CN113627674A - Distributed photovoltaic power station output prediction method and device and storage medium - Google Patents

Distributed photovoltaic power station output prediction method and device and storage medium Download PDF

Info

Publication number
CN113627674A
CN113627674A CN202110925647.5A CN202110925647A CN113627674A CN 113627674 A CN113627674 A CN 113627674A CN 202110925647 A CN202110925647 A CN 202110925647A CN 113627674 A CN113627674 A CN 113627674A
Authority
CN
China
Prior art keywords
prediction
photovoltaic power
data
historical
power station
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.)
Pending
Application number
CN202110925647.5A
Other languages
Chinese (zh)
Inventor
曹传钊
刘明义
曹曦
徐若晨
裴杰
刘大为
李�昊
朱勇
颜云岭
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huaneng Clean Energy Research Institute
Original Assignee
Huaneng Clean Energy Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Huaneng Clean Energy Research Institute filed Critical Huaneng Clean Energy Research Institute
Priority to CN202110925647.5A priority Critical patent/CN113627674A/en
Publication of CN113627674A publication Critical patent/CN113627674A/en
Pending legal-status Critical Current

Links

Images

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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
    • Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Economics (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Computer Hardware Design (AREA)
  • Operations Research (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Quality & Reliability (AREA)
  • Geometry (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Development Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a distributed photovoltaic power station output prediction method, and belongs to the technical field of photovoltaic power station output prediction. Firstly, classifying historical meteorological data of a photovoltaic power station by adopting a system clustering method, and classifying historical generated power according to a mapping relation between the historical meteorological data and historical generated power data; then establishing a corresponding photovoltaic power station power generation power prediction model according to historical meteorological data and historical power generation power data under different classifications; and finally, performing principal component analysis on the prediction sample data, and performing ultra-short-term output prediction of the distributed photovoltaic power station according to the photovoltaic power station power generation power prediction model. The invention can optimize the accuracy of the prediction result, reduce a plurality of environmental factors influencing the output power into a small number of main components with higher correlation, effectively reduce the sample dimension of the input quantity, play the effect of simplifying the network structure and shorten the learning time of the neural network.

