Photovoltaic power generation power prediction method and device and nonvolatile storage medium

Abstract

The invention discloses a photovoltaic power generation power prediction method and device and a nonvolatile storage medium. The method comprises the following steps: acquiring respective historical power generation data of a plurality of photovoltaic power generation devices included in a photovoltaic power station in a historical time period; grouping the plurality of photovoltaic power generation devices according to the respective historical power generation data of the plurality of photovoltaic power generation devices to obtain a plurality of groups of photovoltaic power generation devices; according to the respective historical power generation data of the plurality of photovoltaic power generation devices, determining the respective power generation powers of the plurality of groups of photovoltaic power generation devices; respectively constructing a plurality of feature maps according to the respective generated power of the plurality of groups of photovoltaic power generation equipment; respectively inputting the plurality of feature maps into a plurality of pre-trained convolutional neural network models, and respectively outputting respective predicted power of the plurality of groups of photovoltaic power generation equipment by the plurality of convolutional neural network models; and determining the power generation power of the photovoltaic power station according to the respective predicted power of the plurality of groups of photovoltaic power generation equipment. According to the invention, the technical problem of low efficiency of photovoltaic power generation power prediction in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology, and in particular to a photovoltaic power generation power prediction method, device and non-volatile storage medium. Background Art

[0002] As a new energy source, photovoltaics are clean and environmentally friendly. Therefore, the installed capacity of countries around the world has increased year by year, which has increased the penetration rate of photovoltaic power generation. However, with the changes in photovoltaics, the output of photovoltaic power generation is not stable, which makes the operation of the power system face considerable challenges. The prediction of the output power of photovoltaic generators is crucial to the reasonable dispatch of the power grid system.

[0003] At present, the commonly used prediction methods for photovoltaic power generation prediction include the continuous method, time series method, support vector machine method, etc. The prediction results of the continuous method are unstable and have large errors. The time series method requires a large amount of historical data to build a model, but the final prediction model is not accurate enough for low-order predictions. For high-order predictions, the parameters are difficult to estimate and there are certain defects. The support vector machine method pays more attention to the selection of kernel functions and the determination of related parameters. They need to rely on the designer's experience to determine and rely heavily on light intensity data.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a photovoltaic power generation power prediction method, device and non-volatile storage medium to at least solve the technical problem of low efficiency in photovoltaic power generation power prediction in the related art.

[0006] According to one aspect of an embodiment of the present invention, a photovoltaic power generation power prediction method is provided, comprising: obtaining historical power generation data of each of a plurality of photovoltaic power generation devices included in a photovoltaic power station within a historical time period; grouping the plurality of photovoltaic power generation devices according to the historical power generation data of each of the plurality of photovoltaic power generation devices to obtain a plurality of groups of photovoltaic power generation devices; determining the power generation of each of the plurality of groups of photovoltaic power generation devices according to the historical power generation data of each of the plurality of photovoltaic power generation devices; constructing a plurality of feature graphs according to the power generation of each of the plurality of groups of photovoltaic power generation devices, wherein the plurality of feature graphs correspond one-to-one to the plurality of groups of photovoltaic power generation devices, and any one of the plurality of feature graphs is used to characterize the association between the plurality of photovoltaic devices in a corresponding group of photovoltaic power generation devices; inputting the plurality of feature graphs into a plurality of pre-trained convolutional neural network models, respectively, and outputting the predicted power of each of the plurality of groups of photovoltaic power generation devices respectively by the plurality of convolutional neural network models, wherein the convolutional neural network models are trained by training samples, and the training samples include sample feature graphs and sample powers; determining the power generation of the photovoltaic power station according to the predicted power of each of the plurality of groups of photovoltaic power generation devices.

[0007] Optionally, according to the historical power generation data of each of the multiple photovoltaic power generation devices, the multiple photovoltaic power generation devices are grouped to obtain multiple groups of photovoltaic power generation devices, including: obtaining a preset number of groups; selecting multiple clustering centers corresponding to the number of groups from the multiple photovoltaic power generation devices; according to the historical power generation data of each of the multiple photovoltaic power generation devices, based on the multiple clustering centers, a clustering algorithm is used to group the multiple photovoltaic power generation devices to obtain multiple groups of photovoltaic power generation devices.

[0008] Optionally, multiple feature maps are constructed according to the power generation of each of the multiple groups of photovoltaic power generation equipment, including: adopting the following method to construct a target feature map according to the power generation of the target group of photovoltaic power generation equipment, and constructing multiple feature maps according to the power generation of each of the multiple groups of photovoltaic power generation equipment: determining the power generation of each of the target group of photovoltaic power generation equipment in multiple time periods according to the power generation of each of the target group of photovoltaic power generation equipment; generating a spatial feature map for characterizing the spatial relationship of the devices included in the target group of photovoltaic power generation equipment in each time period according to the power generation of each of the target group of photovoltaic power generation equipment in multiple time periods, wherein the multiple nodes included in the spatial feature map respectively characterize each of the photovoltaic power generation devices in the target group of photovoltaic power generation equipment; constructing a target feature map according to the spatial feature maps of each of the multiple time periods.

[0009] Optionally, a target feature map is constructed based on the spatial feature maps of each of the multiple time periods, including: sorting the spatial feature maps of each of the multiple time periods according to the time sequence of the multiple time periods; and connecting the sorted spatial feature maps of each of the multiple time periods to obtain the target feature map.

[0010] Optionally, the spatial feature maps of the sorted multiple time periods are connected to obtain a target feature map, including: connecting each group of adjacent spatial feature maps in the spatial feature maps of the sorted multiple time periods to obtain a feature map in the following manner: in two adjacent spatial feature maps, matching nodes are identified and the matching nodes are connected, wherein the matching nodes represent the same photovoltaic power generation device.

[0011] Optionally, multiple feature maps are respectively input into multiple pre-trained convolutional neural network models, and the multiple convolutional neural network models respectively output the predicted power of multiple groups of photovoltaic power generation equipment, including: multiple feature maps are respectively input into multiple pre-trained convolutional neural network models, each of the multiple convolutional neural network models segments the input feature map to obtain multiple corresponding local feature maps; each of the multiple convolutional neural network models generates an adjacency matrix corresponding to the multiple local feature maps, wherein the adjacency matrix represents the degree of influence of adjacent nodes in the local feature map on the central node; each of the convolutional neural network models performs convolution operations on the multiple local feature maps according to the adjacency matrices corresponding to the multiple local feature maps to obtain the predicted power of the corresponding group of photovoltaic power generation equipment; and the multiple convolutional neural network models respectively output the predicted power of multiple groups of photovoltaic power generation equipment.

