Distributed photovoltaic cluster power prediction method and device and electronic equipment

Abstract

The invention provides a distributed photovoltaic cluster power prediction method and device and electronic equipment, and belongs to the field of photovoltaic technology, and the method comprises the steps: obtaining the total power generation power of all distributed photovoltaic stations in a distributed photovoltaic cluster; inputting the total power generation power into a power generation power prediction model to obtain a future power generation power prediction result output by the power generation power prediction model; wherein the generated power prediction model is obtained based on training of sample modal components of a plurality of distributed photovoltaic clusters; the sample mode component is obtained by historical generated power of all distributed photovoltaic stations in the distributed photovoltaic cluster based on multivariate variational mode decomposition; the power generation power prediction model performs power prediction based on space-time correlation characteristics among sample mode components of a plurality of distributed photovoltaic clusters. The method is used for overcoming the defects that an existing distributed photovoltaic cluster power prediction model is not high in applicability and the improvement of model prediction precision is limited.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic technology, and in particular to a distributed photovoltaic cluster power prediction method, a distributed photovoltaic cluster power prediction device, an electronic device, a machine-readable storage medium and a computer program product. Background Art

[0002] With the continuous increase in the penetration rate of distributed photovoltaics, its volatility and randomness have become uncontrollable factors in the coordinated dispatch of power grids. Accurate short-term power forecasting of distributed photovoltaics is of great significance to optimizing power system dispatching and ensuring the safe operation of power grids.

[0003] The existing distributed photovoltaic cluster power prediction method divides the clusters according to the geographical location of the photovoltaic power station, and constructs a photovoltaic output prediction model by analyzing the mapping relationship between meteorological data and photovoltaic power generation. However, such a photovoltaic output prediction model is overly dependent on the accuracy of meteorological data collection and prediction. However, the construction of meteorological station equipment in distributed photovoltaic power stations is uneven, resulting in the low applicability of the existing distributed photovoltaic cluster power prediction model and limiting the improvement of the model prediction accuracy. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a distributed photovoltaic cluster power prediction method, a distributed photovoltaic cluster power prediction device, an electronic device, a machine-readable storage medium and a computer program product to solve the defects of the existing distributed photovoltaic cluster power prediction model, which has low applicability and limits the improvement of model prediction accuracy.

[0005] In order to achieve the above object, an embodiment of the present invention provides a distributed photovoltaic cluster power prediction method, comprising:

[0006] Obtain the total power generation of all distributed photovoltaic stations in the distributed photovoltaic cluster;

[0007] Inputting the total power generation into a power generation prediction model to obtain a future power generation prediction result output by the power generation prediction model;

[0008] Among them, the power generation prediction model is trained based on sample modal components of multiple distributed photovoltaic clusters; the sample modal components are the historical power generation of all distributed photovoltaic stations in the distributed photovoltaic cluster based on multivariate variational modal decomposition; the power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between sample modal components of multiple distributed photovoltaic clusters.

[0009] Optionally, the power generation prediction model is trained based on the following steps:

[0010] Obtain historical power generation and historical photovoltaic output data of multiple distributed photovoltaic sites;

[0011] Performing correlation analysis based on the historical photovoltaic output data of all distributed photovoltaic stations to obtain a correlation coefficient between each distributed photovoltaic station;

[0012] Based on the magnitude relationship of the correlation coefficients between each of the distributed photovoltaic stations, the distributed photovoltaic stations are divided into clusters to obtain a plurality of distributed photovoltaic clusters; each of the distributed photovoltaic clusters includes at least one distributed photovoltaic station;

[0013] Perform multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain the sample modal components of each distributed photovoltaic cluster;

[0014] The power generation prediction model is trained based on the sample modal components of all distributed photovoltaic clusters to obtain the trained power generation prediction model.

[0015] Optionally, the correlation coefficient is a Pearson correlation coefficient, which is expressed by the following formula:

[0016]

[0017] Among them, r represents the Pearson correlation coefficient, x i is the i-th data in the first historical photovoltaic output data curve, y i is the i-th data in the second historical photovoltaic output data curve, is the average value in the first historical photovoltaic output data curve, It is the average value in the second historical photovoltaic output data curve, the first historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the first distributed photovoltaic site, and the second historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the second distributed photovoltaic site.

[0018] Optionally, performing multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain sample modal components of each distributed photovoltaic cluster includes:

[0019] Constructing an objective function with constraints based on the historical power generation of all distributed photovoltaic stations in the target cluster; the objective function represents a set of multivariate intrinsic mode functions extracted from the historical power generation of all distributed photovoltaic stations;

[0020] Constructing an augmented Lagrangian function, and removing constraints in the objective function based on the augmented Lagrangian function;

[0021] Repeat the following steps until a preset condition is met: solve the objective function after removing the constraint condition by an alternating direction multiplier method; the solution of the objective function that meets the preset condition is used as the sample modal component of the target cluster;

[0022] The target cluster is any distributed photovoltaic cluster among all distributed photovoltaic clusters.