Description

Distributed photovoltaic power station output prediction method and device and storage medium
Technical Field
The invention belongs to the technical field of photovoltaic power station output prediction, and particularly relates to a distributed photovoltaic power station output prediction method, a distributed photovoltaic power station output prediction device and a storage medium.
Background
In recent years, with the rapid development of renewable energy power generation technology, under the background of a large amount of new energy grid-connected power generation, an energy configuration structure is reasonably improved, and the safe and reliable operation of a power network is ensured, so that the method becomes an important research direction for the development of a power system. The high randomness and the fluctuation of the photovoltaic output can bring challenges to the safe operation and peak shaving scheduling of the power grid, and limit the consumption of renewable energy to a greater extent. Therefore, the reasonable and accurate prediction of the change of the photovoltaic power generation power is beneficial to maintaining the safe and stable operation of the power grid.
The traditional photovoltaic output prediction method mainly comprises the following types:
(1) dividing according to time scale: the ultra-short term prediction (less than 4h) mainly adopts a statistical and physical mixed method, and the main principle is to predict the motion condition of a cloud layer according to a satellite cloud picture shot by a geosynchronous satellite, predict the irradiation intensity reaching the ground and predict the power through a solar irradiation intensity and power conversion efficiency model. The method is generally used for photovoltaic power generation control, power quality evaluation, development and design of photovoltaic power station component parts and the like. The short-term prediction (less than 48h) mainly takes NWP (weather forecast information) data as a main part, and the predicted value of the output power of the photovoltaic power station is obtained by establishing the mapping relation between historical input data and historical output power. The method is generally used for power balance and economic dispatch of a power system, day-ahead power generation planning, power market trading, transient stability evaluation and the like. The medium-long term prediction (more than 1 week) is mainly used for the maintenance scheduling of the system, the prediction of the power generation amount and the like.
(2) According to the prediction mode, the method comprises the following steps: direct prediction: and directly predicting the output power of the photovoltaic power station through the output power curve rule of 8 points earlier to 5 points later. Indirect prediction: the method comprises the steps of predicting the solar irradiation amount, and converting the output of photovoltaic power generation according to the predicted solar irradiation amount.
(3) According to the prediction principle, the method comprises the following steps: the physical method comprises the following steps: the method is used for taking the NWP data as input, researching the characteristics of the photovoltaic power generation equipment, establishing a corresponding mathematical model of the photovoltaic power generation power and the NWP data, and further predicting the photovoltaic power generation power. The prediction accuracy of the physical method is slightly worse than that of the statistical method, but the method has the advantages that the method does not need the support of a large amount of historical data and is suitable for newly-built photovoltaic power stations. The statistical method comprises the following steps: the method is a method for further predicting the photovoltaic power generation power by finding out the internal rule of historical data, eliminating ill-conditioned data points, establishing a function mapping relation between the historical data and the output power. The statistical method needs a large amount of historical photovoltaic power station output data as a modeling basis, is only suitable for photovoltaic power stations with the operation time more than or equal to one year, and is not suitable for newly-built photovoltaic power stations. The learning method comprises the following steps: an input and output mapping relation is established by adopting an artificial intelligence mode, and the method is mainly applied to a nonlinear mapping model.
(4) Dividing according to a mathematical model: the mathematical model includes: a time series prediction method, an autoregressive moving average model method, a neural network method, a support vector machine method, a wavelet analysis method, a grey prediction method, a recursive least square method, a similar day selection algorithm, an intelligent method and the like.
In addition, to make the prediction accuracy higher, it is now more common to combine a plurality of methods. The photovoltaic power generation combined prediction method roughly comprises two types: the first method is that the results of the power prediction of several photovoltaic power generation are comprehensively compared, and a method with small error and simple operation is extracted for power prediction through comparison and analysis; and secondly, weighting results of a plurality of prediction methods according to weight, and then averaging, thereby not only realizing the maximum utilization of prediction information, but also optimizing the defects of all prediction models, eliminating the error of the traditional single model and greatly improving the accuracy of photovoltaic power generation power prediction.
However, due to randomness and uncertainty of photovoltaic power generation, great defects and constraints exist when the conventional method is used for predicting photovoltaic power generation output, such as insufficient prediction accuracy, complex steps, slow calculation speed, poor prediction effect, and the like.
Disclosure of Invention
In order to solve the above problems, the present invention provides a method, an apparatus, and a storage medium for predicting output of a distributed photovoltaic power station, which can optimize stability and accuracy of a prediction result and shorten calculation time.
The invention is realized by the following technical scheme:
a distributed photovoltaic power station output prediction method comprises the following steps:
s1: classifying historical meteorological data of the photovoltaic power station by adopting a system clustering method, and classifying historical generated power according to a mapping relation between the historical meteorological data and the historical generated power data;
s2: establishing a corresponding photovoltaic power station power generation power prediction model according to historical meteorological data and historical power generation power data under different classifications;
s3: and performing principal component analysis on the prediction sample data, and performing ultra-short-term output prediction of the distributed photovoltaic power station according to the photovoltaic power station power generation power prediction model.
Preferably, before S1, a data washing step is further included.
Preferably, in S1, the historical meteorological data includes solar irradiance, air pressure, temperature, humidity, weather, wind direction, and wind speed.
Preferably, in S1, the systematic clustering segmentation method is to perform distance scale segmentation on the longitudinal axis of the result dendrogram according to the clustering result.
Preferably, in S1, the systematic clustering segmentation method is to segment the inter-sample transverse similarity measure of the clustering result.
Preferably, in S2, the photovoltaic power plant generated power prediction model is built through a radial basis kernel function neural network.
Preferably, S3 specifically includes performing principal component analysis on the prediction sample data, extracting principal components as input, and obtaining a prediction result, that is, a power prediction value, corresponding to the input prediction sample data according to the photovoltaic power station power generation prediction model.
The invention discloses computer equipment which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to realize the steps of the distributed photovoltaic power station output prediction method.
