Photovoltaic resource prediction method and device, computer device and storage medium

Abstract

The application relates to a photovoltaic resource prediction method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring regional feature data of a to-be-predicted region in a preset time interval; the regional feature data comprises geographical data and meteorological data; constructing a regional feature image of the to-be-predicted region based on the regional feature data; performing feature extraction processing on the regional feature image to obtain a regional feature vector of the to-be-predicted region; updating a pre-trained sample photovoltaic resource prediction model based on the regional feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in the preset time interval; the sample photovoltaic resource prediction model is trained according to sample regional feature data of a sample region; and the regional feature vector is input into the photovoltaic resource prediction model to obtain photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval. The method can improve the generalization of photovoltaic resource prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid, and in particular, relates to a photovoltaic resource prediction method and device, a computer device, a storage medium, and a computer program product. BACKGROUND

[0002] The photovoltaic resource determines the strength of the power generation capacity of a photovoltaic power station. Therefore, in the construction process of a photovoltaic power station, the prediction of the photovoltaic resource is an essential link.

[0003] In related technologies, when predicting the photovoltaic resource, an experience model is usually relied on, that is, the photovoltaic resource is predicted based on historical photovoltaic data in a specific regional range and a specific time range. However, the experience model is limited to the specific regional range and the specific time range selected when the experience model is constructed, and therefore the generalization of the photovoltaic resource prediction based on the experience model is weak. SUMMARY

[0004] Therefore, it is necessary to provide a photovoltaic resource prediction method, device, computer device, computer readable storage medium, and computer program product capable of improving the generalization of photovoltaic resource prediction.

[0005] In a first aspect, the present application provides a photovoltaic resource prediction method. The method comprises:

[0006] obtaining regional feature data of a to-be-predicted region in a preset time interval; the regional feature data comprises geographical data and meteorological data;

[0007] constructing a regional feature image of the to-be-predicted region based on the regional feature data;

[0008] performing feature extraction processing on the regional feature image to obtain a regional feature vector of the to-be-predicted region;

[0009] updating a pre-trained sample photovoltaic resource prediction model based on the regional feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in the preset time interval; the sample photovoltaic resource prediction model is trained according to sample regional feature data of a sample region;

[0010] inputting the regional feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval.

[0011] In one embodiment, the updating of the pre-trained sample photovoltaic resource prediction model based on the regional feature vector to obtain the photovoltaic resource prediction model of the to-be-predicted region in the preset time interval comprises:

[0012] Based on the sample photovoltaic resource prediction model, a first reference feature vector satisfying a preset distance condition is identified from the sample area feature vector of the sample area, and a second reference feature vector satisfying the preset distance condition is identified from the area feature vector and the sample area feature vector;

[0013] Determine the difference information between the first reference feature vector and the second reference feature vector corresponding to the sample area;

[0014] According to the difference information, update the model parameters of the sample photovoltaic resource prediction model to obtain the photovoltaic resource prediction model of the to-be-predicted area in the preset time interval.

[0015] In one embodiment, the sample photovoltaic resource prediction model is trained by the following method:

[0016] Obtain the sample area feature data of the sample area;

[0017] Based on the sample area feature data, construct a sample area feature image of the sample area;

[0018] Perform feature extraction processing on the sample area feature image to obtain a sample area feature vector of the sample area;

[0019] Take the sample area feature vector of the sample area as input information, and take the actual photovoltaic resource information corresponding to the sample area as supervision information. Train the to-be-trained sample photovoltaic resource prediction model to obtain the sample photovoltaic resource prediction model.

[0020] In one embodiment, the constructing the area feature image of the to-be-predicted area based on the area feature data comprises:

[0021] Sampling in the area map corresponding to the to-be-predicted area to obtain a plurality of sampling points;

[0022] Based on the plurality of sampling points and the area feature data corresponding to the plurality of sampling points, the to-be-predicted area is divided into a plurality of polygonal regions;

[0023] Based on the area feature data corresponding to each polygonal region, the area feature image of the to-be-predicted area is constructed.

[0024] In one embodiment, the constructing the area feature image of the to-be-predicted area based on the area feature data corresponding to each polygonal region comprises:

[0025] respectively, the region feature data corresponding to each polygonal region is subjected to color mapping processing to obtain a color feature vector of each polygonal region;

[0026] Based on the color feature vector of each polygonal region, a color region image of each polygonal region is determined.

[0027] The color region images of the respective polygonal regions are combined to obtain a region feature image of the region to be predicted.

[0028] In one of the embodiments, the inputting of the region feature vector into the photovoltaic resource prediction model to obtain the photovoltaic resource information of the region to be predicted in the future time interval relative to the preset time interval comprises:

[0029] Based on the photovoltaic resource prediction model and the region feature vector, the irradiance and the irradiation duration of each polygonal region of the region to be predicted in the future time interval relative to the preset time interval are predicted to obtain the predicted irradiance and the predicted irradiation duration of each polygonal region in the future time interval.

[0030] Based on the predicted irradiance and the predicted irradiation duration of each polygonal region in the future time interval, the photovoltaic resource information of the region to be predicted in the future time interval relative to the preset time interval is obtained.

[0031] In a second aspect, the present application further provides a photovoltaic resource prediction device. The device comprises:

[0032] A data acquisition module is configured to acquire region feature data of a region to be predicted in a preset time interval.

[0033] An image construction module is configured to construct a region feature image of the region to be predicted based on the region feature data; the region feature data comprises geographic data and meteorological data.

[0034] A feature extraction module is configured to perform feature extraction processing on the region feature image to obtain a region feature vector of the region to be predicted.

[0035] A model updating module is configured to update a sample photovoltaic resource prediction model obtained by pre-training based on the region feature vector to obtain a photovoltaic resource prediction model of the region to be predicted in the preset time interval; the sample photovoltaic resource prediction model is obtained by training based on sample region feature data of a sample region.

[0036] A resource prediction module is configured to input the region feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the region to be predicted in a future time interval relative to the preset time interval.

