Foundation cloud picture-based cluster photovoltaic power prediction method and device

Abstract

The invention provides a cluster photovoltaic power prediction method and device based on a ground-based cloud atlas, and the method comprises the steps: recording the position information of each photovoltaic station in a cluster, collecting the ground-based cloud atlas data and power data of each photovoltaic station, and obtaining the NWP data; nWP data preprocessing: using a maximum correlation minimum redundancy algorithm to screen out meteorological factors with high correlation degree, and obtaining an optimized meteorological feature data set; preprocessing the ground-based cloud picture data, performing registration fusion operation on the ground-based cloud pictures of different point positions, extracting key features of the cloud pictures, and obtaining a cloud picture key feature data set; and training a cluster photovoltaic power prediction model based on the optimized meteorological feature data set and the cloud picture key feature data set, and obtaining cluster photovoltaic prediction power according to the model. According to the method, high-precision feature extraction and image fusion are realized, cluster photovoltaic power prediction is optimized, and the problem that cluster ultra-short-term photovoltaic power prediction cannot be comprehensively, timely and accurately carried out in the prior art is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy electric power, and in particular relates to a cluster photovoltaic power prediction method and device based on ground-based cloud maps. Background Art

[0002] As a clean and renewable energy source, solar energy is increasingly attracting widespread attention. With the continuous expansion of the scale of photovoltaic grid-connected, photovoltaic has become the third largest electric energy source. However, the intensity of sunlight received by the ground is greatly affected by weather conditions, cloud movement, etc. When the photovoltaic penetration power is high, it brings an impact on the power grid. Therefore, in order to ensure the safe and reliable operation of the power grid and reduce the phenomenon of "abandonment of light and power rationing", it is very important to accurately predict the irradiance and photovoltaic power.

[0003] The commonly used photovoltaic cluster power prediction methods currently include the accumulation method and the statistical upscaling method. The accumulation method obtains the power prediction of the photovoltaic cluster by directly accumulating the prediction results of all photovoltaic stations in the cluster, while the statistical upscaling method uses the prediction results of the benchmark photovoltaic station to obtain the power prediction of the photovoltaic cluster through upscaling calculation. These methods are insufficient in the deep mining of power prediction input data, and do not take into account the temporal relationship between multiple photovoltaic stations and the similarity of spatial geographical locations, resulting in the need to improve the prediction accuracy.

[0004] In addition, the data used for cluster ultra-short-term photovoltaic forecasts mainly include numerical weather forecasts and historical power generation. Numerical weather forecasts provide specific values ​​of various meteorological elements, which can greatly reduce system errors, but have low temporal and spatial resolutions and large temporal and spatial scales. They do not have the ability to improve the forecast of intraday cloudy and sudden weather changes, and cannot reflect the details of weather changes at higher temporal resolutions. The update frequency is generally once every 12 hours. It cannot meet the high-precision ultra-short-term prediction needs of each station. If minute-level ground-based cloud map data is used, a more accurate estimate of the forecast error can be given. However, the observation range of the ground-based cloud map only covers the sky information above the local real-time local area, so it is mostly used for single-station power prediction. Since the photovoltaic stations in the cluster may be located in different geographical locations, the spatial distribution of the ground-based cloud map data is different. How to effectively integrate them and use them for cluster power prediction is still a challenge. Summary of the invention

[0005] In view of the problems in the prior art, the present invention proposes a cluster photovoltaic power prediction method and device based on ground-based cloud images, which can achieve high-precision feature extraction and image fusion, optimize cluster photovoltaic power prediction, and overcome the problem that the prior art cannot comprehensively, timely and accurately perform cluster ultra-short-term photovoltaic power prediction.

