CN114676893B - Photovoltaic power station solar irradiance short-term prediction method based on optimal graph structure and storage medium - Google Patents
Photovoltaic power station solar irradiance short-term prediction method based on optimal graph structure and storage medium Download PDFInfo
- Publication number
- CN114676893B CN114676893B CN202210241377.0A CN202210241377A CN114676893B CN 114676893 B CN114676893 B CN 114676893B CN 202210241377 A CN202210241377 A CN 202210241377A CN 114676893 B CN114676893 B CN 114676893B
- Authority
- CN
- China
- Prior art keywords
- station
- irradiance
- space
- data
- nodes
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000000547 structure data Methods 0.000 claims abstract description 43
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 238000012216 screening Methods 0.000 claims abstract description 13
- 239000011159 matrix material Substances 0.000 claims description 22
- 238000004364 calculation method Methods 0.000 claims description 17
- 238000011156 evaluation Methods 0.000 claims description 16
- 238000013507 mapping Methods 0.000 claims description 11
- 238000012937 correction Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 230000002776 aggregation Effects 0.000 claims description 7
- 238000004220 aggregation Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 3
- 241001502050 Acis Species 0.000 claims 1
- 238000012935 Averaging Methods 0.000 claims 1
- 235000006629 Prosopis spicigera Nutrition 0.000 claims 1
- 240000000037 Prosopis spicigera Species 0.000 claims 1
- 238000004458 analytical method Methods 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 abstract description 9
- 238000010586 diagram Methods 0.000 abstract description 3
- 238000004146 energy storage Methods 0.000 abstract description 3
- 238000010276 construction Methods 0.000 abstract 1
- 230000000875 corresponding effect Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 238000010248 power generation Methods 0.000 description 4
- 101150090997 DLAT gene Proteins 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000005855 radiation Effects 0.000 description 3
- 238000010219 correlation analysis Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 108010015046 cell aggregation factors Proteins 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 239000002803 fossil fuel Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Marketing (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a photovoltaic power station solar irradiance short-term prediction method based on an optimal diagram structure, which comprises the following steps: construction by geographic scope and bearing information space-time correlation stations of the target station; acquiring target station and space-time associations station historical irradiance time series data; calculating the correlation between the space-time correlation station and the target station data, and screening space-time correlation stations with high correlation; constructing graph structure data according to the space-time associated stations obtained by screening; and establishing a graph neural network prediction model taking graph structure data as input to realize irradiance short-term prediction. Compared with the prior art, the method and the device can fully consider the influence of the weather irradiance condition of the adjacent range of the target prediction station on the space-time correlation of the target station under the condition of the missing surrounding weather data, and only the historical irradiance data is used for realizing short-term prediction of irradiance, so that the method and the device are easy to realize, are beneficial to improving the short-term prediction precision of irradiance, are beneficial to providing technical support for planning and site selection of photovoltaic power stations, and are also beneficial to carrying out reasonable energy storage guidance on the built photovoltaic power stations.
Description
Technical Field
The invention belongs to the technical field of irradiance prediction, and particularly relates to a photovoltaic power station solar irradiance short-term prediction technology based on an optimal graph structure.
Background
The development of renewable energy sources is increasingly being appreciated and supported by all countries of the world due to the rapid development of global economy and technology, as well as environmental pollution and climate problems caused by traditional fossil fuels. Solar energy is one of renewable energy sources with the most development prospect at present, and photovoltaic power generation is used as a main utilization mode of solar energy and is rapidly developed. The output power of photovoltaic power generation is mainly influenced by solar radiation reaching the earth surface, the earth surface solar radiation is also influenced by the solar radiation intensity, the atmosphere, cloud cluster conditions and the like, and the output power shows the characteristics of intermittence, randomness and the like, so that the prediction essence of the photovoltaic power generation is the prediction of the irradiance of the earth surface, and the improvement of the prediction accuracy of the irradiance of the sun is a major problem to be solved urgently and is concerned.
