoltaic power was determined using the Pearson correlation coefficient method.

[0007] Wavelet packet decomposition is performed on the meteorological factor data with the highest correlation. Based on the wavelet packet decomposition results and the remaining meteorological factor data, multiple similar days are selected from several candidate days.

[0008] Wavelet packet decomposition was performed on the photovoltaic power data of all similar days to obtain multiple power subsequences. Multiple similar day datasets were constructed by combining the multiple power subsequences with meteorological factor data of similar days.

[0009] A gated recurrent unit network is trained using all the aforementioned similar day datasets, and the photovoltaic power of the predicted day is predicted by the trained gated recurrent unit network.

[0010] Secondly, this application also provides a photovoltaic power prediction device, comprising:

[0011] Acquisition module: used to acquire data for the predicted date and multiple candidate dates, wherein the multiple candidate date data includes at least two meteorological impact data and photovoltaic power data;

[0012] Determination module: Used to determine the meteorological factor with the highest correlation to photovoltaic power based on meteorological factor data and photovoltaic power data, using the Pearson correlation coefficient method;

[0013] Selection module: used to perform wavelet packet decomposition on the meteorological factor data with the highest correlation, and select multiple similar days from several candidate days based on the wavelet packet decomposition results and the remaining meteorological factor data;

[0014] Dataset construction module: used to perform wavelet packet decomposition on photovoltaic power data of all similar days to obtain multiple power subsequences, and to construct multiple similar day datasets from several of the power subsequences and meteorological factor data of similar days;

[0015] Prediction module: Used to train a gated recurrent unit network using all the similar day datasets, and the trained gated recurrent unit network is used to predict the photovoltaic power of the predicted day.

[0016] Thirdly, this application also provides a photovoltaic power prediction device, comprising:

[0017] Memory, used to store computer programs;

[0018] A processor is used to implement the steps of the photovoltaic power prediction method when executing the computer program.

[0019] Fourthly, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the photovoltaic power prediction method described above.

[0020] The beneficial effects of this invention are as follows:

[0021] This invention employs wavelet packet decomposition technology to extract high- and low-frequency features from power sequences, and uses gated recurrent unit networks (GRUs) as deep learning to uncover deeper features in time series. Furthermore, the improved beetle whisker algorithm efficiently optimizes the network structure and finds better similarity day factor weight parameters. This invention combines wavelet decomposition, GRUs, and the improved beetle whisker algorithm to solve the problem that single methods cannot effectively predict photovoltaic power, enabling the extraction of more features from the data, improving data quality, and saving computation time.

[0022] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of the photovoltaic power prediction method described in the embodiments of the present invention;

[0025] Figure 2 This is a comparison of the photovoltaic power prediction values ​​obtained by different methods in the embodiments of the present invention. Figure 1 ;

[0026] Figure 3 This is a comparison of the photovoltaic power prediction values ​​obtained by different methods in the embodiments of the present invention. Figure 2 ;

[0027] Figure 4 This is a comparison of the photovoltaic power prediction values ​​obtained by different methods in the embodiments of the present invention. Figure 3 ;

[0028] Figure 5 This is a schematic diagram of the photovoltaic power prediction device described in an embodiment of the present invention;

[0029] Figure 6 This is a schematic diagram of the photovoltaic power prediction device described in an embodiment of the present invention.

[0030] Marked in the image:

[0031] 01. Acquisition Module; 011. Candidate Day Determination Unit; 012. Data Cleaning Unit; 013. Normalization Processing Unit; 02. Determination Module; 03. Selection Module; 031. First Decomposition Unit; 032. First Reconstruction Unit; 033. First Construction Unit; 034. First Calculation Unit; 035. Second Calculation Unit; 0351. First Optimization Unit; 0352. Third Calculation Unit; 036. Selection Unit; 04. Dataset Construction Module; 041. Second Decomposition Unit; 042. Third Decomposition Unit; 043. Fourth Decomposition Unit; 044. Second Reconstruction Unit; 045. First Construction Unit; 05. Prediction Module; 051. Second Construction Unit; 052. Determination Unit; 053. Second Optimization Unit; 054. Training Unit; 055. Prediction Unit;

[0032] 800. Photovoltaic power prediction equipment; 801. Processor; 802. Memory; 803. Multimedia components; 804. I / O interface; 805. Communication components. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0034] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0035] Example 1:

[0036] This embodiment provides a photovoltaic power prediction method.