The invention discloses a computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and when the computer program is executed by a processor, the steps of the output prediction method of the distributed photovoltaic power station are realized.
Compared with the prior art, the invention has the following beneficial technical effects:
the distributed photovoltaic power station output prediction method disclosed by the invention considers the influence of more comprehensive weather information on the photovoltaic power generation prediction precision, and adds a feature extraction method to search the main components of the influence factors of the photovoltaic power generation in the weather data. The invention provides a photovoltaic power station output power prediction model precision improving method of a Radial Basis Function (RBF) neural network on the basis of analyzing two data analysis theories of cluster analysis and principal component analysis, and the method is used for carrying out sectional clustering on modeling data. Firstly, clustering is carried out on historical operating data after data cleaning, a target similar sample set is divided in a clustering segmentation mode, and prediction of output power of a photovoltaic power station is verified by utilizing a neural network algorithm, so that the stability and accuracy of a prediction result can be effectively improved, and the number of samples of input quantity is reduced; on the basis, in order to further improve the prediction precision, the data samples are subjected to principal component analysis, so that the accuracy of a prediction result can be effectively improved, the dimension of an input sample can be reduced, and the effects of simplifying a network structure and shortening the learning time of a neural network are achieved.
Drawings
FIG. 1 is a flow chart of a distributed photovoltaic power plant capacity prediction method of the present invention;
FIG. 2 is a partial enlarged view of the resulting dendrogram of clustering scheme A in the example;
FIG. 3 is a sample distribution diagram of the clustering scheme A in the example;
FIG. 4 is a diagram of a data prediction operation result of the a-group of the clustering scheme A in the embodiment;
FIG. 5 is a diagram of predicted operation results of data of group b of the clustering scheme A in the embodiment;
FIG. 6 is a diagram of predicted operation results of c-group data of the clustering scheme A in the embodiment;
FIG. 7 is a sample distribution diagram of the clustering scheme B in the example;
FIG. 8 is a diagram of a data prediction operation result of the a-group of the clustering scheme B in the embodiment;
FIG. 9 is a B-group data prediction operation result diagram of the clustering scheme B in the embodiment;
FIG. 10 is a diagram of predicted operation results of c-group data of clustering scheme B in the example;
FIG. 11 is a diagram showing the analysis of the principal component composition in the examples;
FIG. 12 is a diagram of the operation results of a photovoltaic power plant generated power prediction model in an embodiment.
Detailed Description
The invention will now be described in further detail with reference to the drawings and specific examples, which are given by way of illustration and not by way of limitation.
Cluster analysis
Introduction to the principle
The system clustering analysis is a clustering method which is used most in various research fields at present, and the basic idea is as follows: the process continues until each sample can be clustered into the appropriate class. The invention adopts a Euclidean distance calculation method for measuring the similarity degree of the samples by the sample distance. The binary Euclidean distance calculation method comprises the following steps:
Figure BDA0003209068760000051
let dij＝d(xi,xj)，D＝(dij) p × p, forming a distance matrix:
Figure BDA0003209068760000052
in the formula: dij＝djiAnd represents the corresponding distance between i and j. The two nearest samples can be clustered into one class and clustering algorithms can be developed according to the sum of squared deviations method (Ward's method). Clustering the two nearest samples into one class according to the size of the distance matrix, and subsequently clustering according to the sum of squared deviations method (Ward's method), when GpAnd GqClustering into GrThis and other classes GkThe distance recurrence formula of (c) is:
Figure BDA0003209068760000053
in the formula: n isp、nk、nr、nqAre each Gp、Gk、Gr、GqThe number of samples.
Clustering analysis-based power prediction segmented modeling method
(1) Power prediction model
The method constructs a distributed photovoltaic power station short-term power prediction model combining cluster analysis and an RBF neural network based on historical environment monitoring data such as solar irradiance, air pressure and temperature and historical power generation data which influence the output value of the photovoltaic power station. Firstly, performing system clustering (segmented clustering) on known photovoltaic power station historical data, analyzing and screening out similar samples corresponding to environmental data at a target moment, and inputting a similar sample set to an RBF neural network to serve as a training set. And then, establishing a power prediction model through an RBF neural network algorithm to realize short-term output prediction of the distributed photovoltaic power station.
(2) Piecewise clustering modeling
In order to select a proper data processing mode and achieve an ideal analysis and prediction result, the invention provides two analysis and screening schemes for historical environmental data samples based on the clustering result of the selected system clustering analysis algorithm. The first segmentation scheme is that distance scale segmentation is carried out on the longitudinal axis of a result dendrogram according to a clustering result, namely, the Euclidean distance value of the longitudinal coordinate in the clustering result dendrogram of an input system is divided; the second segmentation scheme is to segment the inter-sample similarity scale of the clustering result, namely to divide the degree of closeness and sparseness in the process of gathering nodes based on the horizontal axis of the result dendrogram. It is one of the innovative points of the invention.
(3) RBF neural network algorithm
In order to realize the prediction and optimization of the power of the photovoltaic power station, the RBF neural network algorithm is selected to analyze the implicit rule characteristic between the input and the output of the neural network, and the RBF neural network has the advantages of high fitting speed, less weight change and fitting of the nonlinear relation between the input and the output data while being capable of well approaching to the target value. Based on the advantage, the RBF neural network is selected to analyze and process the intrinsic law of the photovoltaic power generation so as to realize short-term prediction of the photovoltaic power.
The RBF neural network consists of 3 layers of structures, namely a signal input layer, a hidden layer for carrying out nonlinear calculation and an output layer for linear transformation. The invention selects the data which is correspondingly analyzed and processed as an input layer, not only meets the input requirement of the system, but also contains the change rule information of the system data. And selecting the predicted power value as an output layer.
For the hidden layer, a gaussian function is selected as a relative activation function, and the relative activation function is realized by using a radial basis function net ═ newrb (P, T, S, M, D). Wherein: p is the input of the neural network, T is the output, S is the radial basis expansion velocity, M is the peak value of the neuron number, and D is the network parameter. The function continuously increases the value of the hidden layer in the process of creating the RBF neural network until the output error meets the requirement.
The output of the ith hidden layer is calculated by the following formula:
Figure BDA0003209068760000071
wherein R isiIs a Euclidean norm; p is input; c. CiIs the central vector of the ith basis function; biIs a reference width; d is the number of nodes in the hidden layer.
Referring to fig. 1, the output prediction method of the distributed photovoltaic power station of the present invention includes the following steps:
