Photovoltaic ultra-short-term power prediction method, device, computer equipment and storage medium

Abstract

The present invention relates to the field of new energy technology, and discloses a photovoltaic ultra-short-term power prediction method, device, computer equipment and storage medium; the photovoltaic ultra-short-term power prediction method comprises: acquiring historical meteorological data at a preset spatial scale; performing meteorological element extrapolation based on the historical meteorological data to obtain extrapolated cloud map data and extrapolated irradiance data; performing spatiotemporal matching on the extrapolated cloud map data and the extrapolated irradiance data based on the geographical location attributes and preset time scale of each distributed photovoltaic site to obtain the extrapolated meteorological data of each distributed photovoltaic site; inputting the extrapolated meteorological data into a pre-trained photovoltaic ultra-short-term power prediction model to obtain the power prediction result of each distributed photovoltaic site; the above method performs spatiotemporal matching on the historical meteorological data to obtain the meteorological data of each distributed photovoltaic site, which can effectively solve the problem of lack of meteorological measurement meteorological data.

Description

Technical Field

[0001] The present invention relates to the field of new energy technology, and in particular to a photovoltaic ultra-short-term power prediction method, device, computer equipment and storage medium. Background Art

[0002] Distributed photovoltaics are usually lacking in meteorological measurement devices due to their scattered installation locations. It is impossible to find meteorological information around most distributed photovoltaic power stations, and it is impossible to obtain local meteorological data such as irradiance. Therefore, they face the challenge of power prediction without meteorological data. In addition, there are complex weather phenomena such as cloudy, foggy, hail, overcast, rainy, and snowy weather, which make it difficult to accurately predict photovoltaic power generation. Summary of the invention

[0003] In view of this, the present invention provides a photovoltaic ultra-short-term power prediction method, device, computer equipment and storage medium to solve the problem that the power of distributed photovoltaic power stations is difficult to accurately predict due to the lack of meteorological conditions.

[0004] In a first aspect, the present invention provides a photovoltaic ultra-short-term power prediction method, which comprises: obtaining historical meteorological data at a preset spatial scale and historical power data of each distributed photovoltaic site; wherein the historical meteorological data comprises historical cloud map data and historical irradiance data, and the preset spatial scale includes the geographical location of each distributed photovoltaic site; performing meteorological element extrapolation based on the historical meteorological data to obtain extrapolated cloud map data and extrapolated irradiance data; performing spatiotemporal matching on the extrapolated cloud map data and the extrapolated irradiance data based on the geographical location attributes and preset time scale of each distributed photovoltaic site to obtain the extrapolated meteorological data of each distributed photovoltaic site; and inputting the extrapolated meteorological data into a pre-trained photovoltaic ultra-short-term power prediction model to obtain a power prediction result for each distributed photovoltaic site.

[0005] As an exemplary embodiment, the extrapolated cloud map data and the extrapolated irradiance data are temporally and spatially matched based on the geographical location attributes and the preset time scale of each of the distributed photovoltaic sites to obtain the extrapolated meteorological data of each of the distributed photovoltaic sites, including: spatially matching the geographical location attributes of each of the distributed photovoltaic sites with the degree of proximity matching with the extrapolated cloud map data and the extrapolated irradiance data to obtain the spatial matching result of each distributed photovoltaic site; numerically interpolating the extrapolated cloud map data and the extrapolated irradiance data based on the preset time scale to obtain the interpolated cloud map data and the interpolated irradiance data; and extracting meteorological data from the interpolated cloud map data and the interpolated irradiance data as the extrapolated meteorological data based on the spatial matching result.

[0006] As an exemplary embodiment, meteorological data is extracted from the interpolated cloud map data and the interpolated irradiance data as the extrapolated meteorological data based on the spatial matching result, including: extracting cloud map grayscale values ​​of the extrapolated cloud map data; performing weather classification on the extrapolated cloud map data based on the cloud map grayscale values ​​to obtain weather type grid data; and extracting interpolated cloud amount data from the interpolated cloud map data, weather type grid data from the weather type grid data, and interpolated irradiance data from the interpolated irradiance data as the extrapolated meteorological data based on the spatial matching result.

[0007] As an exemplary embodiment, the photovoltaic ultra-short-term power prediction model is obtained based on space-time graph neural network training. Before the extrapolated meteorological data is input into the pre-trained photovoltaic ultra-short-term power prediction model, the photovoltaic ultra-short-term power prediction method also includes: extracting location features of each of the distributed photovoltaic sites to obtain location features and relationship features; wherein the location features are used to represent the geographical locations of each of the distributed photovoltaic sites, and the relationship features are used to represent the similarity and distance of each of the distributed photovoltaic sites; determining the nodes constituting the topological graph based on the location features; determining the edges constituting the topological graph based on the relationship features; adding the extrapolated meteorological data to the nodes of the topological graph and the edges of the topological graph to obtain the topological graph; inputting the topological graph into the photovoltaic ultra-short-term power prediction model to obtain a power prediction result.