[0023] Optionally, the power generation prediction model includes:

[0024] A spatiotemporal correlation feature extraction layer is used to extract spatiotemporal correlation features between sample modal components of multiple distributed photovoltaic clusters from sample modal components of multiple distributed photovoltaic clusters;

[0025] The prediction layer is used to perform power prediction based on the spatiotemporal correlation features.

[0026] Optionally, the spatiotemporal correlation feature extraction layer is constructed based on an encoder-decoder of a converter model, and the prediction layer is constructed based on a long short-term memory network.

[0027] On the other hand, an embodiment of the present invention further provides a distributed photovoltaic cluster power prediction device, comprising:

[0028] An acquisition module is used to obtain the total power generation of all distributed photovoltaic sites in a distributed photovoltaic cluster;

[0029] A prediction module, used for inputting the total power generation into a power generation prediction model to obtain a future power generation prediction result output by the power generation prediction model;

[0030] Among them, the power generation prediction model is trained based on sample modal components of multiple distributed photovoltaic clusters; the sample modal components are the historical power generation of all distributed photovoltaic stations in the distributed photovoltaic cluster based on multivariate variational modal decomposition; the power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between sample modal components of multiple distributed photovoltaic clusters.

[0031] Optionally, the power generation prediction model is trained based on the following steps:

[0032] Obtain historical power generation and historical photovoltaic output data of multiple distributed photovoltaic sites;

[0033] Performing correlation analysis based on the historical photovoltaic output data of all distributed photovoltaic stations to obtain a correlation coefficient between each distributed photovoltaic station;

[0034] Based on the magnitude relationship of the correlation coefficients between each of the distributed photovoltaic stations, the distributed photovoltaic stations are divided into clusters to obtain a plurality of distributed photovoltaic clusters; each of the distributed photovoltaic clusters includes at least one distributed photovoltaic station;

[0035] Perform multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain the sample modal components of each distributed photovoltaic cluster;

[0036] The power generation prediction model is trained based on the sample modal components of all distributed photovoltaic clusters to obtain the trained power generation prediction model.

[0037] Optionally, the correlation coefficient is a Pearson correlation coefficient, which is expressed by the following formula:

[0038]

[0039] Among them, r represents the Pearson correlation coefficient, x i is the i-th data in the first historical photovoltaic output data curve, y i is the i-th data in the second historical photovoltaic output data curve, is the average value in the first historical photovoltaic output data curve, It is the average value in the second historical photovoltaic output data curve, the first historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the first distributed photovoltaic site, and the second historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the second distributed photovoltaic site.

[0040] Optionally, performing multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain sample modal components of each distributed photovoltaic cluster includes:

[0041] Constructing an objective function with constraints based on the historical power generation of all distributed photovoltaic stations in the target cluster; the objective function represents a set of multivariate intrinsic mode functions extracted from the historical power generation of all distributed photovoltaic stations;

[0042] Constructing an augmented Lagrangian function, and removing constraints in the objective function based on the augmented Lagrangian function;

[0043] Repeat the following steps until a preset condition is met: solve the objective function after removing the constraint condition by an alternating direction multiplier method; the solution of the objective function that meets the preset condition is used as the sample modal component of the target cluster;

[0044] The target cluster is any distributed photovoltaic cluster among all distributed photovoltaic clusters.

[0045] Optionally, the power generation prediction model includes:

[0046] A spatiotemporal correlation feature extraction layer is used to extract spatiotemporal correlation features between sample modal components of multiple distributed photovoltaic clusters from sample modal components of multiple distributed photovoltaic clusters;

[0047] The prediction layer is used to perform power prediction based on the spatiotemporal correlation features.

[0048] Optionally, the spatiotemporal correlation feature extraction layer is constructed based on an encoder-decoder of a converter model, and the prediction layer is constructed based on a long short-term memory network.

[0049] On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned distributed photovoltaic cluster power prediction method when executing the program.

[0050] On the other hand, the present invention further provides a machine-readable storage medium having a computer program stored thereon, and the computer program, when executed by a processor, implements the above-mentioned distributed photovoltaic cluster power prediction method.

[0051] On the other hand, the present invention further provides a computer program product, including a computer program, which implements the above-mentioned distributed photovoltaic cluster power prediction method when executed by a processor.

