Photovoltaic ultra-short-term power interval prediction method and device based on improved GRU

Abstract

The invention relates to the technical field of photovoltaic power generation power prediction, in particular to a photovoltaic ultra-short-term power interval prediction method and device based on an improved GRU, and the method comprises the steps: carrying out the correlation analysis according to meteorological data, obtaining the meteorological data after the correlation analysis, carrying out the clustering according to the meteorological data after the correlation analysis, and obtaining the photovoltaic ultra-short-term power interval prediction. Clustering the meteorological data to obtain clustered meteorological data, wherein the clustered meteorological data comprises meteorological similar days; constructing a training set and a test set according to the clustered meteorological data and photovoltaic power generation power data; performing training improvement on a pre-constructed GRU model according to the training set to obtain an improved GRU model, and performing photovoltaic ultra-short-term power interval prediction according to the test set and the improved GRU model to obtain a prediction result; and evaluating the prediction result according to a pre-constructed evaluation system to obtain a prediction evaluation result. According to the invention, the prediction accuracy of the photovoltaic power generation power and the reasonability of prediction result evaluation can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation prediction, and in particular to a photovoltaic ultra-short-term power interval prediction method and device based on an improved GRU. Background Art

[0002] With the continuous development of global renewable energy, photovoltaic power generation, as a clean and low-carbon form of power generation, has received attention from countries around the world. However, photovoltaic power generation has the characteristics of randomness and volatility, and is greatly affected by meteorological changes, which has a negative impact on the safe and stable operation of the power system. Accurate photovoltaic power generation power forecasting can provide reliable information support for the planning, construction and operation and dispatch of power systems with large-scale photovoltaic access, and has gradually become an important research content in the development of power systems.

[0003] Photovoltaic ultra-short-term prediction refers to the prediction of photovoltaic power generation in the next few minutes, and has high requirements for prediction accuracy. At present, with the rise of a new generation of artificial intelligence technology, artificial intelligence methods based on machine learning and deep learning have gradually become the mainstream research methods for photovoltaic ultra-short-term power prediction. Neural network methods represented by long short-term memory networks (LSTM) and their variants, gated recurrent units (GRU), have been widely used and have shown strong practicality in the field of ultra-short-term prediction. However, traditional prediction methods are mostly point predictions, and their prediction results are single curves, which contain less data, have strong uncertainty in the results, and are prone to prediction errors. Interval prediction can better solve this problem and can give the possible interval of photovoltaic power generation at the next moment. The prediction results contain more detailed and complete information and have higher accuracy.

[0004] At present, there have been some studies on photovoltaic power generation range prediction, but these methods all have room for improvement. For example, the distributed photovoltaic cluster output interval prediction method disclosed in the application publication number CN 118709827 A does not adequately mine the information of the input data; the short-term photovoltaic power probability interval prediction method disclosed in the application publication number CN 117239732 A, the EEMD-LSTM microgrid photovoltaic short-term power interval prediction method disclosed in the patent with the application publication number CN 115222138 A, and the short-term photovoltaic power integration prediction method based on dual signal decomposition and dual network channels disclosed in the application publication number CN 118035728A use different methods to decompose the input signal to capture feature information, but these studies do not distinguish between meteorologically similar days, which may lead to insufficient prediction accuracy; the quantile regression-based genetic support vector machine photovoltaic power interval prediction method disclosed in the patent with the application publication number CN 108921339 A, the photovoltaic power interval probability prediction method based on deep learning fusion model disclosed in the patent with the application publication number CN115115125 A, and the photovoltaic power interval prediction method based on EEMD-LSTM microgrid photovoltaic short-term power interval prediction method disclosed in the patent with the application publication number CN The photovoltaic power interval prediction method based on multivariate data feature enhancement disclosed in patent 117688362A distinguishes meteorological types, but the solution parameters of its prediction network are still difficult to determine. In addition, none of the above methods proposes a unified photovoltaic power interval prediction accuracy evaluation method, lacking a more reasonable and accurate evaluation index system. Summary of the invention

[0005] In response to the demand for accurate prediction of photovoltaic power generation and the problem of imperfect prediction result evaluation methods, the present invention provides a photovoltaic ultra-short-term power interval prediction method and device based on improved GRU, which can improve the prediction accuracy of photovoltaic power generation and the rationality of prediction result evaluation.