cleaning the acquired data;
s1: classifying historical meteorological data of the photovoltaic power station by adopting a system clustering method, and classifying historical generated power according to a mapping relation between the historical meteorological data and the historical generated power data; historical meteorological data includes solar irradiance, air pressure, temperature, humidity, weather, wind direction, and wind speed; the system clustering segmentation mode has two schemes, one is to segment the distance scale of the longitudinal axis of the result dendrogram according to the clustering result; and the other is to segment the cross similarity scale between the samples of the clustering result.
S2: establishing a corresponding photovoltaic power station power generation power prediction model according to historical meteorological data and historical power generation power data under different classifications; a photovoltaic power station power generation power prediction model is established through a radial basis kernel function (RBF) neural network.
S3: and performing principal component analysis on the prediction sample data, and performing ultra-short-term output prediction of the distributed photovoltaic power station according to the photovoltaic power station power generation power prediction model. The method specifically comprises the following steps: and performing principal component analysis on the prediction sample data, extracting principal components as input, and obtaining a prediction result corresponding to the input prediction sample data, namely a power prediction value according to the photovoltaic power station power generation power prediction model.
Examples analysis
The method selects historical environmental monitoring data of 7:00-18:00 days per day of a certain distributed photovoltaic power station, including meteorological factors such as humidity, environmental temperature, air pressure and wind direction of the photovoltaic power station and corresponding power of the photovoltaic power station. Every 15mins is a time node sample, and the total number of 450 samples are used as input quantity of clustering analysis to prepare for subsequent steps such as system clustering and the like.
Clustering scheme A
The historical data is subjected to systematic clustering (hierarchical clustering), and is displayed in an output result, wherein the horizontal axis represents the distribution of each time node sample, and the vertical axis represents a distance value. As the number of samples is too large, the nodes of the horizontal axis of the dendrogram are too dense, and for intensively searching the clustering rule, the scheme carries out segmentation according to the longitudinal distance scale, namely, the longitudinal coordinate of the dendrogram selects proper numerical values (13 and 22.5) as boundary lines to carry out sample division, as shown in figure 2.
And respectively selecting ordinate distance values (13, 22.5) as boundary lines to divide the nodes into three groups of a, b and c, and selecting the same number of samples in each group for inputting, wherein the type is a group, the type is b group, and the type x is c group. The sample distribution is shown in fig. 3.
The three groups of data a, b and c are compared with historical data at corresponding moments and are respectively input into an RBF neural network prediction program, power prediction is carried out on a two-hour time period (namely 8 time nodes), and the prediction results are shown in fig. 4, 5 and 6.
After observing various groups of data and the operation result of the prediction program, the method can be preliminarily known as follows: the curve fitting degree of the data sets of the group a and the group b predicted by the RBF neural network is higher, and the prediction result is better, wherein the group a is slightly better than the group b, the fitting degree of the group c is poorer, and the prediction result is also poorer. When grouping is carried out according to the Euclidean distance value of the longitudinal axis, the group with smaller Euclidean distance is input into the prediction model and runs to obtain a better prediction result, and the group with larger Euclidean distance is input into the prediction model to obtain a poorer prediction result.
Clustering scheme B
According to the scheme, segmentation is carried out according to the horizontal similar scale, namely segmentation is carried out according to the sample node affinity and sparseness degree in the horizontal axis of the dendrogram, the dendrogram is the same as the scheme A, the local amplification is carried out on the dendrogram, the same number of input samples are selected, and the Euclidean distances among average samples in the a group, the b group and the c group are sequentially increased. The sample distribution is shown in fig. 7.
And (3) respectively inputting the three groups of data a, b and c into the RBF neural network prediction program in comparison with historical data at corresponding moments, and performing power prediction on the RBF neural network in a time range of two hours (namely 8 time nodes). The prediction results are shown in fig. 8, 9 and 10.
The observation and operation results are preliminarily known as follows: after the RBF neural network is input, the fitting degree of a group of samples a, b and c is gradually reduced, wherein the fitting result of the group a is the best, the fitting degree of the group c is the worst, and the prediction result is poor. When the node samples are grouped according to the similarity of the horizontal axis time node samples, the higher the similarity between the samples in the group is, the better the prediction result obtained by the prediction model is.
Analysis of predicted results
RMSE: root Mean square Error (Root Mean Squared Error), which is the square Root of the Mean square Error:
Figure BDA0003209068760000091
MRE: average relative error (meanrelatederror), which is the average of the relative errors:
Figure BDA0003209068760000092
dt: an actual value; n: the number of data samples; f. oft: predicting a value; e.g. of the typet: and (4) error.
Therefore, according to the correlation formula, the correlation calculation is performed for each group of situations of different schemes, and the results are shown in the following table:
Figure BDA0003209068760000093
from the above definitions it follows that: r2The numerical value of the (determining coefficient) represents the proportion of the variation which can be explained after the neural network training to the total variation of the variable and the fitting degree of a prediction curve in a prediction result; RMSE (root mean square error) represents the stability degree of the numerical value of the prediction result; and MRE (homogeneous relative error) reflects the accuracy of the numerical value of the prediction result. According to the above figure, it can be known that the prediction result of the B scheme is generally better than that of the a scheme, and the prediction result of the a group in the B scheme is optimal and the prediction result of the a group in the a scheme is optimal compared with that of the same group. The scheme A is segmented according to the distance of a longitudinal axis in a system clustering result dendrogram, and the results of the scheme A and the scheme B are combined to know that the segmentation in the scheme A is according to the numerical value of the distance of the longitudinal axis, the relativity of Euclidean distances is not considered, the group a with the best prediction effect in the scheme A has the characteristic of the smallest Euclidean distance and can be regarded as a grouping set with high similarity among a plurality of group samples, so that the prediction effect of the group a is the best compared with that of the other two groups, and the group B and the group c cannot reflect the relativity of the Euclidean distance represented by the longitudinal axis, so that the group B and the group c do not have enough reference analysis value; the B scheme is segmented according to the similarity (namely Euclidean distance) among samples, the similarity among the samples of the three groups a, B and c is gradually reduced (namely the Euclidean distance among average samples is gradually increased), and the prediction effect of the three groups is also sequentially reduced, so that the higher the similarity among the samples is, the better the prediction effect is in the equivalent historical data sample set input into the RBF neural network power prediction model.
