CN113609758B - Power prediction method for newly-built wind power plant - Google Patents

Power prediction method for newly-built wind power plant Download PDF

Info

Publication number
CN113609758B
CN113609758B CN202110812849.9A CN202110812849A CN113609758B CN 113609758 B CN113609758 B CN 113609758B CN 202110812849 A CN202110812849 A CN 202110812849A CN 113609758 B CN113609758 B CN 113609758B
Authority
CN
China
Prior art keywords
wind power
power plant
wind
data
input
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
Application number
CN202110812849.9A
Other languages
Chinese (zh)
Other versions
CN113609758A (en
Inventor
胡维昊
张真源
井实
廖启术
曹迪
罗仕华
陈健军
胡家祥
熊康
任曼曼
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.)
University of Electronic Science and Technology of China
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
Original Assignee
University of Electronic Science and Technology of China
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
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 University of Electronic Science and Technology of China, Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd filed Critical University of Electronic Science and Technology of China
Priority to CN202110812849.9A priority Critical patent/CN113609758B/en
Publication of CN113609758A publication Critical patent/CN113609758A/en
Application granted granted Critical
Publication of CN113609758B publication Critical patent/CN113609758B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
    • 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
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/08—Fluids
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
    • 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)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Water Supply & Treatment (AREA)
  • Pure & Applied Mathematics (AREA)
  • Quality & Reliability (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Algebra (AREA)
  • Computing Systems (AREA)
  • Fluid Mechanics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Physics (AREA)
  • Public Health (AREA)
  • Operations Research (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a power prediction method of a new wind power plant, which comprises the steps of respectively collecting historical data of the new wind power plant and historical data of a source wind power plant, assisting the historical data of the new wind power plant to complete training of a multi-task Gaussian process model by utilizing a large amount of historical data of the source wind power plant with long establishment time, and finally realizing prediction of real-time power of the new wind power plant by the trained model, so that the problem that accurate power prediction results are difficult to obtain due to insufficient historical data of the new wind power plant is solved.