[0006] To achieve the above object, the technical solution of the present invention is achieved as follows: A cluster photovoltaic power prediction method based on ground-based cloud images, comprising: S1. Data collection: record the location information of each photovoltaic station in the cluster, collect ground-based cloud map data and power data of each photovoltaic station, and obtain numerical weather forecast NWP data; S2, NWP data preprocessing, using the maximum correlation minimum redundancy algorithm to screen out meteorological factors with high correlation and obtain the optimal meteorological characteristic data set; S3, preprocessing of ground-based cloud map data, performing registration and fusion operations on ground-based cloud maps at different points, extracting key features of cloud maps, and obtaining cloud map key feature data sets; S4. Training a cluster photovoltaic power prediction model based on the preferred meteorological feature data set and the cloud image key feature data set, and obtaining cluster photovoltaic predicted power according to the model.

[0007] Furthermore, in step S1, obtaining numerical weather forecast NWP data includes obtaining numerical weather forecast data sets from different meteorological sources in the same time period, wherein the data sets include irradiance, wind speed, wind direction, temperature, humidity, and air pressure data; and processing abnormal values ​​in the irradiance data that violate the actual lighting laws.

[0008] Further, step S2 includes: S201, performing maximum and minimum normalization processing on the acquired NWP data; S202, using the maximum correlation minimum redundancy algorithm, taking the power data of the cluster photovoltaic as the target variable, calculating the correlation between the NWP data of different meteorological sources and the target variable, and obtaining the optimal meteorological characteristic data set by maximizing the correlation and minimizing the redundancy between the NWP data.

[0009] Further, step S3 includes: S301, retaining the minimum circumscribed rectangle containing the spherical mirror for the ground-based cloud image data of each station, using median filtering to process noise points, and obtaining a pre-processed ground-based cloud image; S302, calculating the distance between photovoltaic stations according to the longitude and latitude data of the location information of each photovoltaic station, and storing it in an adjacency matrix; S303, sorting the photovoltaic stations according to the distances between them to obtain the distance relationship between the locations of the stations, performing image registration of the pre-processed ground cloud images in order from near to far, and fusing the registered images; S304, extracting cloud map key features from the fused ground-based cloud map fusion image to obtain a cloud map key feature data set; the cloud map key features include light intensity, cloud cover percentage, transmittance and zenith distance.

[0010] Furthermore, step S4 includes: S401, treating each photovoltaic station as a node, representing the relationship between nodes through the adjacency matrix, and using the graph attention network to achieve node-level information aggregation and feature embedding for the optimal meteorological feature data set and the cloud map key feature data set, to obtain the spatial features of each node; S402, based on the Scaleformer network model, multi-scale processing is adopted, the spatial characteristics of each node are used as input variables, the power data of the cluster photovoltaic is used as the target variable, and the minimum loss function is used as the goal, and the cluster photovoltaic power prediction model is obtained through training; S403: Predicting according to the cluster photovoltaic power prediction model to obtain a final cluster power prediction value.

[0011] On the other hand, the present invention also proposes a cluster photovoltaic power prediction device based on a ground-based cloud map, comprising: Data collection module: records the location information of each photovoltaic station in the cluster, collects ground-based cloud map data and power data of each photovoltaic station, and obtains numerical weather forecast NWP data; Meteorological feature module: used for NWP data preprocessing, using the maximum correlation minimum redundancy algorithm to screen out meteorological factors with high correlation and obtain the optimal meteorological feature data set; Cloud map feature module: used for ground-based cloud map data preprocessing, registering and fusing ground-based cloud maps at different points, extracting key features of cloud maps, and obtaining cloud map key feature data sets; Prediction model module: training a cluster photovoltaic power prediction model based on the preferred meteorological feature data set and the cloud image key feature data set, and obtaining the cluster photovoltaic predicted power according to the model.

[0012] Furthermore, in the data collection module, obtaining numerical weather forecast NWP data includes obtaining numerical weather forecast data sets from different meteorological sources in the same time period, the data sets including irradiance, wind speed, wind direction, temperature, humidity, and air pressure data; and processing abnormal values ​​in irradiance data that violate the actual lighting laws.