[0012] According to another aspect of an embodiment of the present invention, a photovoltaic power prediction device is also provided, comprising: an acquisition module for acquiring historical power generation data of each of a plurality of photovoltaic power generation devices included in a photovoltaic power station within a historical time period; a grouping module for grouping a plurality of photovoltaic power generation devices according to their respective historical power generation data to obtain a plurality of groups of photovoltaic power generation devices; a determination module for determining the power generation of each of the plurality of groups of photovoltaic power generation devices according to their respective historical power generation data; a composition module for constructing a plurality of feature graphs according to the power generation of each of the plurality of groups of photovoltaic power generation devices, wherein the plurality of feature graphs correspond one-to-one to the plurality of groups of photovoltaic power generation devices, and any one of the plurality of feature graphs is used to characterize the association between the plurality of photovoltaic devices in a corresponding group of photovoltaic power generation devices; a prediction module for inputting the plurality of feature graphs into a plurality of pre-trained convolutional neural network models, respectively, and outputting the predicted power of each of the plurality of groups of photovoltaic power generation devices respectively by the plurality of convolutional neural network models, wherein the convolutional neural network models are trained by training samples, and the training samples include sample feature graphs and sample powers; a determination module for determining the power generation of the photovoltaic power station according to the predicted power of each of the plurality of groups of photovoltaic power generation devices.

[0013] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided, the non-volatile storage medium including a stored program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute any one of the above-mentioned photovoltaic power prediction methods.

[0014] According to another aspect of the embodiments of the present invention, a computer device is provided. The computer device includes a processor, and the processor is used to run a program. When the program is run, any one of the above photovoltaic power prediction methods is executed.

[0015] According to another aspect of the embodiments of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, any one of the above photovoltaic power prediction methods is implemented.

[0016] In an embodiment of the present invention, historical power generation data of each of the multiple photovoltaic power generation devices included in the photovoltaic power station within a historical time period is obtained; according to the historical power generation data of each of the multiple photovoltaic power generation devices, the multiple photovoltaic power generation devices are grouped to obtain multiple groups of photovoltaic power generation devices; according to the historical power generation data of each of the multiple photovoltaic power generation devices, the power generation power of each of the multiple groups of photovoltaic power generation devices is determined; according to the power generation power of each of the multiple groups of photovoltaic power generation devices, multiple feature maps are respectively constructed, wherein the multiple feature maps correspond to the multiple groups of photovoltaic power generation devices one by one, and any one of the multiple feature maps is used to characterize the association between the multiple photovoltaic devices in the corresponding group of photovoltaic power generation devices; the multiple feature maps are respectively input into multiple pre-trained convolutional neural network models In the method, a plurality of convolutional neural network models respectively output the predicted powers of a plurality of groups of photovoltaic power generation equipment, wherein the convolutional neural network model is trained by training samples, and the training samples include sample feature maps and sample powers; the power generation power of the photovoltaic power station is determined according to the predicted powers of the plurality of groups of photovoltaic power generation equipment, and since the feature map can extract the correlation between the outputs of different photovoltaic power generation equipment, the convolutional neural network model can fully learn the hidden features of the output data of the photovoltaic power generation equipment, thereby achieving the purpose of minimizing the parameter adjustment cost while ensuring the prediction accuracy, thereby achieving the technical effect of improving the efficiency of the photovoltaic power generation prediction power, and further solving the technical problem of low efficiency of the photovoltaic power generation prediction power in the related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary 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:

[0018] Figure 1 A hardware structure block diagram of a computer terminal for implementing a photovoltaic power generation power prediction method is shown;

[0019] Figure 2 is a schematic diagram of a process of a photovoltaic power generation prediction method provided according to an embodiment of the present invention;

[0020] Figure 3 It is a schematic flow chart of a photovoltaic power prediction method using adaptive spatiotemporal synchronous graph convolutional neural network and K-means clustering according to an optional embodiment of the present invention;

[0021] Figure 4 2 is a schematic diagram of a photovoltaic power generation prediction device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0024] According to an embodiment of the present invention, a method embodiment of photovoltaic power generation prediction is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0025] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal for implementing a photovoltaic power generation power prediction method is shown. Figure 1 As shown, the computer terminal 10 may include one or more (102a, 102b, ..., 102n are used to illustrate) processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0026] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10. As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0027] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the photovoltaic power generation prediction method in the embodiment of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the photovoltaic power generation prediction method of the above-mentioned application is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0028] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0029] Figure 2 FIG. 1 is a flow chart of a photovoltaic power generation prediction method according to an embodiment of the present invention. Figure 2 As shown, the method comprises the following steps:

[0030] Step S202, obtaining historical power generation data of each of the plurality of photovoltaic power generation devices included in the photovoltaic power station within a historical time period.

[0031] In this step, power generation data of multiple photovoltaic power generation devices over a period of time (such as one day, one week, one month, etc.) can be collected. These data may include environmental parameters such as power generation power, light intensity, temperature, etc. of each device at different time points.

[0032] Step S204: grouping the plurality of photovoltaic power generation devices according to their respective historical power generation data to obtain a plurality of groups of photovoltaic power generation devices.

[0033] In this step, the historical power generation data of multiple photovoltaic power generation devices can be analyzed to find the correlation between them, such as close geographical locations, similar power generation trends, etc. Then similar photovoltaic power generation devices can be grouped together, that is, the devices in each group are relatively close in power generation characteristics, which helps to reduce the complexity of the data and improve the accuracy of subsequent predictions.

[0034] Step S206, determining the power generation power of each of the plurality of photovoltaic power generation devices according to the respective historical power generation data of the plurality of photovoltaic power generation devices.

[0035] In this step, for each group, the historical power generation data of all photovoltaic power generation equipment in the group can be calculated. In the subsequent process, the photovoltaic power generation power can be predicted on a group basis.

[0036] Step S208, constructing a plurality of characteristic graphs respectively according to the power generation powers of the plurality of groups of photovoltaic power generation equipment, wherein the plurality of characteristic graphs correspond one-to-one to the plurality of groups of photovoltaic power generation equipment, and any one of the plurality of characteristic graphs is used to characterize the association between the plurality of photovoltaic devices in the corresponding group of photovoltaic power generation equipment.