[0006] To achieve the above object, the technical solution adopted by the present invention includes the following contents:

[0007] In a first aspect, the present invention provides a photovoltaic ultra-short-term power interval prediction method, comprising:

[0008] Obtain meteorological data and photovoltaic power generation data;

[0009] Performing correlation analysis on the meteorological data to obtain meteorological data after correlation analysis, and clustering the meteorological data after correlation analysis to obtain clustered meteorological data, wherein the clustered meteorological data includes meteorologically similar days;

[0010] Constructing a training set and a test set according to the clustered meteorological data and the photovoltaic power generation data;

[0011] The pre-built GRU model is trained and improved according to the training set to obtain an improved GRU model, and the photovoltaic ultra-short-term power interval prediction is performed according to the test set and the improved GRU model to obtain a prediction result;

[0012] The prediction result is evaluated according to a pre-constructed evaluation system to obtain a prediction evaluation result.

[0013] In a second aspect, the present invention provides a photovoltaic ultra-short-term power interval prediction device, characterized in that it includes:

[0014] A data acquisition module, which is used to acquire meteorological data and photovoltaic power generation data;

[0015] An analysis and clustering module, which is used to perform a correlation analysis on the meteorological data to obtain meteorological data after the correlation analysis, and to perform clustering on the meteorological data after the correlation analysis to obtain clustered meteorological data;

[0016] A data set construction module, which is used to construct a training set and a test set based on the clustered meteorological data and the photovoltaic power generation data;

[0017] An interval prediction module, which is used to train and improve the pre-built GRU model according to the training set to obtain an improved GRU model, and perform photovoltaic ultra-short-term power interval prediction according to the test set and the improved GRU model to obtain a prediction result;

[0018] The evaluation result module is used to evaluate the prediction result according to the pre-built evaluation system to obtain the prediction evaluation result.

[0019] In a third aspect, the present invention further provides an electronic device, including a processor and a memory;

[0020] The memory is used to store programs;

[0021] The processor executes the program to implement the method described above.

[0022] In a fourth aspect, the present invention further provides a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the method described above.

[0023] In a fifth aspect, the present invention further provides a computer program product or a computer program, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the above method.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] (1) Different from the traditional photovoltaic power point prediction method that has been widely studied, the prediction interval is composed of multi-group quantile outputs, which realizes interval prediction. The results contain more information and are more reliable than point prediction. It can effectively avoid the errors caused by small amount of data, making the interval prediction results and the point prediction results obtained by the interval average more accurate.

[0026] (2) Using multiple methods such as SSA, VMD and CNN to improve the single GRU prediction can better extract the input data features and adaptively determine the prediction parameters.

[0027] (3) A new comprehensive evaluation method for interval prediction results is proposed, which can achieve comprehensive quantification of the quality of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0029] Figure 1 A flowchart provided for Example 1 of the present invention;

[0030] Figure 2 This is a schematic diagram of the GRU network structure provided in Example 1 of the present invention;

[0031] Figure 3 This is a flow chart of improved GRU model interval prediction in Example 1 of the present invention;

[0032] Figure 4 The interval prediction results of the SSA-VMD-CNN-GRU model at different confidence levels for a test set of sunny similar days in Example 2 of the present invention;

[0033] Figure 5 This is the interval prediction result of the SSA-VMD-CNN-GRU model at different confidence levels for a test set of similar days on cloudy and rainy days in Example 2 of the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0035] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations of the embodiments of the present invention are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0036] Embodiment 1:

[0037] See also Figures 1 to 5 The present invention provides a photovoltaic ultra-short-term power interval prediction method, which may specifically include the following steps:

[0038] Step 100: Acquire meteorological data and photovoltaic power generation data.