[0008] As an exemplary embodiment, the photovoltaic ultra-short-term power prediction model includes a graph learning layer, a graph convolution module, a time convolution module and an output module. The topological map is input into the photovoltaic ultra-short-term power prediction model to obtain a power prediction result, including: based on the graph learning layer, capturing the hidden relationship between the extrapolated meteorological data and the historical power data contained in each node of the topological map to generate a graph adjacency matrix; based on the graph convolution module and the time convolution module, interleavingly capturing the spatial and temporal dependencies of the graph adjacency matrix, obtaining extrapolated power data based on the captured spatial and temporal dependencies, and outputting the extrapolated meteorological data to the output module; based on the output module, projecting the extrapolated power data to a preset dimension to output the power prediction result of each distributed photovoltaic site.

[0009] As an exemplary embodiment, the method for training the photovoltaic ultra-short-term power prediction model includes: obtaining historical spatiotemporal data of the distributed photovoltaic sites; the historical spatiotemporal data includes historical meteorological data and historical power data; performing location feature extraction on each of the distributed photovoltaic sites to obtain location features and relationship features; constructing a historical topological map based on the location features, the relationship features and the historical spatiotemporal data; inputting the historical topological map into a pre-constructed spatiotemporal graph neural network for model training, and during the model training process, adjusting the parameters of the spatiotemporal graph neural network so that the spatiotemporal graph neural network can learn to capture the hidden relationship between the historical meteorological data and the historical power data contained in each of the nodes to generate a graph adjacent matrix, the capture relationship when capturing the spatial and temporal dependencies of the graph adjacent matrix, the corresponding relationship between the captured spatial and temporal dependencies and the extrapolated meteorological data, and the mapping relationship when projecting the extrapolated meteorological data to a preset dimension.

[0010] In a second aspect, the present invention provides a photovoltaic ultra-short-term power prediction device, the device comprising: an acquisition module, used to acquire historical meteorological data at a preset spatial scale and historical power data of each distributed photovoltaic site; wherein the historical meteorological data includes historical cloud map data and historical irradiance data, and the preset spatial scale includes the geographical location of each distributed photovoltaic site; a meteorological element extrapolation module, used to perform meteorological element extrapolation based on the historical meteorological data to obtain extrapolated cloud map data and extrapolated irradiance data; a spatiotemporal matching module, used to perform spatiotemporal matching on the extrapolated cloud map data and the extrapolated irradiance data based on the geographical location attributes and preset time scale of each distributed photovoltaic site, to obtain the extrapolated meteorological data of each distributed photovoltaic site; a power prediction module, used to input the extrapolated meteorological data into a pre-trained photovoltaic ultra-short-term power prediction model to obtain a power prediction result for each distributed photovoltaic site.

[0011] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the photovoltaic ultra-short-term power prediction method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0012] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the photovoltaic ultra-short-term power prediction method of the first aspect or any corresponding embodiment thereof.

[0013] The present invention provides a photovoltaic ultra-short-term power prediction method, comprising: obtaining historical meteorological data at a preset spatial scale; wherein the historical meteorological data includes historical cloud map data and historical irradiance data, and the preset spatial scale includes the geographical location of each distributed photovoltaic site; meteorological elements are extrapolated based on the historical meteorological data to obtain extrapolated cloud map data and extrapolated irradiance data; based on the geographical location attributes and preset time scale of each distributed photovoltaic site, the extrapolated cloud map data and the extrapolated irradiance data are temporally and spatially matched to obtain the extrapolated meteorological data of each distributed photovoltaic site; the extrapolated meteorological data is input into a pre-trained photovoltaic ultra-short-term power prediction model The power prediction results of each distributed photovoltaic site are obtained in the model; the above method can perform spatiotemporal matching of the meteorological data obtained by extrapolation according to the preset time scale of the geographical location attributes of the distributed photovoltaic site to adapt to the spatial attributes and time attributes of the distributed photovoltaic site, and obtain the meteorological data of each distributed photovoltaic site, which can effectively solve the problem of lack of meteorological measurement data due to the dispersed distribution of photovoltaics; the method of obtaining the power prediction results of each distributed photovoltaic site by inputting the extrapolated meteorological data into a pre-trained photovoltaic ultra-short-term power prediction model can reduce the impact of low accuracy of distributed photovoltaic ultra-short-term power prediction due to lack of measured meteorological data. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 is a schematic flow chart of a photovoltaic ultra-short-term power prediction method according to an embodiment of the present invention;

[0016] Figure 2 is a structural block diagram of a photovoltaic ultra-short-term power prediction device according to an embodiment of the present invention;

[0017] Figure 3 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0019] According to an embodiment of the present invention, an embodiment of a photovoltaic ultra-short-term power prediction method 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.

[0020] In this embodiment, a photovoltaic ultra-short-term power prediction method is provided. Figure 1 is a flow chart of a photovoltaic ultra-short-term power prediction method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0021] Step S101, obtaining historical meteorological data and historical power data of each distributed photovoltaic site at a preset spatial scale; wherein the historical meteorological data includes historical cloud map data and historical irradiance data, and the preset spatial scale includes the geographical location of each distributed photovoltaic site.

[0022] Due to the scattered installation locations of distributed photovoltaics, the relevant photovoltaic sites usually lack meteorological measurement devices, and meteorological information cannot be found around most distributed photovoltaic power stations. To solve this problem, in this embodiment, historical meteorological data at a preset spatial scale including the location of each distributed photovoltaic site is obtained, and then time-space matching is performed according to the location of the distributed photovoltaics and the meteorological data at the preset spatial scale, so as to obtain the meteorological data corresponding to each distributed photovoltaic.

[0023] Among them, historical meteorological data include satellite cloud image data and satellite irradiance data.