[0037] In this step, each photovoltaic power generation device is represented as a node in the feature graph, and each group of photovoltaic power generation devices can correspond to a feature graph. According to certain associations between photovoltaic power generation devices in the group (such as the proximity of geographical locations, similarity of weather conditions, etc.), edges can be added to the feature graph to represent the association between each photovoltaic power generation device and other photovoltaic power generation devices in the group. The weight of the edge can be quantified according to the degree of association. By constructing the feature graph, the mutual influence and dependency between photovoltaic power generation devices can be further explored.

[0038] Step S210, inputting the multiple feature maps into multiple pre-trained convolutional neural network models respectively, and the multiple convolutional neural network models respectively output the predicted power of multiple groups of photovoltaic power generation equipment, wherein the convolutional neural network model is trained by training samples, and the training samples include sample feature maps and sample powers.

[0039] In this step, the constructed feature map can be used as input data and sent to the pre-trained convolutional neural network model. Each convolutional network model processes a certain input feature map. The convolutional network model extracts useful information from the feature map through structures such as convolutional layers and fully connected layers, and outputs multiple groups of predicted powers of photovoltaic power generation equipment. These predicted powers are based on the laws and patterns learned by the model from historical data. By using the feature map as the input of the convolutional neural network model, the correlation information in the graph structure data can be fully utilized to improve the accuracy of the prediction. Prior to this, training the convolutional neural network model through training samples can ensure that the model can learn the changing laws of photovoltaic power generation and make accurate predictions.

[0040] Step S212, determining the power generation of the photovoltaic power station according to the predicted power of each of the multiple groups of photovoltaic power generation equipment. In this step, the predicted power of the multiple groups of photovoltaic power generation equipment can be combined to determine the power generation of the entire photovoltaic power station.

[0041] Through the above steps, since the feature map can extract the correlation between the outputs of different photovoltaic power generation equipment, the convolutional neural network model can fully learn the hidden features of the output data of photovoltaic power generation equipment, thereby achieving the purpose of minimizing the parameter adjustment cost while ensuring the prediction accuracy, thereby achieving the technical effect of improving the efficiency of photovoltaic power generation prediction power, and then solving the technical problem of low efficiency of photovoltaic power generation prediction power in related technologies.

[0042] As an optional embodiment, according to the historical power generation data of each of the multiple photovoltaic power generation devices, the multiple photovoltaic power generation devices are grouped to obtain multiple groups of photovoltaic power generation devices, including: obtaining a preset number of groups; selecting multiple clustering centers corresponding to the number of groups from the multiple photovoltaic power generation devices; according to the historical power generation data of each of the multiple photovoltaic power generation devices, based on the multiple clustering centers, a clustering algorithm is used to group the multiple photovoltaic power generation devices to obtain multiple groups of photovoltaic power generation devices.

[0043] Optionally, cluster analysis is an effective unsupervised learning method in current data information mining, which can mine the similarity of data objects without any prior knowledge to achieve the purpose of data grouping. As a common iterative clustering analysis algorithm, K-means clustering has been widely used in various research fields. This method has the advantages of simple algorithm idea, fast convergence speed, better clustering effect, fewer parameters to be adjusted, and strong algorithm interpretability. The present invention can use the K-means clustering algorithm to group the historical data of photovoltaic power generation equipment.

[0044] First, before starting the grouping, it is necessary to clarify how many groups these photovoltaic power generation devices are to be divided into. This number of groups may be pre-set based on a certain strategy. For example, the devices may be classified according to factors such as geographical location, climate conditions, or power generation characteristics. Secondly, it is necessary to select multiple cluster centers corresponding to the pre-set number of groups from all photovoltaic power generation devices. The cluster center is an important concept in clustering algorithms, representing the center point or representative point of each cluster (or group). The cluster center can be the initial point selected by some random method or heuristic algorithm, and they will be continuously adjusted in the subsequent clustering process to optimize the clustering effect.

[0045] Specifically, a sample point can be randomly selected from the data set as the cluster center; then the shortest distance between each sample and the current cluster center can be calculated, represented by M(x); then the probability N(x) of each sample point being selected as the next cluster center is calculated using formula (1), and finally the sample point corresponding to the maximum probability value is selected as the next cluster center:

[0046]

[0047] Repeating the above steps can initialize K cluster centers, where K can be preset.

[0048] Secondly, a clustering algorithm can be used to group photovoltaic power generation devices according to their historical power generation data. The goal of the clustering algorithm is to cluster similar photovoltaic power generation devices together to form different groups.

[0049] Specifically, the Euclidean space distance between each object and each initial cluster center is calculated, and each object is assigned to the cluster center closest to it; then, the center point of all objects in each cluster can be calculated based on the assignment results of each object, and K corresponding cluster centers can be updated. Repeat the above steps until the assignment results of all objects do not change or the maximum number of iterations is reached, and the entire clustering process is completed.

[0050] As an optional embodiment, multiple feature maps are constructed respectively according to the power generation power of each of the multiple groups of photovoltaic power generation equipment, including: adopting the following method to construct a target feature map according to the power generation power of the target group of photovoltaic power generation equipment, and constructing multiple feature maps respectively according to the power generation power of each of the multiple groups of photovoltaic power generation equipment: determining the power generation power of each of the target group of photovoltaic power generation equipment in multiple time periods according to the power generation power of each of the target group of photovoltaic power generation equipment; generating a spatial feature map for characterizing the spatial relationship of the devices included in the target group of photovoltaic power generation equipment in each time period according to the power generation power of each of the target group of photovoltaic power generation equipment in multiple time periods, wherein the multiple nodes included in the spatial feature map respectively characterize each of the photovoltaic power generation devices in the target group of photovoltaic power generation equipment; constructing a target feature map according to the spatial feature maps of each of the multiple time periods.