Power prediction model input data structure research based on principal component analysis
Principal component analysis
The principal component analysis method is a multivariate statistical analysis method, and can change a plurality of variables into a plurality of comprehensive variables (namely linear combination of initial variables) with small correlation, thereby reducing data dimensionality. The method comprises the following specific steps:
firstly, sample data of an analysis research contains p variables and n samples, and possible adverse effects caused by too large magnitude difference among different variables are avoided.
(1) By normalizing the matrix x1,x2,xpComputing samplesThe formula for calculating the correlation coefficient is:
Figure BDA0003209068760000101
(2) determining corresponding eigenvalues lambda from the matrix of correlation coefficients1,λ2,…,λpAnd a feature vector e1,e2,…,ep；
(3) Calculating the variance contribution rate of each principal component, and calculating the cumulative variance contribution rate, wherein the formula is as follows:
Figure BDA0003209068760000102
the principal components are screened, and when the cumulative variance contribution rate of the m principal components reaches the accuracy index of 85% capable of reflecting the original information, the m principal components y can be obtained1,y2,…,ymTo replace the original variables, the m principal components are used as the input of the RBF neural network model.
The principal component-related expression is as follows:
Figure BDA0003209068760000111
wherein: e.g. of the typei＝[ei1,ei2,…,eip]，eipA p-dimensional eigenvector corresponding to the ith eigenvalue of the initial variable correlation matrix; x is an initial input variable of dimension p, X ═ X1,x2,...,xp]T。
Example analysis
Selecting historical environment monitoring data of a certain distributed photovoltaic power station within 7:00-18:00 time every day within ten days, wherein the historical environment monitoring data comprises factors such as humidity, radiation value, temperature and air pressure of the photovoltaic power station and corresponding power data, and a time node sample is taken every fifteen minutes. As shown in fig. 11, after the systematic clustering is performed, the corresponding similar sample set is selected as the input amount of the principal component analysis, and the subsequent steps such as principal component extraction are performed. Inputting the sorted historical data into a principal component analysis model, and performing principal component analysis to obtain a covariance matrix eigenvalue, a variance contribution rate and an accumulated contribution rate, which are shown in the following table:
Figure BDA0003209068760000112
as can be seen from the above table, since the cumulative contribution rate of the first four feature values has reached 89.56% (greater than 85%), the 4 principal components are selected as the input quantity of the RBF neural network power prediction, and the feature vectors corresponding to the four principal components can be obtained from the table.
Analysis of predicted results
And solving corresponding expressions according to the feature vectors of the four main components, substituting the expressions into historical data for calculation to obtain the four main components, and inputting the main components into the RBF neural network prediction model for power prediction. In order to achieve the comparison effect, raw data which is not subjected to principal component analysis is input into the RBF neural network prediction model as input quantity to be operated. The prediction results are shown in fig. 12.
And (3) carrying out correlation index calculation on the prediction result, wherein the calculation result is shown in the following table:
Figure BDA0003209068760000121
from the above table, after the principal component analysis is performed on the historical data, the principal components are extracted and input into the RBF neural network prediction model, the goodness of fit of the prediction curve, the stability of the numerical values of the prediction results and the accuracy of the numerical values of the prediction results are all superior to those of the results obtained by directly inputting the original data for prediction, and after the principal component analysis is adopted, the input quantity of the analysis of the calculation example is reduced from 8 environmental factors to 4 principal components, so that the dimensionality of input data is remarkably reduced, the redundancy of the data is successfully reduced, and the operation speed is improved.
Comprehensive analysis
Based on the related research and analysis results of system clustering, the grouping a group with the minimum sample distance is selected as a similar sample set according to the sample distance division in the chapter. And substituting the multiple environmental factor variables of each time node sample into the principal component expression to obtain a corresponding principal component, and taking the obtained principal component as the input quantity of the RBF neural network power prediction model.
The correlation index calculation was performed on the predicted results and compared to the predicted results when the data was not processed, the results are shown in the following table:
Figure BDA0003209068760000122
in summary, it can be found that: compared with the power prediction result under the comprehensive influence of principal component analysis, the system clustering has better prediction effect in the initial state before analysis and processing.
The invention also provides computer equipment which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to realize the steps of the distributed photovoltaic power station output prediction method.
The distributed photovoltaic power plant output prediction method of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein. If the output prediction method of the distributed photovoltaic power station is realized in the form of a software functional unit and is sold or used as an independent product, the output prediction method of the distributed photovoltaic power station can be stored in a computer readable storage medium.
Based on such understanding, in the exemplary embodiment, a computer readable storage medium is also provided, all or part of the processes in the method of the above embodiments of the present invention can be realized by a computer program to instruct related hardware, the computer program can be stored in the computer readable storage medium, and when the computer program is executed by a processor, the steps of the above method embodiments can be realized. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. Computer-readable storage media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. It should be noted that the computer readable medium may contain content that is subject to appropriate increase or decrease as required by legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer readable media does not include electrical carrier signals and telecommunications signals as is required by legislation and patent practice. The computer storage medium may be any available medium or data storage device that can be accessed by a computer, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.), optical memory (e.g., CD, DVD, BD, HVD, etc.), and semiconductor memory (e.g., ROM, EPROM, EEPROM, nonvolatile memory (NANDFLASH), Solid State Disk (SSD)), etc.
In an exemplary embodiment, a computer device is also provided, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the distributed photovoltaic power plant contribution prediction method when executing the computer program. The processor may be a Central Processing Unit (CPU), other general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc.
It should be noted that the above description is only a part of the embodiments of the present invention, and equivalent changes made to the system described in the present invention are included in the protection scope of the present invention. Persons skilled in the art to which this invention pertains may substitute similar alternatives for the specific embodiments described, all without departing from the scope of the invention as defined by the claims.