[0039] In this step, meteorological data includes but is not limited to solar irradiance, air temperature, relative humidity, air pressure, photovoltaic module temperature, etc.

[0040] Step 200: performing correlation analysis on the meteorological data to obtain meteorological data after correlation analysis, performing clustering on the meteorological data after correlation analysis to obtain clustered meteorological data, wherein the clustered meteorological data includes meteorologically similar days.

[0041] In this step, correlation analysis is performed on different meteorological data (such as solar irradiance, temperature, relative humidity, etc.) to clarify the correlation between them. Then, based on these correlation analysis results, a clustering algorithm (such as K-means) is used to cluster the meteorological data to distinguish meteorologically similar days. Through the steps provided in this embodiment, meteorological data can be divided into different similar day types (such as sunny days, cloudy days, rainy days, etc.), so that more accurate photovoltaic power generation power prediction can be performed for different types of similar days in the future.

[0042] Step 300: constructing a training set and a test set based on the clustered meteorological data and the photovoltaic power generation data.

[0043] In this step, after distinguishing the days with similar weather, a variety of data including photovoltaic power generation data and meteorological data are used to construct a training set for training the improved GRU model and a test set for verifying the accuracy of the prediction model. Through the steps provided in this embodiment, the training set is used to train the improved GRU model so that it can learn the complex relationship between photovoltaic power generation and meteorological data; the test set is used to evaluate the prediction performance of the model.

[0044] Step 400: training and improving the pre-built GRU model according to the training set to obtain an improved GRU model, and performing photovoltaic ultra-short-term power interval prediction according to the test set and the improved GRU model to obtain a prediction result.

[0045] In this step, the training set is modally decomposed according to the variational mode decomposition algorithm to obtain the training set after mode decomposition, wherein the variational mode decomposition (VMD) is used to perform mode decomposition on the training set to mine its information features, and the signal to be decomposed can be photovoltaic power data or meteorological data. The photovoltaic power data and meteorological data after mining the information features are used as input signals of the subsequent GRU model; then, the convolutional neural network and the pre-built GRU model are coupled to obtain the coupled GRU model, wherein the convolutional neural network (CNN) is used to capture the input signal features more comprehensively; finally, the hyperparameters of the coupled GRU model are optimized according to the sparrow search algorithm to obtain the improved GRU model after optimization, wherein the sparrow search algorithm (SSA) is used to optimize the hyperparameters of the GRU model, and the hyperparameters include but are not limited to the number of GRU hidden units, the maximum number of algorithm iteration cycles, the initial learning rate and the learning rate reduction factor.

[0046] The mathematical expression of the GRU model is:

[0047]

[0048] In the formula, x t ,y t are the input and output vectors respectively; h t-1 and h t They are the state memory variables of the previous moment and the current moment respectively; is the state of the current candidate set; r t is the state of the reset gate; t To update the state of the door; W r , W z , and W oThey are reset gate, update gate, candidate set, output vector and x respectively. t and h t-1 The weight parameter of the connection matrix multiplied by each other; σ represents the Sigmoid activation function

[0049] Through the steps provided in this embodiment, the sparrow search algorithm (SSA) is used to optimize the hyperparameters of the GRU, including the number of hidden units, the maximum number of cycles of algorithm iterations, the initial learning rate, and the learning rate reduction factor. In addition, the variational mode decomposition (VMD) is introduced to perform modal decomposition on the input signal to mine its information features, and the convolutional neural network (CNN) is used to more comprehensively capture the input signal features, so that the improved GRU model can better meet the needs of photovoltaic power prediction and improve the prediction accuracy and stability.