Description

Power prediction method for newly-built wind power plant
Technical Field
The invention belongs to the technical field of new energy, and particularly relates to a power prediction method of a newly-built wind power plant.
Background
With the continuous advance of low-carbon energy system strategies, fossil energy is gradually replaced and new energy becomes main stream energy, and wind power plays an important role in the energy system as a clean renewable energy. Accurate wind power predictions are essential for large-scale wind power integration into the grid.
At present, wind power prediction at home and abroad mainly comprises a physical model and a numerical model. The physical model mainly refers to a prediction method based on Numerical Weather Prediction (NWP), and a differential equation is solved through current meteorological data (initial conditions) to obtain a wind power prediction result. However, in the case of a weather mutation, the prediction accuracy of the physical model may be severely degraded. The numerical model predicts future wind power by using historical data. Conventional numerical models refer mainly to time series models such as autoregressive models, autoregressive moving average models (ARMA), and the like. With the rise of artificial intelligence, a novel numerical model based on various machine learning algorithms, such as an Artificial Neural Network (ANN), a Support Vector Machine (SVM), and the like, has been developed. However, the premise of obtaining accurate prediction results by the numerical model is that there is enough historical data, so that most predictions by the numerical model are based on a large amount of historical data.
For a newly built wind power plant, most numerical models cannot obtain accurate wind power prediction results when enough historical data is not available, however, accurate wind power prediction is critical to grid connection of the newly built wind power plant, and therefore, how to obtain accurate power prediction of the newly built wind power plant becomes a problem to be solved urgently.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a power prediction method of a newly built wind power plant, which utilizes a large amount of historical data of a source wind power plant with relatively long building time to assist the newly built wind power plant to train a multi-task Gaussian process prediction model, thereby realizing the power prediction of the newly built wind power plant.
In order to achieve the aim of the invention, the power prediction method of the newly-built wind power plant is characterized by comprising the following steps:
(1) Acquiring power data and meteorological data of a wind power plant;
(1.1) acquiring wind power data of the newly built wind power plant l at different moments t, and recording the wind power data as
Figure BDA0003168852110000021
And meteorological data of the newly built wind farm l at different moments t, including weft wind components +.>
Figure BDA0003168852110000022
And radial wind component->
Figure BDA0003168852110000023
t＝1,2,…,N l ，N l The sampling time is the number of sampling moments;
(1.2) acquiring wind power data of the source wind power plant s at different moments t, and recording the wind power data as
Figure BDA0003168852110000024
And meteorological data of the source wind farm s at different moments t, including weft wind components +.>
Figure BDA0003168852110000025
And radial wind component->
Figure BDA0003168852110000026
t＝1,2,…,N s ，N s Is the number of sampling moments and satisfies N s ＞＞N l ；
(2) Extracting wind power information and time information;
(2.1) acquiring wind speeds, wind directions and wind energy at different moments of a place according to meteorological data of a newly-built wind power plant and a source wind power plant;
Figure BDA0003168852110000027
Figure BDA0003168852110000028
wherein d is the air density;
Figure BDA0003168852110000029
the wind speed, the wind direction and the wind energy of the newly built wind power plant at the moment t; />
Figure BDA00031688521100000210
The wind speed, the wind direction and the wind energy of the source wind power plant at the moment t;
(2.2) respectively acquiring the hour information corresponding to the time t according to the wind power data of the newly-built wind power plant and the source wind power plant at the time t, and respectively marking as
Figure BDA00031688521100000211
(3) Constructing a data set of the newly built wind power plant and the source wind power plant;
(3.1) constructing input and output data sets of newly-built wind power plant at different moments t
Figure BDA00031688521100000212
Figure BDA00031688521100000213