Claims (9)

1. A distributed photovoltaic power station output prediction method is characterized by comprising the following steps:
s1: classifying historical meteorological data of the photovoltaic power station by adopting a system clustering method, and classifying historical generated power according to a mapping relation between the historical meteorological data and the historical generated power data;
s2: establishing a corresponding photovoltaic power station power generation power prediction model according to historical meteorological data and historical power generation power data under different classifications;
s3: and performing principal component analysis on the prediction sample data, and performing ultra-short-term output prediction of the distributed photovoltaic power station according to the photovoltaic power station power generation power prediction model.
2. The distributed photovoltaic power plant capacity prediction method of claim 1 further comprising a data cleansing step prior to S1.
3. The distributed photovoltaic power plant contribution prediction method of claim 1, wherein in S1, the historical meteorological data comprises solar irradiance, air pressure, temperature, humidity, weather, wind direction, and wind speed.
4. The distributed photovoltaic power plant contribution prediction method of claim 1, wherein in S1, the systematic clustering segmentation means is distance scale segmentation of the longitudinal axis of the result dendrogram based on the clustering results.
5. The distributed photovoltaic power plant output prediction method of claim 1 wherein in S1, the systematic clustering segmentation approach is to segment the inter-sample transverse similarity measure of the clustering results.
6. The distributed photovoltaic power plant output prediction method of claim 1 wherein in S2, the photovoltaic power plant generated power prediction model is built by a radial basis kernel function neural network.
7. The distributed photovoltaic power plant output prediction method of claim 1, wherein S3 is specifically configured to perform principal component analysis on the prediction sample data, extract principal components as input, and obtain a prediction result, i.e., a power prediction value, corresponding to the input prediction sample data according to the photovoltaic power plant power generation prediction model.
8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the distributed photovoltaic power plant contribution prediction method of any of claims 1 to 7 when executing the computer program.
9. A computer-readable storage medium, having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the distributed photovoltaic power plant contribution prediction method of any of claims 1 to 7.
CN202110925647.5A 2021-08-12 2021-08-12 Distributed photovoltaic power station output prediction method and device and storage medium Pending CN113627674A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110925647.5A CN113627674A (en) 2021-08-12 2021-08-12 Distributed photovoltaic power station output prediction method and device and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110925647.5A CN113627674A (en) 2021-08-12 2021-08-12 Distributed photovoltaic power station output prediction method and device and storage medium