[0052] Through the above technical solution, the embodiment of the present invention can complete the training of the power generation prediction model by only using the historical power generation data of the photovoltaic station, and there is no need to install additional meteorological measurement equipment for the distributed photovoltaic station, which saves economic costs and improves the applicability of the model; at the same time, the power generation prediction model is obtained by training based on the sample modal components of multiple distributed photovoltaic clusters, and the sample modal components reflect the spatiotemporal correlation characteristics of each station in the cluster. The power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between the sample modal components of multiple distributed photovoltaic clusters. Therefore, when training the power generation prediction model of the embodiment of the present invention, attention is paid to the spatiotemporal correlation characteristics of each distributed photovoltaic station between clusters and within the cluster, which effectively improves the prediction accuracy of the model.

[0053] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0055] Figure 1 It is a schematic diagram of the flow of the distributed photovoltaic cluster power prediction method provided by the present invention;

[0056] Figure 2 This is one of the flow charts of training the power generation prediction model provided by the present invention;

[0057] Figure 3 This is the second flow chart of training the power generation prediction model provided by the present invention;

[0058] Figure 4 It is a structural schematic diagram of the power generation prediction model provided by the present invention;

[0059] Figure 5 It is a structural schematic diagram of a distributed photovoltaic cluster power prediction device provided by the present invention;

[0060] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0061] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0062] Method Embodiment

[0063] Please refer to Figure 1 The embodiment of the present invention provides a distributed photovoltaic cluster power prediction method, comprising:

[0064] Step 100: Obtain the total power generation of all distributed photovoltaic sites in the distributed photovoltaic cluster.

[0065] The distributed photovoltaic cluster includes multiple distributed photovoltaic stations. The electronic device obtains the total power generation power accumulated from all distributed photovoltaic stations in the distributed photovoltaic cluster to predict the future power generation power of the distributed photovoltaic cluster.

[0066] Step 200: input the total power generation into a power generation prediction model to obtain a future power generation prediction result output by the power generation prediction model.

[0067] The electronic device inputs the total generated power into a generated power prediction model to obtain a future generated power prediction result output by the generated power prediction model.

[0068] The power generation prediction model is obtained by training based on sample modal components of multiple distributed photovoltaic clusters; the sample modal components are obtained by decomposing the historical power generation of all distributed photovoltaic stations in the distributed photovoltaic cluster based on multivariate variational modal decomposition. The historical power generation of each distributed photovoltaic station includes the power generation in the past period of time. For example, the historical power generation of each distributed photovoltaic station can be the power generation recorded every 15 minutes in the past month.

[0069] The embodiment of the present invention finds that the output data of each distributed photovoltaic station in the same geographical and climatic environment presents similar changes, that is, the historical power generation of each distributed photovoltaic station will have a certain correlation with the future power generation of the neighboring station, that is, the time-space correlation. Although the prior art takes into account the meteorological information and the geographical location information of the photovoltaic power station, it does not effectively extract the time-space correlation characteristics of the historical power generation of the distributed photovoltaic station, which limits the improvement of the model prediction accuracy. In order to realize the time-frequency synchronization analysis of the historical power generation of multiple distributed photovoltaic stations in the same cluster, and better extract the time-space correlation characteristics of the historical power generation data of multiple distributed photovoltaic stations in the cluster, the embodiment of the present invention adopts the multivariate variational mode decomposition (Multivariate Variational Mode Decomposition, MVMD) method to extract the characteristics of the historical power. And in order to pay more attention to the time-space correlation characteristics of each distributed photovoltaic station between clusters, the power generation prediction model of the embodiment of the present invention performs power prediction based on the time-space correlation characteristics between the sample modal components of multiple distributed photovoltaic clusters.

[0070] The embodiment of the present invention can complete the training of the power generation prediction model by only using the historical power generation data of the photovoltaic station, and does not need to install additional meteorological measurement equipment for the distributed photovoltaic station, saving economic costs and improving the applicability of the model; at the same time, the power generation prediction model is obtained by training based on the sample modal components of multiple distributed photovoltaic clusters, and the sample modal components reflect the spatiotemporal correlation characteristics of each station in the cluster. The power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between the sample modal components of multiple distributed photovoltaic clusters. Therefore, when training the power generation prediction model of the embodiment of the present invention, attention is paid to the spatiotemporal correlation characteristics of each distributed photovoltaic station between clusters and within the cluster, effectively improving the prediction accuracy of the model.

[0071] For other aspects of the embodiments of the present invention, please refer to Figure 2 and Figure 3 The power generation prediction model is trained based on the following steps:

[0072] Step 10: Obtain historical power generation and historical photovoltaic output data of multiple distributed photovoltaic sites.

[0073] The electronic device obtains the historical power generation and historical photovoltaic output data of multiple distributed photovoltaic sites. The historical photovoltaic output data refers to the photovoltaic output data of distributed photovoltaic sites in the past period of time. The photovoltaic output data may include instantaneous power and daily power generation.