[0013] Furthermore, the meteorological feature module includes: Normalization unit: perform maximum and minimum normalization processing on the acquired NWP data; Correlation unit: Using the maximum correlation minimum redundancy algorithm, with the power data of cluster photovoltaics as the target variable, the correlation between the NWP data of different meteorological sources and the target variable is calculated, and the optimal meteorological characteristic data set is obtained by maximizing the correlation and minimizing the redundancy between NWP data.

[0014] Furthermore, the cloud map feature module includes: Preprocessing unit: For the ground-based cloud image data of each station, the minimum circumscribed rectangle containing the spherical mirror is retained, and the noise points are processed using median filtering to obtain the preprocessed ground-based cloud image; Matrix unit: Calculate the distance between photovoltaic stations based on the longitude and latitude data of the location information of each photovoltaic station and store it in an adjacency matrix; Fusion unit: Sort the photovoltaic stations according to their distances to obtain the distance relationship between the locations of the stations, perform image registration of the pre-processed ground cloud images from near to far, and fuse the registered images; Feature extraction unit: extracts key features of cloud images from the fused ground-based cloud image to obtain a cloud image key feature data set; the key features of the cloud image include light intensity, cloud percentage, transmittance and zenith distance.

[0015] Furthermore, the prediction model module includes: Spatial feature unit: each photovoltaic station is regarded as a node, and the relationship between nodes is represented by the adjacency matrix. For the optimal meteorological feature data set and cloud image key feature data set, the graph attention network is used to realize node-level information aggregation and feature embedding to obtain the spatial features of each node; Model unit: Based on the Scaleformer network model, multi-scale processing is adopted, the spatial characteristics of each node are used as input variables, the power data of cluster photovoltaics is used as the target variable, and the minimum loss function is used as the goal. The cluster photovoltaic power prediction model is obtained through training; Prediction unit: performs prediction according to the cluster photovoltaic power prediction model to obtain a final cluster power prediction value.

[0016] Compared with the prior art, the present invention has the following beneficial effects: Aiming at the situation that photovoltaic stations located in different geographical locations have different spatiotemporal characteristics in cluster photovoltaic power prediction, the present invention effectively integrates the ground-based cloud map data of each station through high-precision feature extraction and image fusion, makes full use of sky image information in different spaces, and provides more reliable and comprehensive regional cloud map observation results, thereby providing more accurate ultra-short-term cluster photovoltaic power prediction, reducing operation and maintenance costs, and enhancing power grid stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of a flow chart of an embodiment of the present invention.

[0018] Figure 2 It is a schematic diagram of the spatial distribution of a certain cluster of photovoltaic stations according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0020] In order to make the purpose and features of the present invention more obvious and understandable, the present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0021] The cluster photovoltaic power prediction method based on ground-based cloud map proposed in the present invention is as follows: Figure 1 As shown, the following steps are included: S1. Data collection: record the location information of all photovoltaic stations in the cluster, use the sky imager to collect the ground-based cloud map data of each photovoltaic station, obtain the power data of each photovoltaic station through the supervisory control and data acquisition system, and obtain the numerical weather forecast NWP data; S2, NWP data preprocessing, using the maximum correlation minimum redundancy algorithm to screen out meteorological factors with high correlation and obtain the optimal meteorological characteristic data set; S3, preprocessing of ground-based cloud image data, performing registration and fusion operations on ground-based cloud images of different stations, extracting key features of cloud images, and obtaining cloud image key feature data sets; S4. Training a cluster photovoltaic power prediction model based on the preferred meteorological feature data set and the cloud image key feature data set, and obtaining cluster photovoltaic predicted power according to the model.

