CN-121983971-A - Distributed photovoltaic short-term generation power prediction method, system, equipment and medium

CN121983971ACN 121983971 ACN121983971 ACN 121983971ACN-121983971-A

Abstract

The invention relates to the technical field of photovoltaic power generation, and provides a method, a system, equipment and a medium for predicting distributed photovoltaic short-term power generation power, which comprise the steps of obtaining foundation cloud picture time sequence data, historical power time sequence data and cluster distributed photovoltaic information of a target photovoltaic power station; the method comprises the steps of extracting characteristics of foundation cloud picture time sequence data to obtain cloud picture characteristic time sequence data, carrying out power prediction based on a first prediction model to obtain a first prediction result according to the cloud picture characteristic time sequence data and historical power time sequence data, carrying out power prediction based on a second prediction model to obtain a second prediction result according to space-time coupling characteristic time sequence data obtained based on cluster distributed photovoltaic information, cloud coverage rate time sequence data and cloud layer change rate time sequence data, and carrying out weighted fusion on the first prediction result and the second prediction result to obtain a target prediction result. The method can remarkably improve the reliability, stability and accuracy of the short-term power prediction result under the sudden complex meteorological scene such as cloudiness, overcast and rainy.

Inventors

Weng Xiuliang
WANG TINGTING
CHE YINFEI
YU XIAOZHANG
YANG RUIJUN
CHEN MENGXIANG
YE HAORAN
Huang Changzhan
WEN QINGKAO
CHEN WEIWEI
XU FANG
HUANG SHAN
LIN DAOZONG
HU CHANGHONG
LI HENAN
LIU JINYUAN