[0074] Step 20: performing correlation analysis based on the historical photovoltaic output data of all distributed photovoltaic stations to obtain the correlation coefficient between each distributed photovoltaic station.

[0075] Step 30: cluster the distributed photovoltaic stations based on the magnitude relationship of the correlation coefficients between each of the distributed photovoltaic stations to obtain a plurality of distributed photovoltaic clusters; each of the distributed photovoltaic clusters includes at least one distributed photovoltaic station.

[0076] Since the power generation prediction of a single distributed photovoltaic station not only has low accuracy and little impact on grid planning and scheduling, but also the models between different distributed photovoltaic stations have poor adaptability and cannot be shared, regional prediction of a cluster composed of multiple distributed photovoltaic power stations can effectively improve accuracy and applicability. Therefore, each distributed photovoltaic station in the photovoltaic system is divided into clusters according to the correlation of the output data. For example, an embodiment of the present invention takes the photovoltaic output data (such as daily power generation) of each distributed photovoltaic station at intervals of 15 minutes in the past month to form a historical photovoltaic output data curve. The historical photovoltaic output data curve is then used to perform correlation analysis to obtain the correlation coefficient of the historical photovoltaic output data curve between each distributed photovoltaic station. For example, the correlation coefficient can be obtained by calculating the Pearson correlation coefficient or the Spearman correlation coefficient. When the correlation coefficient is the Pearson correlation coefficient, the Pearson correlation coefficient is expressed by the following formula:

[0077]

[0078] Among them, r represents the Pearson correlation coefficient, x i is the i-th data in the first historical photovoltaic output data curve, y i is the i-th data in the second historical photovoltaic output data curve, is the average value in the first historical photovoltaic output data curve, It is the average value in the second historical photovoltaic output data curve, the first historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the first distributed photovoltaic site, and the second historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the second distributed photovoltaic site.

[0079] The embodiment of the present invention can divide the distributed photovoltaic stations with a correlation coefficient greater than 0.95 into a cluster, thereby completing the cluster division of the photovoltaic system. The high correlation coefficient of historical output data between different distributed photovoltaic stations in the same cluster indicates that the geographical environment and meteorological conditions where the distributed photovoltaic stations are installed are similar.

[0080] Step 40: Perform multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain sample modal components of each distributed photovoltaic cluster.

[0081] The embodiment of the present invention uses multivariate variational modal decomposition to realize the time-frequency synchronization analysis of the historical power generation data of multiple distributed photovoltaic stations in the same cluster, decomposes it into multiple modal components, and uses the modal components to characterize the spatiotemporal correlation characteristics between distributed photovoltaic stations. The sample modal components of each cluster can be obtained through multivariate variational modal decomposition.

[0082] Step 50: training the power generation prediction model based on the sample modal components of all distributed photovoltaic clusters to obtain the trained power generation prediction model.

[0083] The electronic device trains the power generation prediction model based on the sample modal components of all distributed photovoltaic clusters to obtain the trained power generation prediction model. The power generation prediction model of the embodiment of the present invention performs power prediction based on the spatiotemporal correlation characteristics between the sample modal components of multiple distributed photovoltaic clusters. Among them, the method for training the power generation prediction model through sample modal components can select various existing model training methods, such as various unsupervised training methods, which will not be described in detail here.

[0084] The embodiment of the present invention uses the Pearson correlation coefficient to divide the distributed photovoltaic stations with high correlation in output data into the same distributed photovoltaic cluster; then uses the multivariate variational modal decomposition method to decompose the historical power generation sequence of each distributed photovoltaic station in the same distributed photovoltaic cluster into multiple modal components with the same center frequency, and uses each modal component to characterize the spatiotemporal correlation characteristics between each distributed photovoltaic station; finally, the modal components decomposed from the historical power generation of each distributed photovoltaic station are used to train the power generation prediction model, and after the power generation prediction model is constructed, the future short-term power generation of each of the multiple distributed photovoltaic clusters is predicted.

[0085] In other aspects of the embodiments of the present invention, step 40, performing multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain sample modal components of each distributed photovoltaic cluster, includes:

[0086] Step 41 constructs an objective function with constraints based on the historical power generation of all distributed photovoltaic stations in the target cluster; the objective function represents a set of multivariate intrinsic mode functions extracted from the historical power generation of all distributed photovoltaic stations.

[0087] The embodiment of the present invention adopts a multivariate variational mode decomposition method to extract features from the historical power generation of all distributed photovoltaic stations. As a new multivariate extension of variational mode decomposition (VMD), multivariate variational mode decomposition can be used to process multivariate data containing any number of data channels. The main goal of the multivariate variational mode decomposition in the embodiment of the present invention is to extract a predefined K number of multivariate intrinsic mode function sets from the multivariate original signal x(t) containing C data channels (i.e., the historical power generation of all distributed photovoltaic stations in the cluster).