[0037] In a third aspect, the present application further provides a computer device. The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0038] obtain regional feature data of a to-be-predicted region in a preset time interval;

[0039] construct a regional feature image of the to-be-predicted region based on the regional feature data; the regional feature data comprises geographic data and meteorological data;

[0040] perform feature extraction processing on the regional feature image to obtain a regional feature vector of the to-be-predicted region;

[0041] update a pre-trained sample photovoltaic resource prediction model based on the regional feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in the preset time interval; the sample photovoltaic resource prediction model is trained according to sample regional feature data of a sample region;

[0042] input the regional feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval.

[0043] In a fourth aspect, the present application further provides a computer readable storage medium. The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the following steps:

[0044] obtain regional feature data of a to-be-predicted region in a preset time interval;

[0045] construct a regional feature image of the to-be-predicted region based on the regional feature data; the regional feature data comprises geographic data and meteorological data;

[0046] perform feature extraction processing on the regional feature image to obtain a regional feature vector of the to-be-predicted region;

[0047] update a pre-trained sample photovoltaic resource prediction model based on the regional feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in the preset time interval; the sample photovoltaic resource prediction model is trained according to sample regional feature data of a sample region;

[0048] input the regional feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval.

[0049] In a fifth aspect, the present application also provides a computer program product. The computer program product comprises a computer program which, when executed by a processor, implements the following steps:

[0050] obtaining regional feature data of the to-be-predicted region in a preset time interval;

[0051] constructing a regional feature image of the to-be-predicted region based on the regional feature data; the regional feature data comprises geographical data and meteorological data;

[0052] performing feature extraction processing on the regional feature image to obtain a regional feature vector of the to-be-predicted region;

[0053] updating a sample photovoltaic resource prediction model obtained by pre-training based on the regional feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in the preset time interval; the sample photovoltaic resource prediction model is obtained by training based on sample regional feature data of a sample region;

[0054] inputting the regional feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval.

[0055] The photovoltaic resource prediction method, device, computer device, storage medium and computer program product described above first obtain regional feature data of a to-be-predicted region in a preset time interval; then construct a regional feature image of the to-be-predicted region based on the regional feature data; the regional feature data comprises geographical data and meteorological data; then perform feature extraction processing on the regional feature image to obtain a regional feature vector of the to-be-predicted region; then update a sample photovoltaic resource prediction model obtained by pre-training based on the regional feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in the preset time interval; the sample photovoltaic resource prediction model is obtained by training based on sample regional feature data of a sample region; finally, input the regional feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval. In this way, based on the regional feature vector of the to-be-predicted region, the corresponding photovoltaic resource prediction model of the to-be-predicted region in the preset time interval can be obtained directly based on the sample photovoltaic resource prediction model trained based on the sample region, without the need to re-train the corresponding photovoltaic resource prediction model based on the to-be-predicted region, and thus the photovoltaic resource of the to-be-predicted region can be predicted through the corresponding photovoltaic resource prediction model of the to-be-predicted region in the preset time interval. Compared with photovoltaic resource prediction based on an empirical model, the above photovoltaic resource prediction method does not limit the to-be-predicted region and the sample region to belong to the same specific regional range or the same specific time range, thereby increasing the generalization of photovoltaic resource prediction. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 A flowchart of a photovoltaic resource prediction method in an embodiment;

[0057] Figure 2 A light intensity color map of a region to be predicted in an embodiment;

[0058] Figure 3 A light duration contour map of a region to be predicted in an embodiment;

[0059] Figure 4 A flowchart of steps for obtaining a photovoltaic resource prediction model of a region to be predicted in a preset time interval in an embodiment;

[0060] Figure 5 A flowchart of steps for training a photovoltaic resource prediction model in an embodiment;

[0061] Figure 6 A flowchart of steps for constructing a region feature image of a region to be predicted based on region feature data in an embodiment;

[0062] Figure 7 A flowchart of steps for constructing a region feature image of a region to be predicted based on region feature data corresponding to each polygon region in an embodiment;

[0063] Figure 8 A flowchart of a photovoltaic resource prediction method in another embodiment;

[0064] Figure 9 A flowchart of a complex topography photovoltaic resource prediction method based on an incremental convolutional neural network in an embodiment;

[0065] Figure 10 A structural block diagram of a photovoltaic resource prediction method device in an embodiment;

[0066] Figure 11 An internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0067] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0068] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of the country and region.

[0069] In an example embodiment, as shown in Figure 1 A photovoltaic resource prediction method is provided, and the embodiment is exemplified by the method applied to a server; it can be understood that the method can also be applied to a terminal, and can also be applied to a system including a server and a terminal, and is realized through interaction between the server and the terminal. The server can be realized by an independent server or a server cluster composed of multiple servers; the terminal can be, but is not limited to, various personal computers, notebook computers, smart phones and tablet computers, etc. In the embodiment, the method includes the following steps:

[0070] In step S102, regional feature data of a to-be-predicted region in a preset time interval is acquired.

[0071] The regional feature data includes geographic data and meteorological data; the geographic data at least includes longitude, latitude, terrain height, slope, orientation, solar elevation angle, solar azimuth angle, vegetation index, etc. of the to-be-predicted region; and the meteorological data includes cloud amount, light intensity, light duration, etc. of the to-be-predicted region.

[0072] The preset time interval is a time period for collecting the regional feature data.

[0073] Specifically, the server acquires the geographic data such as longitude, latitude, terrain height, slope, orientation, solar elevation angle, solar azimuth angle, vegetation index, etc. of the to-be-predicted region, and the meteorological data such as cloud amount, light intensity, light duration, etc. of the to-be-predicted region through a geographic information collection device such as a satellite device, and takes the geographic data and the meteorological data as the regional feature data of the to-be-predicted region.

[0074] In step S104, a regional feature image of the to-be-predicted region is constructed based on the regional feature data.

[0075] The regional feature image of the to-be-predicted region is an image representing the regional feature data of the to-be-predicted region by color attributes or numerical attributes, such as a color patch image or an isogram, etc.

[0076] Specifically, the server represents different numerical values in the regional feature data of the to-be-predicted region by color or isogram, so as to draw a color patch image or an isogram of the to-be-predicted region as the regional feature image of the to-be-predicted region.