Publications (1)

Publication Number Publication Date
CN113627674A true CN113627674A (en) 2021-11-09

Family

ID=78385014

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110925647.5A Pending CN113627674A (en) 2021-08-12 2021-08-12 Distributed photovoltaic power station output prediction method and device and storage medium

Country Status (1)

Country Link
CN (1) CN113627674A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114358371A (en) * 2021-11-26 2022-04-15 华能大理风力发电有限公司洱源分公司 Photovoltaic short-term power prediction method and device based on deep learning
CN114676893A (en) * 2022-03-11 2022-06-28 中国长江三峡集团有限公司 Photovoltaic power station solar irradiance short-term prediction method based on optimal graph structure and storage medium
CN116432874A (en) * 2023-06-14 2023-07-14 青岛鼎信通讯科技有限公司 Distributed photovoltaic power prediction method based on characteristic power
CN117613850A (en) * 2023-09-21 2024-02-27 国网江苏省电力有限公司信息通信分公司 Distributed photovoltaic power generation power prediction method, device, equipment and storage medium
CN117634836A (en) * 2023-12-08 2024-03-01 创维互联（北京）新能源科技有限公司 Matrix analysis-based photovoltaic power station cleaning process evaluation method and system

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104200032A (en) * 2014-09-05 2014-12-10 山东大学 Transverse time axis clustering method in generalized load modeling on basis of time periods
CN105204493A (en) * 2015-09-11 2015-12-30 北京电子工程总体研究所 State monitoring and fault diagnosing method for rotating mechanical equipment
CN109583588A (en) * 2018-12-17 2019-04-05 华北电力大学（保定） A kind of short-term wind speed forecasting method and system
CN110619360A (en) * 2019-09-09 2019-12-27 国家电网有限公司 Ultra-short-term wind power prediction method considering historical sample similarity
CN110929953A (en) * 2019-12-04 2020-03-27 国网山东省电力公司电力科学研究院 Photovoltaic power station ultra-short term output prediction method based on cluster analysis
CN111369045A (en) * 2020-02-28 2020-07-03 南京邮电大学 Method for predicting short-term photovoltaic power generation power
CN111680841A (en) * 2020-06-10 2020-09-18 广东电网有限责任公司 Short-term load prediction method and system based on principal component analysis and terminal equipment