(3.2) constructing input and output data sets of the source wind power plant at different moments t
Figure BDA00031688521100000214
Figure BDA00031688521100000215
(4) Constructing a training set;
for input/output data sets at different times t
Figure BDA0003168852110000031
Respectively carrying out normalization treatment;
after normalization is completed, N is added l The input and output data sets at each moment form a training set { X ] l ,Y l }：
Figure BDA0003168852110000032
Will N s The input and output data sets at each moment form a training set { X } for assisting training of newly built wind power plant s ,Y s }：
Figure BDA0003168852110000033
(5) Constructing a multitasking Gaussian process prediction model MTGP;
(5.1) constructing a Gaussian process model of the newly built wind power plant, and marking the Gaussian process model as GP l The model reflects the training set { X } of the newly built wind farm l ,Y l Input-to-output relationship in }:
Figure BDA0003168852110000034
wherein ,εl For Gaussian noise input in newly built wind power plant epsilon l Obeys normal distribution
Figure BDA0003168852110000035
Figure BDA0003168852110000036
Is epsilon l Is a variance of (2); />
Figure BDA0003168852110000037
Finger input->
Figure BDA0003168852110000038
And output->
Figure BDA0003168852110000039
A 0-mean Gaussian process is formed between the two, namely:
Figure BDA00031688521100000310
wherein ,
Figure BDA00031688521100000311
and />
Figure BDA00031688521100000312
Respectively represent training data sets { X } of newly built wind power plant l ,Y l Input +.>
Figure BDA00031688521100000313
Is a data of two different dimensions, +.>
Figure BDA00031688521100000314
Is GP l Is expressed as:
Figure BDA00031688521100000315
wherein B is a positive semi-definite matrix to be trained; k (k) l (. Cndot.) is a kernel function that can be expressed as:
Figure BDA00031688521100000316
wherein ,μl Is a nuclear parameter;
(5.2) constructing a Gaussian process model of the source wind power plant, and marking the Gaussian process model as GP s The model reflects the source wind farm training set { X } s ,Y s Input-to-output relationship in }:
Figure BDA00031688521100000317
wherein ,εs For Gaussian noise input in newly built wind power plant epsilon s Obeys normal distribution
Figure BDA00031688521100000318
Figure BDA00031688521100000319
Is epsilon s Is a variance of (2); />
Figure BDA00031688521100000320
Finger input->
Figure BDA00031688521100000321
And output->
Figure BDA00031688521100000322
A 0-mean Gaussian process is formed between the two, namely:
Figure BDA00031688521100000323
wherein ,
Figure BDA0003168852110000041
and />
Figure BDA0003168852110000042
Respectively represent training data sets { X } of source wind power plant s ,Y s Input +.>
Figure BDA0003168852110000043
Is a data of two different dimensions, +.>
Figure BDA0003168852110000044
Is GP l Is expressed as:
Figure BDA0003168852110000045
wherein B is a positive semi-definite matrix to be trained; k (k) s (. Cndot.) is a kernel function that can be expressed as:
Figure BDA0003168852110000046
wherein ,μs Is a nuclear parameter;
(6) Training a multitasking Gaussian process prediction model MTGP;
(6.1) training data set { X ] of Source wind farm s ,Y s Input to model GP s In (2) training the model GP through optimization s Parameter θ s ＝[B,σ s ,μ s ]Make model GP s Converging;
(6.2) in GP s After training, substituting the obtained positive semi-definite matrix B into the model GP l Then the training data set { X ] of the target wind power plant is used for l ,Y l Input to model GP l In the step B, training is further carried out, and the model GP is waited l After convergence, obtaining the corresponding parameter theta l ＝[B,σ l ,μ l ]；
(7) Predicting the power of a newly built wind power plant in real time;
(7.1) collecting meteorological data of the newly built wind power plant in real time, and processing according to the methods in the steps (2) - (4) to obtain a normalized input data set
Figure BDA00031688521100000413
(7.2), input data set to be processed
Figure BDA00031688521100000414
Inputting trained GP l In the model, the predicted value is thus obtained>
Figure BDA00031688521100000415
Inputting the test set X from the test set { X, Y } of the target wind farm into the trained GP l In the model, the predicted value can be obtained:
Figure BDA0003168852110000047
Figure BDA0003168852110000048
wherein the superscript T denotes a transpose,
Figure BDA0003168852110000049
dacromet product, b l Is column l of B, y= (Y l ,Y s ) T D is a 2×2 diagonal matrix and the diagonal elements are +.>
Figure BDA00031688521100000410
and />
Figure BDA00031688521100000411
I is N l A rank identity matrix;
(7.3) vs. predicted value
Figure BDA00031688521100000412