[0088] The embodiment of the present invention constructs an objective function with constraints based on the historical power generation of all distributed photovoltaic stations in the target cluster. The objective function can be expressed by the following formula:

[0089]

[0090] Among them, {u k,c}——A collection of multivariate intrinsic mode functions; {ω k}——{u k,c} center frequency set; u k,c ——the kth intrinsic mode function of the cth channel; The center frequency of ——Partial derivative operations with respect to time; ——Use the Hilbert transform operator to transform the vector signal u k,c The analytical representation of (t); x c (t)——Input data of the cth channel, the historical power generation of all distributed photovoltaic stations in the cth cluster.

[0091] Step 42: construct an augmented Lagrangian function, and remove the constraints in the objective function based on the augmented Lagrangian function.

[0092] Among them, the augmented Lagrangian function contains two penalty terms, the quadratic term To ensure the reconstruction accuracy, the Lagrangian multiplier term c=1Cλct,xct-k=1Kuk,ct is used to ensure that the constraints are strictly met. The formula for removing the constraints in the objective function based on the augmented Lagrangian function is as follows:

[0093]

[0094] Among them, λ c (t) represents the Lagrange multiplier, and α represents the quadratic penalty factor, which is used to ensure the accuracy of signal reconstruction under the influence of Gaussian noise.

[0095] Step 43, repeatedly perform the following steps until the preset conditions are met: solve the objective function after removing the constraints by the alternating direction multiplier method; the solution result of the objective function that meets the preset conditions is used as the sample modal component of the target cluster; wherein the target cluster is any one of all distributed photovoltaic clusters.

[0096] The embodiment of the present invention solves the formula in step 42 by using the Alternating Direction Method of Multipliers (ADMM). The advantage of the Alternating Direction Method of Multipliers is that the entire complex optimization problem can be converted into multiple sub-optimization problems, thereby reducing the difficulty of solving. The update equation of is shown in formula (4), the update equation of the center frequency is shown in formula (5), and the Lagrange multiplier λ c The update equation of (t) is shown in formula (6).

[0097]

[0098]

[0099]

[0100] In the above formula, α represents the quadratic penalty factor, and τ is the Lagrangian algorithm multiplier update parameter. They are x c (t),u k,c (t), λ c (t) is the Fourier transform of yes The updated value of . Indicates that except for the kth mode Other modes besides . yes The updated value of . Yes k The Fourier transform of .

[0101] yes The updated value of . yes The updated value of .

[0102] Finally, after multiple iterations, the multivariate original signal (the historical power generation of each distributed photovoltaic station in the cluster) x(t)∈R is output N×C The MVMD decomposition result, that is, the modal component u(t)∈R K×N×C , where N represents the signal length.

[0103] The embodiment of the present invention uses multivariate variational mode decomposition (MVMD) to extract the characteristics of spatiotemporal correlation of the historical power generation of distributed photovoltaic stations in a distributed photovoltaic cluster. Multivariate variational mode decomposition can process multivariate data containing any number of data channels, realize the time-frequency synchronization analysis of the historical power generation of multiple distributed photovoltaic stations in the same distributed photovoltaic cluster, better extract the spatiotemporal correlation characteristics of the historical power generation in the distributed photovoltaic cluster, improve the applicability and accuracy of the prediction model, and only use the historical power generation of distributed photovoltaic stations during prediction, without the need to install additional meteorological measurement equipment for distributed photovoltaics, saving economic costs.

[0104] For other aspects of the embodiments of the present invention, please refer to Figure 4 The power generation prediction model includes: a spatiotemporal correlation feature extraction layer 10 and a prediction layer 20. The spatiotemporal correlation feature extraction layer 10 is used to extract the spatiotemporal correlation features between the sample modal components of multiple distributed photovoltaic clusters from the sample modal components of multiple distributed photovoltaic clusters; the prediction layer 20 is used to perform power prediction based on the spatiotemporal correlation features.

[0105] In one embodiment of the present invention, the spatiotemporal correlation feature extraction layer 10 is constructed based on the encoder-decoder of the transformer model (Transformer), and the prediction layer 20 is constructed based on the long short-term memory network (LSTM). In addition, the prediction layer 20 also includes a linear transformation layer and a Softmax layer after the long short-term memory network is constructed. The embodiment of the present invention adds the encoder-decoder of the transformer model on the basis of the construction of the long short-term memory network. The encoder first understands and encodes the information between the sample modal components decomposed by the multivariate variational modal decomposition method, and then the decoder intelligently selects the spatiotemporal correlation features between the modal components and outputs them to the long short-term memory network. Finally, the long short-term memory network learns the changing law of historical power generation and predicts the future power generation. The embodiment of the present invention further pays attention to the spatiotemporal correlation features during model training, and regional prediction of distributed photovoltaic clusters composed of multiple distributed photovoltaic stations can effectively improve the model prediction accuracy. The present invention is conducive to autonomous learning and focuses on the spatiotemporal correlation features of distributed photovoltaic stations between clusters after modal decomposition, improving the model prediction accuracy and calculation efficiency.

