Unobserved point solar irradiance prediction method based on depth map neural network

Abstract

The invention discloses an unobserved point solar irradiance prediction method based on a depth map neural network. The method comprises the following steps: acquiring irradiance historical data of a solar irradiance monitoring station in a target area; assuming the solar irradiance monitoring stations as unobserved points in turn, and extracting irradiance input data from the irradiance historical data of other solar irradiance monitoring stations and extracting irradiance output true value data from the irradiance historical data of the unobserved points according to preset sampling parameters when assuming each unobserved point; when an unobserved point is assumed, determining the position relationship between the assumed unobserved point and other solar irradiance monitoring stations; constructing a graph data training set according to the position relation, the irradiance input data and the irradiance output truth value data; and training a solar irradiance prediction model by using the graph data training set so as to predict the irradiance of a real unobserved point by using the solar irradiance prediction model. According to the invention, accurate prediction of the irradiance of the unobserved point can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and particularly to a method, device, electronic device, and storage medium for predicting solar irradiance at unobserved points based on a deep graph neural network. Background Art

[0002] In a distribution network, most solar photovoltaic systems operate behind the meter, and utility companies cannot directly monitor the power generation situation. Due to the increasing penetration rate of behind-the-meter solar photovoltaic systems, the intermittency and volatility of energy pose a huge challenge to the operation of traditional power grids and the stability of the power market. Smart grid technology aims to optimize the integration of renewable energy and improve the intelligence and efficiency of the power system. The smart grid can utilize its core advantages of real-time monitoring and automatic control to effectively cope with the risks brought by the uncertainty of behind-the-meter solar power generation and provide reliable and safe guarantees for the operation of the power system. Therefore, accurate estimation and prediction of behind-the-meter irradiance play a key role in balancing supply and demand, improving energy utilization efficiency, and supporting the operation of the power market.

[0003] The power output of a solar photovoltaic system depends to a large extent on weather and atmospheric conditions, which can be concretely manifested as the global horizontal irradiance (GHI) received by the photovoltaic power generation system. That is, accurate prediction of solar irradiance is a prerequisite for precisely regulating the output power of the photovoltaic system. Therefore, how to accurately predict solar irradiance has become a technical problem that urgently needs to be solved. Summary of the Invention

[0004] The present invention provides a method, device, electronic device, and storage medium for predicting solar irradiance at unobserved points based on a deep graph neural network.

[0005] According to one aspect of the present invention, there is provided a method for predicting solar irradiance at unobserved points based on a deep graph neural network, including:

[0006] Obtaining historical irradiance data monitored by each solar irradiance monitoring station in a target area;

[0007] Assuming each solar irradiance monitoring station as an unobserved point in turn, and for each assumed unobserved point, extracting irradiance input data from the historical irradiance data of other solar irradiance monitoring stations according to preset sampling parameters, and extracting irradiance output true value data corresponding to the irradiance input data from the historical irradiance data of the assumed unobserved point;

[0008] Assuming each solar irradiance monitoring station as an unobserved point in turn, and for each assumed unobserved point, determining the positional relationship between the assumed unobserved point and other solar irradiance monitoring stations in the target area;

[0009] Construct a graph data training set according to the position relationship and the sampled irradiance input data and irradiance output true value data.

[0010] Use the graph data training set to train a solar irradiance prediction model pre-constructed based on a deep graph neural network, so as to predict the solar irradiance of the true unobserved points in the target area according to the trained solar irradiance prediction model.

[0011] In some embodiments, the method further includes:

[0012] Based on the clear sky index, perform normalization processing on the historical irradiance data to obtain an irradiance time series corresponding to each solar irradiance monitoring station.

[0013] In some embodiments, the preset sampling parameters include sampling time, pre-sampling time scale, and post-sampling time scale;

[0014] The step of assuming all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, extracting irradiance input data from the historical irradiance data of other solar irradiance monitoring stations according to the preset sampling parameters, and extracting the irradiance output true value data corresponding to the irradiance input data from the historical irradiance data of the assumed unobserved point includes:

[0015] After assuming a certain solar irradiance monitoring station as an unobserved point, extract irradiance input data from the irradiance time series corresponding to other solar irradiance monitoring stations according to the sampling time and the pre-sampling time scale;

[0016] According to the sampling time and the post-sampling time scale, extract the irradiance output true value data corresponding to the irradiance input data from the irradiance time series corresponding to the assumed unobserved point.