[0037] See Figure 1 The figure shows that this method includes:

[0038] S1. Obtain forecast date and multiple candidate date data, wherein the multiple candidate date data includes at least two meteorological impact data and photovoltaic power data;

[0039] Specifically, step S1 includes:

[0040] S11. Days that are in the same year and season as the prediction date and are before the prediction date, and days that are in the previous year and are in the same season as the prediction date, are identified as candidate days;

[0041] S12. Obtain meteorological factor data and photovoltaic power data for each candidate day from the preset photovoltaic system, and clean the meteorological factor data and photovoltaic power data to transform outliers into a reasonable range;

[0042] In this embodiment, meteorological factor data and photovoltaic power data for candidate days were obtained from data collected by the Australian Solar Centre in 2014 and 2015, with a data sampling frequency of 5 minutes. The meteorological factors include seven meteorological data points: global horizontal radiation (total solar radiation received on the ground horizontal surface, including direct radiation (DN I) and diffuse radiation (DH I)), diffuse horizontal radiation, wind speed, wind direction, rainfall, ambient temperature, and relative humidity.

[0043] Transforming outliers to a reasonable range includes:

[0044] Convert negative values ​​in photovoltaic power to 0;

[0045] The negative values ​​in the wind direction are converted into the average of the normal values ​​before and after the negative values;

[0046] The box plot method was used to convert the outliers of other meteorological factors into a reasonable range.

[0047] S13. Normalize the cleaned meteorological factor data and photovoltaic power data.

[0048] Based on the above embodiments, this method further includes:

[0049] S2. Based on meteorological factor data and photovoltaic power data, use the Pearson correlation coefficient method to determine the meteorological factor with the highest correlation to photovoltaic power;

[0050] In this embodiment, the correlation between photovoltaic power and other meteorological factors is calculated using the Pearson correlation coefficient method. Meteorological factors with low correlation are removed, and meteorological factors with high correlation are retained, including five meteorological factors: global water radiation, diffuse horizontal radiation, wind speed, ambient temperature, and relative humidity. Among them, global water radiation is the meteorological factor with the highest correlation.

[0051] Based on the above embodiments, this method further includes:

[0052] S3. Perform wavelet packet decomposition on the meteorological factor data with the highest correlation, and select multiple similar days from several candidate days based on the wavelet packet decomposition results and the remaining meteorological factor data;

[0053] Specifically, step S3 includes:

[0054] S31. Perform wavelet packet decomposition on the meteorological factor data with the highest correlation to obtain multiple frequency signals, the frequency signals including high-frequency signals and low-frequency signals;

[0055] S32. Reconstruct the plurality of frequency signals individually into fluctuation components (high-frequency components) and trend components (low-frequency components);

[0056] S33. Construct daily feature vectors for the prediction day and each candidate day based on the remaining meteorological factor data. Specifically, the daily feature vectors consist of daily average temperature, daily maximum temperature, daily minimum temperature, daily average wind speed, daily average relative humidity, daily average diffuse horizontal radiation, daily average rainfall, and the ratio of diffuse horizontal radiation to global horizontal radiation.

[0057] S34. Calculate the Pearson correlation coefficients of the predicted date and each candidate date on the fluctuation component, trend component, and daily eigenvector to obtain the fluctuation factor, trend factor, and comprehensive weather factor for each candidate date;

[0058] In this embodiment, the method for calculating the Pearson correlation coefficient is existing technology and will not be described in detail here;

[0059] Specifically, the Pearson correlation coefficient between the predicted date and the i-th candidate date on the fluctuation component is calculated, thus obtaining the fluctuation factor VF of the predicted date and the i-th candidate date on the fluctuation component of global horizontal radiation. i (Volatility similarity);

[0060] Similarly, the combined weather factor CWF of the predicted day and the i-th candidate day on the daily eigenvector can be obtained. i (Daily eigenvector similarity) and the trend factor TF of the trend component of the global horizontal radiation between the predicted day and the i-th candidate day. i (Trend similarity).