And performing inverse normalization processing to obtain the real-time pre-power of the newly built wind power plant.
The invention aims at realizing the following steps:
according to the power prediction method for the newly built wind power plant, the historical data of the newly built wind power plant and the historical data of the source wind power plant are collected respectively, then the historical data of the newly built wind power plant is assisted by a large amount of historical data of the source wind power plant with long building time to complete training of a multi-task Gaussian process model, and finally the prediction of the real-time power of the newly built wind power plant is achieved through the trained model, so that the problem that an accurate power prediction result is difficult to obtain due to the fact that the historical data of the newly built wind power plant is insufficient is solved.
Meanwhile, the power prediction method of the newly-built wind power plant has the following beneficial effects:
(1) The accurate power prediction result is obtained under the condition that the history data of the newly built wind power plant is insufficient, which is an important basis for grid connection of the newly built wind power plant and power scheduling of new energy, ensures the stability of the power grid when the newly built wind power plant is connected, and provides a reliable basis for power scheduling;
(2) The method has the advantages that the large amount of data of the source wind power plant with long establishment time is utilized to assist in power prediction of the newly established wind power plant, but the prediction result is not interfered and deteriorated due to the large amount of data from the source wind power plant;
(3) Selecting wind power information of a place where a wind power plant is located, such as: the wind speed, the wind direction and the wind energy are taken as input characteristics, and the wind information is a key factor influencing the output of the wind power plant, so that the accuracy of the wind power plant power prediction result can be greatly improved.
Drawings
FIG. 1 is a diagram of a power prediction model of a newly built wind farm according to the present invention;
FIG. 2 is a graph of input characteristics (wind speed) versus output (wind power) for the present invention;
FIG. 3 is a flow chart of a method for predicting power of a newly built wind farm according to the present invention;
FIG. 4 is a block diagram of one embodiment of the central node shown in FIG. 1;
fig. 5 is a block diagram of one embodiment of the user terminal shown in fig. 1.
Detailed Description
The following description of the embodiments of the invention is presented in conjunction with the accompanying drawings to provide a better understanding of the invention to those skilled in the art. It is to be expressly noted that in the description below, detailed descriptions of known functions and designs are omitted here as perhaps obscuring the present invention.
Examples
In this embodiment, taking actual data of a wind farm 1 and a wind farm 2 as an example, a power prediction model of a new wind farm according to the present invention is described, as shown in fig. 1, and a power prediction model structure diagram of the new wind farm is described.
The following describes in detail a power prediction method of a newly built wind farm according to the present invention with reference to fig. 1, as shown in fig. 2, and specifically includes the following steps:
s1, acquiring power data and meteorological data of a wind power plant;
s1.1, acquiring wind power data of a newly built wind power plant l at different moments t, and recording the wind power data as
Figure BDA0003168852110000061
And meteorological data of the newly built wind farm l at different moments t, including weft wind components +.>
Figure BDA0003168852110000062
And radial wind component->
Figure BDA0003168852110000063
t＝1,2,…,N l ，N l The sampling time is the number of sampling moments;
s1.2, acquiring wind power data of a source wind power plant S at different moments t, and recording the wind power data as
Figure BDA0003168852110000064
And meteorological data of the source wind farm s at different moments t, including weft wind components +.>
Figure BDA0003168852110000065
And radial wind component->
Figure BDA0003168852110000066
t＝1,2,…,N s ，N s Is the number of sampling moments and satisfies N s ＞＞N l ；
In this embodiment, at 1h intervals, i.e. once every 1h, 24 points are sampled per day, and if a newly built wind farm has just been built for 5 days, then N l ＝24×5，The source wind power plant collects the data of 5 days and the data of 45 days before the establishment of the new wind power plant, and the total data of 50 days, namely N s ＝24×50。
