A Photovoltaic Power Combination Prediction Method, Device, Equipment and Medium

Abstract

The present invention belongs to the technical field of photovoltaic power generation, and specifically relates to a photovoltaic power combination prediction method, device, equipment and medium. The photovoltaic power combination prediction method of the present invention uses a WOA-Attention-BiLSTM prediction model based on the selection of similar moments to fully exploit time series information, and the combination of the K-means method and the KELM-PSO model more effectively exerts the spatial correlation of photovoltaic power. In addition, a variable weight combination prediction method based on GIOWA is used to more deeply understand and more carefully learn the model, and finally the photovoltaic power is predicted more accurately.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic power generation, and particularly relates to a photovoltaic power combination prediction method, device, equipment and medium. Background Art

[0002] With the development of the photovoltaic industry, higher requirements are put forward for the technical level of photovoltaic power generation. Due to the characteristics of intermittency and volatility of photovoltaic power generation, it has a certain impact on the frequency of the power grid. Therefore, improving the accuracy of photovoltaic power prediction will contribute to the safe and stable operation of the power grid and increase the revenue of photovoltaic power stations.

[0003] Currently, the mainstream prediction methods for photovoltaic output power are divided into physical methods and statistical methods. The physical method constructs a power prediction model based on physical equations such as the solar irradiance transfer equation and the photovoltaic module operation equation. Therefore, this method does not require a large amount of historical data and is suitable for newly built photovoltaic power stations. However, the model usually has poor robustness and is difficult to simulate some extreme abnormal weather conditions, the slow changes of the environment and photovoltaic module parameters over time. Statistical methods mainly include neural network algorithms, classification regression algorithms, time series algorithms, etc. Among them, the research based on the similar day theory is a major hotspot. This method usually establishes a characteristic index system, divides data samples, and obtains similar day samples. However, there is a situation where the daily average error range between the similar day and the day to be predicted is not large, meeting the overall similarity, but the climate conditions in each time period have large differences, resulting in the inability to guarantee the prediction accuracy. Therefore, it is necessary to conduct further refined research on the basis of screening out the similar day data set, so as to accurately measure the similar time period scale and ensure the prediction accuracy under special weather conditions. At the same time, due to the lack of extraction of spatial features, the model accuracy needs to be further improved.

[0004] Each photovoltaic power station has a unique geographical location, and the photovoltaic power generation has spatial correlation. By extracting the information of neighboring stations, the spatial correlation of the reference photovoltaic power station can be effectively utilized to improve the prediction accuracy on the basis of single-station prediction. Currently, the mainstream technical route of this method is to classify photovoltaic power stations by using methods such as clustering and correlation analysis, and then build a model to complete the spatial correlation prediction of the target photovoltaic power station.

[0005] Combined prediction combines various single prediction methods, can combine the advantages of various prediction methods, and complement the error characteristics. Usually, the prediction accuracy will be improved. One of the research directions is the combined prediction method based on weight components. Research shows that the prediction accuracy of photovoltaic output power can be further improved by selecting appropriate sub-models and weighting methods for combined prediction. However, the combined prediction methods proposed currently are relatively complex to implement and have low engineering utilization rate. Therefore, it is necessary to further study more applicable and simple methods. Summary of the Invention

[0006] The object of the present invention is to provide a photovoltaic power combined prediction method, device, equipment and medium, so as to solve the problems that the combined prediction method in the prior art is relatively complex to implement and has a low engineering utilization rate.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect, a photovoltaic power combined prediction method includes the following specific steps:

[0009] Determine a sample sequence for photovoltaic prediction, and select similar moments at the moment to be predicted on the day to be predicted from the sample sequence;

[0010] Input the generated power at the similar moments and numerical weather prediction data into a preset WOA-Attention-BiLSTM model, and the WOA-Attention-BiLSTM model outputs a first photovoltaic power prediction value of the target photovoltaic power station;