[0024] Step S102, extrapolating meteorological elements based on the historical meteorological data to obtain extrapolated cloud map data and extrapolated irradiance data.

[0025] In this embodiment, the extrapolated satellite cloud image data and satellite irradiance data at a preset spatial scale including each distributed photovoltaic site are determined by extrapolating meteorological elements, so as to effectively solve the problem of lack of meteorological measurement data due to the dispersed distribution of photovoltaics.

[0026] Exemplarily, the extrapolation of satellite cloud images can be achieved by implementing at least one of the generative learning algorithms such as optical flow method, PredRNN++, Res-UNet, conditional adversarial network, etc.

[0027] Exemplarily, when adopting the implementation method of the above-mentioned generative learning algorithm, the time series data of the previous 4 hours can be used as features, the time series data of the next 4 hours can be used as labels, and every 10 minutes can be used as the reporting start time. The meteorological element extrapolation model including the irradiance extrapolation sub-model and the cloud map extrapolation sub-model can be trained respectively. Then, the meteorological element extrapolation model can be used to generate the extrapolated irradiance and extrapolated cloud map every 10 minutes in the next 4 hours with the irradiance and cloud map time series data 4 hours before the reporting start time, thereby obtaining the extrapolated cloud map data and the extrapolated irradiance data.

[0028] In one embodiment, the extrapolated cloud image data and the extrapolated irradiance data are grid data.

[0029] In one embodiment, before meteorological elements are extrapolated from historical meteorological data, the historical meteorological data is preprocessed by atmospheric correction, data standardization, background color removal, missing time filling, etc.

[0030] Step S103, performing spatiotemporal matching on the extrapolated cloud map data and the extrapolated irradiance data based on the geographical location attributes and preset time scale of each of the distributed photovoltaic sites, to obtain the extrapolated meteorological data of each of the distributed photovoltaic sites.

[0031] The extrapolated cloud map data and extrapolated irradiance data of each distributed photovoltaic site at a preset spatial scale are discretely distributed. Therefore, it is necessary to spatially match the extrapolated cloud map data and the extrapolated irradiance data based on the geographical location attributes of each distributed photovoltaic site; and, when performing photovoltaic ultra-short-term power forecasting, the time resolution is 15 minutes, and the preset time scale of the 15-minute time resolution does not match the time scale of the extrapolated cloud map data and the extrapolated irradiance data. Therefore, it is necessary to time match the extrapolated cloud map data and the extrapolated irradiance data based on the preset time scale of each distributed photovoltaic site.

[0032] Based on this, in this embodiment, the extrapolated cloud map data and the extrapolated irradiance data are temporally and spatially matched based on the geographical location attributes and preset time scale of each distributed photovoltaic site to obtain the extrapolated meteorological data of each distributed photovoltaic site.

[0033] In one embodiment, the extrapolated cloud map data and extrapolated irradiance data at a preset spatial scale may be spatially downscaled based on the geographical location attributes of each distributed photovoltaic site to obtain meteorological data that matches the space of each distributed photovoltaic site.

[0034] In one embodiment, for the extrapolated cloud map data and the extrapolated irradiance data at a preset spatial scale, the photovoltaic sites can be spatially matched with the grids using a proximity matching method based on the geographical location attributes of each distributed photovoltaic site.

[0035] In one embodiment, when time matching is performed on the extrapolated cloud map data and the extrapolated irradiance data based on a preset time scale of each of the distributed photovoltaic sites, the extrapolated cloud map data and the extrapolated irradiance data can be numerically interpolated based on the preset time scale to obtain interpolated cloud map data and interpolated irradiance data as extrapolated meteorological data.

[0036] Specifically, when performing numerical interpolation, the extrapolated cloud map data and extrapolated irradiance data are interpolated using the average of the previous and next moments to form meteorological data for every 15 minutes in the next 4 hours with a reporting time of every 15 minutes to complete time matching.

[0037] The above implementation mode of this embodiment performs spatiotemporal matching of the extrapolated cloud map data and the extrapolated irradiance data based on the geographical location attributes and preset time scale of each of the distributed photovoltaic sites to obtain the extrapolated meteorological data of each of the distributed photovoltaic sites. It can perform spatiotemporal matching of the meteorological data extrapolated according to the preset time scale of the geographical location attributes of the distributed photovoltaic sites to adapt to the spatial attributes and time attributes of the distributed photovoltaic sites to obtain the meteorological data of each distributed photovoltaic site, which can effectively solve the problem of lack of meteorological measurement and meteorological data due to the dispersed distribution of photovoltaics.

[0038] Step S104: input the extrapolated meteorological data into a pre-trained photovoltaic ultra-short-term power prediction model to obtain power prediction results for each of the distributed photovoltaic sites.

[0039] In this embodiment, the photovoltaic ultra-short-term power prediction model is used to perform ultra-short-term power prediction based on the extrapolated meteorological data of each distributed photovoltaic site.

[0040] In one embodiment, the photovoltaic ultra-short-term power prediction model may be a deep learning algorithm model obtained by training a data set consisting of historical meteorological data and historical power data; specifically, historical meteorological data at a preset spatial scale and historical power data of each distributed photovoltaic site are obtained, and the historical meteorological data are matched based on the geographical location attributes and preset time scale of each distributed photovoltaic site to obtain training meteorological data of each distributed photovoltaic site; further, the training meteorological data and historical power data are input as training sets into a pre-built deep learning algorithm model for training, and during the model training process, the correspondence between the training meteorological data and the historical power data is learned.