The existing solar irradiance prediction method mainly comprises irradiance prediction models based on time sequences, wherein the irradiance prediction models comprise traditional statistical methods and artificial intelligent algorithms, and support vector machines, RNNs, LSTM and other machine learning methods. The traditional statistical method has the advantages that the prediction model is simple in structure, a single deep learning network model is difficult to learn too much useful information from severe fluctuation data, the influence of space-time correlation cannot be considered from the data, and the current deep learning model has the problem of prediction lag generally, so that the prediction accuracy is low. According to the method, irradiance prediction is performed on the earth surface solar irradiance within a logarithmic hour based on an irradiance prediction model of a satellite meteorological image, and the irradiance prediction is realized mainly by extracting image characteristic information from a satellite cloud image or a sky image. Because cloud is the most main meteorological environment factor affecting the solar irradiance reaching the ground, the generation, elimination and movement of the cloud are one of the root causes enabling the solar irradiance on the ground to have random and intermittent changes, and the irradiance prediction model taking the cloud cluster motion effect and weather conditions into consideration grasps the movement rule through monitoring the cloud, predicts the position state of the cloud at a certain moment in the future, and further calculates the irradiance of the area at the moment through a physical modeling method. However, the subsequent physical modeling process is complex, parameter solving is difficult, and the like, so that the prediction accuracy is influenced.
The irradiance prediction methods have some problems to influence the prediction accuracy, have no space-time characteristics of deep mining historical irradiance sequences, and consider the influence of irradiance conditions in the adjacent range of a target station on the target station, so that the prediction accuracy is not high.
Disclosure of Invention
The invention mainly aims to provide a photovoltaic power station solar irradiance short-term prediction technical method based on an optimal graph structure, and aims to solve the problems that the existing irradiance prediction method does not consider the influence of irradiance conditions in the adjacent range of a target station on the target station and the prediction effect is poor due to insufficient mining of space-time information of a historical irradiance sequence.
In order to achieve the above purpose, the invention provides a photovoltaic power station solar irradiance short-term prediction method based on an optimal graph structure, which comprises the following steps:
s10, constructing a space-time associated station of a target station through geographic range and azimuth information;
step S20, acquiring historical irradiance time series data of a target station and each time-space associated station;
step S30, calculating the correlation between the space-time correlation field station and the target field station data, and screening space-time correlation field stations with high correlation;
step S40, constructing graph structure data according to the space-time associated field stations obtained by screening;
and S50, establishing a graph neural network prediction model taking graph structure data as input to realize irradiance short-term prediction.
Further, the step S10 includes:
step S101, intercepting the geographic space range of the area where the target station is located, wherein other stations in the range are space-time associated stations.
Step S102, longitude and latitude coordinate information of the space-time correlation station is obtained.
The longitude and latitude coordinate information of a plurality of time-space associated stations is refined and generated by utilizing an automatic grid generation algorithm.
Further, the step S20 includes:
step S201, using a satellite cloud image-irradiance mapping method, mapping historical irradiance time series data of a target station and each space-time associated station from the satellite cloud image.
Step S202, training a data deviation correction model by using the target station data, and performing deviation correction on irradiance of the time-associated station.
Further, the step S30 includes:
step S301, normalized mutual information N is utilized MI A correlation between the target station and each spatio-temporal associated station historical irradiance data is calculated.
Step S302, each time-space associated station is divided into N MI The values are sorted from big to small, only the top 1/3 of the sorting is retained.
Further, the step S40 includes:
step S401, constructing different graph structure data according to the space-time associated stations obtained by screening, wherein the graph structure data consists of a node characteristic matrix and an adjacent matrix, the node characteristic matrix consists of a target station and historical irradiance data of each space-time associated station, and the adjacent matrix is a connection relation determined according to the correlation among the station data.
In step S402, the graph structure data is optimal graph structure data obtained based on graph connectivity evaluation.
Calculating a complex network topology statistical index according to the adjacency matrix; the complex network topology statistical indexes comprise average degree, average aggregation coefficient, average shortest path distance among nodes, average node betweenness centrality and network connectivity.
And weighting and calculating the statistical indexes of each complex network topology to obtain the comprehensive evaluation index of the graph connectivity.