[0077] For example, taking the regional feature image as an illumination intensity color map, the server divides the illumination intensity of the to-be-predicted region into three levels of strong, medium, and weak, and uses three colors of dark gray, light gray, and white to represent the three illumination intensity levels, respectively. Then, the server determines the illumination intensity level of each sub-region of the to-be-predicted region according to the illumination intensity corresponding to each sub-region of the to-be-predicted region, and further determines the color corresponding to each sub-region, thereby obtaining the illumination intensity color map of the to-be-predicted region as shown in FIG. 6. Figure 2

[0078] For example, taking the regional feature image as an illumination duration contour map, the server divides the illumination duration of the to-be-predicted region according to a certain numerical interval, for example, at an interval of one hour, and draws the illumination duration contour map of the to-be-predicted region as shown in FIG. 7. Figure 3

[0079] Step S106, performing feature extraction processing on the regional feature image to obtain a regional feature vector of the to-be-predicted region.

[0080] Specifically, the server inputs the regional feature image into a convolutional neural network model, and performs dimension reduction processing on the regional feature image through a convolutional layer in the convolutional neural network, so as to extract the regional feature vector of the to-be-predicted region from the regional feature image.

[0081] For example, the server inputs the regional feature image into a VGG16 (Visual Geometry Group Network) model, and extracts the features of the regional feature image through the VGG16 model. The VGG16 model is composed of 1 input layer, 13 convolutional layers, and 5 pooling layers; each convolutional layer is used to perform convolution processing on the feature image input into the convolutional layer through a learnable convolution kernel; and each pooling layer is used to perform down-sampling on the feature image input into the pooling layer through maximum pooling processing, so as to reduce the size of the feature image while retaining key information.

[0082] Step S108, updating a pre-trained sample photovoltaic resource prediction model based on the regional feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in a preset time interval.

[0083] The sample photovoltaic resource prediction model is a support vector machine model trained according to sample regional feature data of a sample region. The sample region can be the to-be-predicted region, or any region other than the to-be-predicted region; and the sample regional feature data is historical regional feature data of the sample region in a historical time interval relative to the preset time interval.

[0084] ​​Specifically, the server performs incremental learning on the pre-trained sample photovoltaic resource prediction model based on the region feature vector of the region to be predicted, to update the sample photovoltaic resource prediction model, thereby obtaining a photovoltaic resource prediction model of the region to be predicted in the preset time interval.

[0085] For example, assuming that the sample region is the region to be predicted, the server performs incremental learning on the sample photovoltaic resource prediction support vector machine model trained based on the historical region feature vector of the region to be predicted in the historical time interval, based on the region feature vector of the region to be predicted in the preset time interval, to obtain a photovoltaic resource prediction incremental support vector machine model capable of predicting photovoltaic resources of the region to be predicted in the future time interval.

[0086] For another example, assuming that the sample region is any one region other than the region to be predicted, the server performs incremental learning on the sample photovoltaic resource prediction support vector machine model trained based on the historical region feature vector of any one region in the historical time interval, based on the region feature vector of the region to be predicted in the preset time interval, to obtain a photovoltaic resource prediction incremental support vector machine model capable of predicting photovoltaic resources of the region to be predicted in the future time interval.

[0087] In the updating process of the sample photovoltaic resource prediction model, through incremental learning of the support vector machine, the server can obtain a photovoltaic resource prediction model corresponding to the region to be predicted based on the sample photovoltaic resource prediction model even if the training data of the sample photovoltaic resource prediction model has no geographical or time correlation with the region to be predicted, so that the subsequent photovoltaic prediction is not limited by a specific geographical range and a specific time range.

[0088] In step S110, the region feature vector is input into the photovoltaic resource prediction model to obtain photovoltaic resource information of the region to be predicted in the future time interval relative to the preset time interval.

[0089] The photovoltaic resource information at least includes the illumination intensity and the illumination duration of the region to be predicted in the future time interval.

[0090] Specifically, the server inputs the region feature vector of the region to be predicted into the photovoltaic resource prediction model, predicts the illumination intensity and the illumination duration of the region to be predicted in the future time interval relative to the preset time interval through the photovoltaic prediction model, obtains the predicted illumination intensity and the predicted illumination duration of the region to be predicted in the future time interval, and confirms the predicted illumination intensity and the predicted illumination duration as the photovoltaic resource information of the region to be predicted in the future time interval.

[0091] In the photovoltaic resource prediction method provided in the above embodiments, the server first acquires regional feature data of the area to be predicted within a preset time interval; then, based on the regional feature data, it constructs a regional feature image of the area to be predicted; the regional feature data includes geographical data and meteorological data; next, it performs feature extraction processing on the regional feature image to obtain a regional feature vector of the area to be predicted; then, based on the regional feature vector, it updates the pre-trained sample photovoltaic resource prediction model to obtain a photovoltaic resource prediction model of the area to be predicted within the preset time interval; the sample photovoltaic resource prediction model is trained based on the sample regional feature data of the sample area; finally, it inputs the regional feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the area to be predicted within a future time interval relative to the preset time interval. In this way, based on the regional feature vector of the region to be predicted, the server can directly obtain the photovoltaic resource prediction model corresponding to the region to be predicted within a preset time interval, based on the sample photovoltaic resource prediction model trained on the sample region, without having to retrain the corresponding photovoltaic resource prediction model based on the region to be predicted. Thus, the photovoltaic resources of the region to be predicted can be predicted through the photovoltaic resource prediction model corresponding to the region to be predicted within the preset time interval. Compared with photovoltaic resource prediction based on empirical models, the above photovoltaic resource prediction method does not limit the region to be predicted and the sample to belong to the same specific regional range or the same specific time range, thereby increasing the generalization of photovoltaic resource prediction.

[0092] like Figure 4 As shown, in an exemplary embodiment, step S108, which updates the pre-trained sample photovoltaic resource prediction model based on the regional feature vector to obtain the photovoltaic resource prediction model for the region to be predicted within a preset time interval, specifically includes the following steps:

[0093] Step S402: Based on the sample photovoltaic resource prediction model, identify a first reference feature vector that meets the preset distance condition from the sample region feature vector of the sample region, and identify a second reference feature vector that meets the preset distance condition from the region feature vector and the sample region feature vector.