[0106] The long short-term memory network can also be replaced by other networks, for example, it can be replaced by a gated recurrent unit or a simple recurrent unit.

[0107] The embodiment of the present invention can complete the training of the power generation prediction model by only using the historical power generation data of the photovoltaic station, and does not need to install additional meteorological measurement equipment for the distributed photovoltaic station, saving economic costs and improving the applicability of the model; at the same time, the power generation prediction model is obtained by training based on the sample modal components of multiple distributed photovoltaic clusters, and the sample modal components reflect the spatiotemporal correlation characteristics of each station in the cluster. The power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between the sample modal components of multiple distributed photovoltaic clusters. Therefore, when training the power generation prediction model of the embodiment of the present invention, attention is paid to the spatiotemporal correlation characteristics of each distributed photovoltaic station between clusters and within the cluster, effectively improving the prediction accuracy of the model.

[0108] Device Embodiment

[0109] Please refer to Figure 5 On the other hand, an embodiment of the present invention further provides a distributed photovoltaic cluster power prediction device, comprising:

[0110] The acquisition module 501 is used to acquire the total power generation of all distributed photovoltaic stations in the distributed photovoltaic cluster;

[0111] A prediction module 502 is used to input the total power generation into a power generation prediction model to obtain a future power generation prediction result output by the power generation prediction model;

[0112] Among them, the power generation prediction model is trained based on sample modal components of multiple distributed photovoltaic clusters; the sample modal components are the historical power generation of all distributed photovoltaic stations in the distributed photovoltaic cluster based on multivariate variational modal decomposition; the power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between sample modal components of multiple distributed photovoltaic clusters.

[0113] Optionally, the power generation prediction model is trained based on the following steps:

[0114] Obtain historical power generation and historical photovoltaic output data of multiple distributed photovoltaic sites;

[0115] Performing correlation analysis based on the historical photovoltaic output data of all distributed photovoltaic stations to obtain a correlation coefficient between each distributed photovoltaic station;

[0116] Based on the magnitude relationship of the correlation coefficients between each of the distributed photovoltaic stations, the distributed photovoltaic stations are divided into clusters to obtain a plurality of distributed photovoltaic clusters; each of the distributed photovoltaic clusters includes at least one distributed photovoltaic station;

[0117] Perform multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain the sample modal components of each distributed photovoltaic cluster;

[0118] The power generation prediction model is trained based on the sample modal components of all distributed photovoltaic clusters to obtain the trained power generation prediction model.

[0119] Optionally, the correlation coefficient is a Pearson correlation coefficient, which is expressed by the following formula:

[0120]

[0121] Among them, r represents the Pearson correlation coefficient, x i and i are the horizontal and vertical coordinates of the i-th data in the historical photovoltaic output data curve, and They are respectively the average value of the horizontal axis and the average value of the vertical axis in the historical photovoltaic output data curve, and the historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the distributed photovoltaic station.

[0122] Optionally, performing multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain sample modal components of each distributed photovoltaic cluster includes:

[0123] Constructing an objective function with constraints based on the historical power generation of all distributed photovoltaic stations in the target cluster; the objective function represents a set of multivariate intrinsic mode functions extracted from the historical power generation of all distributed photovoltaic stations;

[0124] Constructing an augmented Lagrangian function, and removing constraints in the objective function based on the augmented Lagrangian function;

[0125] Repeat the following steps until a preset condition is met: solve the objective function after removing the constraint condition by an alternating direction multiplier method; the solution of the objective function that meets the preset condition is used as the sample modal component of the target cluster;

[0126] The target cluster is any distributed photovoltaic cluster among all distributed photovoltaic clusters.

[0127] Optionally, the power generation prediction model includes:

[0128] A spatiotemporal correlation feature extraction layer is used to extract spatiotemporal correlation features between sample modal components of multiple distributed photovoltaic clusters from sample modal components of multiple distributed photovoltaic clusters;

[0129] The prediction layer is used to perform power prediction based on the spatiotemporal correlation features.

[0130] Optionally, the spatiotemporal correlation feature extraction layer is constructed based on an encoder-decoder of a converter model, and the prediction layer is constructed based on a long short-term memory network.