[0017] In some embodiments, the step of assuming all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, determining the position relationship between the assumed unobserved point and other solar irradiance monitoring stations in the target area includes:

[0018] Determine an initial adjacency matrix according to the number of solar irradiance monitoring stations included in the target area;

[0019] Assume all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, determine the inverse distance weights between the assumed unobserved point and other solar irradiance monitoring stations according to the geographical location information of the assumed unobserved point and the geographical location information of other solar irradiance monitoring stations in the target area;

[0020] Determine the values of the elements in the initial adjacency matrix according to the inverse distance weights between the assumed unobserved points and other solar irradiance monitoring stations, and obtain the final adjacency matrix representing the positional relationship.

[0021] In some embodiments, it further includes:

[0022] In response to the inverse distance weight between the assumed unobserved point and any solar irradiance monitoring station being less than a preset threshold, set the inverse distance weight between the assumed unobserved point and the solar irradiance monitoring station to zero.

[0023] In some embodiments, the predicting the solar irradiance of the true unobserved points in the target area according to the trained solar irradiance prediction model includes:

[0024] Determine the true unobserved points to be predicted for solar irradiance in the target area, and use the current time as the sampling time;

[0025] According to the sampling time and the previous sampling time scale, extract the irradiance input data from the historical irradiance data of the original solar irradiance monitoring stations in the target area;

[0026] According to the geographical location information of the true unobserved points and the geographical location information of the original solar irradiance monitoring stations in the target area, determine the positional relationship between the true unobserved points and the original solar irradiance monitoring stations in the target area;

[0027] Input the irradiance input data, the positional relationship between the original solar irradiance monitoring stations in the target area, and the positional relationship between the true unobserved points and the original solar irradiance monitoring stations in the target area into the trained solar irradiance prediction model for processing, and output the irradiance of the true unobserved points in the future time period.

[0028] In some embodiments, the solar irradiance prediction model is a graph attention network model.

[0029] According to another aspect of the present invention, there is provided a solar irradiance prediction device for unobserved points based on a deep graph neural network, including:

[0030] A data acquisition module for acquiring the historical irradiance data monitored by each solar irradiance monitoring station in the target area;

[0031] The first data processing module is used to assume all solar irradiance monitoring stations as unobserved points in turn. And for each assumed unobserved point, according to the preset sampling parameters, extract irradiance input data from the irradiance historical data of other solar irradiance monitoring stations, and extract the irradiance output true value data corresponding to the irradiance input data from the irradiance historical data of the assumed unobserved point.

[0032] The second data processing module is used to assume all solar irradiance monitoring stations as unobserved points in turn. And for each assumed unobserved point, determine the positional relationship between the assumed unobserved point and other solar irradiance monitoring stations in the target area.

[0033] The training set construction module is used to construct a graph data training set according to the positional relationship, and the sampled irradiance input data and the irradiance output true value data.

[0034] The training module is used to train a solar irradiance prediction model pre-constructed based on a deep graph neural network by using the graph data training set, so as to predict the solar irradiance of the truly unobserved points in the target area according to the trained solar irradiance prediction model.

[0035] In some embodiments, it further includes:

[0036] The normalization processing module is used to perform normalization processing on the irradiance historical data based on the clear sky index to obtain the irradiance time series corresponding to each solar irradiance monitoring station.

[0037] In some embodiments, the preset sampling parameters include sampling time, pre-sampling time scale, and post-sampling time scale.

[0038] The first data processing module is further used for:

[0039] After setting a certain solar irradiance monitoring station as the unobserved point of the prediction target station, according to the sampling time and the pre-sampling time scale, extract irradiance input data from the irradiance time series corresponding to other solar irradiance monitoring stations.

[0040] According to the sampling time and the post-sampling time scale, extract the irradiance output true value data corresponding to the irradiance input data from the irradiance time series corresponding to the assumed unobserved point of the prediction target station.

[0041] In some embodiments, the second data processing module is further used for:

[0042] Determine the initial adjacency matrix according to the number of solar irradiance monitoring stations included in the target area.

[0043] Assume all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, determine the inverse distance weights between the assumed unobserved point and other solar irradiance monitoring stations in the target area according to the geographical location information of the assumed unobserved point and the geographical location information of other solar irradiance monitoring stations in the target area.

[0044] According to the inverse distance weights between the assumed unobserved point and other solar irradiance monitoring stations, determine the values of each element in the initial adjacency matrix to obtain the final adjacency matrix representing the positional relationship.

[0045] In some embodiments, it further includes:

[0046] A verification module, configured to, in response to the inverse distance weight between the assumed unobserved point and any solar irradiance monitoring station being less than a preset threshold, set the inverse distance weight between the assumed unobserved point and the solar irradiance monitoring station to zero.