S2, extracting wind power information and time information;
s2.1, acquiring wind speeds, wind directions and wind energy at different moments of a place according to meteorological data of a newly-built wind power plant and a source wind power plant;
Figure BDA0003168852110000067
Figure BDA0003168852110000068
where d is the air density, taking a constant of 1 in this example;
Figure BDA0003168852110000069
the wind speed, the wind direction and the wind energy of the newly built wind power plant at the moment t; />
Figure BDA00031688521100000610
The wind speed, the wind direction and the wind energy of the source wind power plant at the moment t; in this embodiment, the range of values of the wind directions of the newly built wind power plant and the source wind power plant is 0 ° to 360 °. FIG. 2 is a graph of wind power versus wind speed, and it can be seen that the trend of wind power and wind speed are substantially consistent, reflecting that wind power factors such as wind speed are key factors affecting wind power.
S2.2, respectively acquiring the hour information corresponding to the time t according to the wind power data of the newly-built wind power plant and the source wind power plant at the time t, and respectively marking as
Figure BDA00031688521100000611
S3, constructing a data set of the newly built wind power plant and the source wind power plant;
s3.1, constructing input and output data sets of newly built wind power plant at different moments t
Figure BDA0003168852110000071
/>
Figure BDA0003168852110000072
S3.2, constructing input and output data sets of the source wind power plant at different moments t
Figure BDA0003168852110000073
Figure BDA0003168852110000074
S4, constructing a training set;
for input/output data sets at different times t
Figure BDA0003168852110000075
Respectively carrying out normalization treatment;
after normalization is completed, N is added l The input and output data sets at each moment form a training set { X ] l ,Y l }：
Figure BDA0003168852110000076
Will N s The input and output data sets at each moment form a training set { X } for assisting training of newly built wind power plant s ,Y s }：
Figure BDA0003168852110000077
S5, constructing a multitasking Gaussian process prediction model MTGP;
s5.1, constructing a Gaussian process model of the newly built wind power plant, and marking the Gaussian process model as GP l The model reflects the training set { X } of the newly built wind farm l ,Y l Input-to-output relationship in }:
Figure BDA0003168852110000078
wherein ,εl For Gaussian noise input in newly built wind power plant epsilon l Obeys normal distribution
Figure BDA0003168852110000079
Figure BDA00031688521100000710
Is epsilon l Is a variance of (2); />
Figure BDA00031688521100000711
Finger input->
Figure BDA00031688521100000712
And output->
Figure BDA00031688521100000713
A 0-mean Gaussian process is formed between the two, namely:
Figure BDA00031688521100000714
wherein ,
Figure BDA00031688521100000715
and />
Figure BDA00031688521100000716
Respectively represent training data sets { X } of newly built wind power plant l ,Y l Input +.>
Figure BDA00031688521100000717
Is a data of two different dimensions, +.>
Figure BDA00031688521100000718
Is GP l Is expressed as:
Figure BDA00031688521100000719
wherein B is a positive semi-definite matrix to be trained; k (k) l (. Cndot.) is a kernel function that can be expressed as:
Figure BDA00031688521100000720
wherein ,μl Is a nuclear parameter;
s5.2, constructing a Gaussian process model of the source wind power plant, and marking the Gaussian process model as GP s The model reflects the source wind farm training set { X } s ,Y s Input-to-output relationship in }:
Figure BDA0003168852110000081
/>
wherein ,εs For Gaussian noise input in newly built wind power plant epsilon s Obeys normal distribution
Figure BDA0003168852110000082
Figure BDA0003168852110000083
Is epsilon s Is a variance of (2); />
Figure BDA0003168852110000084
Finger input->
Figure BDA0003168852110000085
And output->
Figure BDA0003168852110000086
A 0-mean Gaussian process is formed between the two, namely:
Figure BDA0003168852110000087
wherein ,
Figure BDA0003168852110000088
and />
Figure BDA0003168852110000089
Respectively represent training data sets { X } of source wind power plant s ,Y s Input +.>
Figure BDA00031688521100000810
Is a data of two different dimensions, +.>
Figure BDA00031688521100000811
Is GP l Is expressed as:
Figure BDA00031688521100000812
wherein B is a positive semi-definite matrix to be trained; k (k) s (. Cndot.) is a kernel function that can be expressed as:
Figure BDA00031688521100000813
wherein ,μs Is a nuclear parameter;
s6, training a multi-task Gaussian process prediction model MTGP;