[0050] In this step, the test set is modally decomposed according to the variational mode decomposition algorithm to obtain the test set after mode decomposition, wherein the variational mode decomposition (VMD) is used to perform mode decomposition on the test set to mine its information features, and the signal to be decomposed can be photovoltaic power data or meteorological data. The photovoltaic power data and meteorological data after mining the information features are used as input signals of the subsequent improved GRU model; then, based on the improved GRU model and the test set after mode decomposition, the quantile prediction method is used to predict the photovoltaic ultra-short-term power interval, and the prediction interval composed of multiple groups of quantile outputs is obtained;

[0051] Among them, the quantile Q is determined according to the following preset formula:

[0052]

[0053] In the formula, Q L is the lower limit of the quantile, Q U is the upper limit of the quantile, and CL is the confidence level, which represents the width of the confidence interval. In order to ensure that the results are as accurate as possible, the difference between the quantiles of two adjacent groups should be as small as possible. This can effectively avoid the error caused by the small amount of data included in the interval, making the interval prediction results more accurate.

[0054] Here is a further explanation of the calculation method of the quantile Q: After confirming the upper and lower limits of the quantile, the experimenter can set the quantile step size s according to actual needs, then Q 1 =Q L , Q 2 =Q 1 +s,Q 3 =Q 2 +s, ..., until Q UIt should be noted that the smaller the step size, the more accurate the prediction, but the longer it takes. For example, when the confidence level is 90% and the step size is 0.05, the quantiles are 0.05, 0.1, 0.15, ..., 0.9, 0.95, and the prediction interval is defined by these quantile prediction results.

[0055] Similarly, the experimenter can also take only the quantile Q value required for the actual experiment, but this may lead to an increase in prediction errors. For example, when the confidence level is 90%, only the 0.05 quantile and the 0.95 quantile are taken, and the prediction interval falls between the two groups of quantile prediction results.

[0056] Through the steps provided in this embodiment, traditional photovoltaic power prediction methods are mostly point predictions, that is, predicting a single power value at a certain moment in the future. However, this method has the problem of strong uncertainty in the prediction results and prone to errors. This embodiment adopts an interval prediction method, which provides more detailed and complete information by predicting the possible power interval at a certain moment in the future. The method of this embodiment can not only avoid the errors caused by the small amount of data, but also make the prediction results more accurate and reliable.

[0057] Step 500: Evaluate the prediction result according to the pre-constructed evaluation system to obtain a prediction evaluation result.

[0058] In this step, the prediction intervals composed of multiple sets of quantile outputs are evaluated according to the interval coverage:

[0059] The value of interval coverage C is determined according to the following preset formula:

[0060] C=CWC 95 +CWC 90 +CWC 85

[0061] Where, CWC 95 、CWC 90 and CWC 85 These are the interval coverage values ​​at 95%, 90% and 85% confidence levels respectively. The smaller the interval coverage C value, the better the interval prediction result.

[0062] Here is a supplementary explanation of the calculation method of CWC:

[0063] For interval prediction, the evaluation of the prediction result quality is mainly considered from two aspects: the coverage rate of the interval to the real signal (PICP) and the width of the interval (PINAW), which are defined as follows:

[0064]

[0065] Where n is the number of samples; c is iis a Boolean variable whose value is determined by the true value y i Decision; L i , U i are the lower and upper bounds of the prediction interval respectively; y max and min are the maximum and minimum true values ​​respectively.

[0066] PICP reflects the ratio of the true value falling within the prediction interval; PINAW defines the narrowness of the prediction interval. There is a certain game relationship between these two values: high coverage means that the interval needs to be wide enough; while a narrow interval may cause the coverage to decrease, or even be less than the confidence level. Generally speaking, when predicting, as long as PICP is not lower than the confidence level, the result is acceptable; under the premise of meeting the confidence level, the smaller the PINAW value, the better. Therefore, the interval coverage (CWC) is used to balance the relationship between the two:

[0067]

[0068] In the formula, η is the penalty parameter, which is set to 50 here.

[0069] This parameter comprehensively considers the interval coverage and width. The smaller its value, the better the prediction effect. When PICP is not lower than the confidence level, its value is equal to PINAW; once PICP is greater than the confidence level, its value will increase significantly. Considering the interval prediction of different methods at 95%, 90% and 85% confidence levels (95%, 90% and 85% are commonly used confidence values ​​in interval prediction), the comprehensive evaluation index C of the interval prediction result is defined according to the method described in this embodiment.