[0047] In some embodiments, it further includes a prediction module, configured to:

[0048] Determine the real unobserved point to be predicted for solar irradiance in the target area, and use the current time as the sampling time;

[0049] According to the sampling time and the previous sampling time scale, extract irradiance input data from the irradiance historical data of the original solar irradiance monitoring stations in the target area;

[0050] According to the geographical location information of the real unobserved point and the geographical location information of the original solar irradiance monitoring stations in the target area, determine the positional relationship between the real unobserved point and the original solar irradiance monitoring stations in the target area;

[0051] Input the irradiance input data, the positional relationship between the original solar irradiance monitoring stations in the target area, and the positional relationship between the real unobserved point and the original solar irradiance monitoring stations in the target area into the trained solar irradiance prediction model for processing, and output the irradiance of the real unobserved point in the future time period.

[0052] In some embodiments, the solar irradiance prediction model is a graph attention network model.

[0053] According to another aspect of the present invention, there is provided an electronic device, which includes:

[0054] At least one processor; and

[0055] A memory communicatively connected to at least one processor; wherein,

[0056] The memory stores a computer program that can be executed by at least one processor. The computer program is executed by at least one processor to enable the at least one processor to execute the method for predicting solar irradiance at unobserved points based on a deep graph neural network according to an embodiment of the present invention.

[0057] According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to implement the method for predicting solar irradiance at unobserved points based on a deep graph neural network according to an embodiment of the present invention when executed.

[0058] The technical solution of the embodiment of the present invention can accurately predict the solar irradiance at unobserved points.

[0059] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0061] Figure 1 is a schematic flowchart of a method for predicting solar irradiance at unobserved points based on a deep graph neural network provided by an embodiment of the present invention;

[0062] Figure 2 is a schematic flowchart of another method for predicting solar irradiance at unobserved points based on a deep graph neural network provided by an embodiment of the present invention;

[0063] Figure 3 is a schematic structural diagram of a device for predicting solar irradiance at unobserved points based on a deep graph neural network provided by an embodiment of the present invention;

[0064] Figure 4 is a schematic structural diagram of an electronic device for implementing the method for predicting solar irradiance at unobserved points based on a deep graph neural network according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0065] To enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0066] Embodiment 1

[0067] Figure 1 The figure is a flowchart of a method for predicting solar irradiance at unobserved points based on a deep graph neural network provided by an embodiment of the present invention. This embodiment is applicable to scenarios of predicting solar irradiance in the field of photovoltaic power generation. This method can be executed by a device for predicting solar irradiance at unobserved points based on a deep graph neural network. The device for predicting solar irradiance at unobserved points based on a deep graph neural network can be implemented in the form of hardware and / or software, and the device for predicting solar irradiance at unobserved points based on a deep graph neural network can be configured in an electronic device.

[0068] As Figure 1 shown, the method for predicting solar irradiance at unobserved points based on a deep graph neural network includes:

[0069] S101. Obtain the historical irradiance data monitored by each solar irradiance monitoring station in the target area.

[0070] In the embodiment of the present invention, the target area refers to a geographical area where multiple photovoltaic power stations are deployed. A photovoltaic power station refers to a power generation system that converts solar energy into electrical energy. For example, a photovoltaic power station can be a photovoltaic system installed at home that can convert light energy into electrical energy and output electrical energy. The size of the target area can be set according to actual needs. A solar irradiance monitoring station can be composed of devices capable of detecting the solar radiation intensity. For example, it can be composed of irradiance meter devices; each solar irradiance monitoring station can be connected to a photovoltaic power station to monitor the solar irradiance of the photovoltaic power station. The number of solar irradiance monitoring stations in the target area can be multiple. The solar irradiance monitoring stations in the target area can be represented by the following data set: Among them, S i represents a certain solar irradiance monitoring station in the target area, and N is the number of distributed solar irradiance monitoring stations in the target area.

[0071] In the embodiment of the present invention, the historical irradiance data monitored by each solar irradiance monitoring station can include the real-time irradiance monitored by each solar irradiance monitoring station. In addition, the geographical location information of each solar irradiance monitoring station in the target area is known.