[0061] S35. Calculate the similarity between the predicted day and each candidate day using the trend factor, fluctuation factor, and comprehensive weather factor;

[0062] Specifically, step S35 includes:

[0063] S351. Obtain a preset beetle whisker search mathematical model, and use the trend factor, fluctuation factor, comprehensive weather factor, and beetle whisker search mathematical model to optimize and obtain the weights of the trend factor, fluctuation factor, and comprehensive weather factor:

[0064] In this embodiment, the beetle whisker search mathematical model is an improved model that combines the beetle whisker search algorithm with the Levy flight strategy and linearly decreasing inertial weights.

[0065] The formula for calculating the similarity between the predicted date and the candidate date is:

[0066]

[0067] In the formula, R i This represents the similarity between the predicted date and the i-th candidate date. This represents the weights of the comprehensive weather factors. The weights of the trend factor are represented. The weights of the volatility factor, the trend factor, the volatility factor, and the comprehensive weather factor are the parameters to be optimized in the longhorn beetle search mathematical model;

[0068] Specifically, the optimization method for the longhorn beetle search mathematical model is as follows:

[0069] 1) Initialize the longhorn beetle's step size, sensing range, and position, and use these three weights as the beetle's position. Will Initialize to and initial fitness As the optimal fitness best ;

[0070] 2) Randomize the beetle orientation vector and normalize it:

[0071]

[0072] In the formula, This represents the normalized direction vector of the longhorn beetle, k is the optimization dimension, i.e. the number of parameters to be optimized, and rands() is a random function;

[0073] 3) Calculate the positions of the longhorn beetle's left and right whiskers:

[0074]

[0075] Where, x r Indicates the position of the longhorn beetle's left whisker, x l Indicates the position of the longhorn beetle's right whiskers. This indicates the location of the longhorn beetle, and d represents the sensing range of the longhorn beetle's whiskers;

[0076] 4) Use the beetle whisker algorithm to update the beetle position and obtain the first candidate position:

[0077]

[0078] Where, x ′ The first candidate position is represented by δ, the beetle step size is represented by sign(), and the fitness function is represented by fit().

[0079] 5) Using the Levy flight strategy, further update the beetle's position to obtain a second candidate position:

[0080]

[0081] In the formula, x levy 's' represents the second candidate position, and 's' represents the Levy random number.

[0082] 6) Compare the fitness functions of the first candidate position and the second candidate position, and select the beetle position with the larger fitness function value as the new beetle position. The expression is:

[0083]

[0084] Wherein, fit() represents the fitness function. Specifically, the similarity between each candidate day and the predicted day is calculated using formula (1). All similarities are sorted in descending order, and the candidate days corresponding to the top 20 similarities in the sort are selected. The Euclidean distance between the photovoltaic power of any two days among the 20 candidate days is calculated, and the sum of all Euclidean distances is used as the fitness function.

[0085] 7) If the fitness function of the new longhorn beetle location is fit(x n (Better than optimal fitness) best Then update the optimal fitness and the position of the longhorn beetle to complete one iteration:

[0086]

[0087] 8) Check if the number of iterations is less than the preset number of iterations. If so, repeat step 2); otherwise, output the optimal position x. best and the optimal fitness function fit best .

[0088] 9) Based on the optimal position x best get

[0089] S352. The similarity between the predicted day and the candidate day is calculated by weighting the trend factor, fluctuation factor and comprehensive weather factor using the corresponding weights of the trend factor, fluctuation factor and comprehensive weather factor; specifically, R is calculated using formula (1). i .

[0090] S36. Select multiple similar days from several candidate days based on their similarity.

[0091] Sort all similarities in descending order and select the top 20 candidate days as similar days.

[0092] Based on the above embodiments, this method further includes:

[0093] S4. Wavelet packet decomposition is performed on the photovoltaic power data of all similar days to obtain multiple power subsequences. Multiple similar day datasets are constructed from the multiple power subsequences and meteorological factor data of similar days.