[0051] Optionally, the feature map can be a spatiotemporal feature map, which can simultaneously represent the temporal correlation and spatial correlation of data. A feature map can be constructed for each group of photovoltaic power generation equipment, and the spatiotemporal feature map corresponding to a group of photovoltaic power generation equipment can be constructed by the following steps: First, it is necessary to collect or obtain the power generation data of each photovoltaic power generation equipment in the group of photovoltaic power generation equipment in multiple different time periods (such as different time periods of each day). For each time period, it needs to be processed separately. Then a node can be established for each photovoltaic power generation equipment, and each node can carry the power generation characteristics of the photovoltaic power generation equipment, such as power generation, location, etc., and a spatial feature map is generated according to the spatial relationship between the photovoltaic power generation equipment included in the group of photovoltaic power generation equipment. Among them, the spatial feature map can be presented in the form of a network map or a topological map, in which the node represents the photovoltaic power generation equipment, and the edge or connection relationship represents the spatial relationship between the equipment (such as the proximity of the geographical location, etc.). Then, based on the spatial feature maps of multiple time periods, the spatial feature maps of multiple time periods can be combined according to the time relationship between the multiple time periods to construct a comprehensive spatiotemporal feature map. The spatiotemporal feature map can be a three-dimensional graph, in which two dimensions represent space (i.e. the distribution of photovoltaic power generation equipment) and the other dimension represents time (i.e. the spatial feature map of different time periods). Such a feature map can fully display the spatial distribution and change relationship of photovoltaic power generation equipment at different time points.

[0052] As an optional embodiment, a target feature map is constructed based on the spatial feature maps of each of the multiple time periods, including: sorting the spatial feature maps of each of the multiple time periods according to the time sequence of the multiple time periods; and connecting the sorted spatial feature maps of each of the multiple time periods to obtain the target feature map.

[0053] Optionally, the time sequence of multiple time periods can be determined, and then the spatial feature graphs corresponding to each time period can be sorted according to the determined time sequence. After the sorting is completed, these spatial feature graphs can be connected in time sequence, where the connection can be a simple sequential arrangement, or the spatial feature graphs of adjacent time periods can be superimposed, fused, or smoothly transitioned. After the connection is completed, the result is a comprehensive spatiotemporal feature graph, which contains the spatial distribution and relationship information of photovoltaic power generation equipment in all selected time periods. The spatiotemporal feature graph can be expanded along the time axis to show the spatial distribution and change relationship of photovoltaic power generation equipment at different time points.

[0054] As an optional embodiment, the spatial feature maps of the sorted multiple time periods are connected to obtain a target feature map, including: connecting each group of adjacent spatial feature maps in the spatial feature maps of the sorted multiple time periods to obtain a feature map in the following manner: in two adjacent spatial feature maps, matching nodes are identified and the matching nodes are connected, wherein the matching nodes represent the same photovoltaic power generation device.

[0055] Optionally, a spatiotemporal feature map can be obtained by connecting adjacent spatial feature maps in time sequence. Specifically, nodes representing the same photovoltaic power generation device can be identified in two adjacent spatial feature maps. These nodes may have different positions and different feature values ​​in spatial feature maps of different time periods, but they represent the same physical entity - a specific photovoltaic power generation device. After identifying the matching nodes, the matching nodes are connected. This connection not only represents the continuity of the photovoltaic power generation device at different time points, but also implies the change information of the device status or position. By capturing the spatial distribution and change relationship of photovoltaic power generation equipment at different time points, the model can learn richer information, thereby improving the accuracy of prediction and optimization.

[0056] As an optional embodiment, multiple feature maps are respectively input into multiple pre-trained convolutional neural network models, and the multiple convolutional neural network models respectively output the predicted power of multiple groups of photovoltaic power generation equipment, including: multiple feature maps are respectively input into multiple pre-trained convolutional neural network models, each of the multiple convolutional neural network models segments the input feature map to obtain multiple corresponding local feature maps; each of the multiple convolutional neural network models generates an adjacency matrix corresponding to the multiple local feature maps, wherein the adjacency matrix represents the degree of influence of adjacent nodes in the local feature map on the central node; each of the convolutional neural network models in the convolutional neural network model performs convolution operations on the multiple local feature maps according to the adjacency matrices corresponding to the multiple local feature maps to obtain the predicted power of the corresponding group of photovoltaic power generation equipment; and the multiple convolutional neural network models respectively output the predicted power of multiple groups of photovoltaic power generation equipment.

[0057] Optionally, the convolutional neural network model can be a spatiotemporal synchronous graph convolutional network. The spatiotemporal synchronous graph convolutional network is a graph convolutional network that combines spatial and temporal information, and is particularly suitable for processing time series problems with graph-structured data. First, the spatiotemporal feature graph containing the spatial distribution and relationship information of photovoltaic power generation equipment is input into a pre-trained spatiotemporal synchronous graph convolutional network. Among them, each convolutional neural network inputs a feature graph, and the convolutional neural network processes the spatiotemporal graph to generate multiple local feature graphs and an adjacency matrix, and finally predicts the power of the corresponding group of photovoltaic power generation equipment. Therefore, multiple convolutional neural networks can output the predicted power of multiple groups of photovoltaic power generation equipment respectively. The spatiotemporal synchronous graph convolutional network segments the input spatiotemporal feature graph to obtain multiple local feature graphs, and performs feature extraction and analysis on each local feature graph.

[0058] For each local feature map, the spatiotemporal synchronous graph convolutional network generates a corresponding adjacency matrix. In the local feature map, the adjacency matrix represents the degree of influence of the adjacent nodes (i.e., photovoltaic power generation equipment) in the local feature map on the central node. Through the adjacency matrix, the network can capture the spatial relationship between devices and understand the mutual influence between devices. The spatiotemporal synchronous graph convolutional network performs graph convolution operations on multiple local feature maps according to the adjacency matrices corresponding to the multiple local feature maps. After the graph convolution operation, the spatiotemporal synchronous graph convolutional network outputs the predicted power of multiple groups of photovoltaic power generation equipment. These predicted powers are based on the input feature map, the spatial relationship between the equipment, and the time information. Such prediction results can help power stations achieve more efficient and reliable operation, improve energy utilization efficiency, and reduce operating costs.