[0072] In an embodiment of the present invention, after obtaining the historical irradiance data monitored by each solar irradiance monitoring station, in order to facilitate data processing and subsequent model training, it is necessary to standardize the historical irradiance data. In an alternative implementation, based on the clearness index, the historical irradiance data is standardized to obtain the irradiance time series corresponding to each solar irradiance monitoring station; wherein, the clearness index can be a preset value. Exemplarily, for the irradiance values measured by all solar irradiance monitoring stations in the target area Using the clearness index for standardization and calibration, a normalized irradiance time series can be obtained Specifically, it can be calculated according to the following formula:

[0073] S102. Assume all solar irradiance monitoring stations as unobserved points in turn. And for each assumed unobserved point, according to the preset sampling parameters, extract the irradiance input data from the historical irradiance data of other solar irradiance monitoring stations, and extract the irradiance output true value data corresponding to the irradiance input data from the historical irradiance data of the assumed unobserved point.

[0074] In an embodiment of the present invention, the preset sampling parameters include sampling time, pre-sampling time scale, and post-sampling time scale; wherein, there can be multiple sampling times, and the sampling time can be determined by means of a sliding window, that is, each time the sliding window slides, a sampling time is determined; the pre-sampling time scale refers to the fixed time window length for collecting data before the sampling time; the post-sampling time scale refers to the fixed time window length for collecting data after the sampling time. In this way, for each determined sampling time, the collection of the irradiance input data and its corresponding irradiance output true value data can be completed in combination with the pre-sampling time scale and the post-sampling time scale.

[0075] In an alternative implementation, assume all solar irradiance monitoring stations as unobserved points in turn. And for each assumed unobserved point, according to the sampling time and the pre-sampling time scale, extract the irradiance input data from the irradiance time series corresponding to other solar irradiance monitoring stations; according to the sampling time and the post-sampling time scale, extract the irradiance output true value data corresponding to the irradiance input data from the irradiance time series corresponding to the assumed unobserved point; the unobserved point refers to an unknown point in the target area that only knows the geographical location information but does not know the irradiance data.

[0076] Exemplarily, the target area includes N solar irradiance monitoring stations numbered from 1 to N; the sampling time is t, the pre-sampling time scale is n, and the post-sampling time scale is m; m and n can represent the number of sampling points. On this basis, assuming the solar irradiance monitoring station numbered 1 as the unobserved point, the collected irradiance input data can be expressed as: Among them, the first row in the matrix represents the n irradiances before time t obtained from the solar irradiance monitoring station numbered 2; similarly, the last row in the matrix represents the n irradiances before time t obtained from the solar irradiance monitoring station numbered N. And the true irradiance output data collected from the solar irradiance station numbered 1 can be expressed as: It can be understood that by continuously updating the sampling time, a large amount of irradiance input data and the corresponding true irradiance output data of the irradiance input data can be obtained in the above manner.

[0077] In another alternative implementation, the sampling time values are determined sequentially by means of a sliding window, and after each sampling time is determined, the following operations are performed: according to the sampling time and the pre-sampling time scale, extract the candidate irradiance input data from the irradiance time series corresponding to each solar irradiance monitoring station; according to the sampling time and the post-sampling time scale, extract the candidate true irradiance output data corresponding to the candidate irradiance input data from the irradiance time series corresponding to each solar irradiance monitoring station; assume each solar irradiance monitoring station as the unobserved point in turn, and for each assumed unobserved point, delete the irradiance time series corresponding to this unobserved point from the candidate irradiance input data to obtain the final irradiance input data; at the same time, only retain the irradiance time series corresponding to the unobserved point in the candidate true irradiance output data, and delete the others to obtain the final true irradiance output data.

[0078] Exemplarily, the target area includes N solar irradiance monitoring stations numbered from 1 to N; the sampling time is t, the pre-sampling time scale is n, and the post-sampling time scale is m; m and n can represent the number of sampling points. On this basis, the candidate irradiance input data extracted from the irradiance time series corresponding to each solar irradiance monitoring station according to the sampling time t and the pre-sampling time scale n can be expressed as: The first row in the matrix represents the n irradiances before time t obtained from the solar irradiance monitoring station numbered 1; the other rows are similar and will not be elaborated here. The candidate true irradiance output data extracted from the irradiance time series corresponding to each solar irradiance monitoring station according to the sampling time t and the post-sampling time scale m can be expressed as: The first row in the matrix represents the m irradiances after time t obtained from the solar irradiance monitoring site numbered 1; the other rows are similar and will not be elaborated here. On this basis, the solar irradiance monitoring sites numbered 1 - N can be assumed to be unobserved points in turn. For example, when the solar irradiance monitoring site numbered 1 is assumed to be an unobserved point, only the x in [x 1 …x N needs to be removed, and the remaining 1 is the final irradiance input data; at the same time, only y in [y 1 ……y N is retained, and the rest are deleted. At this time, the remaining 1 is which is the corresponding true value data of the irradiance output.