[0131] The distributed photovoltaic cluster power prediction device includes a processor and a memory. The acquisition module 501 and the prediction module 502 are stored in the memory as program units. The processor executes the program units stored in the memory to implement corresponding functions.

[0132] The processor includes a kernel, which calls the corresponding program unit from the memory. There can be one or more kernels.

[0133] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0134] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the distributed photovoltaic cluster power prediction method, the method comprising: obtaining the total power generation of all distributed photovoltaic sites in the distributed photovoltaic cluster; inputting the total power generation into the power generation prediction model to obtain the future power generation prediction result output by the power generation prediction model; wherein the power generation prediction model is obtained by training based on the sample modal components of multiple distributed photovoltaic clusters; the sample modal components are the historical power generation of all distributed photovoltaic sites in the distributed photovoltaic cluster based on multivariate variational modal decomposition; the power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between the sample modal components of multiple distributed photovoltaic clusters.

[0135] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0136] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a machine-readable storage medium. When the computer program is executed by a processor, the computer can execute a distributed photovoltaic cluster power prediction method, which includes: obtaining the total power generation of all distributed photovoltaic sites in a distributed photovoltaic cluster; inputting the total power generation into a power generation prediction model to obtain a future power generation prediction result output by the power generation prediction model; wherein the power generation prediction model is trained based on sample modal components of multiple distributed photovoltaic clusters; the sample modal components are historical power generation of all distributed photovoltaic sites in the distributed photovoltaic cluster based on multivariate variational modal decomposition; the power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between sample modal components of multiple distributed photovoltaic clusters.

[0137] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, which is implemented when a processor executes a distributed photovoltaic cluster power prediction method, the method comprising: obtaining the total power generation of all distributed photovoltaic sites in a distributed photovoltaic cluster; inputting the total power generation into a power generation prediction model to obtain a future power generation prediction result output by the power generation prediction model; wherein the power generation prediction model is trained based on sample modal components of multiple distributed photovoltaic clusters; the sample modal components are historical power generation of all distributed photovoltaic sites in the distributed photovoltaic cluster based on multivariate variational modal decomposition; the power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between sample modal components of multiple distributed photovoltaic clusters.

[0138] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0139] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0140] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distributed photovoltaic cluster power prediction method, characterized in that: include: Obtain the total power generation of all distributed photovoltaic stations in the distributed photovoltaic cluster; Inputting the total power generation into a power generation prediction model to obtain a future power generation prediction result output by the power generation prediction model; Among them, the power generation prediction model is trained based on sample modal components of multiple distributed photovoltaic clusters; the sample modal components are the historical power generation of all distributed photovoltaic stations in the distributed photovoltaic cluster based on multivariate variational modal decomposition; the power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between sample modal components of multiple distributed photovoltaic clusters.

2. The distributed photovoltaic cluster power prediction method according to claim 1, characterized in that: The power generation prediction model is trained based on the following steps: Obtain historical power generation and historical photovoltaic output data of multiple distributed photovoltaic sites; Performing correlation analysis based on the historical photovoltaic output data of all distributed photovoltaic stations to obtain a correlation coefficient between each distributed photovoltaic station; Based on the magnitude relationship of the correlation coefficients between each of the distributed photovoltaic stations, the distributed photovoltaic stations are divided into clusters to obtain a plurality of distributed photovoltaic clusters; each of the distributed photovoltaic clusters includes at least one distributed photovoltaic station; Perform multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain the sample modal components of each distributed photovoltaic cluster; The power generation prediction model is trained based on the sample modal components of all distributed photovoltaic clusters to obtain the trained power generation prediction model.

3. The distributed photovoltaic cluster power prediction method according to claim 2, characterized in that: The correlation coefficient is the Pearson correlation coefficient, which is expressed by the following formula: Among them, r represents the Pearson correlation coefficient, x i is the i-th data in the first historical photovoltaic output data curve, y i is the i-th data in the second historical photovoltaic output data curve, is the average value in the first historical photovoltaic output data curve, It is the average value in the second historical photovoltaic output data curve, the first historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the first distributed photovoltaic site, and the second historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the second distributed photovoltaic site.

4. The distributed photovoltaic cluster power prediction method according to claim 2, characterized in that: The multivariate variational modal decomposition is performed on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain the sample modal components of each distributed photovoltaic cluster, including: Constructing an objective function with constraints based on the historical power generation of all distributed photovoltaic stations in the target cluster; the objective function represents a set of multivariate intrinsic mode functions extracted from the historical power generation of all distributed photovoltaic stations; Constructing an augmented Lagrangian function, and removing constraints in the objective function based on the augmented Lagrangian function; Repeat the following steps until a preset condition is met: solve the objective function after removing the constraint condition by an alternating direction multiplier method; the solution of the objective function that meets the preset condition is used as the sample modal component of the target cluster; The target cluster is any distributed photovoltaic cluster among all distributed photovoltaic clusters.