[0041] In one embodiment, the photovoltaic ultra-short-term power prediction model can be a spatiotemporal graph neural network (MTGNN) trained with a data set consisting of historical meteorological data and historical power data. It is good at predicting irregular graph distributions of similar distributed photovoltaic points and can capture the time and space dependencies of data. Specifically, the spatiotemporal graph neural network can be trained based on a topological graph constructed from historical spatiotemporal data of each distributed photovoltaic site.

[0042] In this embodiment, by inputting the extrapolated meteorological data into a pre-trained photovoltaic ultra-short-term power prediction model, a method for obtaining power prediction results for each distributed photovoltaic site can reduce the impact of low accuracy of distributed photovoltaic ultra-short-term power prediction due to lack of measured meteorological data.

[0043] The photovoltaic ultra-short-term power prediction method of this embodiment obtains historical meteorological data at a preset spatial scale and historical power data of each distributed photovoltaic site; wherein the historical meteorological data includes historical cloud map data and historical irradiance data, and the preset spatial scale includes the geographical location of each distributed photovoltaic site; meteorological elements are extrapolated based on the historical meteorological data to obtain extrapolated cloud map data and extrapolated irradiance data; based on the geographical location attributes and preset time scale of each distributed photovoltaic site, the extrapolated cloud map data and the extrapolated irradiance data are temporally and spatially matched to obtain the extrapolated meteorological data of each distributed photovoltaic site; the extrapolated meteorological data is input into a pre-trained photovoltaic ultra-short-term power prediction method. The power prediction model of the distributed photovoltaic site is used to obtain the power prediction results of each distributed photovoltaic site; the above method can perform spatiotemporal matching of the meteorological data obtained by extrapolating according to the preset time scale of the geographical location attributes of the distributed photovoltaic site to adapt to the spatial attributes and time attributes of the distributed photovoltaic site, so as to obtain the meteorological data of each distributed photovoltaic site, which can effectively solve the problem of lack of meteorological measurement data due to the dispersed distribution of photovoltaic power plants; the method of obtaining the power prediction results of each distributed photovoltaic site by inputting the extrapolated meteorological data into the pre-trained photovoltaic ultra-short-term power prediction model can reduce the impact of low accuracy of distributed photovoltaic ultra-short-term power prediction due to lack of measured meteorological data.

[0044] As an exemplary embodiment, the extrapolated cloud map data and the extrapolated irradiance data are temporally and spatially matched based on the geographical location attributes and the preset time scale of each of the distributed photovoltaic sites to obtain the extrapolated meteorological data of each of the distributed photovoltaic sites, including: spatially matching the geographical location attributes of each of the distributed photovoltaic sites with the proximity matching degree of the extrapolated cloud map data and the extrapolated irradiance data to obtain the spatial matching result of each distributed photovoltaic site; numerically interpolating the extrapolated cloud map data and the extrapolated irradiance data based on the preset time scale to obtain the interpolated cloud map data and the interpolated irradiance data; and extracting meteorological data from the interpolated cloud map data and the interpolated irradiance data as the extrapolated meteorological data based on the spatial matching result.

[0045] Exemplarily, when extracting meteorological data from the interpolated cloud map data and the interpolated irradiance data based on the spatial matching result, the meteorological data of each distributed photovoltaic site and the meteorological data between each distributed photovoltaic site may be extracted based on the spatial matching result.

[0046] In this embodiment, the photovoltaic sites are spatially matched with the grids using a proximity matching method based on the geographical location attributes of each distributed photovoltaic site.

[0047] In this embodiment, the extrapolated cloud map data and the extrapolated irradiance data are numerically interpolated based on the preset time scale to obtain interpolated cloud map data and interpolated irradiance data; specifically, when performing numerical interpolation, the average of the previous and next moments is used to interpolate the extrapolated cloud map data and the extrapolated irradiance data into interpolated cloud map data and interpolated irradiance data for every 15 minutes in the next 4 hours with a starting time of every 15 minutes.

[0048] When performing ultra-short-term photovoltaic power forecasting, complex weather phenomena such as non-sunny weather such as cloudy, foggy, hail, overcast, rainy, snowy, etc. also have an impact on the ultra-short-term photovoltaic power forecasting. Failure to consider weather phenomena when performing power forecasting will lead to difficulty in accurately predicting photovoltaic power generation. In the present invention, the impact of weather phenomena on the ultra-short-term photovoltaic power forecasting is considered through satellite cloud map data.

[0049] Specifically, as an exemplary embodiment, meteorological data is extracted from the interpolated cloud map data and the interpolated irradiance data as the extrapolated meteorological data based on the spatial matching result, including: extracting the cloud map grayscale value of the extrapolated cloud map data; performing weather classification on the extrapolated cloud map data based on the cloud map grayscale value to obtain weather type grid data; and extracting interpolated cloud amount data from the interpolated cloud map data, weather type grid data from the weather type grid data, and interpolated irradiance data from the interpolated irradiance data as the extrapolated meteorological data based on the spatial matching result.