[0022] In step S1, the cluster photovoltaic consists of multiple photovoltaic stations distributed in different spaces, such as Figure 2 As shown, 1, 2, 3, 4, 5, and 6 represent photovoltaic stations in different spaces, so data collection includes the following steps: (1-1) Record the location information of each photovoltaic station; prepare the historical observed power of each photovoltaic station and obtain the historical power data set; (1-2) Prepare numerical weather prediction (NWP) datasets of different meteorological sources for the same time period as the historical power dataset. The data in the NWP dataset include six meteorological characteristics: irradiance, wind speed, wind direction, temperature, humidity, and air pressure. (1-3) Processing abnormal values ​​of irradiance data in NWP datasets that violate the real lighting law, where the violation of the real lighting law includes two situations: the night value is not 0 and the day value has no change for 2 consecutive hours. The abnormal values ​​that meet these two situations are deleted, and the data values ​​of the irradiance data at the same time within the previous and next 7 days are averaged, and the abnormal values ​​are replaced by the average value; (1-4) Obtain ground-based cloud map data of each photovoltaic station collected by a sky imager in the same time period as the historical data set.

[0023] The step S2 comprises the following steps: (2-1) The data sets of numerical weather forecast (NWP) data from different meteorological sources are normalized to the maximum and minimum values ​​to eliminate the influence of different dimensions; ; in is the kth value of the jth variable at the ith station in the NWP dataset time series after maximum and minimum normalization. is the kth value of the jth variable at the ith station in the original NWP dataset time series, and are the minimum and maximum values ​​of the jth variable at the ith station in the original NWP dataset time series, respectively.

[0024] (2-2) The maximum correlation minimum redundancy algorithm is used to calculate the meteorological characteristics and cluster photovoltaic power values ​​in NWP data sets from different sources. The cluster photovoltaic power value is the power sum of all photovoltaic stations in the cluster in the power data set that observe the power in the same time period. The cluster photovoltaic power value is used as the target variable. By maximizing the correlation between the meteorological characteristics in the NWP data set and the target variable and minimizing the redundancy between the meteorological characteristics in the NWP data set, the generalization ability of the model is improved, the risk of overfitting is reduced, and the optimal meteorological characteristic data set is obtained.

[0025] The step S3 comprises the following steps: (3-1) Preprocessing operations are performed on the ground-based cloud image data of each photovoltaic station collected by the sky imager. Crop the unnecessary information around and only keep the minimum circumscribed rectangle containing the spherical mirror. In the acquisition and processing process of the ground-based cloud image data, the whole sky imager will have disturbances of the transmission medium or defects of the electronic equipment when collecting images, which will affect the imaging of the image, and the shooting, recording and transmission process will inevitably be disturbed by various noises. In the image, it is usually a single pulse signal, and in the graph it is expressed as a pixel block or an isolated pixel point, which affects the target display with useless information. The basic principle of the median filter is a nonlinear signal processing method designed based on the sorting statistical theory that can effectively control the signal-to-noise ratio. It can be effectively used in the case of pulse noise and salt and pepper noise, and can better retain the texture information and edge details of the image. Therefore, for some noise points in the image of the ground-based cloud map, the median filtering method is used to solve them.

[0026] ; in is the filtered image at position The pixel value of is the original image at position The pixel value of and is the index of the pixels in the neighborhood. represents the median function, taking all The median value of .

[0027] (3-2) Based on the location information of each photovoltaic station, the longitude and latitude data are extracted and the distances between each other are calculated. In this embodiment, the Euclidean distance is used to represent the distance relationship. After the Euclidean distance between each photovoltaic station is calculated, the calculation result is stored in an adjacency matrix. The adjacency matrix is ​​constructed by treating each photovoltaic station as a node, and the Euclidean distance between each photovoltaic station represents the edge between the nodes, and is constructed based on the KNN algorithm.