[0011] Determine the reference power station of the target photovoltaic power station, calculate the photovoltaic power prediction value of the reference power station, optimize the kernel parameters of the KELM model according to the predicted power of the reference photovoltaic power station and the real power data of the target photovoltaic power station, and substitute the optimal kernel parameters into the KELM model to obtain a KELM-PSO model;

[0012] Use the KELM-PSO model to predict the power of the target photovoltaic power station, and obtain a second photovoltaic power prediction value of the target photovoltaic power station;

[0013] Based on the variable weight combined prediction method of generalized induced weighted average with the minimum sum of squared errors as the optimization criterion, fuse the first photovoltaic power prediction value and the second photovoltaic power prediction value to obtain the final prediction value of the photovoltaic power.

[0014] Further, the step of determining a sample sequence for photovoltaic prediction and selecting similar moments at the moment to be predicted on the day to be predicted from the sample sequence includes:

[0015] According to the data of the day to be predicted, select a sample sequence of the same season and weather type from the pre-divided data set;

[0016] Based on the temperature characteristics, use the grey relational degree to calculate the relational degree R between the day to be predicted and the sample sequence i , and select several days with the top-ranked relational degree R i as the similar day sample set M1;

[0017] According to the moment t to be predicted on the day to be predicted, select several moment data in the similar day sample set M1 as the similar moment sample set M2;

[0018] Based on the temperature characteristics, calculate the temperature T at time t of the day to be predicted. t Calculate the Pearson correlation coefficient r between the temperature characteristics of the data at the similar time samples in set M2, and select several moments with the top r rankings as the similar times for the time to be predicted on the day to be predicted.

[0019] Further, in the step of selecting the sample sequences of the same season and weather type from the pre-divided data set, the division method of the data set is as follows:

[0020] Obtain the original photovoltaic data samples and perform data processing;

[0021] Classify the original photovoltaic data samples after data processing according to the season type;

[0022] Further classify the data of different season types according to the weather type to obtain data sets of different weather types in different seasons.

[0023] Further, the weather types include sunny, cloudy, overcast, and rainy, and the weather types are divided according to the interval of the average daily irradiance level.

[0024] Further, the step of determining the reference power station of the target photovoltaic power station specifically includes:

[0025] Use the historical output power data of the power stations as feature variables and perform cluster analysis by calculating the Euclidean distance between different power stations;

[0026] Classify the power stations with highly correlated output power characteristics, so as to select the reference photovoltaic power stations of the target photovoltaic power station.

[0027] In a second aspect, a photovoltaic power combined prediction device includes:

[0028] A similar time selection module, configured to determine a sample sequence for photovoltaic prediction and select a similar time for the time to be predicted on the day to be predicted from the sample sequence;

[0029] A first prediction module, configured to input the generated power at the similar time and the numerical weather forecast data into a preset WOA-Attention-BiLSTM model, and the WOA-Attention-BiLSTM model outputs a first photovoltaic power prediction value of the target photovoltaic power station;

[0030] A model optimization module, which is used to determine a reference power station for the target photovoltaic power station, calculate the predicted photovoltaic power generation value of the reference power station, optimize the kernel parameters of the KELM model using the PSO algorithm according to the predicted power of the reference photovoltaic power station and the real power data of the target photovoltaic power station, and substitute the optimal kernel parameters into the KELM model to obtain the KELM-PSO model;

[0031] A second prediction module, which is used to predict the power of the target photovoltaic power station using the KELM-PSO model to obtain the second predicted photovoltaic power generation value of the target photovoltaic power station;

[0032] A fusion prediction module, which is used to fuse the first predicted photovoltaic power generation value and the second predicted photovoltaic power generation value based on the variable weight combination prediction method of the generalized induced weighted average with the minimum sum of squared errors as the optimization criterion to obtain the final predicted value of the photovoltaic power.