[0050] In this embodiment, after obtaining the extrapolated cloud map data, the cloud map grayscale value of the extrapolated cloud map data is extracted; the extrapolated cloud map data is divided into weather types based on the cloud map grayscale value to obtain weather type grid data; specifically, the cloud map grayscale value is mapped to a numerical value of 0-1 through the extrapolated extrapolated cloud map data, and the cloud type is divided according to the different cloud amount thresholds at which the numerical value lies; in one embodiment, the cloud types include 11 cloud types, including clear sky cumulus, cirrocumulus, cirrus, stratus, thin stratocumulus, high-level clouds, thick stratocumulus, cumulus, cirrostratus, cumulonimbus and nimbostratus, as well as clear sky, and different weather types are mapped according to the cloud types.

[0051] Specifically, Table 1 is a mapping table of cloud types, cloud ranges, and weather phenomena:

[0052]

[0053] Table 1. Mapping table of cloud types, cloud ranges and weather phenomena

[0054] In the above implementation mode, weather type characteristics are mapped out by extrapolating cloud image data.

[0055] Furthermore, based on the spatial matching result, interpolated cloud amount data are extracted from the interpolated cloud map data, weather type grid data are extracted from the weather type grid data, and interpolated irradiance data are extracted from the interpolated irradiance data as the extrapolated meteorological data, so as to solve the problem that photovoltaic power generation power is difficult to accurately predict when power prediction is performed without considering weather phenomena.

[0056] As an exemplary embodiment, the photovoltaic ultra-short-term power prediction model is obtained based on space-time graph neural network training. Before the extrapolated meteorological data is input into the pre-trained photovoltaic ultra-short-term power prediction model, the photovoltaic ultra-short-term power prediction method also includes: extracting location features of each of the distributed photovoltaic sites to obtain location features and relationship features; wherein the location features are used to represent the geographical locations of each of the distributed photovoltaic sites, and the relationship features are used to represent the similarities and distances of each of the distributed photovoltaic sites; determining the nodes constituting a topological graph based on the location features; determining the edges constituting the topological graph based on the relationship features; adding the extrapolated meteorological data to the nodes of the topological graph and the edges of the topological graph to obtain the topological graph; inputting the topological graph into the photovoltaic ultra-short-term power prediction model to obtain a power prediction result.

[0057] Since MTGNN is good at predicting irregular graph distributions of similar distributed photovoltaic points and can capture the temporal and spatial dependencies of meteorological data, a topological map is first constructed based on the location characteristics, relationship characteristics, meteorological data, and power data of each distributed photovoltaic site, and the topological map is further input into MTGNN to obtain the power prediction results.

[0058] Exemplarily, when constructing a topological map, location features are extracted for each of the distributed photovoltaic sites to obtain location features and relationship features; wherein the location features are used to represent the geographical locations of each of the distributed photovoltaic sites, and the relationship features are used to represent the similarities and distances of each of the distributed photovoltaic sites; the nodes constituting the topological map are determined based on the location features; the edges constituting the topological map are determined based on the relationship features; the extrapolated meteorological data are added to the nodes of the topological map and the edges of the topological map to obtain the topological map.

[0059] In one embodiment, the extrapolated meteorological data is added to the nodes of the topological map and the edges of the topological map to obtain a topological map of a preset spatial scale; the nodes of the topological map include the location characteristics of each of the distributed photovoltaic sites and the first meteorological data of each of the distributed photovoltaic sites, and the edges of the topological map include the similarity characteristics, distance characteristics, second meteorological data based on the relationship characteristics, and meteorological change characteristics based on the relationship characteristics of each of the distributed photovoltaic stations; wherein the first meteorological data is used to represent the meteorological status of each of the distributed photovoltaic sites, the second meteorological data is used to represent the meteorological status between at least two of the distributed photovoltaic sites, and the meteorological change characteristics are used to represent the meteorological change status between at least two of the distributed photovoltaic sites.

[0060] As described above, since MTGNN is good at predicting irregular graph distributions of similar distributed photovoltaic points, it can capture the temporal and spatial dependencies of meteorological data; in one embodiment, the photovoltaic ultra-short-term power prediction model is used to extract the time series dependency relationship of the first meteorological data and the second meteorological data to obtain the first time series change feature and the second time series change feature; the photovoltaic ultra-short-term power prediction model is used to extract the spatial dependency relationship of the meteorological change feature to obtain the spatial dependency feature; power prediction is performed based on the first time series change feature, the second time series change feature and the spatial dependency feature to obtain a power prediction result.

[0061] In the above implementation mode, the photovoltaic ultra-short-term power prediction model captures the first time series variation characteristics of the first meteorological data changing with time, the second time series variation characteristics of the second meteorological data and the spatial dependence characteristics of the meteorological variation characteristics changing in space relative to the relationship characteristics, and performs power prediction based on the first time series variation characteristics, the second time series variation characteristics and the spatial dependence characteristics. On the one hand, it is possible to consider the time variation of the meteorological characteristics of each distributed photovoltaic itself for power prediction, and it is possible to consider the time variation of the meteorological characteristics between each distributed photovoltaic for power prediction; on the other hand, it is possible to consider the relationship characteristics of the meteorological characteristics of distributed photovoltaics with spatial distance and / or similarity for power prediction, and considering the time and space dependence of meteorological data has the advantage of more accurate prediction.