[0094] Step S404: Determine the difference information between the first reference feature vector and the second reference feature vector corresponding to the sample region.

[0095] Step S406: Based on the difference information, update the model parameters of the sample photovoltaic resource prediction model to obtain the photovoltaic resource prediction model for the region to be predicted within the preset time interval.

[0096] The reference feature vector is a support vector in the support vector machine model (a sample photovoltaic resource prediction model). It can be understood that the support vector refers to a training sample whose distance from a hyperplane in the support vector machine model is less than or equal to a preset distance threshold.

[0097] The first reference feature vector is a support vector obtained by taking the sample region feature vector of the sample region as a training sample of the sample photovoltaic resource prediction model.

[0098] The second reference feature vector is a support vector obtained by taking the region feature vector of the region to be predicted and the sample region feature vector of the sample region as training samples of the sample photovoltaic resource prediction model.

[0099] The difference information between the first reference feature vector and the second reference feature vector corresponding to the sample region can be represented by the distance between the first reference feature vector and the second reference feature vector.

[0100] The model parameters of the sample photovoltaic resource prediction model refer to the Lagrange multiplier term and the constant bias term in the output function of the sample photovoltaic resource prediction model.

[0101] Specifically, the server first identifies the support vector of the sample photovoltaic resource prediction model in the initial training sample set of the sample photovoltaic resource prediction model as the first reference feature vector that meets the preset distance condition. Then, the server adds the region feature vector to be predicted to the initial training sample set of the sample photovoltaic resource prediction model to obtain a target training sample set, and then updates the support vector of the sample photovoltaic resource prediction model based on the target training sample set, and the updated support vector is taken as the second reference feature vector that meets the preset distance condition. Then, the server determines the difference information between the first reference feature vector and the second reference feature vector corresponding to the sample region according to the kernel function of the sample photovoltaic resource prediction model, and updates the new Lagrange multiplier term corresponding to the region feature vector to be predicted and the constant bias term in the output function of the sample photovoltaic resource prediction model based on the difference information. Finally, the server updates the Lagrange multiplier term in the output function of the sample photovoltaic resource prediction model based on the new Lagrange multiplier term, thereby obtaining the photovoltaic resource prediction model of the region to be predicted in the preset time interval.

[0102] It should be noted that the objective function of the sample photovoltaic resource prediction model is shown in formula 1:

[0103]

[0104] wherein f(x) represents an output function of the sample photovoltaic resource prediction model, based on which a photovoltaic resource prediction vector can be obtained; L(x) represents an objective function of the sample photovoltaic resource prediction model; J represents the number of output variables (i.e. the vector length of the photovoltaic resource prediction vector) in the photovoltaic resource prediction vector output by the output function; I represents the number of samples input into the sample photovoltaic resource prediction model, i.e. the number of sample regional feature vectors input into the sample photovoltaic resource prediction model; y ij represents the true value of the jth output variable of the photovoltaic resource prediction vector corresponding to the ith sample regional feature vector; f j (x i ) represents the predicted value of the jth output variable of the photovoltaic resource prediction vector corresponding to the ith sample regional feature vector; β represents a regularization parameter.

[0105] Based on the objective function, the output function of the sample photovoltaic resource prediction model can be obtained as shown in Formula 2:

[0106]

[0107] wherein α i and respectively represent the ith sample regional feature vector corresponding to the Lagrange multiplier item and the optimal Lagrange multiplier item; x represents all sample regional feature vectors; x i represents the ith sample regional feature vector; k(·) represents a kernel function, k(x i , x) represents the kernel function value between the ith sample regional feature vector and all sample regional feature vectors; b is a constant bias item.

[0108] After the server identifies the first reference feature vector based on the target training sample set, the sequence minimum optimization method is used to update the new Lagrange multiplier item and the constant bias item based on Formula 3:

[0109]

[0110] wherein t represents the number of iterations, represents the new Lagrange multiplier of the tth iteration, represents the new Lagrange multiplier of the t+1th iteration; y new represents the output vector corresponding to the regional feature vector of the new to-be-predicted region; represents the prediction error of the tth iteration; k new is the difference information between the first reference feature vector and the second reference feature vector, i.e. the kernel function value between the first reference feature vector and the second reference feature vector.

[0111] wherein b t represents the constant bias item of the tth iteration, bt+1 represents the constant bias term of the t+1th iteration; P represents the number of samples in the initial training sample set, i.e., the total number of sample region feature vectors; P S represents the number of samples of the support vector set obtained according to the initial training sample set; SV represents the size of the support vector set obtained according to the initial training sample set; y i represents the output vector corresponding to the i th sample region feature vector in the initial training sample set; a i represents the Lagrange multiplier term corresponding to the i th sample region feature vector in the initial training sample set; k(x i , x l ) represents the kernel function value between the i th sample region feature vector and the l th sample region feature vector in the initial training sample set.

[0112] wherein a i represents the Lagrange multiplier term corresponding to the i th sample region feature vector; x new represents the region feature vector of the newly added to-be-predicted region; k(x i , x new ) represents the kernel function value between the i th sample region feature vector and the region feature vector of the newly added to-be-predicted region.

[0113] Finally, the server obtains the updated Lagrange multiplier term through formula 4:

[0114]

[0115] wherein, the Lagrange multiplier term corresponding to the h th feature vector in the target training sample set of the t th iteration, represents the Lagrange multiplier term corresponding to the h th feature vector (including the sample region feature vector of the sample region and the region feature vector of the to-be-predicted region) in the target training sample set of the t+1th iteration; represents the kernel function value of the t th iteration; k h represents the kernel function value between the h th feature vector in the target training sample set and itself.

[0116] The server substitutes the updated Lagrange multiplier term in formula 4 and the updated constant bias term b t+1 in formula 3 into formula 2, i.e., the photovoltaic resource prediction model of the to-be-predicted region in the preset time interval can be obtained; as shown in formula 5, which is the output function of the photovoltaic resource prediction model of the to-be-predicted region in the preset time interval:

[0117]

[0118] wherein, represents the optimal Lagrange multiplier term corresponding to the h-th feature vector; x' represents all feature vectors; x h represents the h-th feature vector; k(x h , x') represents the kernel function value between the h-th feature vector and all feature vectors.