[0059] Specifically, the adaptive spatiotemporal synchronous graph convolutional neural network does not rely on special time components and space components. It uses local modeling to simultaneously mine the temporal and spatial correlations of spatiotemporal graph data. The local spatiotemporal graph (local feature graph) refers to connecting each current moment node with its own nodes in the two moments before and after it to construct a spatiotemporal graph containing the three moments before, during and after. The essence of the graph convolutional neural network is to aggregate the features of adjacent nodes into the nodes themselves. For the spatiotemporal synchronous graph convolutional neural network, the adjacency matrix of the local spatiotemporal graph determines the contribution of each adjacent node to the central node. In order to improve the learning effect of the graph convolutional neural network, the adjacency matrix of the local spatiotemporal graph is improved using a learnable adaptive mask matrix. The mask matrix is ​​multiplied by the corresponding elements of the local adjacency matrix of the local spatiotemporal graph to obtain the local adaptive adjacency matrix:

[0060]

[0061] Where P' is the local adaptive adjacency matrix, P is the local adjacency matrix, and L mask is the local mask matrix, defined as follows:

[0062]

[0063] Where, L mask represents the element in the i-th row and j-th column of the local mask matrix; P[i,j] represents the element in the i-th row and j-th column of the adjacency matrix of the local spatiotemporal graph.

[0064] The operation of the spatio-temporal synchronous graph convolution layer of the adaptive spatio-temporal synchronous graph convolutional neural network is realized through sliding windows and spatio-temporal synchronous graph convolutional modules (Spatio-temporal Synchronous Graph Convolutional Module, STSGCM). For a spatio-temporal graph sequence of a certain time length, time periods are sequentially intercepted, and independent local spatio-temporal graphs and corresponding graph signal matrix sequences and adjacency matrices are constructed in each time period. Through the spatio-temporal synchronous graph convolution module, a new spatio-temporal graph sequence is calculated. Among them, the spatio-temporal synchronous graph convolution module contains multiple spatio-temporal synchronous graph convolution operations, which realizes residual splicing by aggregating the outputs of multiple graph convolution operations, and then uses the shearing operation to delete the corresponding features extracted at the previous moment and the next moment, and only retains the aggregated features extracted at the intermediate moment, thereby realizing feature extraction at a specific moment.

[0065] As a specific embodiment, Figure 3 As shown, the photovoltaic power generation prediction based on adaptive spatiotemporal synchronous graph convolutional neural network and K-means clustering provided by the present invention includes the following parts.

[0066] Step 1: Data information mining based on K-means clustering.

[0067] Cluster analysis is an effective unsupervised learning method in current data information mining. It can mine the similarities of data objects without any prior knowledge to achieve the purpose of data grouping. As a common iterative clustering analysis algorithm, K-means clustering has been widely used in many research fields. This method has the advantages of simple algorithm idea, fast convergence speed, good clustering effect, fewer parameters to be adjusted, and strong algorithm interpretability.

[0068] The clustering process in the present invention mainly consists of the following six sub-steps:

[0069] Step 11: Randomly select a sample point from the data set as the cluster center.

[0070] Step 12: Next, calculate the shortest distance between each sample and the current cluster center, represented by M(x); then calculate the probability N(x) of each sample point being selected as the next cluster center, and finally select the sample point corresponding to the maximum probability value as the next cluster center.

[0071]

[0072] Step 13: Repeat steps 11 to 12 to initialize K cluster centers.

[0073] Step 14: Calculate the Euclidean distance between each object and each initial cluster center, and assign each object to the cluster center closest to it. The Euclidean distance between two points x and y in n-dimensional space is calculated as follows:

[0074]

[0075] Where n is the spatial dimension; x k and k are the kth attribute feature values ​​of x and y respectively.

[0076] Step 15: According to the allocation results of each object, calculate the center points of all objects in each cluster and update K corresponding cluster centers.

[0077] Step 16: Repeat steps 14 to 15 until the allocation results of all objects do not change or the maximum number of iterations is reached, and the entire clustering process is completed.

[0078] The K-means clustering algorithm is used to group the historical output sequences of photovoltaic generator groups. Each group of photovoltaic generators is used as a node in the spatiotemporal graph data, and the aggregated historical output sequence of each group of photovoltaic generators is used as the feature sequence of each node, which helps the graph neural network to better mine the output pattern of each group of photovoltaic generators.

[0079] Step 2: Power prediction based on adaptive spatiotemporal synchronized graph convolutional neural network.

[0080] Graph neural networks are neural network models that specifically process graph data. The adaptive spatiotemporal synchronous graph convolutional neural network does not rely on special time components and space components. It uses local modeling to simultaneously mine the temporal and spatial correlations of spatiotemporal graph data. The local spatiotemporal graph refers to connecting each current moment node with its own nodes in the two moments before and after it to construct a spatiotemporal graph containing the three moments before, during, and after. The essence of a graph convolutional neural network is to aggregate the features of adjacent nodes into the node itself. For the spatiotemporal synchronous graph convolutional neural network, the adjacency matrix of the local spatiotemporal graph determines the contribution of each adjacent node to the central node. In order to improve the learning effect of the graph convolutional neural network, a learnable adaptive mask matrix is ​​used to improve the adjacency matrix of the local spatiotemporal graph. The mask matrix is ​​multiplied by the corresponding elements of the local adjacency matrix of the local spatiotemporal graph to obtain a local adaptive adjacency matrix:

[0081]

[0082] Where P' is the local adaptive adjacency matrix, P is the local adjacency matrix, and L mask is the local mask matrix.

[0083]

[0084] Where, L mask [i,j] represents the element in the i-th row and j-th column of the local mask matrix; P[i,j] represents the element in the i-th row and j-th column of the adjacency matrix of the local spatiotemporal graph.

[0085] Step 21: Mining the input features of the adaptive spatiotemporal synchronized graph convolutional neural network.

[0086] In order to effectively distinguish the time information and spatial information of each node in the local spatiotemporal graph, the adaptive spatiotemporal synchronous graph convolutional neural network uses spatiotemporal embedding to integrate the spatiotemporal correlation information into the graph signal matrix. The graph signal matrix sequence of the original local spatiotemporal graph is denoted as X. For each moment and each node of the local spatiotemporal graph, a vector is generated by random initialization. The former is called the time embedding representation, denoted as T, and the latter is called the space embedding representation, denoted as S. Then, a sharing method is adopted to make different nodes at the same time share the same time embedding vector and the same node at different times share the same space embedding vector, so that the spatiotemporal embedding representation and the graph signal matrix sequence are added by broadcasting to obtain the graph signal matrix sequence X' with spatiotemporal embedding, that is:

[0087] X'=X+T+S

[0088] Step 22: Graph convolution operation of adaptive spatiotemporal synchronous graph convolutional neural network.