[0028] (3-3) Sort the distance values ​​of each station in the cluster to obtain the distance relationship of each node (each photovoltaic station), perform image registration from near to far, and fuse the registered images; Wherein step (3-3) includes the following sub-steps: (3-3-a) Sort the distance values ​​(Euclidean distance) of each station in the cluster, and first take the ground-based cloud image of the two stations closest to each other; (3-3-b) Combine the efficiency of SURF (Speeded Up Robust Features) feature extraction with the rotation invariance and robustness of ORB (Oriented FAST and Rotated BRIEF) descriptors to perform image registration of the two closest station foundation cloud images. SURF (Speeded Up Robust Features) is used to detect feature points and generate descriptors, and then the ORB algorithm is used for fast matching. After image registration, it is necessary to project each image in the image sequence onto the reference image according to the transformation matrix between images to complete the panoramic stitching. The traditional bundle adjustment method is used to optimize the transformation matrix to improve the accuracy of the transformation matrix, and a mosaic of the two images is obtained; (3-3-c) In the foundation cloud images of the remaining photovoltaic stations, take out the closest one in turn, splice it with the fused image, and repeat step (3-3-b) until the foundation cloud images of all stations are spliced ​​together to obtain a fused image of the foundation cloud image; The above image registration algorithm can also be replaced by its homologous algorithm such as SIFT.

[0029] (3-4) Extract key features such as light intensity, cloud percentage, transmittance and zenith distance from the ground-based cloud image fusion images of each time period to obtain a cloud image key feature data set; Wherein step (3-4) includes the following sub-steps: (3-4-a) The ground-based cloud image is fused into an RGB image, and the RGB image is converted into the color space model HSV to obtain the light intensity characteristics. The calculation formulas for each component of the actual ground-based cloud image AHSV model are as follows: ; ; ; Where, AR, AG, and AB are the red, green, and blue color channel values ​​in the ARGB model of the actual ground-based cloud image; max = max(AR, AG, AB), which indicates the largest of the three color channel values; min = min(AR, AG, AB), which indicates the smallest of the three color channel values; AH is the hue component value, which represents the position information of the image color, that is, the spectral color, expressed in degrees, where 0 is the maximum value. o , 120 o , 240 o They correspond to red, green and blue respectively; AS is the saturation component value, which represents the numerical characteristics of color depth; AV is the lightness component value, that is, the color brightness level characteristics, and its value corresponds to the interval [0,1], indicating that the value range is 0.0 (black) to 1.0 (white).

[0030] The calculated AH, AS, and AV are used as the illumination intensity features of the ground-based cloud image fusion image.

[0031] (3-4-b) Extract key features of sky cloud cover from ground-based cloud image fusion images; ; Among them, F represents the cloud cover in the sky; Represents the number of cloud pixels, The number of pixels representing the sky area; (3-4-c) Transmittance affects the solar energy received by the photovoltaic panel and can reflect the clarity of the ground cloud map. The actual ground cloud map transmittance is calculated as follows: ; In the formula, is the estimated value of the actual cloud image transmittance, J C is a color channel of J, is a square area with its center point x a , A is the global sunlight component; (3-4-d) The zenith distance reflects the relative distance between the sun and the all-sky imager. The zenith distance Ad of the actual cloud image and the predicted cloud image is as follows: ; Where (x0, y0) is the position of the all-sky imager in the ground-based cloud map, (xA ,y A ) is the corresponding position of the sun in the actual ground-based cloud map.

[0032] The step S4 comprises the following steps: (4-1) Each photovoltaic station is regarded as a node in the adjacency matrix. The distance between photovoltaic stations represents the edge between nodes. For the preferred meteorological features in the preferred meteorological feature dataset and the fused cloud map features, a graph attention network is used to realize node-level information aggregation and feature embedding to capture the potential spatial structure information in the data.

[0033] The input of the graph attention network is the positional relationship between photovoltaic stations and the characteristic data of each station. The node information of different stations is aggregated by introducing the attention mechanism. For the neighboring station information around each station in the adjacency matrix, the weights learned by the model are used for weighted aggregation. First, the node features are linearly transformed, and then the attention score between nodes is calculated and normalized using the attention mechanism. Finally, the station information around each station is weightedly aggregated to obtain high-dimensional features that integrate the surrounding station information.