[0094] Specifically, step S4 includes:

[0095] S41. Use low-pass and high-pass filters to decompose photovoltaic power data from similar days to obtain low-frequency and high-frequency signals;

[0096] S42. The low-frequency signal is decomposed using a low-pass filter and a high-pass filter to obtain a first low-frequency sub-signal and a first high-frequency sub-signal;

[0097] S43. Use a low-pass filter and a high-pass filter to decompose the high-frequency signal to obtain a second low-frequency sub-signal and a second high-frequency sub-signal;

[0098] S44. Perform single-branch reconstruction on the first low-frequency sub-signal, the first high-frequency sub-signal, the second low-frequency sub-signal, and the second high-frequency sub-signal to obtain a preset number of power sub-sequences; in this embodiment, perform single-branch reconstruction on one sub-signal to obtain one power sub-sequence, thereby obtaining 4 power sub-sequences.

[0099] S45. A similar day dataset is constructed from a power subsequence and meteorological factor data of all similar days, thus obtaining 4 similar day datasets.

[0100] Based on the above embodiments, this method further includes:

[0101] S5. Train a gated recurrent unit network using all the similar day datasets, and predict the photovoltaic power of the predicted day using the trained gated recurrent unit network;

[0102] Specifically, step S5 includes:

[0103] S51. Construct a gated recurrent unit network for each similar day dataset. The gated recurrent unit network includes an input layer, a preset number of hidden layers, and an output layer. In this embodiment, two hidden layers are provided.

[0104] S52. Determine the hyperparameters of the gated recurrent unit network to be optimized, including batch size, random dropout ratio of hidden layers (dropout ratio of two hidden layers), and number of neurons;

[0105] S53. Substitute the hyperparameters of the network to be optimized into the preset beetle whisker search mathematical model and iterate to optimize them to obtain the optimized network hyperparameters;

[0106] In this embodiment, the mean square error between the predicted and actual values ​​of the gated recurrent unit network is used as the fitness function. A smaller mean square error indicates a better fitness function. The network hyperparameters are optimized using a beetle whisker search mathematical model. Because the search process for network hyperparameters is similar to... The optimization process is the same, so it will not be repeated here.

[0107] S54. Substitute the optimized network hyperparameters into the corresponding gated recurrent unit network and train it using the corresponding similar daily dataset;

[0108] S55. After summing all the output values ​​and performing inverse normalization, the photovoltaic power of the predicted day is obtained. The output value is the result of a gated recurrent unit network after training, that is, the final value obtained by summing the output values ​​of 4 gated recurrent unit networks and performing inverse normalization is the photovoltaic prediction value.

[0109] Specifically, such as Figure 2-4 As shown, the figures illustrate the photovoltaic power prediction for three different prediction days using different prediction methods. In the figures, MLP represents the photovoltaic power prediction value of the multilayer sensor, GRU represents the photovoltaic power prediction value of the gated loop unit, SG represents the photovoltaic power prediction value of the similar day-gated loop unit, SPG represents the photovoltaic power prediction value of the similar day-particle swarm optimization-gated loop unit, SLG represents the photovoltaic power prediction value of the similar day-improved beetle search algorithm-gated loop unit, and SWLG represents the photovoltaic power prediction value of the method provided in this embodiment. Figures 2-4 As can be seen, the photovoltaic power predicted using SWLG is closest to the actual value, proving that the method provided in this embodiment has the highest prediction accuracy.

[0110] Example 2:

[0111] like Figure 5 As shown, this embodiment provides a photovoltaic power prediction device, the device comprising:

[0112] Acquisition Module 01: Used to acquire data for the predicted date and multiple candidate dates, wherein the multiple candidate date data includes at least two meteorological impact data and photovoltaic power data;

[0113] Module 02: Used to determine the meteorological factor with the highest correlation to photovoltaic power based on meteorological factor data and photovoltaic power data, using the Pearson correlation coefficient method;

[0114] Selection module 03: is used to perform wavelet packet decomposition on the meteorological factor data with the highest correlation, and select multiple similar days from several candidate days based on the wavelet packet decomposition results and the remaining meteorological factor data;

[0115] Dataset Construction Module 04: This module is used to perform wavelet packet decomposition on the photovoltaic power data of all similar days to obtain multiple power subsequences. Multiple similar day datasets are constructed from several of these power subsequences and meteorological factor data of similar days.