Assignees

国网浙江省电力有限公司平阳县供电公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. A method for predicting short-term generated power of a distributed photovoltaic, the method comprising: Acquiring foundation cloud picture time sequence data, historical power time sequence data and cluster distributed photovoltaic information of a target photovoltaic power station; The cloud picture characteristic time sequence data comprises edge characteristic time sequence data, texture characteristic time sequence data, cloud cover rate time sequence data and cloud change rate time sequence data; According to the cloud picture characteristic time sequence data and the historical power time sequence data, carrying out power prediction based on a first prediction model constructed in advance to obtain a first prediction result; Acquiring time-space coupling characteristic time sequence data according to the cluster distributed photovoltaic information, the cloud cover rate time sequence data and the cloud cover change rate time sequence data, and performing power prediction based on a second prediction model constructed in advance according to the time-space coupling characteristic time sequence data to obtain a second prediction result; and carrying out weighted fusion on the first prediction result and the second prediction result to obtain a target prediction result.
2. The method for predicting the short-term generated power of a distributed photovoltaic system according to claim 1, wherein the step of extracting features from the cloud image time series data of the foundation to obtain the cloud image feature time series data comprises the steps of: Preprocessing each foundation cloud picture in the foundation cloud picture time sequence data respectively to obtain foundation cloud picture time sequence data to be analyzed; performing edge feature extraction on the foundation cloud picture time sequence data to be analyzed based on a dynamic scale direction gradient histogram technology to obtain the edge feature time sequence data; Extracting texture features of the foundation cloud picture time sequence data to be analyzed based on a multidirectional weighting gray level co-occurrence matrix technology to obtain the texture feature time sequence data; and carrying out cloud characteristic analysis on the time sequence data of the foundation cloud picture based on the Rayleigh scattering principle to obtain the time sequence data of the cloud coverage rate and the time sequence data of the cloud change rate.
3. The method for predicting the short-term generated power of a distributed photovoltaic system according to claim 2, wherein the step of extracting edge features of the time series data of the foundation cloud image to be analyzed based on the dynamic scale direction gradient histogram to obtain the time series data of the edge features comprises the steps of: according to candidate unit scales in a preset unit scale set, respectively carrying out pixel gradient energy variation analysis on each foundation cloud image to be analyzed in the foundation cloud image time sequence data to be analyzed to obtain target scales of different pixel point positions in each foundation cloud image to be analyzed; According to the target scales of different pixel point positions in each foundation cloud image to be analyzed, carrying out directional gradient histogram calculation on the foundation cloud image to be analyzed to obtain corresponding image histogram features; And summarizing the image histogram features corresponding to all the foundation cloud pictures to be analyzed, and generating the edge feature time sequence data.
4. The method for predicting the short-term generated power of a distributed photovoltaic system according to claim 2, wherein the step of extracting texture features from the time series data of the foundation cloud image to be analyzed based on the multidirectional weighted gray level co-occurrence matrix technique to obtain the time series data of the texture features comprises the steps of: respectively carrying out gray level co-occurrence matrix calculation on each foundation cloud picture to be analyzed in the foundation cloud picture time sequence data to be analyzed according to a preset direction angle set and a preset pixel interval to obtain a corresponding direction gray level co-occurrence matrix set; respectively extracting texture features of each direction gray level co-occurrence matrix in the direction gray level co-occurrence matrix set of each foundation cloud picture to be analyzed to obtain a corresponding texture feature set, wherein the texture feature set comprises texture features of each direction angle; Respectively carrying out structure tensor analysis on each foundation cloud image to be analyzed to obtain corresponding global dominant direction angles, and respectively carrying out angle distance similarity analysis on the global dominant direction angles and each direction angle in the preset direction angle set to obtain corresponding direction weight sets; According to the direction weight set of each foundation cloud image to be analyzed, carrying out weighted fusion on the texture features of each direction angle in the texture feature set to obtain corresponding image texture features; And summarizing the image texture features corresponding to all the foundation cloud pictures to be analyzed, and generating the texture feature time sequence data.
5. The method for predicting the short-term generated power of a distributed photovoltaic system according to claim 4, wherein the step of performing a structure tensor analysis on each of the to-be-analyzed ground cloud patterns to obtain a corresponding global dominant direction angle comprises: respectively obtaining image gradient fields of each foundation cloud image to be analyzed, and calculating corresponding pixel point gradient outer product matrixes according to gradient components of different pixel point positions in the image gradient fields; Accumulating corresponding elements in all pixel point gradient outer product matrixes of the foundation cloud pictures to be analyzed to obtain corresponding structure tensor matrixes; and carrying out eigenvalue decomposition on the structure tensor matrix of each foundation cloud picture to be analyzed, obtaining an eigenvector corresponding to the minimum eigenvalue as a texture dominant eigenvector, and carrying out direction information extraction on the texture dominant eigenvector to obtain the global dominant direction angle.
6. The method for predicting the short-term generated power of a distributed photovoltaic system according to claim 2, wherein the step of performing cloud characteristic analysis on the time series data of the base cloud map based on the rayleigh scattering principle to obtain the time series data of the cloud coverage rate and the time series data of the cloud change rate comprises the following steps: calculating pixel brightness attenuation coefficients corresponding to all pixel points in the foundation cloud images according to the coordinates of the solar pixel points of all the foundation cloud images in the foundation cloud images to be analyzed, and carrying out pixel binarization processing according to the brightness attenuation coefficients and the corresponding red-blue channel ratios to obtain corresponding binarized gray level cloud images; Obtaining corresponding cloud cover coverage according to the binarization gray scale accumulated values of all pixel points in the binarization gray scale cloud maps of each foundation cloud map and the foundation cloud map size, and summarizing the cloud cover coverage of all the foundation cloud maps to generate the cloud cover coverage time sequence data; And carrying out change processing on the cloud cover coverage time sequence data to obtain corresponding cloud cover change time sequence data, and obtaining the cloud cover change rate time sequence data according to the cloud cover change time sequence data and time sequence data duration intervals.
7. The distributed photovoltaic short-term generated power prediction method according to claim 1, wherein the clustered distributed photovoltaic information includes geographic locations, power timing data and weather timing data of all photovoltaic power stations in a cluster where the target photovoltaic power station is located; The step of obtaining time-space coupling characteristic time sequence data according to the cluster distributed photovoltaic information, the cloud coverage time sequence data and the cloud change rate time sequence data comprises the following steps: Performing cluster analysis according to the geographic position and power time sequence data of each photovoltaic power station in the cluster distributed photovoltaic information to obtain a target photovoltaic power station subgroup; constructing corresponding graph node feature matrix time sequence data according to the power time sequence data and the meteorological time sequence data of all photovoltaic power stations in the target photovoltaic power station subgroup; Calculating corresponding side weight time sequence data according to the graph node characteristic matrix time sequence data, the cloud cover rate time sequence data and the cloud change rate time sequence data, wherein the side weight time sequence data comprises side weights between photovoltaic power station pairs in the target photovoltaic power station subgroup; and generating weighted graph adjacent matrix time sequence data according to the edge weight time sequence data, and generating the space-time coupling characteristic time sequence data according to the weighted graph adjacent matrix time sequence data and the graph node characteristic time sequence data.
8. A distributed photovoltaic short-term generated power prediction system, the system comprising: the data acquisition module is used for acquiring foundation cloud picture time sequence data, historical power time sequence data and cluster distributed photovoltaic information of the target photovoltaic power station; The characteristic extraction module is used for carrying out characteristic extraction on the foundation cloud picture time sequence data to obtain cloud picture characteristic time sequence data, wherein the cloud picture characteristic time sequence data comprises edge characteristic time sequence data, texture characteristic time sequence data, cloud coverage rate time sequence data and cloud layer change rate time sequence data; The first prediction module is used for carrying out power prediction based on a first prediction model constructed in advance according to the cloud picture characteristic time sequence data and the historical power time sequence data to obtain a first prediction result; The second prediction module is used for acquiring time-space coupling characteristic time sequence data according to the cluster distributed photovoltaic information, the cloud coverage rate time sequence data and the cloud layer change rate time sequence data, and performing power prediction based on a second prediction model constructed in advance according to the time-space coupling characteristic time sequence data to obtain a second prediction result; And the result generation module is used for carrying out weighted fusion on the first prediction result and the second prediction result to obtain a target prediction result.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.