[0034] The graph attention network uses the self-attention mechanism to calculate the attention weight between each photovoltaic station node in the adjacency matrix and its neighboring station nodes, so that the state update of each photovoltaic station node will take into account the state of its neighboring station nodes; the node features are aggregated by weighted summation, and the weight is determined by the attention coefficient. In this way, the model can automatically learn which neighboring station nodes around the photovoltaic station are more important to the state update of the current node; using a multi-head attention mechanism, each attention head independently calculates the attention score, and finally splices the results of all heads, so that the model can pay attention to different features or patterns, thereby capturing more information. Through these characteristics, the graph attention network can effectively capture the potential spatial structure information in cluster photovoltaic data, providing strong support for graph data analysis tasks such as cluster photovoltaic power prediction.

[0035] (4-2) The Scaleformer network model adopts a multi-scale processing method. According to the spatial features and target cluster power of the previous step, the loss function is minimized and trained to generate the final cluster power prediction value. While using multiple scales to ensure high accuracy, it reduces the computational overhead by sharing weights and optimizing the architecture design, which is conducive to improving the accuracy, real-time and robustness of the prediction.

[0036] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A cluster photovoltaic power prediction method based on ground-based cloud map, characterized in that: include: S1. Data collection: record the location information of each photovoltaic station in the cluster, collect ground-based cloud map data and power data of each photovoltaic station, and obtain numerical weather forecast NWP data; S2, NWP data preprocessing, using the maximum correlation minimum redundancy algorithm to screen out meteorological factors with high correlation and obtain the optimal meteorological characteristic data set; S3, preprocessing of ground-based cloud map data, performing registration and fusion operations on ground-based cloud maps at different points, extracting key features of cloud maps, and obtaining cloud map key feature data sets; S4. Training a cluster photovoltaic power prediction model based on the preferred meteorological feature data set and the cloud image key feature data set, and obtaining cluster photovoltaic predicted power according to the model.

2. The cluster photovoltaic power prediction method based on ground-based cloud map according to claim 1 is characterized in that: In step S1, obtaining numerical weather forecast NWP data includes obtaining numerical weather forecast data sets from different meteorological sources in the same time period, the data sets including irradiance, wind speed, wind direction, temperature, humidity, and air pressure data; and processing abnormal values ​​in irradiance data that violate the actual lighting laws.

3. The cluster photovoltaic power prediction method based on ground-based cloud map according to claim 1 is characterized in that: Step S2 includes: S201, performing maximum and minimum normalization processing on the acquired NWP data; S202, using the maximum correlation minimum redundancy algorithm, taking the power data of the cluster photovoltaic as the target variable, calculating the correlation between the NWP data of different meteorological sources and the target variable, and obtaining the optimal meteorological characteristic data set by maximizing the correlation and minimizing the redundancy between the NWP data.

4. The cluster photovoltaic power prediction method based on ground-based cloud map according to claim 1 is characterized in that: Step S3 includes: S301, retaining the minimum circumscribed rectangle containing the spherical mirror for the ground-based cloud image data of each station, using median filtering to process noise points, and obtaining a pre-processed ground-based cloud image; S302, calculating the distance between photovoltaic stations according to the longitude and latitude data of the location information of each photovoltaic station, and storing it in an adjacency matrix; S303, sorting the photovoltaic stations according to the distances between them to obtain the distance relationship between the locations of the stations, performing image registration of the pre-processed ground cloud images in order from near to far, and fusing the registered images; S304, extracting cloud map key features from the fused ground-based cloud map fusion image to obtain a cloud map key feature data set; the cloud map key features include light intensity, cloud cover percentage, transmittance and zenith distance.