[0062] As an exemplary embodiment, the extrapolated meteorological data is added to the nodes of the topological map and the edges of the topological map to obtain the topological map, including: extracting the extrapolated meteorological data based on the location characteristics to obtain the first extrapolated meteorological data corresponding to each of the distributed photovoltaic sites; extracting the extrapolated meteorological data based on the relationship characteristics to obtain the second extrapolated meteorological data corresponding to the edges constituting the topological map; adding the first extrapolated meteorological data to the nodes of the topological map; extracting the meteorological change characteristics of the second extrapolated meteorological data that changes with the relationship characteristics; and adding the second extrapolated meteorological data and the meteorological change characteristics to the edges of the topological map.

[0063] As an exemplary embodiment, the photovoltaic ultra-short-term power prediction model includes a graph learning layer, a graph convolution module, a time convolution module and an output module. The topological map is input into the photovoltaic ultra-short-term power prediction model to obtain a power prediction result, including: based on the graph learning layer, capturing the hidden relationship between the extrapolated meteorological data and the historical power data contained in each node of the topological map to generate a graph adjacency matrix; based on the graph convolution module and the time convolution module, interleavingly capturing the spatial and temporal dependencies of the graph adjacency matrix, obtaining extrapolated power data based on the captured spatial and temporal dependencies, and outputting the extrapolated meteorological data to the output module; based on the output module, projecting the extrapolated power data to a preset dimension to output the power prediction result of each distributed photovoltaic site.

[0064] In this embodiment, the photovoltaic ultra-short-term power prediction model includes a graph learning layer, a graph convolution module, a time convolution module and an output module; wherein the graph learning layer is used to adaptively learn the graph adjacency matrix to capture the hidden relationship between time series data, and only calculate the pairwise relationship between node subsets to reduce the computational complexity; the graph convolution layer is used to integrate node and neighbor node information and extract spatial features; the time convolution layer is used to extract timing features; specifically, the graph convolution module and the time convolution module interleave to capture the spatial and temporal dependencies of the graph adjacency matrix, and obtain the extrapolated power data based on the captured spatial and temporal dependencies.

[0065] In one embodiment, the graph convolution layer includes two mix-hop propagation layers to solve the problem that the standard graph convolutional neural network can only learn the information between adjacent nodes but cannot learn the neighborhood mixed relationship; wherein the two mix-hops respectively process the information propagation and information selection of a single node, and finally the outputs of the two mix-hop modules are added to obtain the output of the graph convolution layer.

[0066] In one embodiment, the temporal convolution layer is composed of two 1-dimensional expanded Inception layers to extract temporal features. The output of an expanded Inception layer is followed by a tanh function as a filter, and a sigmoid branch is a gate circuit to control information weights. The design of the expanded Inception layer has two aspects, namely, using filters of various sizes and applying expanded convolution. The purpose of this is to discover temporal patterns of different ranges and to process temporal convolution modules of very long sequences.

[0067] Specifically, when the topological map is input into the photovoltaic ultra-short-term power prediction model, the graph learning layer captures the hidden relationship between the extrapolated meteorological data and the historical power data contained in each node of the topological map based on the graph learning layer to generate a graph adjacency matrix, which is later used as the input of all graph convolution modules, and the graph convolution modules are interleaved with the time convolution modules to capture the spatial and temporal dependencies respectively; wherein, in order to avoid the problem of gradient vanishing, a residual connection is added from the input of the time convolution module to the output of the graph convolution module, and a skip connection is added after each time convolution module; in order to obtain the final output, the output module projects the hidden features to the required output dimension to obtain the extrapolated power data based on the captured spatial and temporal dependencies.

[0068] In one embodiment, the output module projects the extrapolated power data to a preset dimension to obtain an ultra-short-term photovoltaic power prediction topology diagram.

[0069] As an exemplary embodiment, the method for training the photovoltaic ultra-short-term power prediction model includes: obtaining historical spatiotemporal data of the distributed photovoltaic sites; the historical spatiotemporal data includes historical meteorological data and historical power data; extracting location features of each distributed photovoltaic site to obtain location features and relationship features; constructing a historical topological map based on the location features, the relationship features and the historical spatiotemporal data; inputting the historical topological map into a pre-constructed spatiotemporal graph neural network for model training, and during the model training process, adjusting the parameters of the spatiotemporal graph neural network so that the spatiotemporal graph neural network can learn to capture the hidden relationship reflected by the first time series change feature contained in each of the nodes, learn to capture the hidden relationship of the second time series change feature and the spatial dependency feature contained in each of the edges to generate a capture generation relationship when generating a graph adjacent matrix, capture the capture relationship when capturing the spatial and temporal dependency of the graph adjacent matrix, the corresponding relationship between the captured spatial and temporal dependency and the extrapolated meteorological data, and the mapping relationship when projecting the extrapolated meteorological data to a preset dimension.

[0070] In this embodiment, supervised learning can be used to implement the training of the photovoltaic ultra-short-term power prediction model; specifically, after obtaining the historical spatiotemporal data of the distributed photovoltaic site, a historical topological map is constructed based on the historical spatiotemporal data, and the historical topological map is input into a pre-constructed MTGNN for model training; in the process of model training, the historical power data in the historical spatiotemporal data is used as a label, and the model parameters are optimized by minimizing the difference between the predicted value and the true value. The ultra-short-term predicted power and the actual power are calculated respectively by the root mean square error accuracy and the harmonic mean accuracy, and the prediction model performance is evaluated until the model converges.