[0119] In this embodiment, the server can update the Lagrange multiplier term and the constant bias term in the output function of the sample photovoltaic resource prediction model by using the support vectors obtained based on the sample region feature vectors and the support vectors obtained after adding the region feature vectors, thereby realizing the update of the sample photovoltaic resource prediction model, obtaining the photovoltaic resource prediction model corresponding to the region to be predicted in the preset time interval, so that the subsequent photovoltaic prediction is not limited by a specific geographic range and a specific time range, and the generalization of photovoltaic resource prediction is increased.

[0120] As shown in Figure 5 , in an exemplary embodiment, the sample photovoltaic resource prediction model in any of the above embodiments is trained in the following manner:

[0121] Step S502, obtaining sample region feature data of a sample region.

[0122] Step S504, constructing a sample region feature image of the sample region based on the sample region feature data.

[0123] Step S506, performing feature extraction processing on the sample region feature image to obtain a sample region feature vector of the sample region.

[0124] Step S508, training the sample photovoltaic resource prediction model to be trained by taking the sample region feature vector of the sample region as input information and taking the actual photovoltaic resource information corresponding to the sample region as supervision information, to obtain the sample photovoltaic resource prediction model.

[0125] Specifically, the server first acquires sample area feature data of the sample area through geographic information acquisition equipment, such as satellite equipment. Next, the server uses color or contour lines to represent different values ​​in the sample area feature data, thereby drawing a color map or contour map of the sample area as its sample area feature image. Then, the server inputs the sample area feature image into a convolutional neural network model. Through the convolutional layers in the convolutional neural network, the sample area feature image is dimensionality reduced to extract the sample area feature vector. Then, the server inputs the sample area feature vector into a photovoltaic resource prediction model to be trained. Through this model, predicted photovoltaic resource information for the sample area is obtained. Finally, a loss function is used to calculate the loss between the actual and predicted photovoltaic resource information of the sample area. If the loss value is greater than a preset loss threshold, the model parameters of the photovoltaic resource prediction model are optimized to retrain the model until the corresponding loss value is less than the preset loss threshold, thus obtaining the final photovoltaic resource prediction model.

[0126] In this embodiment, the server trains the photovoltaic resource prediction model to be trained through the feature vector of the sample region. This can obtain a sample photovoltaic resource prediction model that is not limited by a specific region range or feature time range. This makes it easier to generate corresponding photovoltaic resource prediction models for different regions to be predicted based on the sample photovoltaic resource prediction model. Based on the sample photovoltaic resource prediction model, the operation of retraining a prediction model for each additional region to be predicted is avoided, thereby increasing the generalization of photovoltaic prediction.

[0127] like Figure 6 As shown, in an exemplary embodiment, step S104 above, which constructs a regional feature image of the region to be predicted based on regional feature data, specifically includes the following steps:

[0128] Step S602: Sample the area map corresponding to the area to be predicted to obtain multiple sampling points.

[0129] Step S604: Based on multiple sampling points and the regional feature data corresponding to the multiple sampling points, the region to be predicted is divided into multiple polygonal regions.

[0130] Step S606: Construct a regional feature image of the region to be predicted based on the regional feature data corresponding to each polygonal region.

[0131] Specifically, the server first randomly samples in the region map corresponding to the region to be predicted to obtain a plurality of sampling points; then, the server converts the latitude and longitude coordinates of each sampling point into a planar rectangular coordinate, thereby obtaining a planar rectangular coordinate point set composed of the plurality of sampling points; next, the server performs Delaunay triangle division in the planar rectangular coordinate point set based on the region feature data corresponding to each sampling point, thereby dividing the region to be predicted into a plurality of non-overlapping and non-missing triangular regions; finally, the server represents the region feature data corresponding to each triangular region, such as terrain height, slope, orientation, solar elevation angle, solar azimuth angle, vegetation index, cloud cover, light intensity, and light duration, by color, thereby obtaining a color region corresponding to each triangular region, and the server combines the color regions to obtain the region feature image of the region to be predicted.

[0132] For example, the server converts the latitude and longitude coordinates of each sampling point into a planar rectangular coordinate based on formula 6:

[0133]

[0134] wherein a represents the horizontal coordinate of the sampling point in the planar rectangular coordinate system, b represents the vertical coordinate of the sampling point in the planar rectangular coordinate system, R represents the radius of the earth, wherein a represents the longitude of the sampling point in the latitude and longitude coordinate system, and b represents the latitude of the sampling point in the latitude and longitude coordinate system.

[0135] In this embodiment, the server performs Delaunay division on the region to be predicted through the sampling points, which can divide the region to be predicted into a plurality of non-overlapping and non-missing polygonal regions, thereby facilitating subsequent prediction of photovoltaic resource information of each polygonal region to obtain the photovoltaic resource information of the region to be predicted; based on the above process, the accuracy of the photovoltaic resource prediction result is ensured.

[0136] As shown in FIG. 6, Figure 7 In an exemplary embodiment, the step S606 of constructing the region feature image of the region to be predicted based on the region feature data corresponding to each polygonal region specifically includes the following steps:

[0137] Step S702: Color mapping processing is respectively performed on the region feature data corresponding to each polygonal region to obtain a color feature vector of each polygonal region.

[0138] Step S704: Based on the color feature vector of each polygonal region, a color region image of each polygonal region is determined.

[0139] Step S706: The color region images of the polygonal regions are combined to obtain the region feature image of the region to be predicted.

[0140] Specifically, the server maps the region feature data corresponding to each polygonal region into a color space through multiple channels to obtain a color feature vector of each polygonal region, according to the region feature data corresponding to each polygonal region, wherein one element in the color feature vector corresponds to one color channel; the multiple channels are RGB channels (red, green, and blue channels), then the server determines a target color corresponding to each polygonal region based on the color feature vector of each polygonal region to obtain a color region image with the target color corresponding to each polygonal region; finally, the server combines the color region images to obtain the region feature image of the to-be-predicted region.