[0033] Specifically, the similar moment selection module is specifically used for: according to the data of the day to be predicted, selecting a sample sequence of the same season and weather type from the pre-divided dataset; based on the temperature feature, calculating the correlation degree R between the day to be predicted and the sample sequence using the grey correlation degree i and selecting several days with the top-ranked correlation degree R i as the similar day sample set M1; according to the moment t to be predicted of the day to be predicted, selecting the data of several moments in the similar day sample set M1 as the similar moment sample set M2; based on the temperature feature, calculating the Pearson correlation coefficient r between the temperature T at the moment t of the day to be predicted and the temperature features of the data in the similar moment sample set M2, and selecting several moments with the top-ranked r as the similar moments of the moment t to be predicted of the day to be predicted. t

[0034] Specifically, in the model optimization module, the method for determining the reference power station of the target photovoltaic power station is as follows:

[0035] Using the historical output power data of the power station as a feature variable, performing cluster analysis by calculating the Euclidean distance between different power stations; classifying the power stations with highly correlated output power characteristics, so as to select the reference photovoltaic power station of the target photovoltaic power station.

[0036] In a third aspect, an electronic device includes a processor and a memory, and the processor is used to execute a computer program stored in the memory to implement the above-mentioned photovoltaic power combination prediction method.

[0037] In a fourth aspect, a computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the above-mentioned photovoltaic power combination prediction method is implemented.

[0038] ​Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] 1) The present invention uses the WOA-Attention-BiLSTM prediction model based on the selection of similar moments to fully exploit the time series information, combines the K-means method and the KELM-PSO model to more effectively utilize the spatial correlation of photovoltaic power, and uses the variable weight combination prediction method based on GIOWA to more deeply understand and more carefully learn the model, and finally accurately predicts the photovoltaic power.

[0040] 2) On the basis of selecting similar days using the gray correlation degree, the present invention further selects similar moments, improves the prediction accuracy of the model from the perspective of similar reference, and uses the WOA-Attention-BiLSTM model to predict the photovoltaic power generation, so as to fully utilize the time series information and ensure the prediction accuracy under special weather conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings forming a part of this application are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0042] Figure 1 is a flowchart of a photovoltaic power combination prediction method according to an embodiment of the present invention;

[0043] Figure 2 is a structural block diagram of a photovoltaic power combination prediction device according to an embodiment of the present invention;

[0044] Figure 3 is a structural block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other.

[0046] The following detailed descriptions are all exemplary descriptions, aiming to provide further details of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention.

[0047] Embodiment 1

[0048] Abbreviation and definition of key terms:

[0049] NWP: Numerical Weather Prediction;

[0050] WOA: Whale Optimization Algorithm;

[0051] BiLSTM: Bidirectional Long Short-Term Memory;

[0052] Attention: Attention mechanism;

[0053] KELM: Kernel Extreme Learning Machine;

[0054] PSO: Particle Swarm Optimization Algorithm;

[0055] GIOWA: Generalized Induced Ordered Weighted Averaging;

[0056] As Figure 1 shown, the present invention proposes a photovoltaic power combination prediction method, including the following specific steps:

[0057] Step1: First, perform data processing on the original photovoltaic data samples; then, according to the classification standard of the national meteorological bureau, divide the data according to four seasons; finally, divide the generalized weather types into 4 categories: sunny, cloudy, overcast, and rainy.

[0058] Specifically, calculate the daily irradiance average level F according to Equation (1) average , and divide the data according to different weather types based on the classification indicators listed in Table 1.

[0059]

[0060] In the formula: n is the length of the power generation time; F i is the per-unit value of the solar irradiance at the i-th moment.

[0061] Table 1 Weather types and corresponding model parameters

[0062]

[0063] Step2: Complete the selection of similar times for the day to be predicted.