And (3) calculating a graph connectivity comprehensive evaluation index of the constructed graph structure data in the step S401, when the comprehensive evaluation index is larger than a threshold value, considering the graph connectivity comprehensive evaluation index as the optimal graph structure data, otherwise, returning to the step S401 for circulation until the optimal graph structure data is constructed.
Further, the step S50 includes:
step S501, performing normalization processing on the historical irradiance data of each station in the node feature matrix in the graph structure data.
Step S502, a graph neural network prediction model is built by taking graph structure data as input and irradiance as output.
Step S503, the graph structure data is input into a graph neural network prediction model, and the output irradiance prediction result is subjected to inverse normalization processing to obtain a final target station irradiance prediction result.
The invention also provides a storage medium for storing a computer program which is executed by a processor to realize the photovoltaic power station solar irradiance short-term prediction method based on the optimal graph structure.
The beneficial effects of the invention are as follows:
(1) The solar irradiance short-term prediction method can fully consider the space-time correlation influence of the weather irradiance condition of the adjacent range of the target prediction station on the target station under the condition of missing surrounding weather data, and the influence of surrounding factors on the target station is often ignored by the person skilled in the art.
(2) The solar irradiance short-term prediction method disclosed by the invention only needs to utilize the historical irradiance time sequence to realize short-term prediction of irradiance, is easy to realize, is beneficial to improving the short-term prediction precision of irradiance, is beneficial to providing technical support for planning and site selection of photovoltaic power stations, and is also beneficial to carrying out reasonable energy storage guidance on the built photovoltaic power stations.
(3) The solar irradiance short-term prediction method considers irradiance of the adjacent range stations, compared with other consideration methods, the space-time correlation stations in the method are purposefully and autonomously constructed and screened, have stronger space-time correlation with the target stations, and provide graph connectivity indexes based on complex network topology indexes to select optimal graph structure data.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
FIG. 1 is a schematic flow chart of a first embodiment of a short-term solar irradiance prediction method based on an optimal graph structure of the present invention;
FIG. 2 is a graph of historical irradiance data correlation analysis results for each space-time associated station and target station constructed by the solar irradiance short-term prediction method based on the optimal graph structure of the present invention;
FIG. 3 is a graph showing irradiance prediction results of the short-term solar irradiance prediction method based on the optimal graph structure of the present invention.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Referring to fig. 1, the invention provides a photovoltaic power station solar irradiance prediction method based on an optimal graph structure, which comprises the following steps:
s10, constructing a space-time associated station of a target station through geographic range and azimuth information;
step S20, acquiring historical irradiance time series data of a target station and each time-space associated station;
step S30, calculating the correlation between the space-time correlation field station and the target field station data, and screening space-time correlation field stations with high correlation;
step S40, constructing graph structure data according to the space-time associated field stations obtained by screening;
and S50, establishing a graph neural network prediction model taking graph structure data as input to realize irradiance short-term prediction.
The constructing the space-time associated field station of the target field station through the geographic range and the azimuth information in the step S10 includes:
step S101, intercepting the geographic space range of the area where the target station is located, wherein other stations in the range are space-time associated stations. Taking a target station as a center, intercepting a rectangular area, calculating four vertex coordinates of the rectangular area, and calculating the rectangular area according to the following formulas (1) - (6).
left-top＝(lat+dlat,lon-dlon) (3)
left-bottom＝(lat-dlat,lon-dlon) (4)
right-top＝(lat+dlat,lon+dlon) (5)
right-bottom＝(lat-dlat,lon+dlon) (6)
Wherein R represents the geographical distance between the vertex and the target station, R=150 km is taken in the invention, R is taken as Ground (floor) = 6378.137km denotes the earth radius, lon is the target station longitude, lat is the target station latitude, in degrees,the radian is converted into an angle, dlon and dlat represent the offset of longitude and latitude, left-top, left-bottom, right-top and right-bottom represent the coordinates of four vertexes of a rectangular frame respectively, and other field stations in the rectangular frame are space-time switchesThe station is a real photovoltaic power station or a virtual power station, because the station can be assumed to be a photovoltaic power station as long as the station has coordinate information and corresponding irradiance time series data.