[0141] For example, assuming that the terrain height corresponds to the red channel in the multiple channels, for a polygonal region with a high terrain height, the value of the vector element corresponding to the red color in the color feature vector corresponding to the region is higher, and the red color is more obvious in the target color corresponding to the region.

[0142] It can be understood that after obtaining the region feature image of the to-be-predicted region, the region feature image of the to-be-predicted region can also be processed for color correction, sharpening, contrast enhancement, noise removal, and the like.

[0143] In this embodiment, the server maps the region feature data corresponding to the polygonal region based on the RGB channels to obtain the target color corresponding to the polygonal region, and then can visualize the region feature data corresponding to the polygonal region, which is beneficial to subsequent feature extraction of the to-be-predicted region based on the region feature image.

[0144] In an example embodiment, the step S110 of inputting the region feature vector into the photovoltaic resource prediction model to obtain the photovoltaic resource information of the to-be-predicted region in the future time interval relative to the preset time interval includes the following content: based on the photovoltaic resource prediction model and the region feature vector, the irradiance and the irradiation duration of each polygonal region of the to-be-predicted region in the future time interval relative to the preset time interval are predicted to obtain the predicted irradiance and the predicted irradiation duration of each polygonal region in the future time interval; and based on the predicted irradiance and the predicted irradiation duration of each polygonal region in the future time interval, the photovoltaic resource information of the to-be-predicted region in the future time interval relative to the preset time interval is obtained.

[0145] Specifically, the server inputs the region feature vector of the to-be-predicted region into a photovoltaic resource prediction model of the to-be-predicted region in a preset time interval, obtains a corresponding output vector of the to-be-predicted region through the photovoltaic resource prediction model, each vector element in the output vector corresponds to the predicted light intensity and the predicted light duration of each polygonal region of the to-be-predicted region in a future time interval, and then determines the predicted light intensity and the predicted light duration of the corresponding polygonal region in the future time interval based on each vector element of the output vector. Finally, the server combines the predicted light intensity and the predicted light duration of each polygonal region in the future time interval to obtain the photovoltaic resource information of the to-be-predicted region in the future time interval.

[0146] In this embodiment, the photovoltaic resource prediction model of the to-be-predicted region of the server in the preset time interval can predict the photovoltaic resource information of the to-be-predicted region in the future time interval, which is convenient for subsequent decision-making of the site selection and power generation output of the photovoltaic power station of the to-be-predicted region.

[0147] In an exemplary embodiment, as shown in Figure 8 Another photovoltaic resource prediction method is provided, which is described by taking a server as an example and includes the following steps:

[0148] In step S801, region feature data of a to-be-predicted region in a preset time interval is obtained.

[0149] In step S802, sampling is performed in a region map corresponding to the to-be-predicted region to obtain a plurality of sampling points.

[0150] In step S803, the to-be-predicted region is divided into a plurality of polygonal regions based on the plurality of sampling points and the region feature data corresponding to the plurality of sampling points.

[0151] In step S804, a region feature image of the to-be-predicted region is constructed based on the region feature data corresponding to each polygonal region.

[0152] In step S805, color mapping processing is respectively performed on the region feature data corresponding to each polygonal region to obtain a color feature vector of each polygonal region.

[0153] In step S806, a color region image of each polygonal region is determined based on the color feature vector of each polygonal region.

[0154] In step S807, color region images of the polygonal regions are combined to obtain a region feature image of the to-be-predicted region.

[0155] In step S808, feature extraction processing is performed on the region feature image to obtain a region feature vector of the to-be-predicted region.

[0156] In step S809, based on the sample photovoltaic resource prediction model, a first reference feature vector meeting the preset distance condition is identified from the sample region feature vector of the sample region, and a second reference feature vector meeting the preset distance condition is identified from the region feature vector and the sample region feature vector.

[0157] In step S810, difference information between the first reference feature vector and the second reference feature vector corresponding to the sample region is determined.

[0158] In step S811, the model parameters of the sample photovoltaic resource prediction model are updated according to the difference information, and a photovoltaic resource prediction model of the to-be-predicted region in the preset time interval is obtained.

[0159] In step S812, the region feature vector is input into the photovoltaic resource prediction model, and photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval is obtained.

[0160] In this embodiment, first, the server can update the Lagrange multiplier term and the constant bias term in the output function of the sample photovoltaic resource prediction model by using the support vectors obtained based on the sample region feature vector and the support vectors obtained after adding the region feature vector, and then update the sample photovoltaic resource prediction model to obtain a photovoltaic resource prediction model corresponding to the to-be-predicted region in the preset time interval, so that the subsequent photovoltaic prediction is not limited by a specific geographic range and a specific time range. Second, the to-be-predicted region can be divided into a plurality of non-overlapping and non-missing polygonal regions by using the sampling points, so that the photovoltaic resource information of each polygonal region can be predicted to obtain the photovoltaic resource information of the to-be-predicted region, thereby ensuring the accuracy of the photovoltaic resource prediction result. Third, the server can map the region feature data corresponding to the polygonal region based on the RGB channel to obtain the target color corresponding to the polygonal region, and then visualize the region feature data corresponding to the polygonal region, which is beneficial to the subsequent feature extraction of the to-be-predicted region based on the region feature image. Compared with the photovoltaic resource prediction based on the empirical model, the above photovoltaic resource prediction method does not limit the to-be-predicted region and the sample to belong to the same specific region range or the same specific time range, thereby increasing the generalization of the photovoltaic resource prediction.

[0161] In order to more clearly illustrate the photovoltaic resource prediction method provided by the embodiments of the present application, the photovoltaic resource prediction method will be specifically described below with one specific embodiment, but it should be understood that the embodiments of the present application are not limited thereto. As shown in Figure 9 As shown in the example embodiment, the present application also provides a complex topography photovoltaic resource prediction method based on an incremental convolutional neural network, which specifically includes the following steps:

[0162] Step 1: Acquisition of geographic information data.

[0163] The server acquires geographic information data of the to-be-predicted region through a geographic information system, including topographic parameters and image parameters.