5. The cluster photovoltaic power prediction method based on ground-based cloud map according to claim 4 is characterized in that: Step S4 includes: S401, treating each photovoltaic station as a node, representing the relationship between nodes through the adjacency matrix, and using the graph attention network to achieve node-level information aggregation and feature embedding for the optimal meteorological feature data set and the cloud map key feature data set, to obtain the spatial features of each node; S402, based on the Scaleformer network model, multi-scale processing is adopted, the spatial characteristics of each node are used as input variables, the power data of the cluster photovoltaic is used as the target variable, and the minimum loss function is used as the goal, and the cluster photovoltaic power prediction model is obtained through training; S403: Predicting according to the cluster photovoltaic power prediction model to obtain a final cluster power prediction value.

6. A cluster photovoltaic power prediction device based on ground-based cloud map, characterized in that: include: Data collection module: records the location information of each photovoltaic station in the cluster, collects ground-based cloud map data and power data of each photovoltaic station, and obtains numerical weather forecast NWP data; Meteorological feature module: used for NWP data preprocessing, using the maximum correlation minimum redundancy algorithm to screen out meteorological factors with high correlation and obtain the optimal meteorological feature data set; Cloud map feature module: used for ground-based cloud map data preprocessing, registering and fusing ground-based cloud maps at different points, extracting key features of cloud maps, and obtaining cloud map key feature data sets; Prediction model module: training a cluster photovoltaic power prediction model based on the preferred meteorological feature data set and the cloud image key feature data set, and obtaining the cluster photovoltaic predicted power according to the model.

7. The cluster photovoltaic power prediction device based on ground-based cloud map according to claim 6, characterized in that: In the data collection module, obtaining numerical weather forecast NWP data includes obtaining numerical weather forecast data sets from different meteorological sources in the same time period, which include irradiance, wind speed, wind direction, temperature, humidity, and air pressure data; and processing abnormal values ​​in irradiance data that violate the actual lighting laws.

8. The cluster photovoltaic power prediction device based on ground-based cloud map according to claim 6, characterized in that: Meteorological feature modules include: Normalization unit: perform maximum and minimum normalization processing on the acquired NWP data; Correlation unit: Using the maximum correlation minimum redundancy algorithm, with the power data of cluster photovoltaics as the target variable, the correlation between the NWP data of different meteorological sources and the target variable is calculated, and the optimal meteorological characteristic data set is obtained by maximizing the correlation and minimizing the redundancy between NWP data.

9. The cluster photovoltaic power prediction device based on ground-based cloud map according to claim 6, characterized in that: The cloud map feature modules include: Preprocessing unit: For the ground-based cloud image data of each station, the minimum circumscribed rectangle containing the spherical mirror is retained, and the noise points are processed using median filtering to obtain the preprocessed ground-based cloud image; Matrix unit: Calculate the distance between photovoltaic stations based on the longitude and latitude data of the location information of each photovoltaic station and store it in an adjacency matrix; Fusion unit: Sort the photovoltaic stations according to their distances to obtain the distance relationship between the locations of the stations, perform image registration of the pre-processed ground cloud images from near to far, and fuse the registered images; Feature extraction unit: extracts key features of cloud images from the fused ground-based cloud image to obtain a cloud image key feature data set; the key features of the cloud image include light intensity, cloud percentage, transmittance and zenith distance.

10. The cluster photovoltaic power prediction device based on ground-based cloud map according to claim 9, characterized in that: The prediction model module includes: Spatial feature unit: each photovoltaic station is regarded as a node, and the relationship between nodes is represented by the adjacency matrix. For the optimal meteorological feature data set and cloud image key feature data set, the graph attention network is used to realize node-level information aggregation and feature embedding to obtain the spatial features of each node; Model unit: Based on the Scaleformer network model, multi-scale processing is adopted, the spatial characteristics of each node are used as input variables, the power data of cluster photovoltaics is used as the target variable, and the minimum loss function is used as the goal. The cluster photovoltaic power prediction model is obtained through training; Prediction unit: performs prediction according to the cluster photovoltaic power prediction model to obtain a final cluster power prediction value.

Images