s6.1, training data set { X ] of source wind power plant s ,Y s Input to model GP s In (2) training the model GP through optimization s Parameter θ s ＝[B,σ s ,μ s ]Make model GP s Converging;
s6.2, GP s After training, substituting the obtained positive semi-definite matrix B into the model GP l Then the training data set { X ] of the target wind power plant is used for l ,Y l Input to model GP l In the step B, training is further carried out, and the model GP is waited l After convergence, obtaining the corresponding parameter theta l ＝[B,σ l ,μ l ]；
S7, predicting the power of the newly-built wind power plant in real time;
s7.1, collecting meteorological data of the newly built wind power plant l in real time, and processing according to the method described in the steps S2-S4 to obtain a normalized input data set
Figure BDA00031688521100000814
S7.2 input data set to be processed
Figure BDA00031688521100000815
Inputting trained GP l In the model, the predicted value is thus obtained>
Figure BDA00031688521100000816
Inputting the test set X from the test set { X, Y } of the target wind farm into the trained GP l In the model, the predicted value can be obtained:
Figure BDA00031688521100000817
Figure BDA00031688521100000818
wherein the superscript T denotes a transpose,
Figure BDA0003168852110000091
dacromet product, b l Is column l of B, y= (Y l ,Y s ) T D is a 2×2 diagonal matrix and the diagonal elements are +.>
Figure BDA0003168852110000092
and />
Figure BDA0003168852110000093
I is N l =24×5 rank identity matrix;
S7.3、for predicted values
Figure BDA0003168852110000094
And performing inverse normalization processing to obtain the real-time pre-power of the newly built wind power plant.
Method comparison and index evaluation:
in order to verify that the invention can obtain accurate power prediction results under the condition of insufficient history data of a newly built wind farm, other comparison methods, namely a Gaussian Process (GP), a Back Propagation Neural Network (BPNN), linear Regression (LR), quantile Regression (QR) and Support Vector Regression (SVR), are used. These comparison methods use only training data from the target wind farm, N l Training was performed with =24×5 sets of data without using large amounts of data from the source wind farm for auxiliary prediction.
To compare the power prediction effect of different schemes, two evaluation indexes are Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), respectively, and the specific form is:
Figure BDA0003168852110000095
Figure BDA0003168852110000096
wherein ,
Figure BDA0003168852110000097
representing the predicted value; y represents an actual value; n represents the total number of predictions. Both of which are smaller values represent higher accuracy and better prediction effect.
Prediction scenario 1:
in the scene 1, taking the wind power plant 1 as a target wind power plant, taking the wind power plant 2 as a source wind power plant, and predicting results are shown in fig. 4; compared with the linear regression of the reference method, the method improves the RMSE and the MAE by approximately 36.00 percent and 23.71 percent respectively, and improves the QR comparison with the best prediction effect in the comparison method, and improves the RMSE and the MAE by 24.71 percent and 20.00 percent respectively. The result shows that the effect of the method for carrying out auxiliary training by using a large amount of data of the source wind power plant in the scene 1 is far more than that of the method for carrying out auxiliary training without using the data of the source wind power plant;
prediction scenario 2:
in the scene 2, taking the wind power plant 1 as a source wind power plant, taking the wind power plant 2 as a target wind power plant, and predicting results are shown in fig. 5; compared with linear regression of a reference method, the method improves the RMSE and the MAE by approximately 25.91 percent and 29.15 percent respectively, and improves the GP comparison with the best prediction effect in a comparison method by 11.50 percent and 13.57 percent respectively. The above results show that the method for performing auxiliary training by using a large amount of data of the source wind farm has far more effect in the scene 2 than all the methods for performing auxiliary training without using the data of the source wind farm.
While the foregoing describes illustrative embodiments of the present invention to facilitate an understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, but is to be construed as protected by the accompanying claims insofar as various changes are within the spirit and scope of the present invention as defined and defined by the appended claims.