Description

Distributed photovoltaic short-term generation power prediction method, system, equipment and medium Technical Field The invention relates to the technical field of photovoltaic power generation, in particular to a method, a system, equipment and a medium for predicting short-term power generation power of distributed photovoltaic. Background Along with the large-scale grid connection of the distributed photovoltaic, the short-term power prediction result becomes an important basis for power grid optimization scheduling, and important support is provided for safe and stable operation of the power grid. However, the actual output of the distributed photovoltaic is obviously affected by factors such as weather mutation, dynamic shielding of cloud layers and the like, so that short-term power fluctuation is severe, and the difficulty of accurate short-term power prediction is also increased. In the prior art, although the influence of cloud layer information on short-term power prediction accuracy is recognized, only shallow sub-image static characteristics of a cloud layer are used, the effects of deep image characteristics and cloud layer dynamic trends are not focused, cloud image information cannot be fully utilized, the relevance of output force among distributed photovoltaic power stations in a cluster is not considered, and reliability and accuracy of prediction results in mutation complex meteorological scenes such as cloudiness, overcast and rainy are not effectively ensured. Disclosure of Invention The invention aims to provide a distributed photovoltaic short-term generation power prediction method, which is based on a multi-level cloud image characteristic analysis mechanism for extracting edge texture characteristics and dynamic change characteristics of a foundation cloud image, combines a moment coupling characteristic extraction analysis mechanism between photovoltaic power stations based on the dynamic change characteristics, realizes effective integration and full utilization of multi-mode data, can remarkably improve reliability, stability and accuracy of short-term prediction results under mutation complex meteorological scenes such as cloudiness, overcast and rainy, and provides high-precision and high-robustness scheduling analysis support for safe and stable operation of a high-proportion new energy grid and efficient consumption of new energy. In order to achieve the above object, it is necessary to provide a distributed photovoltaic short-term generated power prediction method, system, computer device, and storage medium. In a first aspect, an embodiment of the present invention provides a method for predicting a short-term generated power of a distributed photovoltaic, the method including the steps of: Acquiring foundation cloud picture time sequence data, historical power time sequence data and cluster distributed photovoltaic information of a target photovoltaic power station; The cloud picture characteristic time sequence data comprises edge characteristic time sequence data, texture characteristic time sequence data, cloud cover rate time sequence data and cloud change rate time sequence data; According to the cloud picture characteristic time sequence data and the historical power time sequence data, carrying out power prediction based on a first prediction model constructed in advance to obtain a first prediction result; Acquiring time-space coupling characteristic time sequence data according to the cluster distributed photovoltaic information, the cloud cover rate time sequence data and the cloud cover change rate time sequence data, and performing power prediction based on a second prediction model constructed in advance according to the time-space coupling characteristic time sequence data to obtain a second prediction result; and carrying out weighted fusion on the first prediction result and the second prediction result to obtain a target prediction result. Further, the step of extracting the characteristics of the cloud image time sequence data of the foundation to obtain the cloud image characteristic time sequence data comprises the following steps: Preprocessing each foundation cloud picture in the foundation cloud picture time sequence data respectively to obtain foundation cloud picture time sequence data to be analyzed; performing edge feature extraction on the foundation cloud picture time sequence data to be analyzed based on a dynamic scale direction gradient histogram technology to obtain the edge feature time sequence data; Extracting texture features of the foundation cloud picture time sequence data to be analyzed based on a multidirectional weighting gray level co-occurrence matrix technology to obtain the texture feature time sequence data; and carrying out cloud characteristic analysis on the time sequence data of the foundation cloud picture based on the Rayleigh scattering principle to obtain the time sequence data of the cloud coverage rate and the tim