[0116] Prediction Module 05: Used to train a gated recurrent unit network using all the similar day datasets, and the trained gated recurrent unit network is used to predict the photovoltaic power of the predicted day.

[0117] Based on the above embodiments, the acquisition module 01 includes:

[0118] Candidate Day Determination Unit 011: Used to determine the days that are in the same year and season as the prediction date and are before the prediction date, and the days that are in the previous year and are in the same season as the prediction date, as candidate days;

[0119] Data cleaning unit 012: used to obtain meteorological factor data and photovoltaic power data for each candidate day from the preset photovoltaic system, and clean the meteorological factor data and photovoltaic power data to transform outliers into a reasonable range;

[0120] Normalization processing unit 013: Used to normalize the cleaned meteorological factor data and photovoltaic power data.

[0121] Based on the above embodiments, the selection module 03 includes:

[0122] First decomposition unit 031: used to perform a wavelet packet decomposition on the meteorological factor data with the highest correlation to obtain multiple frequency signals;

[0123] First reconstruction unit 032: used to reconstruct the plurality of frequency signals into fluctuation components and trend components individually;

[0124] First construction unit 033: used to construct the daily feature vector of the prediction day and each candidate day based on the remaining meteorological factor data;

[0125] First calculation unit 034: used to calculate the Pearson correlation coefficient of the prediction day and each candidate day on the fluctuation component, trend component, and daily eigenvector, to obtain the fluctuation factor, trend factor, and comprehensive weather factor of each candidate day;

[0126] Second calculation unit 035: used to calculate the similarity between the predicted day and each candidate day using the trend factor, fluctuation factor and comprehensive weather factor;

[0127] Selection Unit 036: Used to select multiple similar days from several candidate days based on similarity.

[0128] Based on the above embodiments, the second computing unit 035 includes:

[0129] First optimization unit 0351: used to obtain a preset beetle whisker search mathematical model, and use the trend factor, fluctuation factor, comprehensive weather factor and beetle whisker search mathematical model to optimize and obtain the weight of the trend factor, the weight of the fluctuation factor and the weight of the comprehensive weather factor.

[0130] The third calculation unit 0352 is used to calculate the similarity between the predicted day and the candidate day by using the weights corresponding to the trend factor, fluctuation factor and comprehensive weather factor.

[0131] Based on the above embodiments, the dataset construction module 04 includes:

[0132] Second decomposition unit 041: used to decompose photovoltaic power data of similar days using low-pass and high-pass filters to obtain low-frequency and high-frequency signals;

[0133] Third decomposition unit 042: used to decompose the low-frequency signal using a low-pass filter and a high-pass filter to obtain a first low-frequency sub-signal and a first high-frequency sub-signal;

[0134] Fourth decomposition unit 043: used to decompose high-frequency signals using low-pass and high-pass filters to obtain a second low-frequency sub-signal and a second high-frequency sub-signal;

[0135] Second reconstruction unit 044: used to perform single-branch reconstruction on the first low-frequency sub-signal, the first high-frequency sub-signal, the second low-frequency sub-signal and the second high-frequency sub-signal to obtain a preset number of power sub-sequences;

[0136] First building unit 045: Used to construct a similar day dataset from a power subsequence and meteorological factor data of all similar days.

[0137] Based on the above embodiments, the prediction module 05 includes:

[0138] The second construction unit 051 is used to construct a gated recurrent unit network for each similar day dataset. The gated recurrent unit network includes an input layer, a preset number of hidden layers, and an output layer.

[0139] Determining Unit 052: Used to determine the hyperparameters of the gated recurrent unit network to be optimized, the hyperparameters of which include batch size, random deactivation rate of hidden layers, and number of neurons;

[0140] The second optimization unit 053 is used to substitute the hyperparameters of the network to be optimized into the preset beetle whisker search mathematical model for iterative optimization, so as to obtain the optimized network hyperparameters.

[0141] Training unit 054: used to substitute the optimized network hyperparameters into the corresponding gated recurrent unit network and train it using the corresponding similar daily dataset;

[0142] Prediction Unit 055: This unit sums all the output values ​​and then performs inverse normalization to obtain the photovoltaic power on the predicted day. The output values ​​are the result of a gated recurrent unit network after training.