Claims (1)

1. The power prediction method of the newly-built wind power plant is characterized by comprising the following steps of:
(1) Acquiring power data and meteorological data of a wind power plant;
(1.1) acquiring wind power data of the newly built wind power plant l at different time t, and recording the wind power data as { P } t l -a }; and meteorological data of newly built wind farm l at different time t, including latitudinal wind components
Figure QLYQS_1
And radial wind component->
Figure QLYQS_2
Figure QLYQS_3
N l The sampling time is the number of sampling moments;
(1.2) acquiring wind power data of the source wind power plant s at different time points t, and recording the wind power data as { P } t s -a }; and meteorological data of the source wind farm s at different times t, including latitudinal wind components
Figure QLYQS_4
And radial wind component->
Figure QLYQS_5
Figure QLYQS_6
N s Is the number of sampling moments and satisfies N s ＞＞N l ；
(2) Extracting wind power information and time information;
(2.1) acquiring wind speeds, wind directions and wind energy at different moments of a place according to meteorological data of a newly-built wind power plant and a source wind power plant;
Figure QLYQS_7
Figure QLYQS_8
wherein ,
Figure QLYQS_9
the wind speed, the wind direction and the wind energy of the newly built wind power plant at the moment t; />
Figure QLYQS_10
Figure QLYQS_11
Wind speed at time t for source wind farmWind direction and wind energy;
(2.2) respectively acquiring the hour information corresponding to the time t according to the wind power data of the newly-built wind power plant and the source wind power plant at the time t, and respectively marking as
Figure QLYQS_12
(3) Constructing a data set of the newly built wind power plant and the source wind power plant;
(3.1) constructing input and output data sets of newly-built wind power plant at different moments t
Figure QLYQS_13
Figure QLYQS_14
(3.2) constructing input and output data sets of the source wind power plant at different moments t
Figure QLYQS_15
Figure QLYQS_16
(4) Constructing a training set;
for input/output data sets at different times t
Figure QLYQS_17
Respectively carrying out normalization treatment; />
After normalization is completed, N is added l The input and output data sets at each moment form a training set { X ] l ,Y l }：
Figure QLYQS_18
Will N s The input and output data sets at each moment form a training set { X } for assisting training of newly built wind power plant s ,Y s }：
Figure QLYQS_19
(5) Constructing a multitasking Gaussian process prediction model MTGP;
(5.1) constructing a Gaussian process model of the newly built wind power plant, and marking the Gaussian process model as GP l The model reflects the training set { X } of the newly built wind farm l ,Y l Input-to-output relationship in }:
Figure QLYQS_20
wherein ,εl For Gaussian noise input in newly built wind power plant epsilon l Obeys normal distribution
Figure QLYQS_21
Figure QLYQS_22
Is epsilon l Is a variance of (2);
Figure QLYQS_23
finger input->
Figure QLYQS_24
And output Y t l A 0-mean Gaussian process is formed between the two, namely:
Figure QLYQS_25
wherein ,
Figure QLYQS_26
and />
Figure QLYQS_27
Respectively represent training data sets { X } of newly built wind power plant l ,Y l Input +.>
Figure QLYQS_28
Is a data of two different dimensions, +.>
Figure QLYQS_29
Is GP l Is expressed as:
Figure QLYQS_30
wherein B is a positive semi-definite matrix to be trained; k (k) l (. Cndot.) is a kernel function that can be expressed as:
Figure QLYQS_31
wherein ,μl Is a nuclear parameter;
(5.2) constructing a Gaussian process model of the source wind power plant, and marking the Gaussian process model as GP s The model reflects the source wind farm training set { X } s ,Y s Input-to-output relationship in }:
Figure QLYQS_32
wherein ,εs For Gaussian noise input in newly built wind power plant epsilon s Obeys normal distribution
Figure QLYQS_33
Figure QLYQS_34
Is epsilon s Is a variance of (2);
Figure QLYQS_35
finger input->
Figure QLYQS_36
And output Y t s A 0-mean Gaussian process is formed between the two, namely:
Figure QLYQS_37
wherein ,
Figure QLYQS_38
and />
Figure QLYQS_39
Respectively represent training data sets { X } of source wind power plant s ,Y s Input +.>
Figure QLYQS_40
Is a data of two different dimensions, +.>
Figure QLYQS_41
Is GP l Is expressed as:
Figure QLYQS_42
wherein B is a positive semi-definite matrix to be trained; k (k) s (. Cndot.) is a kernel function that can be expressed as:
Figure QLYQS_43
wherein ,μs Is a nuclear parameter;
(6) Training a multitasking Gaussian process prediction model MTGP;
(6.1) training data set { X ] of Source wind farm s ,Y s Input to model GP s In (2) training the model GP through optimization s Parameter θ s ＝[B,σ s ,μ s ]Make model GP s Converging;
(6.2) in GP s After training, substituting the obtained positive semi-definite matrix B into the model GP l Then the training data set { X ] of the target wind power plant is used for l ,Y l Input to model GP l In the step B, training is further carried out, and the model GP is waited l After convergence, obtaining the corresponding parameter theta l ＝[B,σ l ,μ l ]；
(7) Predicting the power of a newly built wind power plant in real time;
(7.1) collecting meteorological data of the newly built wind power plant in real time, and processing according to the methods in the steps (2) - (4) to obtain a normalized input data set
Figure QLYQS_44
(7.2), input data set to be processed
Figure QLYQS_45
Inputting trained GP l In the model, the predicted value is thus obtained>
Figure QLYQS_46
Inputting the test set X from the test set { X, Y } of the target wind farm into the trained GP l In the model, the predicted value can be obtained:
Figure QLYQS_47
Figure QLYQS_48
wherein the superscript T denotes a transpose,
Figure QLYQS_49
dacromet product, b l Is column l of B, y= (Y l ,Y s ) T D is a 2×2 diagonal matrix and the diagonal elements are +.>
Figure QLYQS_50
and />
Figure QLYQS_51
I is N l A rank identity matrix;
(7.3) vs. predicted value
Figure QLYQS_52
And performing inverse normalization processing to obtain the real-time pre-power of the newly built wind power plant. />
CN202110812849.9A 2021-07-19 2021-07-19 Power prediction method for newly-built wind power plant Active CN113609758B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110812849.9A CN113609758B (en) 2021-07-19 2021-07-19 Power prediction method for newly-built wind power plant