Step S102, longitude and latitude coordinate information of the space-time correlation station is obtained. And (3) generating longitude and latitude coordinate information of a plurality of space-time associated stations in a refining mode by using an automatic grid generation algorithm in the geographic space range of the area generated in the step S101 and taking the target station as the center according to different distances and different azimuth angles.
The step of obtaining historical irradiance time series data of the target station and each space-time associated station in the step S20 includes:
step S201, using a satellite cloud image-irradiance mapping method, mapping historical irradiance time series data of a target station and each space-time associated station from the satellite cloud image. By adopting the satellite cloud image-irradiance mapping method in the patent application 202111336372.8, longitude and latitude coordinates of the target station and the space-time correlation station obtained in the step S102 are input, and historical irradiance time series data of the corresponding position is output.
Because the data obtained by mapping has deviation from the actual value, deviation correction and data quality control are required.
Step S202, training a data deviation correction model by using the target station data, and performing deviation correction on irradiance of the time-associated station. Taking the irradiance mapping value and the temperature information of the target station as input, taking the irradiance actual value of the target station as output, and training by using an SVM model to obtain a data deviation correction model; and correcting the deviation of irradiance mapping values of other space-time related stations by using a data deviation correction model obtained by training, and further carrying out an abnormal data analysis preprocessing process on the irradiance mapping values to remove data with larger deviation and complement missing values in order to improve the prediction accuracy.
The step S30 of calculating the correlation between the space-time correlation station and the target station data and screening the space-time correlation station with high correlation includes:
step S301, normalized mutual information N is utilized MI Calculation ofCorrelation between the target station historical irradiance data obtained in step S20 and each spatiotemporal correlated station historical irradiance data. The calculation formula is as follows:
p r (N MI (X,Y)≥τ)＝2/3
where X and Y are input data of historical irradiance of the target station and the spatio-temporal associated station, pr (·) is a probability function, and pr (X, Y) is a joint probability. N (N) MI The higher the value, the stronger the correlation.
Step S302, each time-space associated station is divided into N MI The values are sorted from big to small, only the top 1/3 of the sorting is retained. Or choose N MI The time-space correlation station with a value greater than the threshold value can be valued according to the actual situation and the experience of the person skilled in the art.
Referring to fig. 2, the result of correlation analysis of historical irradiance data of a space-time correlation station and a target station constructed based on a solar irradiance short-term prediction model of an optimal diagram structure is shown. The abscissa indicates the reference numerals of the respective stations, wherein 0 indicates the target station, and the color in each cell represents the correlation calculation result between the two stations, and the deeper the color, the stronger the correlation.
The step S40 of constructing the optimal diagram structural data according to the space-time associated field station obtained by the screening includes:
step S401, constructing different graph structure data according to the space-time associated stations obtained by screening, wherein the graph structure data consists of a node characteristic matrix and an adjacent matrix, the node characteristic matrix consists of a target station and historical irradiance data of each space-time associated station, and the adjacent matrix is a connection relation determined according to the correlation among the station data.
A graph neural network is a neural network that can process graph structure data. A graph typically contains nodes and edges, each node having its own attributes or characteristics, while edges represent relationships between nodes. Only if irradiance data is converted into map structure data can irradiance be predicted using a map neural network model. The stations are thus considered nodes, with the correlation of irradiance data between stations as an attribute of the edge.
By selecting different data and setting different thresholds, different graph structure data are constructed, and the adjacency matrix is a connection relation determined according to the correlation between the station data, which may not be limited to the following definition:
in the formula (8), N MI (i, j) represents N between historical irradiance data for photovoltaic field station i and photovoltaic field station j MI The correlation coefficient epsilon represents the threshold value of the correlation coefficient, in the invention, half of the average value of the correlation matrix between stations is used as the threshold value, and can be set according to practical conditions, when the correlation coefficient between two photovoltaic power stations is smaller than the threshold value, the two photovoltaic power stations are considered to be not connected, and at the moment, the element A in the adjacent matrix is set ij Is 0.
In step S402, the graph structure data is optimal graph structure data obtained based on graph connectivity evaluation.