[0164] Step 2: Division of the to-be-predicted region.

[0165] The server divides the to-be-predicted region into multiple triangular regions using the Delaunay triangulation method.

[0166] Step 3: Feature image of the to-be-predicted region.

[0167] The server maps the corresponding topographic height, slope, orientation, solar elevation angle, solar azimuth angle, cloud cover, vegetation index, and light intensity information of each triangular region into a three-color space of red, green, and blue, obtaining a color feature vector of the triangular region and determining the target color of the triangular region. Then, the server fills the target color into the corresponding triangular region to form a color image of the triangular region, and splices the color images of multiple triangular regions to form a regional feature image of the to-be-predicted region.

[0168] Step 4: Extraction of regional features.

[0169] The server inputs the regional feature image of the to-be-predicted region into a convolutional neural network model to extract the features of each triangular region, obtaining a regional feature vector of the to-be-predicted region.

[0170] Step 5: Training of support vector machine model.

[0171] The server trains the support vector machine model based on the regional feature vector of the to-be-predicted region.

[0172] Step 6: Construction of incremental support vector machine.

[0173] The server performs incremental learning on the support vector machine model based on geographic information data after the to-be-predicted region or geographic information data of other regions, and constructs an incremental support vector machine.

[0174] Step 7: Prediction of photovoltaic resources.

[0175] The server predicts the geographic information data after the to-be-predicted region or the geographic information data of other regions through the corresponding incremental support vector machine, obtaining a photovoltaic resource prediction result of the to-be-predicted region or a photovoltaic resource prediction result of other regions.

[0176] In this embodiment, the photovoltaic prediction based on the incremental support vector machine can adaptively learn the photovoltaic resource prediction strategy under different landforms and meteorological conditions online, has strong spatiotemporal generalization ability, can also learn the historical landform and meteorological data in batches online, reduces the training time of the model, and improves the utilization rate of the model; secondly, the method can efficiently and accurately predict the illumination intensity and illumination time according to the geographical feature information and meteorological information of different landforms, reduce the differentiation degree of photovoltaic resources and meteorological parameters under complex landforms, improve the evaluation accuracy of photovoltaic resources, provide reliable data for the budget and planning of photovoltaic power station construction projects, and provide a reference for long-term operation and maintenance of photovoltaic power stations.

[0177] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0178] Based on the same inventive concept, the embodiments of the present application also provide a photovoltaic resource prediction device for implementing the photovoltaic resource prediction method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more photovoltaic resource prediction device embodiments provided below can refer to the limitations of the photovoltaic resource prediction method described above, and will not be repeated here.

[0179] In an exemplary embodiment, as shown in Figure 10 A photovoltaic resource prediction device is provided, comprising: a data acquisition module 1002, an image construction module 1004, a feature extraction module 1006, a model update module 1008, and a resource prediction module 1010, wherein:

[0180] The data acquisition module 1002 is configured to acquire regional feature data of a to-be-predicted region in a preset time interval.

[0181] The image construction module 1004 is configured to construct a regional feature image of the to-be-predicted region based on the regional feature data; the regional feature data includes geographical data and meteorological data.

[0182] The feature extraction module 1006 is configured to perform feature extraction processing on the region feature image to obtain a region feature vector of the to-be-predicted region.

[0183] The model updating module 1008 is configured to update a pre-trained sample photovoltaic resource prediction model based on the region feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in a preset time interval.

[0184] The resource prediction module 1010 is configured to input the region feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval.

[0185] In an example embodiment, the model updating module 1008 is further configured to identify, from sample region feature vectors of the sample region, a first reference feature vector that satisfies a preset distance condition based on the sample photovoltaic resource prediction model, and identify, from the region feature vector and the sample region feature vector, a second reference feature vector that satisfies the preset distance condition; determine difference information between the first reference feature vector and the second reference feature vector corresponding to the sample region; and update model parameters of the sample photovoltaic resource prediction model according to the difference information to obtain the photovoltaic resource prediction model of the to-be-predicted region in the preset time interval.

[0186] In an example embodiment, the photovoltaic resource prediction device further includes a model training module configured to obtain sample region feature data of a sample region; construct a sample region feature image of the sample region based on the sample region feature data; perform feature extraction processing on the sample region feature image to obtain a sample region feature vector of the sample region; and train a to-be-trained sample photovoltaic resource prediction model using the sample region feature vector of the sample region as input information and using actual photovoltaic resource information corresponding to the sample region as supervision information to obtain the sample photovoltaic resource prediction model.

[0187] In an example embodiment, the image construction module 1004 is configured to sample in a region map corresponding to the to-be-predicted region to obtain a plurality of sampling points; divide the to-be-predicted region into a plurality of polygonal regions based on the plurality of sampling points and region feature data corresponding to the plurality of sampling points; and construct a region feature image of the to-be-predicted region based on region feature data corresponding to each polygonal region.

[0188] In an exemplary embodiment, the image construction module 1004 is configured to perform color mapping processing on the regional feature data corresponding to each polygonal region to obtain a color feature vector for each polygonal region; determine a color region image for each polygonal region based on the color feature vector of each polygonal region; and combine the color region images of each polygonal region to obtain a regional feature image of the region to be predicted.

[0189] In an exemplary embodiment, the resource prediction module 1010 is further configured to predict the light intensity and duration of each polygonal region of the region to be predicted in a future time interval relative to a preset time interval, based on the photovoltaic resource prediction model and regional feature vector, to obtain the predicted light intensity and predicted duration of each polygonal region in the future time interval; and to obtain the photovoltaic resource information of the region to be predicted in the future time interval relative to the preset time interval based on the predicted light intensity and predicted duration of each polygonal region in the future time interval.

[0190] Each module in the aforementioned photovoltaic resource prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0191] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 11 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores regional characteristic data for various areas. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a photovoltaic resource prediction method.

[0192] Those skilled in the art will understand that Figure 11The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0193] In an exemplary embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor implementing the steps in the above method embodiments when executing the computer program.

[0194] In an exemplary embodiment, a computer readable storage medium is provided, storing a computer program, and the computer program implementing the steps in the above method embodiments when executed by a processor.