[0064] First, based on the data set divided in Step1, according to the data of the day to be predicted, select a sample sequence of the same season and weather type for it to form an initial sample;

[0065] Secondly, based on the temperature characteristics, use the grey relational degree to calculate the relational degree R i between the day to be predicted and the sample sequence, as shown in Equation (3). And select the top 10 days with the ranking of R i as the similar day sample set M1;

[0066] Then, according to the time t to be predicted of the day to be predicted, select the data of three moments (t-1, t, t+1) in M1 as the similar time sample set M2;

[0067] Finally, based on the temperature characteristics, according to Equation (4), calculate the temperature T at time t of the day to be predicted. t Calculate the Pearson correlation coefficient r between the temperature characteristics of the data in M2 and select the top 3 moments with the highest r rankings as the similar moments for the time to be predicted on the day to be predicted.

[0068]

[0069] In the formula: t'0(k) is the k-th component of the temperature of the day to be predicted after normalization; t i '(k) is the k-th temperature component of the i-th day of the sample after normalization; ρ is a constant; ε i (k) is the correlation coefficient.

[0070]

[0071] In the formula: D is the total number of temperature characteristic components.

[0072]

[0073] In the formula: T m is the temperature characteristic data in M2.

[0074] Step3: Input the power generation power at the similar moments of the time to be predicted on the day to be predicted, as well as the NWP data such as the weather type, season identifier, temperature, and wind speed of the day to be predicted into the WOA-Attention-BiLSTM model, and the predicted value of the photovoltaic power generation can be obtained through calculation.

[0075] Among them, the Attention mechanism in the Attention-BiLSTM model is used to analyze the correlation between each input feature and the photovoltaic power generation and dynamically adjust the weights, thereby changing the contribution rate of different features to the output; the WOA algorithm performs intelligent optimization on important parameters such as the batch size, learning rate, number of hidden layers, and number of neurons in each layer of the BiLSTM model to obtain the optimal parameters; BiLSTM can better capture bidirectional dependency information, thereby obtaining better prediction results.

[0076] Step4: Use the K-means clustering algorithm to obtain the reference photovoltaic power station of the target photovoltaic power station.

[0077] First, use the historical output power data of the power station as the feature variables, perform clustering analysis by calculating the Euclidean distance between different power stations, and its feature vector is shown in Equation (5). Classify the power stations with highly correlated output power characteristics, and thus select the reference photovoltaic power station of the target photovoltaic power station.

[0078] c i ={p1,p2,…p j} (5)

[0079] where: c i is the eigenvector; p j is the actual output power of the photovoltaic at a certain moment.

[0080] Step5: First, according to the content described in Step1~Step3, obtain the predicted power of the reference photovoltaic power station;

[0081] Then, according to the predicted power of the reference photovoltaic power station and the true power data of the target photovoltaic power station, use the PSO algorithm to optimize the kernel parameters of the KELM model, substitute the optimal kernel parameters into KELM to obtain the KELM-PSO model, and according to Equation (6), use this model to perform deterministic prediction of the power of the target photovoltaic power station;

[0082]

[0083] where: w is the predicted power of the reference photovoltaic power station; k is the kernel function; N is the number of historical data; T is the predicted target value vector; I is the identity matrix; C is the kernel parameter; Ω EML is the kernel matrix.

[0084] Step6: Taking the minimum sum of squared errors as the optimization criterion, according to Equation (7), based on the variable weight combination prediction method of the generalized induced weighted average (GIOWA), effectively fuse the photovoltaic power prediction results obtained by the WOA-Attention-BiLSTM model and the KELM-PSO model to obtain the final predicted value of the photovoltaic power, so as to achieve the effective fusion of spatio-temporal information.

[0085]

[0086] where: Q is the sum of squared errors of the combined prediction model; A is the weight vector of the single prediction method in the combined prediction method; G is the induced error information matrix of the GIOWA operator; n is the number of data within the time scale; R n =(1, 1, …, 1) T .