And calculating a complex network topology statistical index according to the adjacency matrix. The complex network topology statistical indexes include, but are not limited to, average degree, average aggregation coefficient, average shortest path distance between nodes, average node betweenness centrality and network connectivity, and the calculation formulas of the indexes are as follows:
equation (9) is an average degree calculation formula, where N is the total number of nodes in the network, and the number k of adjacent sides connected with node i i Called the degree of the node. Intuitively, the greater the degree of a node, the more important that node is. The degree of all nodes in the network is averaged to obtain the average degree DC of the network.
Equation (10) is an average aggregation coefficient calculation formula assuming nodes i and k i The individual nodes are directly connected, then for an undirected network this k i The maximum number of edges that may exist between individual nodes is k i While the number of actually existing edges is M i Whereby the aggregation factor is defined as C i AC is the average concentration factor.
Formula (11) is an average shortest path distance calculation formula among nodes, wherein G is a set of all nodes, N is the total number of nodes in the network, and the distance d between two nodes i and j in the network ij Is defined as the number of edges of the shortest path connecting the two nodes. For an undirected network, defining the average shortest path distance L between nodes as the distance d between node pairs in the network ij Average value of (2):
equation (12) is an average node betweenness centrality calculation formula, and the node betweenness is the proportion of the number of the nodes passing through all the shortest paths in the network. Wherein N is jl Representing the shortest path number, N, between nodes j and l jl (i) Representing the number of shortest paths between nodes j and l through node i. BC (BC) type i Represents node bets, and BC represents average node bets centrality.
Equation (13) is a calculation formula of network connectivity,for node-connectivity of node pair (i, j) in the network, the minimum number of nodes (except i, j) in the network that are to be removed such that node pair (i, j) (i+notejand i, j are not adjacent) are not connected. Connectivity of the graph network is represented by the average of node-connectivity for all non-adjacent node pairs in the network.
And weighting and calculating the statistical indexes of each complex network topology to obtain the comprehensive evaluation index of the graph connectivity. And (3) performing dimension reduction by using a principal component analysis method, representing all connectivity evaluation indexes by using a plurality of effective principal components with smaller quantity and higher contribution rate, determining weight coefficients corresponding to each effective principal component item by using an entropy weight method, and finally obtaining the graph connectivity comprehensive evaluation indexes according to each weight coefficient. The calculation formula is as follows, wherein k 1 +k 2 +k 3 +k 4 +k 5 ＝1。
GC＝k 1 ·DC+k 2 ·AC+k 3 ·L+k 4 ·BC+k 5 ·K (14)
And (3) calculating a graph connectivity comprehensive evaluation index of the constructed graph structure data in the step S401, when the comprehensive evaluation index is larger than a threshold value, considering the graph connectivity comprehensive evaluation index as the optimal graph structure data, otherwise, returning to the step S401 for circulation until the optimal graph structure data is constructed. For example, the threshold may be taken as 10.
In the step S50, the step of establishing a graph neural network prediction model using graph structure data as input to implement irradiance prediction includes:
step S501, performing normalization processing on the historical irradiance data of each station in the node feature matrix in the graph structure data.
In order to accelerate the training speed of the prediction model and improve the prediction precision, the historical irradiance data of each corresponding station in the node characteristic matrix in the graph structure data is required to be normalized according to the following formula, and the data is kept between 0 and 1:
wherein,X i and X i ' is the original data and normalized data, X min And X max Is the minimum and maximum values in the sample. The normalization process includes both irradiance data for the target site and irradiance data for the spatio-temporal associated site for use in subsequently building the predictive model.
Step S502, a graph neural network prediction model is built by taking graph structure data as input and irradiance as output. Dividing the graph structure data sequence in the step S401 or the optimal graph structure data sequence in the step S402 into a training set and a test set according to the proportion of 7:3, taking time sequence data of three days of graph structure history as input on the training set, taking irradiance data of a target station and each space-time associated station in one future day as output, setting initialization parameters for a short-term irradiance prediction model based on a graph neural network, and training, wherein the prediction model has four layers of networks, including a graph learning layer, two graph convolution layers and one output layer.