[0079] S103. Assume all the solar irradiance monitoring sites in the target area to be unobserved points in turn, and for each assumed unobserved point, determine the positional relationship between the assumed unobserved point and other solar irradiance monitoring sites in the target area.

[0080] In the embodiment of the present invention, the geographical location information of each solar irradiance monitoring site in the target area is known. Therefore, the initial adjacency matrix can be determined first according to the number of solar irradiance monitoring sites included in the target area; assume all the solar irradiance monitoring sites in the target area to be unobserved points in turn, and for each assumed unobserved point, determine the inverse distance weight between the assumed unobserved point and other solar irradiance monitoring sites in the target area according to the geographical location information of the assumed unobserved point and the geographical location information of other solar irradiance monitoring sites in the target area. The inverse distance weight represents a statistical index of the information transfer relationship between the two monitoring sites. Optionally, the inverse distance weight w between the assumed unobserved point and any other solar irradiance monitoring site is calculated according to the following formula i，j i，j

[0081] wherein, i，j is the distance determined according to the geographical location of the assumed unobserved point i and the geographical location of the solar irradiance monitoring site j; according to the inverse distance weights between the assumed unobserved point and other different solar irradiance monitoring sites, determine the values of each element in the initial adjacency matrix to obtain the final adjacency matrix representing the positional relationship. To ensure the accuracy of the final adjacency matrix, in response to the inverse distance weight between the assumed unobserved point and any solar irradiance monitoring site being less than the preset threshold, the inverse distance weight between the assumed unobserved point and the solar irradiance monitoring site is set to zero. That is, any element A of the adjacency matrix i，j is determined according to the following formula: where C is the preset threshold.S104. Construct a graph data training set based on the positional relationship, as well as the sampled irradiance input data and the true irradiance output data.

[0082] In the embodiments of the present invention, the graph data training set includes multiple graph data. Each graph data includes the positional relationship irradiance input data of the assumed unobserved point and other solar irradiance monitoring stations in the target area, as well as the true irradiance output data corresponding to the irradiance input data. Among them, the true irradiance output data is essentially the true label of a graph data.

[0083] S105. Use the graph data training set to train a solar irradiance prediction model pre-constructed based on a deep graph neural network, so as to predict the solar irradiance of the true unobserved point in the target area according to the trained solar irradiance prediction model.

[0084] In some embodiments, the solar irradiance prediction model to be trained based on the deep graph neural network may optionally be a graph attention network model. On this basis, the data in the graph data training set determined in step S104 can be used for model training. Taking the graph attention network layer (Graph Attention Networks, GAT) as an example of the solar irradiance prediction model, the solar irradiance prediction model can assign accurate weight values to the neighbor nodes (another solar irradiance monitoring station) of each node (solar irradiance monitoring station) through a graph theory mechanism. By alternately assuming unobserved points, the current unobserved node can be expressed as the aggregation of the information of its neighbor nodes; its node expression is updated as:

[0085]

[0086] where h′ i is the feature expression vector of node i in the current GAT layer; σ is a non-linear activation function; α ij is the attention weight value calculated and assigned by the GAT network layer between node i and its neighbor node j; W is a linear transformation matrix, representing the linear mapping of the features between the nodes participating in the calculation; h j is the input feature of the neighbor node j of the current node i.

[0087] Furthermore, GAT calculates and updates the weights between different nodes and their neighbor nodes through an attention mechanism, that is: where a T is a learnable attention vector; ∥ represents the connection between nodes; LeakyReLU is an activation function.

[0088] After completing one round of training using the graph data training set through the above steps, the validation set is used to evaluate the performance of the model during the training process, and the training results and evaluation results are recorded; the above steps are repeated to perform multiple trainings and validations on the prediction model to obtain the optimal model parameters.

[0089] Based on the trained solar irradiance prediction model, if there are real unobserved points in the target area, the solar irradiance of the real unobserved points can be predicted based on the trained solar irradiance prediction model; among them, the real unobserved points can be the location points where new photovoltaic power stations are added in the target area.

[0090] Through the solution of the present invention, accurate prediction of solar irradiance for any unobserved point in the target area can be achieved.