[0143] It should be noted that the specific manner in which each module performs its operation in the apparatus described in the above embodiments has been described in detail in the embodiments of the method, and will not be elaborated here.

[0144] Example 3:

[0145] Corresponding to the above method embodiments, this embodiment also provides a photovoltaic power prediction device. The photovoltaic power prediction device described below and the photovoltaic power prediction method described above can be referred to each other.

[0146] Figure 6 This is a block diagram illustrating a photovoltaic power prediction device 800 according to an exemplary embodiment. Figure 6 As shown, the photovoltaic power prediction device 800 may include a processor 801 and a memory 802. The photovoltaic power prediction device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0147] The processor 801 controls the overall operation of the photovoltaic power prediction device 800 to complete all or part of the steps in the photovoltaic power prediction method described above. The memory 802 stores various types of data to support the operation of the photovoltaic power prediction device 800. This data may include, for example, instructions for any application or method used to operate on the photovoltaic power prediction device 800, as well as application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons can be virtual or physical buttons. Communication component 805 is used for wired or wireless communication between the photovoltaic power prediction device 800 and other devices. Wireless communication includes, for example, Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof; therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0148] In an exemplary embodiment, the photovoltaic power prediction device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the photovoltaic power prediction method described above.

[0149] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the photovoltaic power prediction method described above. For example, the computer-readable storage medium may be the memory 802 including program instructions described above, which may be executed by the processor 801 of the photovoltaic power prediction device 800 to complete the photovoltaic power prediction method described above.

[0150] Example 4:

[0151] Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the photovoltaic power prediction method described above.

[0152] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the photovoltaic power prediction method described in the above method embodiments.

[0153] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0154] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0155] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A photovoltaic power prediction method, characterized in that, include: Acquire data for the predicted date and multiple candidate dates, wherein the multiple candidate date data includes at least two meteorological factor data and photovoltaic power data; Based on meteorological factor data and photovoltaic power data, the meteorological factor data with the highest correlation to photovoltaic power was determined using the Pearson correlation coefficient method. Wavelet packet decomposition is performed on the meteorological factor data with the highest correlation. Based on the wavelet packet decomposition results and the remaining meteorological factor data, multiple similar days are selected from several candidate days, including: The meteorological factor data with the highest correlation is subjected to a single wavelet packet decomposition to obtain multiple frequency signals; The multiple frequency signals are individually reconstructed into fluctuation components and trend components; Construct daily feature vectors for the predicted date and each candidate date based on the remaining meteorological data; Calculate the Pearson correlation coefficients of the predicted date and each candidate date on the fluctuation component, trend component, and daily eigenvector to obtain the fluctuation factor, trend factor, and comprehensive weather factor for each candidate date; The similarity between the predicted day and each candidate day is calculated using the trend factor, fluctuation factor, and comprehensive weather factor. Select multiple similar days from several candidate days based on their similarity. Wavelet packet decomposition was performed on the photovoltaic power data of all similar days to obtain multiple power subsequences. Multiple similar day datasets were constructed by combining the multiple power subsequences with meteorological factor data of similar days. A gated recurrent unit network is trained using all the aforementioned similar day datasets, and the photovoltaic power of the predicted day is predicted by the trained gated recurrent unit network.

2. The photovoltaic power prediction method according to claim 1, characterized in that... The calculation of the similarity between the predicted day and each candidate day using the trend factor, fluctuation factor, and comprehensive weather factor includes: Obtain a preset beetle whisker search mathematical model, and use the trend factor, fluctuation factor, comprehensive weather factor and the preset beetle whisker search mathematical model to optimize and obtain the weights of the trend factor, fluctuation factor and comprehensive weather factor. The similarity between the predicted day and the candidate day is obtained by weighting the trend factor, fluctuation factor and comprehensive weather factor using the corresponding weights.