5. The distributed photovoltaic cluster power prediction method according to claim 1, characterized in that: The power generation prediction model includes: A spatiotemporal correlation feature extraction layer is used to extract spatiotemporal correlation features between sample modal components of multiple distributed photovoltaic clusters from sample modal components of multiple distributed photovoltaic clusters; The prediction layer is used to perform power prediction based on the spatiotemporal correlation features.

6. The distributed photovoltaic cluster power prediction method according to claim 5, characterized in that: The spatiotemporal correlation feature extraction layer is constructed based on the encoder-decoder of the converter model, and the prediction layer is constructed based on the long short-term memory network.

7. A distributed photovoltaic cluster power prediction device, characterized in that: include: An acquisition module is used to obtain the total power generation of all distributed photovoltaic sites in a distributed photovoltaic cluster; A prediction module, used for inputting the total power generation into a power generation prediction model to obtain a future power generation prediction result output by the power generation prediction model; Among them, the power generation prediction model is trained based on sample modal components of multiple distributed photovoltaic clusters; the sample modal components are the historical power generation of all distributed photovoltaic stations in the distributed photovoltaic cluster based on multivariate variational modal decomposition; the power generation prediction model performs power prediction based on the spatiotemporal correlation characteristics between sample modal components of multiple distributed photovoltaic clusters.

8. The distributed photovoltaic cluster power prediction device according to claim 7, characterized in that: The power generation prediction model is trained based on the following steps: Obtain historical power generation and historical photovoltaic output data of multiple distributed photovoltaic sites; Performing correlation analysis based on the historical photovoltaic output data of all distributed photovoltaic stations to obtain a correlation coefficient between each distributed photovoltaic station; Based on the magnitude relationship of the correlation coefficients between each of the distributed photovoltaic stations, the distributed photovoltaic stations are divided into clusters to obtain a plurality of distributed photovoltaic clusters; each of the distributed photovoltaic clusters includes at least one distributed photovoltaic station; Perform multivariate variational modal decomposition on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain the sample modal components of each distributed photovoltaic cluster; The power generation prediction model is trained based on the sample modal components of all distributed photovoltaic clusters to obtain the trained power generation prediction model.

9. The distributed photovoltaic cluster power prediction device according to claim 8, characterized in that: The correlation coefficient is the Pearson correlation coefficient, which is expressed by the following formula: Among them, r represents the Pearson correlation coefficient, x i is the i-th data in the first historical photovoltaic output data curve, y i is the i-th data in the second historical photovoltaic output data curve, is the average value in the first historical photovoltaic output data curve, It is the average value in the second historical photovoltaic output data curve, the first historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the first distributed photovoltaic site, and the second historical photovoltaic output data curve is constructed based on the historical photovoltaic output data of the second distributed photovoltaic site.

10. The distributed photovoltaic cluster power prediction device according to claim 8, characterized in that: The multivariate variational modal decomposition is performed on the historical power generation of all distributed photovoltaic stations in each distributed photovoltaic cluster to obtain the sample modal components of each distributed photovoltaic cluster, including: Constructing an objective function with constraints based on the historical power generation of all distributed photovoltaic stations in the target cluster; the objective function represents a set of multivariate intrinsic mode functions extracted from the historical power generation of all distributed photovoltaic stations; Constructing an augmented Lagrangian function, and removing constraints in the objective function based on the augmented Lagrangian function; Repeat the following steps until a preset condition is met: solve the objective function after removing the constraint condition by an alternating direction multiplier method; the solution of the objective function that meets the preset condition is used as the sample modal component of the target cluster; The target cluster is any distributed photovoltaic cluster among all distributed photovoltaic clusters.

11. The distributed photovoltaic cluster power prediction device according to claim 7, characterized in that: The power generation prediction model includes: A spatiotemporal correlation feature extraction layer is used to extract spatiotemporal correlation features between sample modal components of multiple distributed photovoltaic clusters from sample modal components of multiple distributed photovoltaic clusters; The prediction layer is used to perform power prediction based on the spatiotemporal correlation features.

12. The distributed photovoltaic cluster power prediction device according to claim 11, characterized in that: The spatiotemporal correlation feature extraction layer is constructed based on the encoder-decoder of the converter model, and the prediction layer is constructed based on the long short-term memory network.

13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the distributed photovoltaic cluster power prediction method according to any one of claims 1 to 6 is implemented.

14. A machine-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the distributed photovoltaic cluster power prediction method according to any one of claims 1 to 6 is implemented.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the distributed photovoltaic cluster power prediction method according to any one of claims 1 to 6 is implemented.

Images