[0091] Embodiment 2

[0092] Figure 2 It is a flowchart of a method for predicting solar irradiance of unobserved points based on a deep graph neural network provided by an embodiment of the present invention. This embodiment mainly describes the process of predicting the solar irradiance of real unobserved points in the target area according to the trained solar irradiance prediction model. Refer to Figure 2 The method for predicting solar irradiance of unobserved points based on a deep graph neural network includes the following steps:

[0093] S201. Determine the real unobserved points in the target area for which solar irradiance is to be predicted, and use the current time as the sampling time.

[0094] Among them, the real unobserved points in the target area can optionally be the location points where new photovoltaic power stations are added in the target area. When it is determined that there are real unobserved points in the target area for which solar irradiance is to be predicted, the current time can be used as the sampling time, and then the solar irradiance of the real unobserved points can be predicted according to the subsequent steps.

[0095] S202. Extract the irradiance input data from the historical irradiance data of the original solar irradiance monitoring stations in the target area according to the sampling time and the previous sampling time scale.

[0096] For the specific implementation process, reference can be made to the description of the above embodiment, which will not be elaborated here.

[0097] S203. Determine the positional relationship between the real unobserved points and the original solar irradiance monitoring stations in the target area according to the geographical location information of the real unobserved points and the geographical location information of the original solar irradiance monitoring stations in the target area.

[0098] In the embodiments of the present invention, the positional relationship between the original solar irradiance monitoring stations in the target area can be represented by the adjacency matrix determined in the above embodiments. The specific determination process can be referred to the above embodiments and will not be elaborated here. Further, according to the number of real unobserved points, the existing adjacency matrix can be extended. For example, first, according to the geographical location information of the real unobserved points and the geographical location information of the original solar irradiance monitoring stations in the target area, calculate the inverse distance weights between each original solar irradiance monitoring station in the target area and the real unobserved points; then update the element values of the extended adjacency matrix according to the calculated inverse distance weights, and a latest adjacency matrix for characterizing the positional relationship of each position point where solar irradiance needs to be monitored in the target area can be obtained.

[0099] S204. Input the irradiance input data, the positional relationship between the original solar irradiance monitoring stations in the target area, and the positional relationship between the real unobserved points and the original solar irradiance monitoring stations in the target area into the trained solar irradiance prediction model for processing, and output the irradiance of the real unobserved points in the future time period.

[0100] Among them, the positional relationship between the original solar irradiance monitoring stations in the target area and the positional relationship between the real unobserved points and the original solar irradiance monitoring stations in the target area can be represented by the latest adjacency matrix obtained in the above steps. Therefore, the latest adjacency matrix and the collected irradiance input data can be input into the trained solar irradiance prediction model for processing, and the irradiance of the real unobserved points in the future time period can be output.

[0101] In the embodiments of the present invention, the trained solar irradiance prediction model can be used to estimate the irradiance of the real unobserved points in the future time period according to the latest adjacency matrix and the irradiance input data. Furthermore, the output power of the photovoltaic power station at the real unobserved points can be estimated according to the predicted irradiance, so as to regulate the smart grid.

[0102] Embodiment III

[0103] Figure 3 It is a schematic structural diagram of a solar irradiance prediction device for unobserved points based on a deep graph neural network provided by the embodiments of the present invention. This embodiment is applicable to the scenario of predicting solar irradiance in the field of photovoltaic power generation. As Figure 3 shown, the device includes:

[0104] A data acquisition module 301, configured to acquire the historical irradiance data monitored by each solar irradiance monitoring station in the target area;

[0105] The first data processing module 302 is configured to assume all solar irradiance monitoring stations as unobserved points in turn. And for each assumed unobserved point, according to the preset sampling parameters, extract irradiance input data from the irradiance historical data of other solar irradiance monitoring stations, and extract irradiance output true value data corresponding to the irradiance input data from the irradiance historical data of the assumed unobserved point;

[0106] The second data processing module 303 is configured to assume all solar irradiance monitoring stations as unobserved points in turn. And for each assumed unobserved point, determine the positional relationship between the assumed unobserved point and other solar irradiance monitoring stations in the target area;

[0107] The training set construction module 304 is configured to construct a graph data training set according to the positional relationship, and the sampled irradiance input data and the irradiance output true value data;

[0108] The training module 305 is configured to use the graph data training set to train a solar irradiance prediction model pre-constructed based on a deep graph neural network, so as to predict the solar irradiance of the truly unobserved points in the target area according to the trained solar irradiance prediction model.

[0109] In some embodiments, it further includes:

[0110] The normalization processing module is configured to perform normalization processing on the irradiance historical data based on the clearness index to obtain an irradiance time series corresponding to each solar irradiance monitoring station.