[0195] In an exemplary embodiment, a computer program product is provided, including a computer program, and the computer program implementing the steps in the above method embodiments when executed by a processor.

[0196] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0197] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0198] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A photovoltaic resource prediction method, characterized in that, The method comprises: acquiring regional feature data of a to-be-predicted region in a preset time interval; the regional feature data comprises geographical data and meteorological data; constructing a regional feature image of the to-be-predicted region based on the regional feature data; performing feature extraction processing on the regional feature image to obtain a regional feature vector of the to-be-predicted region; updating a pre-trained sample photovoltaic resource prediction model based on the regional feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in the preset time interval; further comprising: identifying a support vector of the sample photovoltaic resource prediction model from an initial training sample set of the sample photovoltaic resource prediction model as a first reference feature vector; adding the regional feature vector to the initial training sample set to obtain a target training sample set of the sample photovoltaic resource prediction model; updating the support vector of the sample photovoltaic resource prediction model based on the target training sample set, and taking the updated support vector as a second reference feature vector; determining difference information between the first reference feature vector and the second reference feature vector; determining a new Lagrange multiplier term corresponding to the regional feature vector and updating a constant bias term in an output function of the sample photovoltaic resource prediction model according to the difference information; updating a Lagrange multiplier term in the output function of the sample photovoltaic resource prediction model based on the new Lagrange multiplier term; obtaining the photovoltaic resource prediction model of the to-be-predicted region in the preset time interval according to the updated Lagrange multiplier term and the updated constant bias term; the sample photovoltaic resource prediction model is trained based on sample regional feature data of a sample region; the sample region is any region; inputting the regional feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval; the photovoltaic resource information at least comprises illumination intensity and illumination duration.

2. The method of claim 1, wherein, The sample photovoltaic resource prediction model is trained in the following way: acquiring sample regional feature data of the sample region; constructing a sample regional feature image of the sample region based on the sample regional feature data; performing feature extraction processing on the sample regional feature image to obtain a sample regional feature vector of the sample region; training a to-be-trained sample photovoltaic resource prediction model by taking the sample regional feature vector of the sample region as input information and taking actual photovoltaic resource information corresponding to the sample region as supervision information to obtain the sample photovoltaic resource prediction model.

3. The method of claim 1, wherein, The method comprises: sampling in a region map corresponding to the to-be-predicted region to obtain a plurality of sampling points; dividing the to-be-predicted region into a plurality of polygonal regions based on the plurality of sampling points and regional feature data corresponding to the plurality of sampling points; constructing a regional feature image of the to-be-predicted region based on regional feature data corresponding to each polygonal region.

4. The method of claim 3, wherein, The region feature image of the to-be-predicted region is constructed based on the region feature data corresponding to each polygon region, and the region feature image of the to-be-predicted region is constructed based on the region feature data corresponding to each polygon region, comprising: respectively, the color feature vector of each polygon region is obtained by performing color mapping processing on the region feature data corresponding to each polygon region; determine the color region image of each polygon region based on the color feature vector of each polygon region; combining the color region image of each polygon region, obtaining the region feature image of the to-be-predicted region.

5. The method according to any one of claims 1 to 4, characterized in that, The region feature vector is input into the photovoltaic resource prediction model to obtain the photovoltaic resource information of the to-be-predicted region in the future time interval relative to the preset time interval, comprising: based on the photovoltaic resource prediction model and the region feature vector, the illumination intensity and the illumination duration of each polygon region of the to-be-predicted region in the future time interval relative to the preset time interval are predicted, and the predicted illumination intensity and the predicted illumination duration of each polygon region in the future time interval are obtained; based on the predicted illumination intensity and the predicted illumination duration of each polygon region in the future time interval, the photovoltaic resource information of the to-be-predicted region in the future time interval relative to the preset time interval is obtained.

6. A photovoltaic resource prediction apparatus characterized by comprising: The device comprises: a data acquisition module for acquiring region feature data of a to-be-predicted region in a preset time interval; an image construction module for constructing a region feature image of the to-be-predicted region based on the region feature data; the region feature data comprises geographic data and meteorological data; a feature extraction module for performing feature extraction processing on the region feature image to obtain a region feature vector of the to-be-predicted region; a model updating module for updating a pre-trained sample photovoltaic resource prediction model based on the region feature vector to obtain a photovoltaic resource prediction model of the to-be-predicted region in the preset time interval; further used for identifying support vectors of the sample photovoltaic resource prediction model from an initial training sample set of the sample photovoltaic resource prediction model as a first reference feature vector; adding the region feature vector to the initial training sample set to obtain a target training sample set of the sample photovoltaic resource prediction model; updating the support vectors of the sample photovoltaic resource prediction model based on the target training sample set, and taking the updated support vectors as a second reference feature vector; determining the difference information between the first reference feature vector and the second reference feature vector; determining the newly added Lagrange multiplier term corresponding to the region feature vector and updating the constant bias term in the output function of the sample photovoltaic resource prediction model according to the difference information; updating the Lagrange multiplier term in the output function of the sample photovoltaic resource prediction model based on the newly added Lagrange multiplier term; obtaining the photovoltaic resource prediction model of the to-be-predicted region in the preset time interval according to the updated Lagrange multiplier term and the updated constant bias term; The resource prediction module is configured to input the region feature vector into the photovoltaic resource prediction model to obtain photovoltaic resource information of the to-be-predicted region in a future time interval relative to the preset time interval. The sample photovoltaic resource prediction model is trained according to sample region feature data of a sample region, and the sample region is any region; and the photovoltaic resource information at least includes illumination intensity and illumination duration.

7. The apparatus of claim 6, wherein, The device further includes a model training module configured to acquire sample region feature data of the sample region; construct a sample region feature image of the sample region based on the sample region feature data; perform feature extraction processing on the sample region feature image to obtain a sample region feature vector of the sample region; train a to-be-trained sample photovoltaic resource prediction model by taking the sample region feature vector of the sample region as input information and taking actual photovoltaic resource information corresponding to the sample region as supervision information, and obtain the sample photovoltaic resource prediction model.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method in any one of claims 1 to 5.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 5.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 5. The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 5.

Images