[0071] This embodiment provides a photovoltaic ultra-short-term power prediction device, such as Figure 2 As shown, including:

[0072] The acquisition module 501 is used to acquire historical meteorological data at a preset spatial scale; wherein the historical meteorological data includes historical cloud map data and historical irradiance data, and the preset spatial scale includes the geographical location of each distributed photovoltaic site;

[0073] A meteorological element extrapolation module 502 is used to perform meteorological element extrapolation based on the historical meteorological data to obtain extrapolated cloud image data and extrapolated irradiance data;

[0074] A time-space matching module 503 is used to perform time-space matching on the extrapolated cloud image data and the extrapolated irradiance data based on the geographical location attributes and the preset time scale of each distributed photovoltaic site to obtain the extrapolated meteorological data of each distributed photovoltaic site;

[0075] The power prediction module 504 is used to input the extrapolated meteorological data into a pre-trained photovoltaic ultra-short-term power prediction model to obtain the power prediction results of each distributed photovoltaic site.

[0076] It should be noted here that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments.

[0077] It should be noted that the above modules as part of the device can be implemented by software or hardware, wherein the hardware environment includes a network environment.

[0078] An embodiment of the present invention also provides a computer device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus, the memory is used to store a computer program; the processor is used to execute a method in any of the above embodiments by running the computer program stored in the memory.

[0079] Figure 3 is a structural block diagram of an optional computer device according to an embodiment of the present application, such as Figure 3 As shown, it includes a processor 10, a communication interface 20, a memory 30 and a communication bus 40, wherein the processor 10, the communication interface 20 and the memory 30 communicate with each other through the communication bus 40, wherein:

[0080] A memory 30, for storing computer programs;

[0081] The processor 10 is used to implement the method of any of the above embodiments when executing the computer program stored in the memory 30.

[0082] Optionally, in this embodiment, the communication bus may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0083] The communication interface is used for communication between the above-mentioned computer device and other devices.

[0084] The memory may include RAM, or may include non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0085] The above-mentioned processor can be a general-purpose processor, which can include but not be limited to: CPU (Central Processing Unit), NP (Network Processor), etc.; it can also be DSP (Digital Signal Processing), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0086] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment will not be described in detail here.

[0087] It can be understood by those skilled in the art that Figure 3 The structure shown is for illustration only. The device for implementing any one of the methods in the above embodiments may be a terminal device, which may be a smart phone (such as an Android phone, an IOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, or other terminal devices. Figure 3 It does not limit the structure of the above electronic device. For example, the terminal device may also include Figure 3 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Figure 3 Different configurations shown.

[0088] 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 computer-readable storage medium, which can include: a flash drive, ROM, RAM, a magnetic disk or an optical disk, etc.

[0089] As an exemplary embodiment, the present application also provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute any one of the method steps of the present embodiment when running.

[0090] Optionally, in this embodiment, the above-mentioned storage medium can be used to execute the program code of the method steps of the embodiment of the present application.

[0091] Optionally, in this embodiment, the storage medium may be located on at least one network device among a plurality of network devices in the network shown in the above embodiment.

[0092] Optionally, in this embodiment, the storage medium is configured to store data for executing the method in the above embodiment.

[0093] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, which will not be described in detail in this embodiment.

[0094] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media that can store program codes, such as a USB flash drive, a ROM, a RAM, a mobile hard disk, a magnetic disk, or an optical disk.

[0095] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0096] If the integrated units in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in the above computer-readable storage medium. Based on this understanding, the technical solution of the present application, 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, which is stored in a storage medium and includes several instructions for enabling one or more computer devices (which may be personal computers, servers or network devices, etc.) to execute all or part of the steps of the method in the above embodiments.

[0097] In the several embodiments provided in the present application, it should be understood that the disclosed client can be implemented in other ways. Among them, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and 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.

[0098] 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 network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution provided in this embodiment.

[0099] In addition, each functional unit in each embodiment of the present application 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.

[0100] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0101] The above are only preferred implementations of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A photovoltaic ultra-short-term power prediction method, characterized in that: The photovoltaic ultra-short-term power prediction method comprises: Acquire historical meteorological data at a preset spatial scale; wherein the historical meteorological data includes historical cloud map data and historical irradiance data, and the preset spatial scale includes the geographical location of each distributed photovoltaic site; Extrapolating meteorological elements based on the historical meteorological data to obtain extrapolated cloud image data and extrapolated irradiance data; Based on the geographical location attributes and preset time scale of each distributed photovoltaic site, the extrapolated cloud map data and the extrapolated irradiance data are temporally and spatially matched to obtain the extrapolated meteorological data of each distributed photovoltaic site; The extrapolated meteorological data is input into a pre-trained photovoltaic ultra-short-term power prediction model to obtain power prediction results for each of the distributed photovoltaic sites.

2. The photovoltaic ultra-short-term power prediction method according to claim 1, characterized in that: The extrapolated cloud map data and the extrapolated irradiance data are temporally and spatially matched based on the geographical location attributes and the preset time scale of each distributed photovoltaic site to obtain the extrapolated meteorological data of each distributed photovoltaic site, including: Perform spatial matching based on the geographical location attributes of each of the distributed photovoltaic sites and the proximity matching degree of the extrapolated cloud map data and the extrapolated irradiance data to obtain spatial matching results of each of the distributed photovoltaic sites; Performing numerical interpolation on the extrapolated cloud image data and the extrapolated irradiance data based on the preset time scale to obtain interpolated cloud image data and interpolated irradiance data; Based on the spatial matching result, meteorological data is extracted from the interpolated cloud image data and the interpolated irradiance data as the extrapolated meteorological data.