[0089] The operation of the spatio-temporal synchronous graph convolution layer of the adaptive spatio-temporal synchronous graph convolutional neural network is realized through sliding windows and spatio-temporal synchronous graph convolutional modules (Spatio-temporal Synchronous Graph Convolutional Module, STSGCM). For a spatio-temporal graph sequence of a certain time length, time periods are sequentially intercepted, and independent local spatio-temporal graphs and corresponding graph signal matrix sequences and adjacency matrices are constructed in each time period. Through the spatio-temporal synchronous graph convolution module, a new spatio-temporal graph sequence is calculated. Among them, the spatio-temporal synchronous graph convolution module contains multiple spatio-temporal synchronous graph convolution operations, which realizes residual splicing by aggregating the outputs of multiple graph convolution operations, and then uses the shearing operation to delete the corresponding features extracted at the previous moment and the next moment, and only retains the aggregated features extracted at the intermediate moment, thereby realizing feature extraction at a specific moment.

[0090] The specific steps of the spatiotemporal synchronous graph convolution operation are as follows: Given a local spatiotemporal graph G, the local adaptive adjacency matrix L mask , the graph signal sequence X' of the local spatiotemporal graph G represented by spatiotemporal embedding, then the spatiotemporal synchronous graph convolution operation is defined as:

[0091] m s =α(Pm s-1 W+b)

[0092] In the formula, m s is the input matrix of the s-th layer of spatiotemporal synchronous graph convolution method, W and b are the training parameters of spatiotemporal synchronous graph convolution, and α is the activation function Relu:

[0093]

[0094] Step 23: Build the adaptive spatiotemporal synchronous graph convolutional neural network structure.

[0095] The adaptive spatiotemporal synchronous graph convolutional neural network adopts an end-to-end approach, taking historical spatiotemporal sequences as input and future spatiotemporal sequences as output. First, the input historical spatiotemporal sequence is transformed through a fully connected layer to map the original low-dimensional features to high-dimensional features. Then, the time and space information of each node is added through the spatiotemporal embedding layer. Next, the spatiotemporal synchronous graph convolution layer uses a sliding window to automatically segment the spatiotemporal sequence to form a local spatiotemporal graph and a graph signal matrix, and uses graph convolution operations to mine the spatiotemporal correlation of each local spatiotemporal graph, and then aggregates the output results of each spatiotemporal synchronous graph convolution operation. Finally, the results of the graph convolution operation are mapped to the final prediction result through two fully connected layers.

[0096] After the above two steps of historical output sequence clustering, spatiotemporal graph data construction, and photovoltaic generator output model training, a photovoltaic power generation prediction model based on adaptive spatiotemporal synchronous graph convolutional neural network and K-means clustering is formed, thereby realizing accurate prediction of the total power of the photovoltaic power field.

[0097] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0098] Through the description of the above implementation methods, those skilled in the art can clearly understand that the photovoltaic power generation power prediction method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0099] According to an embodiment of the present invention, there is also provided a photovoltaic power prediction device for implementing the photovoltaic power prediction method. Figure 4 is a structural block diagram of a photovoltaic power prediction device provided according to an embodiment of the present invention. Figure 4As shown, the photovoltaic power generation prediction device includes: an acquisition module 41, a grouping module 42, a determination module 44, a composition module 45 and a prediction module 45. The photovoltaic power generation prediction device is described below.

[0100] The acquisition module 41 is used to acquire historical power generation data of each of the multiple photovoltaic power generation devices included in the photovoltaic power station within a historical time period.

[0101] The grouping module 42 is used to group the multiple photovoltaic power generation devices according to their respective historical power generation data to obtain multiple groups of photovoltaic power generation devices.

[0102] The determination module 22 is used to determine the power generation power of each of the multiple groups of photovoltaic power generation equipment according to the historical power generation data of each of the multiple photovoltaic power generation equipment.

[0103] The composition module 44 is used to construct multiple feature maps according to the power generation power of each of the multiple groups of photovoltaic power generation equipment, wherein the multiple feature maps correspond one-to-one to the multiple groups of photovoltaic power generation equipment, and any one of the multiple feature maps is used to characterize the relationship between the multiple photovoltaic devices in the corresponding group of photovoltaic power generation equipment.

[0104] The prediction module 45 is used to input the multiple feature maps into multiple pre-trained convolutional neural network models respectively, and the multiple convolutional neural network models respectively output the predicted power of multiple groups of photovoltaic power generation equipment, wherein the convolutional neural network model is trained by training samples, and the training samples include sample feature maps and sample powers.

[0105] The determination module 46 is used to determine the power generation power of the photovoltaic power station according to the predicted power of each of the multiple groups of photovoltaic power generation equipment.

[0106] It should be noted that the acquisition module 41, the grouping module 42, the determination module 44, the composition module 44, the prediction module 45 and the determination module 46 correspond to steps S202 to S212 in the embodiment, and the examples and application scenarios implemented by the multiple modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above modules, as part of the device, can be run in the computer terminal 10 provided in the embodiment.

[0107] An embodiment of the present invention may provide a computer device. Optionally, in this embodiment, the computer device may be located in at least one network device among multiple network devices of a computer network. The computer device includes a memory and a processor.

[0108] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the photovoltaic power generation prediction method and device in the embodiment of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned photovoltaic power generation prediction method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories can be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0109] The processor can call the information and application program stored in the memory through the transmission device to perform the following steps: obtain the historical power generation data of each of the multiple photovoltaic power generation devices included in the photovoltaic power station within the historical time period; group the multiple photovoltaic power generation devices according to the historical power generation data of each of the multiple photovoltaic power generation devices to obtain multiple groups of photovoltaic power generation devices; determine the power generation power of each of the multiple groups of photovoltaic power generation devices according to the historical power generation data of each of the multiple photovoltaic power generation devices; construct multiple feature maps according to the power generation power of each of the multiple groups of photovoltaic power generation devices, wherein the multiple feature maps correspond to the multiple groups of photovoltaic power generation devices one by one, and any one of the multiple feature maps is used to characterize the association between the multiple photovoltaic devices in the corresponding group of photovoltaic power generation devices; input the multiple feature maps into the multiple pre-trained convolutional neural network models respectively, and the multiple convolutional neural network models output the predicted power of each of the multiple groups of photovoltaic power generation devices respectively, wherein the convolutional neural network model is trained by training samples, and the training samples include sample feature maps and sample powers; determine the power generation power of the photovoltaic power station according to the predicted power of each of the multiple groups of photovoltaic power generation devices.