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110812849.9A CN113609758B (en) 2021-07-19 2021-07-19 Power prediction method for newly-built wind power plant

Publications (2)

Publication Number Publication Date
CN113609758A CN113609758A (en) 2021-11-05
CN113609758B true CN113609758B (en) 2023-06-02

Family

ID=78337889

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110812849.9A Active CN113609758B (en) 2021-07-19 2021-07-19 Power prediction method for newly-built wind power plant

Country Status (1)

Country Link
CN (1) CN113609758B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102184337A (en) * 2011-06-07 2011-09-14 中国电力科学研究院 Dynamic combination analysis method of new energy generating capacity influenced by meteorological information
EP2853731A1 (en) * 2013-09-27 2015-04-01 Korea Electric Power Corporation Apparatus for simulating wind power farm
CN112160879A (en) * 2020-07-15 2021-01-01 姜庆超 FDA and SOM based fan blade icing visual state monitoring
CN112288164A (en) * 2020-10-29 2021-01-29 四川大学 Wind power combined prediction method considering spatial correlation and correcting numerical weather forecast
CN112651576A (en) * 2021-01-07 2021-04-13 云南电力技术有限责任公司 Long-term wind power prediction method and device

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150278405A1 (en) * 2014-03-31 2015-10-01 Vestas Wind Systems A/S Method for evaluating a performance prediction for a wind farm

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102184337A (en) * 2011-06-07 2011-09-14 中国电力科学研究院 Dynamic combination analysis method of new energy generating capacity influenced by meteorological information
EP2853731A1 (en) * 2013-09-27 2015-04-01 Korea Electric Power Corporation Apparatus for simulating wind power farm
CN112160879A (en) * 2020-07-15 2021-01-01 姜庆超 FDA and SOM based fan blade icing visual state monitoring
CN112288164A (en) * 2020-10-29 2021-01-29 四川大学 Wind power combined prediction method considering spatial correlation and correcting numerical weather forecast
CN112651576A (en) * 2021-01-07 2021-04-13 云南电力技术有限责任公司 Long-term wind power prediction method and device

Also Published As

Publication number Publication date
CN113609758A (en) 2021-11-05

Similar Documents

Publication Publication Date Title
CN109802430B (en) Wind power grid control method based on LSTM-Attention network
CN109002915B (en) Photovoltaic power station short-term power prediction method based on Kmeans-GRA-Elman model
CN112733462B (en) Ultra-short-term wind power plant power prediction method combining meteorological factors
CN110009135B (en) Wind power prediction method based on width learning
CN113822418A (en) Wind power plant power prediction method, system, device and storage medium
Chen et al. Research on wind power prediction method based on convolutional neural network and genetic algorithm
CN110991737A (en) Ultra-short-term wind power prediction method based on deep belief network
Dong et al. Short-term wind-speed forecasting based on multiscale mathematical morphological decomposition, K-means clustering, and stacked denoising autoencoders
CN112421631A (en) New energy consumption capacity assessment method and system
CN112149905A (en) Photovoltaic power station short-term power prediction method based on wavelet transformation and wavelet neural network
CN116070769A (en) Ultra-short-term wind power plant power multi-step interval prediction modularization method and device thereof
Wang et al. Hybrid deep learning model for short-term wind speed forecasting based on time series decomposition and gated recurrent unit
CN113627685B (en) Wind driven generator power prediction method considering wind power internet load limit
CN116629416A (en) Photovoltaic power station power prediction method and device
He et al. Similar day selecting based neural network model and its application in short-term load forecasting
CN111612242A (en) Motor state parameter prediction method based on LSTM deep learning model
CN111080001A (en) Deep neural network prediction method applied to wind speed of wind power plant
CN113344283B (en) Energy internet new energy consumption capability assessment method based on edge intelligence
CN113609758B (en) Power prediction method for newly-built wind power plant
CN117172097A (en) Power distribution network dispatching operation method based on cloud edge cooperation and multi-agent deep learning
CN112734073A (en) Photovoltaic power generation short-term prediction method based on long and short-term memory network
Xiaosheng et al. A deep learning approach for wind power prediction based on stacked denoising auto encoders optimized by bat algorithm
CN111179110B (en) Virtual power plant variable order aggregation equivalent robust dynamic model modeling method and device
Liao et al. Wind power prediction based on periodic characteristic decomposition and multi-layer attention network
Wang et al. Ultra-short-term forecast of photovoltaic power based on vmd error correction and cnn-gru-am

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