[0111] In some embodiments, the preset sampling parameters include sampling time, pre-sampling time scale, and post-sampling time scale;

[0112] The first data processing module 302 is further configured to:

[0113] After setting a certain solar irradiance monitoring station as the unobserved point of the prediction target station, according to the sampling time and the pre-sampling time scale, extract irradiance input data from the irradiance time series corresponding to other solar irradiance monitoring stations;

[0114] According to the sampling time and the post-sampling time scale, extract irradiance output true value data corresponding to the irradiance input data from the irradiance time series corresponding to the assumed unobserved point of the prediction target station.

[0115] In some embodiments, the second data processing module 303 is further configured to:

[0116] Determine an initial adjacency matrix according to the number of solar irradiance monitoring stations included in the target area;

[0117] Assume all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, determine the inverse distance weights between the assumed unobserved point and other solar irradiance monitoring stations in the target area according to the geographical location information of the assumed unobserved point and the geographical location information of other solar irradiance monitoring stations in the target area.

[0118] According to the inverse distance weights between the assumed unobserved point and other solar irradiance monitoring stations, determine the values of each element in the initial adjacency matrix to obtain the final adjacency matrix representing the position relationship.

[0119] In some embodiments, it further includes:

[0120] A verification module, configured to, in response to the inverse distance weight between the assumed unobserved point and any solar irradiance monitoring station being less than a preset threshold, set the inverse distance weight between the assumed unobserved point and the solar irradiance monitoring station to zero.

[0121] In some embodiments, it further includes a prediction module, configured to:

[0122] Determine a real unobserved point for which solar irradiance prediction is to be performed in the target area, and use the current time as the sampling time;

[0123] Extract irradiance input data from the historical irradiance data of the original solar irradiance monitoring stations in the target area according to the sampling time and the previous sampling time scale;

[0124] Determine the position relationship between the real unobserved point and the original solar irradiance monitoring stations in the target area according to the geographical location information of the real unobserved point and the geographical location information of the original solar irradiance monitoring stations in the target area;

[0125] Input the irradiance input data, the position relationship between the original solar irradiance monitoring stations in the target area, and the position relationship between the real unobserved point and the original solar irradiance monitoring stations in the target area into the trained solar irradiance prediction model for processing, and output the irradiance of the real unobserved point in the future time period.

[0126] In some embodiments, the solar irradiance prediction model is a graph attention network model.

[0127] The solar irradiance prediction device for unobserved points based on a deep graph neural network provided by the embodiments of the present invention can execute the solar irradiance prediction method for unobserved points based on a deep graph neural network provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0128] Embodiment 4

[0129] Figure 4 The structural schematic diagram of the electronic device 10 that can be used to implement the embodiments of the present invention is shown. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0130] As Figure 4 shown, the electronic device 10 includes at least one processor 11 and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. Among them, the memory stores a computer program executable by the at least one processor. The processor 11 can execute various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. The input / output (I / O) interface 15 is also connected to the bus 14.

[0131] A plurality of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disc, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0132] The processor 11 can be various general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as executing the method for predicting solar irradiance of unobserved points based on a depth graph neural network.

[0133] In some embodiments, the solar irradiance prediction method may be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the solar irradiance prediction method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform the solar irradiance prediction method for unobserved points based on a deep graph neural network in any other suitable manner (e.g., by means of firmware).

[0134] The various implementations of the systems and techniques described above in this document may be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), systems on a chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implemented in one or more computer programs that may be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a special-purpose or general-purpose programmable processor that receives data and instructions from a storage system, at least one input device, and at least one output device, and transmits the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0135] The computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable solar irradiance prediction device, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The computer programs may be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine, or entirely on a remote machine or server.

[0136] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0137] In order to provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0138] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by any form or medium of digital data communication (e.g., a communication network). Examples of the communication network include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0139] A computing system may include a client and a server. The client and the server are generally far from each other and usually interact via a communication network. The relationship between the client and the server is created by computer programs running on respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, and solves the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0140] It should be understood that various forms of the processes shown above can be used, steps can be reordered, added or deleted. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is made herein.