[0087] In summary, the photovoltaic power combination prediction method provided by the present invention fully considers time series information, proposes a WOA-Attention-BiLSTM photovoltaic power prediction model based on the selection of similar moments, realizes further precise scaling of the similar time period on the basis of screening out the similar-day data set, so as to ensure the prediction accuracy under special weather conditions; effectively utilizes the spatial correlation of photovoltaic power, realizes the clustering of power stations based on the K-means method, and gives full play to the characteristics of strong generalization ability and stability of the KELM-PSO model to establish the mapping relationship between the reference photovoltaic power station and the target photovoltaic power station, and then completes the prediction; fuses spatio-temporal information, and based on the variable weight combination prediction method of GIOWA, effectively combines the above two prediction models to obtain the output power of the photovoltaic.

[0088] On the basis of selecting similar days by using the grey relational degree, the present invention further selects similar moments, improves the prediction accuracy of the model from the perspective of similar reference, and uses the WOA-Attention-BiLSTM model to predict the photovoltaic power generation power, so as to make full use of time series information and ensure the prediction accuracy under special weather conditions.

[0089] Embodiment 2

[0090] Based on the same inventive concept as the above embodiment, the present solution also provides a photovoltaic power combination prediction device, including:

[0091] A similar moment selection module, configured to determine a sample sequence for photovoltaic prediction, and select a similar moment of the moment to be predicted on the day to be predicted from the sample sequence;

[0092] A first prediction module, configured to input the power generation power at the similar moment and the numerical weather forecast data into a preset WOA-Attention-BiLSTM model, and the WOA-Attention-BiLSTM model outputs a first photovoltaic power generation power prediction value of the target photovoltaic power station;

[0093] A model optimization module, configured to determine a reference power station of the target photovoltaic power station, calculate a predicted value of the photovoltaic power generation power of the reference power station, optimize the kernel parameters of the KELM model by using the PSO algorithm according to the predicted power of the reference photovoltaic power station and the real power data of the target photovoltaic power station, and substitute the optimal kernel parameters into the KELM model to obtain a KELM-PSO model;

[0094] A second prediction module, configured to predict the power of the target photovoltaic power station by using the KELM-PSO model to obtain a second photovoltaic power generation power prediction value of the target photovoltaic power station;

[0095] The fusion prediction module is used to fuse the first photovoltaic power prediction value and the second photovoltaic power prediction value based on the variable weight combination prediction method of generalized induced weighted average with the minimum sum of squared errors as the optimization criterion, so as to obtain the final prediction value of the photovoltaic power.

[0096] Specifically, the similar moment selection module is specifically used for: according to the data of the day to be predicted, selecting the sample sequences of the same season and weather type from the pre-divided data set; based on the temperature characteristics, calculating the correlation degree R between the day to be predicted and the sample sequences by using the grey relational degree. i , and selecting several days with the top-ranked correlation degree R i as the similar day sample set M1; according to the moment t to be predicted of the day to be predicted, selecting the data of several moments in the similar day sample set M1 as the similar moment sample set M2; based on the temperature characteristics, calculating the temperature T at the moment t of the day to be predicted. t and the Pearson correlation coefficient r of the temperature characteristics of the data in the similar moment sample set M2, and selecting several moments with the top-ranked r as the similar moments of the moment to be predicted of the day to be predicted.

[0097] Specifically, in the model optimization module, the method for determining the reference power station of the target photovoltaic power station is as follows:

[0098] Using the historical output power data of the power station as the characteristic variables, performing clustering analysis by calculating the Euclidean distance between different power stations; classifying the power stations with highly correlated output power characteristics, so as to select the reference photovoltaic power station of the target photovoltaic power station.

[0099] Embodiment 3

[0100] The present invention also provides an electronic device 100 for implementing the photovoltaic power combination prediction method; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104. The memory 101 can be used to store the computer program 103, and the processor 102 realizes the steps of a photovoltaic power combination prediction method in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0101] The memory 101 may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created according to the use of the electronic device 100 (such as audio data, etc.). In addition, the memory 101 may include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0102] At least one processor 102 may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or the processor 102 may also be any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and connects various parts of the entire electronic device 100 through various interfaces and lines.