[0110] Optionally, according to the historical power generation data of each of the multiple photovoltaic power generation devices, the multiple photovoltaic power generation devices are grouped to obtain multiple groups of photovoltaic power generation devices, including: obtaining a preset number of groups; selecting multiple clustering centers corresponding to the number of groups from the multiple photovoltaic power generation devices; according to the historical power generation data of each of the multiple photovoltaic power generation devices, based on the multiple clustering centers, a clustering algorithm is used to group the multiple photovoltaic power generation devices to obtain multiple groups of photovoltaic power generation devices.

[0111] Optionally, multiple feature maps are constructed according to the power generation of each of the multiple groups of photovoltaic power generation equipment, including: adopting the following method to construct a target feature map according to the power generation of the target group of photovoltaic power generation equipment, and constructing multiple feature maps according to the power generation of each of the multiple groups of photovoltaic power generation equipment: determining the power generation of each of the target group of photovoltaic power generation equipment in multiple time periods according to the power generation of each of the target group of photovoltaic power generation equipment; generating a spatial feature map for characterizing the spatial relationship of the devices included in the target group of photovoltaic power generation equipment in each time period according to the power generation of each of the target group of photovoltaic power generation equipment in multiple time periods, wherein the multiple nodes included in the spatial feature map respectively characterize each of the photovoltaic power generation devices in the target group of photovoltaic power generation equipment; constructing a target feature map according to the spatial feature maps of each of the multiple time periods.

[0112] Optionally, a target feature map is constructed based on the spatial feature maps of each of the multiple time periods, including: sorting the spatial feature maps of each of the multiple time periods according to the time sequence of the multiple time periods; and connecting the sorted spatial feature maps of each of the multiple time periods to obtain the target feature map.

[0113] Optionally, the spatial feature maps of the sorted multiple time periods are connected to obtain a target feature map, including: connecting each group of adjacent spatial feature maps in the spatial feature maps of the sorted multiple time periods to obtain a feature map in the following manner: in two adjacent spatial feature maps, matching nodes are identified and the matching nodes are connected, wherein the matching nodes represent the same photovoltaic power generation device.

[0114] Optionally, multiple feature maps are respectively input into multiple pre-trained convolutional neural network models, and the multiple convolutional neural network models respectively output the predicted power of multiple groups of photovoltaic power generation equipment, including: multiple feature maps are respectively input into multiple pre-trained convolutional neural network models, each of the multiple convolutional neural network models segments the input feature map to obtain multiple corresponding local feature maps; each of the multiple convolutional neural network models generates an adjacency matrix corresponding to the multiple local feature maps, wherein the adjacency matrix represents the degree of influence of adjacent nodes in the local feature map on the central node; each of the convolutional neural network models performs convolution operations on the multiple local feature maps according to the adjacency matrices corresponding to the multiple local feature maps to obtain the predicted power of the corresponding group of photovoltaic power generation equipment; and the multiple convolutional neural network models respectively output the predicted power of multiple groups of photovoltaic power generation equipment.

[0115] By adopting the embodiment of the present invention, a scheme for predicting photovoltaic power generation is provided. The historical power generation data of each of the multiple photovoltaic power generation devices included in the photovoltaic power station within the historical time period is obtained; the multiple photovoltaic power generation devices are grouped according to the historical power generation data of each of the multiple photovoltaic power generation devices to obtain multiple groups of photovoltaic power generation devices; the power generation of each of the multiple groups of photovoltaic power generation devices is determined according to the historical power generation data of each of the multiple photovoltaic power generation devices; multiple feature maps are constructed according to the power generation of each of the multiple groups of photovoltaic power generation devices, wherein the multiple feature maps correspond to the multiple groups of photovoltaic power generation devices one by one, and any one of the multiple feature maps is used to characterize the association between the multiple photovoltaic devices in the corresponding group of photovoltaic power generation devices; the multiple feature maps are respectively input into multiple pre-trained convolutional neural network models, and ... The convolutional neural network models respectively output the predicted power of multiple groups of photovoltaic power generation equipment, wherein the convolutional neural network model is trained by training samples, and the training samples include sample feature maps and sample powers; according to the predicted powers of multiple groups of photovoltaic power generation equipment, the power generation power of the photovoltaic power station is determined. Since the feature map can extract the correlation between the outputs of different photovoltaic power generation equipment, the convolutional neural network model can fully learn the hidden features of the output data of the photovoltaic power generation equipment, thereby achieving the purpose of minimizing the parameter adjustment cost while ensuring the prediction accuracy, thereby achieving the technical effect of improving the efficiency of photovoltaic power generation prediction power, and then solving the technical problem of low efficiency of photovoltaic power generation prediction power in related technologies.

[0116] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a non-volatile storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0117] The embodiment of the present invention further provides a non-volatile storage medium. Optionally, in this embodiment, the non-volatile storage medium can be used to store the program code executed by the photovoltaic power prediction method provided in the above embodiment.

[0118] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0119] Optionally, in this embodiment, the non-volatile storage medium is configured to store program codes for executing the following steps: obtaining historical power generation data of each of the multiple photovoltaic power generation devices included in the photovoltaic power station within a historical time period; grouping the multiple photovoltaic power generation devices according to the historical power generation data of each of the multiple photovoltaic power generation devices to obtain multiple groups of photovoltaic power generation devices; determining the power generation power of each of the multiple groups of photovoltaic power generation devices according to the historical power generation data of each of the multiple photovoltaic power generation devices; constructing multiple feature maps according to the power generation power of each of the multiple groups of photovoltaic power generation devices, wherein the multiple feature maps correspond one-to-one to the multiple groups of photovoltaic power generation devices, and any one of the multiple feature maps is used to characterize the association between the multiple photovoltaic devices in the corresponding group of photovoltaic power generation devices; inputting the multiple feature maps into multiple pre-trained convolutional neural network models, respectively, and outputting the predicted power of each of the multiple groups of photovoltaic power generation devices respectively by the multiple convolutional neural network models, wherein the convolutional neural network models are trained by training samples, and the training samples include sample feature maps and sample powers; determining the power generation power of the photovoltaic power station according to the predicted power of each of the multiple groups of photovoltaic power generation devices.

[0120] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the photovoltaic power prediction method in each embodiment of the present application.