3. The photovoltaic ultra-short-term power prediction method according to claim 2, characterized in that: Extracting meteorological data from the interpolated cloud image data and the interpolated irradiance data as the extrapolated meteorological data based on the spatial matching result includes: Extracting the cloud image grayscale value of the extrapolated cloud image data; Performing weather classification on the extrapolated cloud image data based on the grayscale value of the cloud image to obtain weather type grid data; Based on the spatial matching result, interpolated cloud amount data are extracted from the interpolated cloud map data, weather type grid data are extracted from the weather type grid data, and interpolated irradiance data are extracted from the interpolated irradiance data as the extrapolated meteorological data.

4. The photovoltaic ultra-short-term power prediction method according to claim 1, characterized in that: The photovoltaic ultra-short-term power prediction model is obtained based on the training of the spatiotemporal graph neural network. Before the extrapolated meteorological data is input into the pre-trained photovoltaic ultra-short-term power prediction model, the photovoltaic ultra-short-term power prediction method further includes: Extracting location features of each of the distributed photovoltaic sites to obtain location features and relationship features; wherein the location features are used to represent the geographical locations of each of the distributed photovoltaic sites, and the relationship features are used to represent the similarities and distances of each of the distributed photovoltaic sites; Determine the nodes constituting the topological graph based on the location features; Determining edges constituting the topological graph based on the relationship features; Adding the extrapolated meteorological data to the nodes of the topological graph and the edges of the topological graph to obtain the topological graph; The topological diagram is input into the photovoltaic ultra-short-term power prediction model to obtain a power prediction result.

5. The photovoltaic ultra-short-term power prediction method according to claim 4, characterized in that: The step of adding the extrapolated meteorological data to the nodes of the topological graph and the edges of the topological graph to obtain the topological graph comprises: Extracting the extrapolated meteorological data based on the location features to obtain first extrapolated meteorological data corresponding to each of the distributed photovoltaic sites; Extracting the extrapolated meteorological data based on the relationship feature to obtain second extrapolated meteorological data corresponding to the edge constituting the topological graph; adding the first extrapolated meteorological data to the node of the topological map; extracting meteorological change characteristics of the second extrapolated meteorological data as the relationship characteristics change; The second extrapolated meteorological data and the meteorological change characteristics are added to the edges of the topological graph.

6. The photovoltaic ultra-short-term power prediction method according to claim 4, characterized in that: The photovoltaic ultra-short-term power prediction model includes a graph learning layer, a graph convolution module, a time convolution module and an output module. The topological map is input into the photovoltaic ultra-short-term power prediction model to obtain a power prediction result, including: Capturing the hidden relationship between the extrapolated meteorological data included in each of the nodes of the topological graph based on the graph learning layer to generate a graph adjacency matrix; Capturing the spatial and temporal dependencies of the graph adjacent matrix based on the graph convolution module and the temporal convolution module, obtaining extrapolated power data based on the captured spatial and temporal dependencies, and outputting the extrapolated meteorological data to the output module; The extrapolated power data is projected to a preset dimension based on the output module to output the power prediction result of each distributed photovoltaic site.

7. The photovoltaic ultra-short-term power prediction method according to claim 6, characterized in that: The method for training the photovoltaic ultra-short-term power prediction model includes: Acquire historical spatiotemporal data of the distributed photovoltaic site; the historical spatiotemporal data includes historical meteorological data and historical power data; Extracting location features of each of the distributed photovoltaic sites to obtain location features and relationship features; Constructing a historical topological map based on the location features, the relationship features and the historical spatiotemporal data; The historical topological graph is input into a pre-built space-time graph neural network for model training. During the model training process, the parameters of the space-time graph neural network are adjusted so that the space-time graph neural network can learn to capture the hidden relationship between the historical meteorological data and the historical power data contained in each node to generate a graph adjacency matrix, the capture relationship when capturing the spatial and temporal dependencies of the graph adjacency matrix, the correspondence between the captured spatial and temporal dependencies and the extrapolated meteorological data, and the mapping relationship when projecting the extrapolated meteorological data to a preset dimension.

8. A photovoltaic ultra-short-term power prediction device, characterized in that: The device comprises: An acquisition module, used to acquire historical meteorological data at a preset spatial scale; wherein the historical meteorological data includes historical cloud map data and historical irradiance data, and the preset spatial scale includes the geographical location of each distributed photovoltaic site; A meteorological element extrapolation module is used to perform meteorological element extrapolation based on the historical meteorological data to obtain extrapolated cloud image data and extrapolated irradiance data; A time-space matching module, used to perform time-space matching on the extrapolated cloud map data and the extrapolated irradiance data based on the geographical location attributes and preset time scale of each distributed photovoltaic site, so as to obtain the extrapolated meteorological data of each distributed photovoltaic site; The power prediction module is used to input the extrapolated meteorological data into a pre-trained photovoltaic ultra-short-term power prediction model to obtain power prediction results for each of the distributed photovoltaic sites.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the photovoltaic ultra-short-term power prediction method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the photovoltaic ultra-short-term power prediction method according to any one of claims 1 to 7.

Images