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104200032A (en) * 2014-09-05 2014-12-10 山东大学 Transverse time axis clustering method in generalized load modeling on basis of time periods
CN105204493A (en) * 2015-09-11 2015-12-30 北京电子工程总体研究所 State monitoring and fault diagnosing method for rotating mechanical equipment
CN109583588A (en) * 2018-12-17 2019-04-05 华北电力大学（保定） A kind of short-term wind speed forecasting method and system
CN110619360A (en) * 2019-09-09 2019-12-27 国家电网有限公司 Ultra-short-term wind power prediction method considering historical sample similarity
CN110929953A (en) * 2019-12-04 2020-03-27 国网山东省电力公司电力科学研究院 Photovoltaic power station ultra-short term output prediction method based on cluster analysis
CN111369045A (en) * 2020-02-28 2020-07-03 南京邮电大学 Method for predicting short-term photovoltaic power generation power
CN111680841A (en) * 2020-06-10 2020-09-18 广东电网有限责任公司 Short-term load prediction method and system based on principal component analysis and terminal equipment

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
刘志祥: "《中国上市公司财务预警模型的构建与应用研究》", 31 December 2016, pages: 61 - 64 *
成珂等: "聚类分析在光伏发电量预测中的应用研究", 《可再生能源》, vol. 35, no. 5, pages 697 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114358371A (en) * 2021-11-26 2022-04-15 华能大理风力发电有限公司洱源分公司 Photovoltaic short-term power prediction method and device based on deep learning
CN114676893A (en) * 2022-03-11 2022-06-28 中国长江三峡集团有限公司 Photovoltaic power station solar irradiance short-term prediction method based on optimal graph structure and storage medium
CN114676893B (en) * 2022-03-11 2024-02-02 中国长江三峡集团有限公司 Photovoltaic power station solar irradiance short-term prediction method based on optimal graph structure and storage medium
CN116432874A (en) * 2023-06-14 2023-07-14 青岛鼎信通讯科技有限公司 Distributed photovoltaic power prediction method based on characteristic power
CN117613850A (en) * 2023-09-21 2024-02-27 国网江苏省电力有限公司信息通信分公司 Distributed photovoltaic power generation power prediction method, device, equipment and storage medium
CN117634836A (en) * 2023-12-08 2024-03-01 创维互联（北京）新能源科技有限公司 Matrix analysis-based photovoltaic power station cleaning process evaluation method and system

Similar Documents

Publication Publication Date Title
Hong et al. A hybrid deep learning-based neural network for 24-h ahead wind power forecasting
Lipu et al. Artificial intelligence based hybrid forecasting approaches for wind power generation: Progress, challenges and prospects
Wu et al. A data mining approach combining $ K $-means clustering with bagging neural network for short-term wind power forecasting
CN113627674A (en) Distributed photovoltaic power station output prediction method and device and storage medium
CN111860982A (en) Wind power plant short-term wind power prediction method based on VMD-FCM-GRU
CN106503867A (en) A kind of genetic algorithm least square wind power forecasting method
CN110717610A (en) Wind power prediction method based on data mining
CN111369045A (en) Method for predicting short-term photovoltaic power generation power
CN114792156A (en) Photovoltaic output power prediction method and system based on curve characteristic index clustering
Wang et al. Wind speed forecasting based on hybrid model with model selection and wind energy conversion
CN114021483A (en) Ultra-short-term wind power prediction method based on time domain characteristics and XGboost
CN112580874A (en) Short-term wind power prediction method based on random forest algorithm and TCN
CN111242355A (en) Photovoltaic probability prediction method and system based on Bayesian neural network
Aliberti et al. Forecasting Short-term Solar Radiation for Photovoltaic Energy Predictions.
CN115759389A (en) Day-ahead photovoltaic power prediction method based on weather type similar day combination strategy
CN115995810A (en) Wind power prediction method and system considering weather fluctuation self-adaptive matching
CN116701868A (en) Short-term wind power section probability prediction method
CN115660182A (en) Photovoltaic output prediction method based on maximum expected sample weighted neural network model
CN115600500A (en) Ultrashort-term probability wind power prediction method based on space-time multiscale and K-SDW
CN114298132A (en) Wind power prediction method and device and electronic equipment
CN113610285A (en) Power prediction method for distributed wind power
Brusca et al. Wind farm power forecasting: New algorithms with simplified mathematical structure
CN117458437A (en) Short-term wind power prediction method, system, equipment and medium
CN116070458A (en) New wind power plant scene generation method based on RAC-GAN
CN116435987A (en) Ultra-short-term wind power prediction method for large-scale wind power clusters

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