[0121] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0122] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0123] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0124] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0125] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0126] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or 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, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0127] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A photovoltaic power generation power prediction method, characterized in that: include: Obtaining historical power generation data of each of the multiple photovoltaic power generation devices included in the photovoltaic power station within a historical time period; According to the historical power generation data of each of the plurality of photovoltaic power generation devices, the plurality of photovoltaic power generation devices are grouped to obtain a plurality of groups of photovoltaic power generation devices; Determining the power generation of each of the plurality of groups of photovoltaic power generation equipment according to the respective historical power generation data of the plurality of photovoltaic power generation equipment; According to the power generation powers of the multiple groups of photovoltaic power generation equipment, respectively construct multiple characteristic graphs, wherein the multiple characteristic graphs correspond to the multiple groups of photovoltaic power generation equipment one by one, and any one of the multiple characteristic graphs is used to characterize the association between the multiple photovoltaic devices in the corresponding group of photovoltaic power generation equipment; Inputting the plurality of feature maps into a plurality of pre-trained convolutional neural network models respectively, and having the plurality of convolutional neural network models output the respective predicted powers of the plurality of groups of photovoltaic power generation devices respectively, wherein the convolutional neural network models are trained by training samples, and the training samples include sample feature maps and sample powers; The power generation power of the photovoltaic power station is determined according to the predicted power of each of the multiple groups of photovoltaic power generation equipment.

2. The method according to claim 1, characterized in that The method of grouping the plurality of photovoltaic power generation devices according to the respective historical power generation data of the plurality of photovoltaic power generation devices to obtain a plurality of groups of photovoltaic power generation devices comprises: Get the preset number of groups; Selecting a plurality of cluster centers corresponding to the number of groups from the plurality of photovoltaic power generation devices; According to the historical power generation data of each of the plurality of photovoltaic power generation devices and based on the plurality of cluster centers, a clustering algorithm is adopted to group the plurality of photovoltaic power generation devices to obtain a plurality of groups of photovoltaic power generation devices.

3. The method according to claim 1, characterized in that: The method of constructing a plurality of characteristic graphs according to the power generation powers of the plurality of groups of photovoltaic power generation devices respectively comprises: The following method of constructing a target characteristic graph according to the power generation power of the target group of photovoltaic power generation equipment is adopted, and a plurality of characteristic graphs are respectively constructed according to the power generation power of each of the plurality of groups of photovoltaic power generation equipment: Determining the power generation of each photovoltaic power generation device in the target group in multiple time periods according to the power generation of each photovoltaic power generation device in the target group; According to the power generation of each of the target group of photovoltaic power generation devices in multiple time periods, a spatial feature graph for characterizing the spatial relationship of the devices included in the target group of photovoltaic power generation devices is generated in each time period, wherein the multiple nodes included in the spatial feature graph respectively characterize each photovoltaic power generation device in the target group of photovoltaic power generation devices; The target feature map is constructed according to the spatial feature maps of each of the multiple time periods.

4. The method according to claim 3, characterized in that The step of constructing the target feature map according to the spatial feature maps of the multiple time periods includes: According to the time sequence of the multiple time periods, the spatial feature graphs of the multiple time periods are sorted; The spatial feature maps of the sorted multiple time periods are connected to obtain the target feature map.

5. The method according to claim 4, characterized in that The step of connecting the spatial feature maps of the sorted multiple time periods to obtain the target feature map includes: The feature map is obtained by connecting each group of adjacent spatial feature maps in the spatial feature maps of the multiple time periods after sorting: In two adjacent spatial feature graphs, matching nodes are identified and connected, wherein the matching nodes represent the same photovoltaic power generation device.

6. The method according to any one of claims 1 to 5, characterized in that The step of inputting the plurality of feature maps into a plurality of pre-trained convolutional neural network models respectively, and having the plurality of convolutional neural network models respectively output the respective predicted powers of the plurality of groups of photovoltaic power generation equipment comprises: The multiple feature maps are respectively input into the multiple pre-trained convolutional neural network models, and each of the multiple convolutional neural network models segments the input feature map to obtain a corresponding multiple local feature maps; each of the multiple convolutional neural network models generates an adjacency matrix corresponding to the multiple local feature maps, wherein the adjacency matrix represents the degree of influence of adjacent nodes in the local feature map on the central node; each of the convolutional neural network models performs convolution operations on the multiple local feature maps according to the adjacency matrices corresponding to the multiple local feature maps to obtain the predicted power of each of the corresponding groups of photovoltaic power generation equipment; The multiple convolutional neural network models respectively output the predicted powers of the multiple groups of photovoltaic power generation equipment.

7. A photovoltaic power generation power prediction device, characterized in that: include: An acquisition module, used to acquire historical power generation data of each of a plurality of photovoltaic power generation devices included in a photovoltaic power station within a historical time period; A grouping module, used for grouping the plurality of photovoltaic power generation devices according to the respective historical power generation data of the plurality of photovoltaic power generation devices to obtain a plurality of groups of photovoltaic power generation devices; A determination module, configured to determine the power generation power of each of the plurality of groups of photovoltaic power generation equipment according to the respective historical power generation data of the plurality of photovoltaic power generation equipment; A mapping module, used to construct a plurality of characteristic graphs according to the power generation powers of the plurality of groups of photovoltaic power generation devices, wherein the plurality of characteristic graphs correspond to the plurality of groups of photovoltaic power generation devices one by one, and any one of the plurality of characteristic graphs is used to characterize the association between the plurality of photovoltaic devices in the corresponding group of photovoltaic power generation devices; A prediction module, used for inputting the plurality of feature maps into a plurality of pre-trained convolutional neural network models respectively, and having the plurality of convolutional neural network models output the respective predicted powers of the plurality of groups of photovoltaic power generation equipment respectively, wherein the convolutional neural network models are trained by training samples, and the training samples include sample feature maps and sample powers; The determination module is used to determine the power generation power of the photovoltaic power station according to the predicted power of each of the multiple groups of photovoltaic power generation equipment.

8. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute the photovoltaic power generation power prediction method according to any one of claims 1 to 6.

9. A computer device, characterized in that: include: Memory and processor, The memory stores a computer program; The processor is used to execute the computer program stored in the memory, and when the computer program is running, the processor executes the photovoltaic power prediction method according to any one of claims 1 to 6.

10. A computer program product comprising computer instructions, characterized in that: The computer instructions are executed by the processor to implement the photovoltaic power prediction method described in any one of claims 1 to 6.

Images