3. The photovoltaic power prediction method according to claim 1, characterized in that... The step of training a gated recurrent unit network using all the similar day datasets, and then predicting the photovoltaic power for the predicted day using the trained gated recurrent unit network, includes: A gated recurrent unit network is constructed for each similar day dataset. The gated recurrent unit network includes an input layer, a preset number of hidden layers, and an output layer. Determine the hyperparameters to be optimized in the gated recurrent unit network, including batch size, random deactivation rate of hidden layers, and number of neurons; The hyperparameters of the network to be optimized are substituted into the preset beetle whisker search mathematical model for iterative optimization to obtain the optimized network hyperparameters. The optimized network hyperparameters are substituted into the corresponding gated recurrent unit network and trained using the corresponding similar daily dataset; After summing all the output values ​​and performing inverse normalization, the photovoltaic power for the predicted day is obtained. The output value is the result of a gated recurrent unit network after training.

4. A photovoltaic power prediction device, characterized in that, include: Acquisition module: used to acquire data for the predicted date and multiple candidate dates, wherein the multiple candidate date data includes at least two meteorological factor data and photovoltaic power data; Determination module: Used to determine the meteorological factor with the highest correlation to photovoltaic power based on meteorological factor data and photovoltaic power data, using the Pearson correlation coefficient method; Selection module: Used to perform wavelet packet decomposition on the meteorological factor data with the highest correlation, and select multiple similar days from several candidate days based on the wavelet packet decomposition results and the remaining meteorological factor data, including: First decomposition unit: used to perform wavelet packet decomposition on the meteorological factor data with the highest correlation to obtain multiple frequency signals; First reconstruction unit: used to reconstruct the plurality of frequency signals into fluctuation components and trend components individually; The first building unit is used to construct the daily feature vectors for the prediction day and each candidate day based on the remaining meteorological factor data. The first calculation unit is used to calculate the Pearson correlation coefficient of the prediction day and each candidate day on the fluctuation component, trend component, and daily eigenvector, and to obtain the fluctuation factor, trend factor, and comprehensive weather factor for each candidate day. The second calculation unit is used to calculate the similarity between the predicted day and each candidate day using the trend factor, fluctuation factor and comprehensive weather factor. Selection unit: used to select multiple similar days from several candidate days based on similarity; Dataset construction module: used to perform wavelet packet decomposition on photovoltaic power data of all similar days to obtain multiple power subsequences, and to construct multiple similar day datasets from several power subsequences and meteorological factor data of similar days; Prediction module: Used to train a gated recurrent unit network using all the similar day datasets, and the trained gated recurrent unit network is used to predict the photovoltaic power of the predicted day.

5. The photovoltaic power prediction device according to claim 4, characterized in that, The second computing unit includes: The first optimization unit is used to obtain a preset beetle whisker search mathematical model, and to optimize the trend factor, fluctuation factor and comprehensive weather factor by using the trend factor, fluctuation factor, comprehensive weather factor and beetle whisker search mathematical model. The third calculation unit is used to calculate the similarity between the predicted day and the candidate day by using the weights corresponding to the trend factor, fluctuation factor and comprehensive weather factor.

6. The photovoltaic power prediction device according to claim 5, characterized in that, The prediction module includes: The second building unit is used to build a gated recurrent unit network for each similar day dataset. The gated recurrent unit network includes an input layer, a preset number of hidden layers, and an output layer. Determining Unit: Used to determine the hyperparameters of the gated recurrent unit network to be optimized, including batch size, random deactivation rate of hidden layers, and number of neurons; The second optimization unit is used to substitute the hyperparameters of the network to be optimized into a preset beetle whisker search mathematical model for iterative optimization, so as to obtain the optimized network hyperparameters. Training unit: used to input the optimized network hyperparameters into the corresponding gated recurrent unit network and train it using the corresponding similar daily dataset; Prediction Unit: Used to sum all output values ​​and then perform inverse normalization to obtain the photovoltaic power on the predicted day. The output value is the result of a gated recurrent unit network after training.

7. A photovoltaic power prediction device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the photovoltaic power prediction method as described in any one of claims 1 to 3 when executing the computer program.

8. A readable storage medium, characterized in that: The readable storage medium stores a computer program that, when executed by a processor, implements the steps of the photovoltaic power prediction method as described in any one of claims 1 to 3.

Images