Step S503, the graph structure data is input into a graph neural network prediction model, and the output irradiance prediction result is subjected to inverse normalization processing to obtain a final target station irradiance prediction result.
To illustrate the effectiveness of the method, the irradiance value of a certain photovoltaic power station 2020.5.1-2020.5.5 of inner Mongolia is predicted by the method, and the comparison result of the predicted value and the actual value is shown in fig. 3. As can be seen from FIG. 3, the irradiance prediction method of the invention has ideal effect and higher prediction accuracy through practical tests.
The embodiment of the invention also provides a storage medium for storing a computer program which is executed by a processor to realize the solar irradiance short-term prediction method of the photovoltaic power station.
In summary, the irradiance short-term prediction method of the embodiment of the invention is beneficial to providing technical support for planning and locating photovoltaic power stations, meanwhile, the method is also beneficial to reasonable energy storage guidance for the built photovoltaic power station, promotes benign healthy development in the field of photovoltaic power generation and ensures safe and stable operation of a power system.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.
Claims (3)
1. The solar irradiance short-term prediction method of the photovoltaic power station based on the optimal graph structure is characterized by comprising the following steps of:
s10, constructing a space-time associated station of a target station through geographic range and azimuth information;
step S20, acquiring historical irradiance time series data of a target station and each time-space associated station;
step S30, calculating the correlation between the space-time correlation field station and the target field station data, and screening space-time correlation field stations with high correlation;
step S40, constructing graph structure data according to the space-time associated field stations obtained by screening;
step S50, a graph neural network prediction model taking graph structure data as input is established, so that irradiance short-term prediction is realized;
wherein, step S10 includes:
s101, intercepting the geographic space range of the area where the target station is located, wherein other stations in the range are space-time associated stations;
step S102, acquiring longitude and latitude coordinate information of a space-time associated station;
the step S101 specifically includes:
taking a target station as a center, intercepting a rectangular area, and calculating four vertex coordinates of the rectangular area, wherein the calculation process is as follows:
wherein,Rrepresenting the geographic distance of the vertex from the target station,R ground (floor) Representing the radius of the earth,lonfor the target site longitude to be,latfor the target station latitude, in degrees,dlonanddlotrepresenting the offset of the longitude and latitude,left-top、left-bottom、 right-top、right-bottomrespectively representing coordinates of four vertexes of the rectangular frame;
the step S102 specifically includes: utilizing an automatic grid generation algorithm in the geographic space range of the area generated in the step S101, and generating longitude and latitude coordinate information of a plurality of time-space associated stations in a refinement mode according to different distances and different azimuth angles by taking a target station as a center;
step S20 includes:
step S201, mapping historical irradiance time series data of a target station and each space-time associated station from a satellite cloud picture by adopting a satellite cloud picture-irradiance mapping method;
step S202, training a data deviation correction model by using target station data, and performing deviation correction on irradiance of a time-associated station;
step S30 includes:
step S301, normalized mutual information is utilizedN MI The correlation between the target station historical irradiance data obtained in step S20 and each space-time associated station historical irradiance data is calculated as follows:
wherein the method comprises the steps ofXAndYis the input data of the historical irradiance of the target station and the spatio-temporal associated station,p r (·)as a function of the probability,p r (x,y)is a joint probability;
step S40 includes:
step S401, constructing different graph structure data according to the space-time associated stations obtained by screening, wherein the graph structure data consists of a node characteristic matrix and an adjacent matrix, the node characteristic matrix consists of a target station and historical irradiance data of each space-time associated station, and the adjacent matrix is a connection relation determined according to the correlation among the station data;