[0141] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for predicting solar irradiance at unobserved points based on a deep graph neural network, characterized in that Including: Obtaining the historical irradiance data monitored by each solar irradiance monitoring station within the target area; Assuming all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, extracting irradiance input data from the historical irradiance data of other solar irradiance monitoring stations according to preset sampling parameters, and extracting irradiance output true value data corresponding to the irradiance input data from the historical irradiance data of the assumed unobserved point; Assuming all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, determining the positional relationship between the assumed unobserved point and other solar irradiance monitoring stations within the target area; Constructing a graph data training set according to the positional relationship, and the sampled irradiance input data and the irradiance output true value data; Using the graph data training set to train a solar irradiance prediction model pre-constructed based on a deep graph neural network, so as to predict the solar irradiance of the truly unobserved points in the target area according to the trained solar irradiance prediction model.

2. The method according to claim 1, wherein Also including: Based on the clearness index, performing normalization processing on the historical irradiance data to obtain an irradiance time series corresponding to each solar irradiance monitoring station.

3. The method according to claim 2, wherein The preset sampling parameters include sampling time, pre-sampling time scale, and post-sampling time scale; The step of assuming all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, extracting irradiance input data from the historical irradiance data of other solar irradiance monitoring stations according to preset sampling parameters, and extracting irradiance output true value data corresponding to the irradiance input data from the historical irradiance data of the assumed unobserved point includes: After assuming a certain solar irradiance monitoring station as an unobserved point, extracting irradiance input data from the irradiance time series corresponding to other solar irradiance monitoring stations according to the sampling time and the pre-sampling time scale; Extracting irradiance output true value data corresponding to the irradiance input data from the irradiance time series corresponding to the assumed unobserved point according to the sampling time and the post-sampling time scale.

4. The method according to claim 1, wherein The step of assuming all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, determining the positional relationship between the assumed unobserved point and other solar irradiance monitoring stations within the target area includes: Determining an initial adjacency matrix according to the number of solar irradiance monitoring stations included in the target area; Assuming all solar irradiance monitoring stations as unobserved points in turn, and for each assumed unobserved point, determining the inverse distance weights between the assumed unobserved point and other solar irradiance monitoring stations according to the geographical location information of the assumed unobserved point and the geographical location information of other solar irradiance monitoring stations within the target area; Determining the values of each element in the initial adjacency matrix according to the inverse distance weights between the assumed unobserved point and other solar irradiance monitoring stations to obtain a final adjacency matrix characterizing the positional relationship.

5. The method according to claim 4, wherein Also including: In response to the inverse distance weight between the assumed unobserved point and any solar irradiance monitoring site being less than a preset threshold, the inverse distance weight between the assumed unobserved point and the solar irradiance monitoring site is set to zero.

6. The method according to claim 1, wherein The method of predicting solar irradiance of real unobserved points in the target area according to the trained solar irradiance prediction model includes: Determine a real unobserved point in the target area for solar irradiance prediction, and use the current time as the sampling time; Extracting irradiance input data from irradiance historical data of existing solar irradiance monitoring stations in the target area according to the sampling time and the previous sampling time scale; Determine the positional relationship between the real unobserved point and the original solar irradiance monitoring station in the target area according to the real geographical location information of the unobserved point and the geographical location information of the original solar irradiance monitoring station in the target area; The irradiance input data, the positional relationship between the original solar irradiance monitoring stations in the target area, and the positional relationship between the actual unobserved points and the original solar irradiance monitoring stations in the target area are input into the trained solar irradiance prediction model for processing, and the irradiance of the actual unobserved points in the future time period is output.

7. The method according to claim 1 or 6, characterized in that, The solar irradiance prediction model is a graph attention network model.

8. A solar irradiance prediction device for unobserved points based on a deep graph neural network, characterized in that, include: A data acquisition module is used to obtain historical irradiance data monitored by each solar irradiance monitoring station in the target area; A first data processing module is used to assume all solar irradiance monitoring sites as unobserved points in turn, and for each assumed unobserved point, extract irradiance input data from the irradiance historical data of other solar irradiance monitoring sites according to preset sampling parameters, and extract irradiance output true value data corresponding to the irradiance input data from the irradiance historical data of the assumed unobserved point; The second data processing module is used to assume all solar irradiance monitoring sites as unobserved points in turn, and for each assumed unobserved point, determine the positional relationship between the assumed unobserved point and other solar irradiance monitoring sites in the target area; A training set construction module, used to construct a graph data training set according to the positional relationship, and the sampled irradiance input data and the irradiance output true value data; A training module is used to use the graph data training set to train a solar irradiance prediction model pre-constructed based on a deep graph neural network, so as to predict the solar irradiance of real unobserved points in the target area according to the trained solar irradiance prediction model.

9. An electronic device, characterized in that, include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the method according to any one of claims 1 to 7.

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 processor to implement the method according to any one of claims 1 to 7 when executed.

Images