[0070] Through the steps provided in this embodiment, the two factors of interval coverage and width are comprehensively considered, and the accuracy of the prediction results can be improved by calculating indicators such as interval coverage (CWC) to evaluate the quality of the prediction results.

[0071] Embodiment 2:

[0072] The photovoltaic power generation record data of a photovoltaic power station in Ningxia is used as an example to verify the implementation example. The data content includes photovoltaic power generation power data and historical meteorological data. The recorded data is from January 1, 2019 to June 30, 2019, a total of 181 days; the sampling time interval is 15 minutes, and there are 96 sampling points per day. In addition, since photovoltaic power generation does not occur at night, for the sake of simplicity of calculation, only the data from 5:00 to 20:00 is retained, with 61 sampling points per day.

[0073] According to the results of correlation analysis, solar radiation data is used as the standard to distinguish meteorological similar days. First, the permutation entropy value of daily solar radiation data is calculated, and then the K-means clustering algorithm is used to distinguish 181 groups of data into sunny similar days and cloudy and rainy similar days. Among them, there are 102 sunny similar days and 79 cloudy and rainy similar days. The last day of the two groups of similar days is used as the test set for photovoltaic short-term power prediction, where the last day of the sunny similar day is June 18, and the last day of the cloudy and rainy similar day is June 30; the other data is used as the training set to train the neural network.

[0074] For two types of similar days, the interval prediction results of three single models, LSTM, Bi-LSTM, and GRU, and three improved models, VMD-GRU, VMD-CNN-GRU, and SSA-VMD-CNN-GRU, are compared. Tables 1 and 2 below respectively give the C value comparisons of six different prediction models under different meteorological similar days.

[0075] A comprehensive comparison of the two tables shows that: on similar sunny days, the C value is generally smaller; on the contrary, on similar rainy days, the C value is larger. This is because the photovoltaic power generation power is relatively smoother and easier to predict on sunny days. In addition, the SSA-VMD-CNN-GRU method proposed in the present invention has the highest prediction accuracy. On similar sunny days, its C value is 54.3% of the single GRU method; on similar rainy days, its C value is 37.4% of the single GRU method. Compared with several other methods, the method proposed in the present invention also has different degrees of performance improvement.

[0076] Table 1 C values ​​of different prediction methods (similar sunny days)

[0077]

[0078] Table 2 C values ​​of different prediction methods (similar days of cloudy and rainy days)

[0079]

[0080] Figure 4 , Figure 5 The interval prediction results of the SSA-VMD-CNN-GRU model under two similar days at different confidence levels are given respectively. It can be seen that as the confidence level decreases, the prediction interval gradually narrows, but the model always maintains a good interval coverage.

[0081] Embodiment 3:

[0082] Based on the same inventive concept, an embodiment of the present invention further provides a photovoltaic ultra-short-term power interval prediction device, which is characterized by comprising:

[0083] A data acquisition module, which is used to acquire meteorological data and photovoltaic power generation data;

[0084] An analysis and clustering module, which is used to perform a correlation analysis on the meteorological data to obtain meteorological data after the correlation analysis, and to perform clustering on the meteorological data after the correlation analysis to obtain clustered meteorological data;

[0085] A data set construction module, which is used to construct a training set and a test set based on the clustered meteorological data and the photovoltaic power generation data;

[0086] An interval prediction module, which is used to train and improve the pre-built GRU model according to the training set to obtain an improved GRU model, and perform photovoltaic ultra-short-term power interval prediction according to the test set and the improved GRU model to obtain a prediction result;

[0087] The evaluation result module is used to evaluate the prediction result according to the pre-built evaluation system to obtain the prediction evaluation result.

[0088] Since the device is a device corresponding to the photovoltaic ultra-short-term power interval prediction method of the embodiment of the present invention, and the principle of solving the problem by the device is similar to that of the method, the implementation of the device can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0089] Embodiment 4:

[0090] Based on the same inventive concept, an embodiment of the present invention also provides an electronic device, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the photovoltaic ultra-short-term power interval prediction method based on the improved GRU as described above.

[0091] It is understandable that the memory may include a random access memory (RAM) or a read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory may be used to store instructions, programs, codes, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function, instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data created according to the use of the server, etc.

[0092] The processor may include one or more processing cores. The processor uses various interfaces and lines to connect the various parts of the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory, and calling data stored in the memory. Optionally, the processor can be implemented in at least one hardware form of digital signal processing (DSP), field programmable gate array (FPGA), and programmable logic array (PLA). The processor can integrate one or a combination of a central processing unit (CPU) and a modem. Among them, the CPU mainly processes the operating system and application programs; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor, but may be implemented separately through a chip.

[0093] Since the electronic device is the electronic device corresponding to the photovoltaic ultra-short-term power interval prediction method based on improved GRU in the embodiment of the present invention, and the principle of solving the problem by the electronic device is similar to that of the method, the implementation of the electronic device can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0094] Example 5

[0095] Based on the same inventive concept, an embodiment of the present invention also provides a computer-readable storage medium, in which at least one instruction, at least one program, a code set or an instruction set is stored. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the photovoltaic ultra-short-term power interval prediction method based on the improved GRU as described above.

[0096] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, the storage medium including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0097] Since the storage medium is the storage medium corresponding to the photovoltaic ultra-short-term power interval prediction method based on improved GRU in the embodiment of the present invention, and the principle of solving the problem by the storage medium is similar to that of the method, the implementation of the storage medium can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0098] Embodiment 6:

[0099] In some possible implementations, various aspects of the method of the embodiment of the present invention may also be implemented in the form of a program product, which includes a program code. When the program product is run on a computer device, the program code is used to enable the computer device to execute the steps of the photovoltaic ultra-short-term power interval prediction method based on the improved GRU according to various exemplary embodiments of the present application described above in this specification. Among them, the executable computer program code or "code" for executing each embodiment can be written in a high-level programming language such as C, C++, C#, Smalltalk, Java, JavaScript, Visual Basic, structured query language (e.g., Transact-SQL), Perl, or in various other programming languages.

[0100] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0101] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0102] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable ordinary technicians in the field to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made based on the essence of the content of the present invention should be included in the protection scope of the present invention.

Claims

1. A photovoltaic ultra-short-term power interval prediction method, characterized in that: include: Obtain meteorological data and photovoltaic power generation data; Performing correlation analysis on the meteorological data to obtain meteorological data after correlation analysis, and clustering the meteorological data after correlation analysis to obtain clustered meteorological data, wherein the clustered meteorological data includes meteorologically similar days; Constructing a training set and a test set according to the clustered meteorological data and the photovoltaic power generation data; The pre-built GRU model is trained and improved according to the training set to obtain an improved GRU model, and the photovoltaic ultra-short-term power interval prediction is performed according to the test set and the improved GRU model to obtain a prediction result; The prediction result is evaluated according to a pre-constructed evaluation system to obtain a prediction evaluation result.

2. The photovoltaic ultra-short-term power interval prediction method according to claim 1 is characterized in that: The step of training and improving the pre-built GRU model according to the training set to obtain an improved GRU model specifically includes: Performing modal decomposition on the training set according to the variational modal decomposition algorithm to obtain a modally decomposed training set; According to the convolutional neural network and the pre-built GRU model, a coupled GRU model is obtained; The hyperparameters of the coupled GRU model are optimized according to the sparrow search algorithm to obtain an improved GRU model after optimization.

3. The photovoltaic ultra-short-term power interval prediction method according to claim 2 is characterized in that: The photovoltaic ultra-short-term power interval prediction is performed according to the test set and the improved GRU model to obtain the prediction results, which specifically include: Perform modal decomposition on the test set according to the variational modal decomposition algorithm to obtain a modally decomposed test set; Based on the improved GRU model and the test set after modal decomposition, the quantile prediction method is used to predict the ultra-short-term power interval of photovoltaic power, and the prediction interval composed of multiple groups of quantile outputs is obtained; Among them, the quantile Q is determined according to the following preset formula: In the formula, Q L is the lower limit of the quantile, Q U is the upper limit of the quantile, and CL is the confidence level, which represents the width of the confidence interval.

4. The photovoltaic ultra-short-term power interval prediction method according to claim 3 is characterized in that: The step of evaluating the prediction result according to the pre-constructed evaluation system to obtain the prediction evaluation result specifically includes: The prediction intervals composed of multiple sets of quantile outputs are evaluated based on interval coverage: The value of interval coverage C is determined according to the following preset formula: C=CWC 95 + CWC 90 + CWC 85 ; Where, CWC 95 、CWC 90 and CWC 85 These are the interval coverage values ​​at 95%, 90% and 85% confidence levels respectively. The smaller the interval coverage C value, the better the interval prediction result.

5. A photovoltaic ultra-short-term power interval prediction device, characterized in that: include: A data acquisition module, which is used to acquire meteorological data and photovoltaic power generation data; An analysis and clustering module, which is used to perform a correlation analysis on the meteorological data to obtain meteorological data after the correlation analysis, and to perform clustering on the meteorological data after the correlation analysis to obtain clustered meteorological data, wherein the clustered meteorological data includes meteorologically similar days; A data set construction module, which is used to construct a training set and a test set based on the clustered meteorological data and the photovoltaic power generation data; An interval prediction module, which is used to train and improve the pre-built GRU model according to the training set to obtain an improved GRU model, and perform photovoltaic ultra-short-term power interval prediction according to the test set and the improved GRU model to obtain a prediction result; The evaluation result module is used to evaluate the prediction result according to the pre-built evaluation system to obtain the prediction evaluation result.

6. The photovoltaic ultra-short-term power interval prediction device according to claim 5, characterized in that: The step of training and improving the pre-built GRU model according to the training set to obtain an improved GRU model specifically includes: Performing modal decomposition on the training set according to the variational modal decomposition algorithm to obtain a modally decomposed training set; According to the convolutional neural network and the pre-built GRU model, a coupled GRU model is obtained; The hyperparameters of the coupled GRU model are optimized according to the sparrow search algorithm to obtain an improved GRU model after optimization.

7. The photovoltaic ultra-short-term power interval prediction device according to claim 6, characterized in that: The photovoltaic ultra-short-term power interval prediction is performed according to the test set and the improved GRU model to obtain the prediction results, which specifically include: Perform modal decomposition on the test set according to the variational modal decomposition algorithm to obtain a modally decomposed test set; Based on the improved GRU model and the test set after modal decomposition, the quantile prediction method is used to predict the ultra-short-term power interval of photovoltaic power, and the prediction interval composed of multiple groups of quantile outputs is obtained; Among them, the quantile Q is determined according to the following preset formula: In the formula, Q L is the lower limit of the quantile, Q U is the upper limit of the quantile, and CL is the confidence level, which represents the width of the confidence interval.

8. The photovoltaic ultra-short-term power interval prediction device according to claim 7, characterized in that: The step of evaluating the prediction result according to the pre-constructed evaluation system to obtain the prediction evaluation result specifically includes: The prediction intervals composed of multiple sets of quantile outputs are evaluated based on interval coverage: The value of interval coverage C is determined according to the following preset formula: C=CWC 95 + CWC 90 + CWC 85 ; Where, CWC 95 、CWC 90 and CWC 85 These are the interval coverage values ​​at 95%, 90% and 85% confidence levels respectively. The smaller the interval coverage C value, the better the interval prediction result.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the photovoltaic ultra-short-term power interval prediction method based on the improved GRU as described in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by the processor to implement the photovoltaic ultra-short-term power interval prediction method based on the improved GRU as described in any one of claims 1 to 4.

Images