step S402, the graph structure data is optimal graph structure data obtained based on graph connectivity evaluation;
calculating complex network topology statistical indexes according to the adjacency matrix, wherein the complex network topology statistical indexes comprise, but are not limited to, average degree, average aggregation coefficient, average shortest path distance among nodes, average node betweenness centrality and network connectivity, and the calculation formulas of the indexes are as follows:
the average degree calculation formula:
wherein the method comprises the steps ofNIs the total number of nodes in the network, and the nodeiNumber of adjacent edges with connectionk i A degree called the node; averaging the degrees of all nodes in the network to obtain the average degree of the networkDC；
The average aggregation coefficient calculation formula:
wherein the method comprises the steps ofMiIs a nodeiThe number of the actually existing edges is connected, and the aggregation coefficient isCi，ACIs the average aggregation coefficient;
the average shortest path distance between nodes is calculated by the formula:
wherein the method comprises the steps ofGFor a set of all the nodes,Nfor the total number of nodes in the network, two nodes in the networkiAndjdistance betweend ij Defining the number of edges of the shortest path connecting the two nodes, defining the average shortest path distance between the nodesLFor the distance between node pairs in a networkd ij Average value of (2):
the mean node betweenness centrality calculation formula:
wherein the method comprises the steps ofN jl Representing nodesjAndlthe number of shortest paths between the two paths,N jl (i)representing nodesjAndlthe shortest path between the nodesiIs used for the number of the strips,BC i the node bets are represented by the numbers of nodes,BCrepresenting average node betweenness centrality;
the calculation formula of the network connectivity:
wherein,k ij V is node pair #)i，j) Node-connectivity in a network, representing the degree of node pairing in the networki，j) The minimum number of nodes to be removed is not communicated,i、jexcept for; node pairi，j) In,i≠jand is also provided withi、jNon-adjacent;
weighting and calculating each complex network topology statistical index to obtain a graph connectivity comprehensive evaluation index:
the main component analysis method is utilized to reduce the dimension, a plurality of effective main components with smaller quantity and higher contribution rate are used for representing all connectivity evaluation indexes, then the entropy weight method is utilized to determine the weight coefficient corresponding to each effective main component item, finally, the graph connectivity comprehensive evaluation index is obtained according to each weight coefficient, and the calculation formula is as follows:
GC=k 1 .DC+k 2 .AC+k 3 .L+k 4 .BC+k 5 .K
wherein k is 1 +k 2 +k 3 +k 4 +k 5 =1；
Step S50 includes:
step S501, carrying out normalization processing on historical irradiance data of each station in a node characteristic matrix in the graph structure data;
step S502, a graph neural network prediction model is built by taking graph structure data as input and irradiance as output;
step S503, inputting the graph structure data into a graph neural network prediction model, and performing inverse normalization processing on the output irradiance prediction result to obtain a final target station irradiance prediction result.
2. The method according to claim 1, wherein step S30 further comprises:
step S302, each time-space associated station is pressedN MI The values are sorted from big to small, only the first 1/3 of the sorting is kept, or the values are selectedN MI Space-time associated stations having values greater than a threshold.
3. A storage medium storing a computer program, which when executed by a processor implements the method for short-term prediction of solar irradiance of a photovoltaic power plant of any of claims 1-2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210241377.0A CN114676893B (en) | 2022-03-11 | 2022-03-11 | Photovoltaic power station solar irradiance short-term prediction method based on optimal graph structure and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210241377.0A CN114676893B (en) | 2022-03-11 | 2022-03-11 | Photovoltaic power station solar irradiance short-term prediction method based on optimal graph structure and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114676893A CN114676893A (en) | 2022-06-28 |
| CN114676893B true CN114676893B (en) | 2024-02-02 |
Family
ID=82072948
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210241377.0A Active CN114676893B (en) | 2022-03-11 | 2022-03-11 | Photovoltaic power station solar irradiance short-term prediction method based on optimal graph structure and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114676893B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111489028A (en) * | 2020-04-09 | 2020-08-04 | 国网江苏省电力有限公司苏州供电分公司 | Thundercloud trajectory tracking-based photovoltaic power prediction method under lightning condition |