[0103] The memory 101 in the electronic device 100 stores multiple instructions to implement a photovoltaic power combination prediction method. The processor 102 can execute the multiple instructions to implement:

[0104] Determine a sample sequence for photovoltaic prediction, and select a similar moment of the moment to be predicted on the day to be predicted from the sample sequence;

[0105] Input the generated power at the similar moment and the numerical weather forecast data into a preset WOA-Attention-BiLSTM model, and the WOA-Attention-BiLSTM model outputs a first photovoltaic power prediction value of the target photovoltaic power station;

[0106] Determine the reference power station of the target photovoltaic power station, calculate the predicted value of the photovoltaic power generation of the reference power station, and use the PSO algorithm to optimize the kernel parameters of the KELM model according to the predicted power of the reference photovoltaic power station and the real power data of the target photovoltaic power station, and substitute the optimal kernel parameters into the KELM model to obtain the KELM-PSO model;

[0107] Use the KELM-PSO model to predict the power of the target photovoltaic power station, and obtain the second predicted photovoltaic power value of the target photovoltaic power station;

[0108] Taking the minimum sum of squared errors as the optimization criterion, based on the variable weight combination prediction method of generalized induced weighted average, fuse the first predicted photovoltaic power value and the second predicted photovoltaic power value to obtain the final predicted value of photovoltaic power.

[0109] Embodiment 4

[0110] If the modules / units integrated in the electronic device 100 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above embodiment methods of the present invention, it can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory and read-only memory (ROM, Read-Only Memory).

[0111] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0112] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0113] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one process Figure 1 or a plurality of processes and / or blocks Figure 1 or a plurality of blocks.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process Figure 1 or a plurality of processes and / or blocks Figure 1 or a plurality of blocks.

[0115] In the description of this specification, the descriptions referring to the terms "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection 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 can be combined in any one or more embodiments or examples in a suitable manner.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent replacements can still be made to the specific embodiments of the present invention, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.

Claims

1. A photovoltaic power combination prediction method, characterized in that, Including specific steps: Determine a sample sequence for photovoltaic prediction, and select similar moments at the moment to be predicted on the day to be predicted from the sample sequence; Input the generated power at the similar moment and numerical weather prediction data into a preset WOA-Attention-BiLSTM model, and the WOA-Attention-BiLSTM model outputs a first photovoltaic power prediction value of the target photovoltaic power station; Determine the reference power station of the target photovoltaic power station, calculate the predicted photovoltaic power value of the reference power station, optimize the kernel parameters of the KELM model according to the predicted power of the reference power station and the real power data of the target photovoltaic power station, and substitute the optimal kernel parameters into the KELM model to obtain the KELM-PSO model; Use the KELM-PSO model to predict the power of the target photovoltaic power station, and obtain a second photovoltaic power prediction value of the target photovoltaic power station; Based on the optimization criterion of the minimum sum of squared errors, and based on the variable weight combination prediction method of the generalized induced weighted average, fuse the first photovoltaic power prediction value and the second photovoltaic power prediction value to obtain the final prediction value of the photovoltaic power; The step of determining the sample sequence for photovoltaic prediction and selecting similar moments at the moment to be predicted on the day to be predicted from the sample sequence includes: According to the data of the day to be predicted, select a sample sequence of the same season and weather type from the pre-divided dataset; Based on the temperature characteristics, calculate the correlation degree between the day to be predicted and the sample sequence using the grey correlation degree R i , and select the correlation degree R i of the top-ranked several days as the similar-day sample set M 1; According to the moment to be predicted on the day to be predicted t , select several moment data from the similar day sample set M 1 as the similar moment sample set M 2; Based on the temperature characteristics, calculate the temperature at the t time of the day to be predicted T t and the Pearson correlation coefficient of the temperature characteristics of the data in the similar time sample set M 2, and select r a number of moments with the top rankings as the similar times of the time to be predicted on the day to be predicted. r ​ 2. The photovoltaic power combination prediction method according to claim 1, wherein In the step of selecting a sample sequence of the same season and weather type from the pre-divided dataset, the division method of the dataset is as follows: Obtain the original photovoltaic data sample and perform data processing; Classify the original photovoltaic data sample after data processing according to the season; Further classify the data of different season types according to the weather type to obtain datasets of different weather types in different seasons.

3. The photovoltaic power combination prediction method according to claim 2, wherein, The weather types include sunny, cloudy, overcast, and rainy, and the weather types are classified according to the interval of the average daily irradiance level.

4. The photovoltaic power combination prediction method according to claim 1, characterized in that The step of determining the reference power station of the target photovoltaic power station specifically includes: Use the historical output power data of the power station as a feature variable, and perform clustering analysis by calculating the Euclidean distance between different power stations; Classify the power stations with highly correlated output power characteristics, so as to select the reference photovoltaic power station of the target photovoltaic power station.

5. A photovoltaic power combination prediction device, characterized in that, Including: A similar moment selection module, used to determine a sample sequence for photovoltaic prediction, and select similar moments at the moment to be predicted on the day to be predicted from the sample sequence; A first prediction module, used to input the generated power at the similar moment and numerical weather prediction data into a preset WOA-Attention-BiLSTM model, and the WOA-Attention-BiLSTM model outputs a first photovoltaic power prediction value of the target photovoltaic power station; A model optimization module, which is used to determine a reference power station for the target photovoltaic power station, calculate a predicted value of the photovoltaic power generation of the reference power station, optimize the kernel parameters of the KELM model according to the predicted power of the reference power station and the actual power data of the target photovoltaic power station by using the PSO algorithm, and substitute the optimal kernel parameters into the KELM model to obtain a KELM-PSO model; A second prediction module, which is used to predict the power of the target photovoltaic power station by using the KELM-PSO model to obtain a second predicted value of the photovoltaic power generation of the target photovoltaic power station; A fusion prediction module, which is used to fuse the first predicted value of the photovoltaic power generation and the second predicted value of the photovoltaic power generation based on a variable weight combination prediction method of generalized induced weighted average with the minimum sum of squared errors as the optimization criterion to obtain a final predicted value of the photovoltaic power; The similar time selection module is specifically used for: According to the data of the day to be predicted, selecting a sample sequence of the same season and weather type from a pre-divided data set; Based on temperature characteristics, calculate the correlation degree between the day to be predicted and the sample sequence using grey relational degree R i , and select several days with the R i highest-ranked correlation degrees as the similar-day sample set M 1; According to the moment to be predicted on the day to be predicted t , select several moment data from the similar day sample set M 1 as the similar moment sample set M 2; Based on the temperature characteristics, calculate the temperature at the time of the day to be predicted t at the moment T t and the Pearson correlation coefficient of the temperature characteristics of the data in the similar moment sample set M in 2 r , and select r several moments with the top rankings as the similar moments of the moments to be predicted on the day to be predicted 6. The photovoltaic power combination prediction device according to claim 5, characterized in that, In the model optimization module, the method for determining the reference power station of the target photovoltaic power station is: Using the historical output power data of the power station as a feature variable, and performing clustering analysis by calculating the Euclidean distance between different power stations; Classifying the power stations with highly correlated output power characteristics, so as to select a reference photovoltaic power station for the target photovoltaic power station.

7. An electronic device, characterized in that, It includes a processor and a memory. The processor is used to execute a computer program stored in the memory to implement the photovoltaic power combination prediction method according to any one of claims 1 to 4.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by the processor, it implements the photovoltaic power combination prediction method according to any one of claims 1 to 4.

Images