| CN112215478A (en) * | 2020-09-27 | 2021-01-12 | 珠海博威电气股份有限公司 | Power coordination control method and device for optical storage power station and storage medium |
| CN112507793A (en) * | 2020-11-05 | 2021-03-16 | 上海电力大学 | Ultra-short-term photovoltaic power prediction method |
| CN113627674A (en) * | 2021-08-12 | 2021-11-09 | 中国华能集团清洁能源技术研究院有限公司 | Distributed photovoltaic power station output prediction method and device and storage medium |
| CN114091317A (en) * | 2021-11-16 | 2022-02-25 | 国网甘肃省电力公司电力科学研究院 | Photovoltaic power station power prediction method based on NWP irradiance correction and error prediction |
-
2022
- 2022-03-11 CN CN202210241377.0A patent/CN114676893B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111489028A (en) * | 2020-04-09 | 2020-08-04 | 国网江苏省电力有限公司苏州供电分公司 | Thundercloud trajectory tracking-based photovoltaic power prediction method under lightning condition |
| CN112215478A (en) * | 2020-09-27 | 2021-01-12 | 珠海博威电气股份有限公司 | Power coordination control method and device for optical storage power station and storage medium |
| CN112507793A (en) * | 2020-11-05 | 2021-03-16 | 上海电力大学 | Ultra-short-term photovoltaic power prediction method |
| CN113627674A (en) * | 2021-08-12 | 2021-11-09 | 中国华能集团清洁能源技术研究院有限公司 | Distributed photovoltaic power station output prediction method and device and storage medium |
| CN114091317A (en) * | 2021-11-16 | 2022-02-25 | 国网甘肃省电力公司电力科学研究院 | Photovoltaic power station power prediction method based on NWP irradiance correction and error prediction |
Non-Patent Citations (1)
| Title |
|---|
| 基于图机器学习的分布式光伏发电预测;阚博文、刘广一等;《供用电》;第36卷(第11期);第20-27页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114676893A (en) | 2022-06-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Probabilistic solar irradiation forecasting based on variational Bayesian inference with secure federated learning | |
| Zhong et al. | A city-scale estimation of rooftop solar photovoltaic potential based on deep learning | |
| Jalalkamali | Using of hybrid fuzzy models to predict spatiotemporal groundwater quality parameters | |
| CN111027175A (en) | Method for evaluating social and economic influences of flood based on coupling model integrated simulation | |
| Dai et al. | Satellite constellation design with multi-objective genetic algorithm for regional terrestrial satellite network | |
| Plonski et al. | Environment and solar map construction for solar-powered mobile systems | |
| CN110781458B (en) | Method for predicting surface solar irradiance based on mixed regression model | |
| CN112801332B (en) | Short-term wind speed prediction method based on gray level co-occurrence matrix | |
| CN115640956A (en) | Future water resource supply and demand balance analysis method | |
| Li et al. | Spiral-based chaotic chicken swarm optimization algorithm for parameters identification of photovoltaic models | |
| Gómez et al. | Evaluation of a particle swarm optimization based method for optimal location of photovoltaic grid-connected systems | |
| CN114462316A (en) | Air quality optimization simulation method for homeland space planning | |
| CN114676893B (en) | Photovoltaic power station solar irradiance short-term prediction method based on optimal graph structure and storage medium | |
| CN117374956A (en) | Short-term prediction method for photovoltaic power generation of comprehensive energy station | |
| Reinoso et al. | Photovoltaic generation model as a function of weather variables using artificial intelligence techniques | |
| CN116861348A (en) | Photovoltaic power station anomaly detection method based on actual distance establishment cluster analysis | |
| CN112684519B (en) | Weather forecasting method and device, computer equipment and storage medium | |
| CN114418243B (en) | Distributed new energy cloud grid prediction method and system | |
| CN108510097B (en) | Power transmission line corridor snow disaster monitoring method and system based on satellite remote sensing | |
| Zhang et al. | Graph Neural Network Based Short-Term Solar Irradiance Forcasting Model Considering Surrounding Meteorological Factors | |
| Gerards et al. | Pv predictions made easy: Flexibility through simplicity | |
| Alcañiz et al. | Photovoltaic system monitoring and fault detection using peer systems | |
| CN107506949B (en) | Photovoltaic data clustering method, photovoltaic power prediction method, storage medium and device | |
| Uzhinskiy et al. | Management of environmental monitoring data: UNECE ICP Vegetation case | |
| Núñez-Reyes et al. | Spatial Estimation of Solar Radiation Using Geostatistics and Machine Learning Techniques |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |