CN114297947A - Data-driven wind power system twinning method and system based on deep learning network - Google Patents

Data-driven wind power system twinning method and system based on deep learning network Download PDF

Info

Publication number
CN114297947A
CN114297947A CN202210213632.0A CN202210213632A CN114297947A CN 114297947 A CN114297947 A CN 114297947A CN 202210213632 A CN202210213632 A CN 202210213632A CN 114297947 A CN114297947 A CN 114297947A
Authority
CN
China
Prior art keywords
deep learning
wind power
power system
learning network
model
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.)
Granted
Application number
CN202210213632.0A
Other languages
Chinese (zh)
Other versions
CN114297947B (en
Inventor
武昕
刘宇航
余昊杨
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.)
North China Electric Power University
Original Assignee
North China Electric Power University
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 North China Electric Power University filed Critical North China Electric Power University
Priority to CN202210213632.0A priority Critical patent/CN114297947B/en
Publication of CN114297947A publication Critical patent/CN114297947A/en
Application granted granted Critical
Publication of CN114297947B publication Critical patent/CN114297947B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • Y04S10/12—Monitoring or controlling equipment for energy generation units, e.g. distributed energy generation [DER] or load-side generation
    • Y04S10/123—Monitoring or controlling equipment for energy generation units, e.g. distributed energy generation [DER] or load-side generation the energy generation units being or involving renewable energy sources
    • 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
    • Y04S10/40—Display of information, e.g. of data or controls
    • 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

  • Wind Motors (AREA)

Abstract

The invention relates to a data-driven wind power system twinning method and system based on a deep learning network, belonging to the technical field of digital twinning, firstly generating a physical entity static model based on a monocular vision three-dimensional reconstruction method, then constructing a dual deep learning network model, obtaining a dynamic operation simulation process of a physical entity in a wind power system by using the dual deep learning network model, constructing a long-time and short-time memory neural network rule model for predicting the power generation power of the wind power system, predicting the power generation power by using the long-time and short-time memory neural network rule model, and finally combining the dynamic operation simulation process of the physical entity in the wind power system, the predicted power generation power and the physical entity static model by using a digital twinning platform to realize the three-dimensional visualization of the physical entity in the wind power generation system, realize the remote monitoring and management of wind power system equipment and improve the operation and maintenance level thereof, and short-term wind power prediction is realized.

Description

Data-driven wind power system twinning method and system based on deep learning network
Technical Field
The invention relates to the technical field of digital twinning, in particular to a data-driven wind power system twinning method and system based on a deep learning network.
Background
Wind power generation is a clean energy power generation mode which is popular in the world at present and also occupies an important position in the electric power production market in China. Therefore, the operation, maintenance, state monitoring and management, power generation amount prediction and the like of the wind power system also become key problems which need to be processed urgently. The operation, maintenance and state detection of the wind power system aim to ensure safe and reliable operation of wind power system equipment, so that the wind power system is ensured to have higher investment income in the full-period production process.
The traditional wind power system operation and maintenance and state detection mainly comprises two aspects of planning and non-planning. The wind turbine generator system is maintained and defects are eliminated regularly by site operation and maintenance personnel in a planning mode according to an operation and maintenance plan, the method has the problems of discontinuity and hysteresis, and the fault condition of the wind turbine generator system is difficult to find in time. The method is characterized in that the fault warning prompt of a wind turbine monitoring system is depended on in an unplanned mode, operation and maintenance personnel carry out fault processing and maintenance after the wind turbine monitoring system sends a fault report, the operation and maintenance mode is influenced by factors such as spare parts, field environment, fault occurrence time and the like, and the problems of slow fault processing response, long shutdown time of the wind turbine and the like easily occur. Therefore, how to detect the running state and the short-term wind power generation level of the wind generating set in real time, realize the continuous improvement of the operation and maintenance quality of the wind generating set with low cost and high reliability, and improve the running stability of the wind generating set becomes a problem to be solved urgently for wind power enterprises.
Disclosure of Invention
The invention aims to provide a data-driven wind power system twinning method and system based on a deep learning network, so as to realize remote monitoring and management of wind power system equipment, improve the operation and maintenance level of the wind power system equipment and realize short-term wind power prediction.
In order to achieve the purpose, the invention provides the following scheme:
a data-driven wind power system twinning method based on a deep learning network comprises the following steps:
acquiring a multi-view image of a physical entity in a wind power system;
generating a static model of the physical entity by adopting a three-dimensional reconstruction method based on monocular vision according to the multi-view image of the physical entity;
constructing a dual deep learning network model; the dual deep learning network model comprises a first dual deep learning network and a second dual deep learning network; the first heavy deep learning network is used for determining physical attributes, motion rules and operation characteristics of the physical entity according to the operation data of the wind power system; the second deep learning network is used for obtaining an operation behavior model of the physical entity according to the physical attribute, the motion rule and the operation characteristic of the physical entity determined by the first deep learning network;
acquiring the running data of a physical entity in the wind power system in real time by utilizing an OPC UA technology;
inputting the real-time acquired operation data of the physical entity into the dual deep learning network model, and outputting the dynamic operation simulation process of the physical entity in the wind power system to realize data driving;
selecting indexes influencing the power generation power of a physical entity in the wind power system by using a grey correlation analysis method;
according to the selected indexes, constructing a long-term memory neural network rule model for predicting the power generation power of the wind power system based on a particle swarm algorithm;
acquiring the running data of the index by utilizing an OPC UA technology;
inputting the operation data of the index into the long-time memory neural network rule model, and outputting the predicted power generation power of a physical entity in the wind power system;
and combining the dynamic operation simulation process and the predicted power generation power of the physical entity in the wind power system with the physical entity static model by using the digital twin platform to realize the three-dimensional visualization of the physical entity in the twin wind power system.
Optionally, the generating a static model of the physical entity by using a monocular vision-based three-dimensional reconstruction method according to the multi-view image of the physical entity specifically includes:
respectively extracting feature points from the multi-view images by adopting an SIFT feature extraction algorithm;
matching the extracted characteristic points of the multi-view images by using a fast approximate nearest neighbor algorithm;
calculating a transformation matrix between the successfully matched feature points;
calculating rotation and translation information between images according to the transformation matrix;
converting the pixel point coordinates of the multi-view images in the world coordinate system into the pixel point coordinates of the multi-view images in the camera coordinate system according to the rotation and translation information among the images, so as to realize camera calibration;
projecting pixel points of multi-view images calibrated by a camera into the same three-dimensional coordinate system, and constructing sparse point clouds of physical entities in a wind power system;
using a multi-view clustering and a patch model-based dense matching algorithm to densify sparse point clouds of physical entities in a wind power system to obtain dense point clouds;
performing surface reconstruction on the dense point cloud by adopting a Poisson surface reconstruction algorithm;
and performing texture mapping on the dense point cloud after surface reconstruction to generate a physical entity static model.
Optionally, the obtaining, in real time, the operation data of the physical entity in the wind power system by using the OPC UA technology specifically includes:
constructing an OPC UA information model for describing attributes and relations of production elements of the whole wind power system;
receiving heterogeneous data of a wind power system through an OPC UA server;
based on an OPC UA information model, fusing and processing the heterogeneous data through an OPC UA server;
and subscribing the required type of wind power system operation data processed by the OPC UA server through the OPC UA client by utilizing a subscription mechanism.
Optionally, the constructing a dual deep learning network model specifically includes:
establishing a dual deep learning network model consisting of a first dual deep learning network and a second dual deep learning network;
respectively determining the expression relations of input and output of the first heavy deep learning network and the second heavy deep learning network; the first heavy deep learning network takes wind power system operation data as input and takes physical attributes, motion rules and operation characteristics of physical entities as output; the second deep learning network takes the physical attributes, the motion rules and the operation characteristics of the physical entities output by the first deep learning network as input and takes the operation behavior model of the physical entities as output;
and respectively configuring parameters of the first heavy deep learning network and the second heavy deep learning network by adopting a mean square error loss function and utilizing a random gradient descent method.
Optionally, the establishing, according to the selected index, a long-term and short-term memory neural network rule model for predicting the generated power of the wind power system based on a particle swarm algorithm specifically includes:
optimizing the long-term and short-term memory neural network by adopting a particle swarm algorithm to obtain an optimized long-term and short-term memory neural network model;
acquiring historical operating data of the indexes and historical generating power of corresponding physical entities according to the selected indexes to form a historical data set;
dividing the historical data set into a training set and a test set;
training the optimized long-time and short-time memory neural network model by using the training set to obtain the trained long-time and short-time memory neural network model;
carrying out simulation test on the trained long-time and short-time memory neural network model by using the test set to obtain a prediction error;
if the prediction error is larger than the error threshold, returning to the step of obtaining historical operating data of the index and historical generating power of the corresponding physical entity according to the selected index to form a historical data set;
and if the prediction error is smaller than or equal to the error threshold, outputting the tested long-time and short-time memory neural network model as a long-time and short-time memory neural network rule model for predicting the generated power of the wind power system.
A data-driven wind power system twinning system based on a deep learning network, the system comprising:
the multi-view image acquisition module is used for acquiring multi-view images of physical entities in the wind power system;
the three-dimensional reconstruction module is used for generating a physical entity static model by adopting a monocular vision-based three-dimensional reconstruction method according to the multi-view image of the physical entity;
the dual deep learning network model building module is used for building a dual deep learning network model; the dual deep learning network model comprises a first dual deep learning network and a second dual deep learning network; the first heavy deep learning network is used for determining physical attributes, motion rules and operation characteristics of the physical entity according to the operation data of the wind power system; the second deep learning network is used for obtaining an operation behavior model of the physical entity according to the physical attribute, the motion rule and the operation characteristic of the physical entity determined by the first deep learning network;
the operation data acquisition module is used for acquiring the operation data of the physical entity in the wind power system in real time by utilizing an OPC UA technology;
the dynamic operation simulation process output module is used for inputting the operation data of the physical entity acquired in real time into the dual deep learning network model and outputting the dynamic operation simulation process of the physical entity in the wind power system to realize data driving;
the index selection module is used for selecting indexes which influence the power generation power of the physical entity in the wind power system by utilizing a grey correlation analysis method;
the long-time memory neural network rule model building module is used for building a long-time memory neural network rule model for predicting the power generation power of the wind power system based on a particle swarm algorithm according to the selected index;
the index operation data acquisition module is used for acquiring the operation data of the index by utilizing an OPC UA technology;
the generating power prediction module is used for inputting the operation data of the index into the long-time memory neural network rule model and outputting the predicted generating power of a physical entity in the wind power system;
and the three-dimensional visualization module is used for combining the dynamic operation simulation process and the predicted power generation power of the physical entity in the wind power system with the physical entity static model by utilizing the digital twin platform to realize the three-dimensional visualization of the physical entity in the twin wind power system.
Optionally, the three-dimensional reconstruction module specifically includes:
the characteristic point extraction submodule is used for respectively extracting characteristic points from the multi-view images by adopting an SIFT characteristic extraction algorithm;
the matching submodule is used for matching the extracted characteristic points of the multi-view images by utilizing a fast approximate nearest neighbor algorithm;
the transformation matrix calculation submodule is used for calculating a transformation matrix between the successfully matched characteristic points;
the transformation submodule is used for calculating rotation and translation information between images according to the transformation matrix;
the camera calibration submodule is used for converting the pixel point coordinates of the multi-view images in the world coordinate system into the pixel point coordinates of the multi-view images in the camera coordinate system according to the rotation and translation information among the images so as to realize camera calibration;
the sparse point cloud construction sub-module is used for projecting pixel points of the multi-view images after the camera calibration to the same three-dimensional coordinate system and constructing sparse point cloud of a physical entity in the wind power system;
the denseness submodule is used for using a multi-view clustering and a dense matching algorithm based on a patch model to denseness the sparse point cloud of the physical entity in the wind power system to obtain the dense point cloud;
the surface reconstruction submodule is used for performing surface reconstruction on the dense point cloud by adopting a Poisson surface reconstruction algorithm;
and the texture mapping submodule is used for performing texture mapping on the dense point cloud after the surface reconstruction to generate a physical entity static model.
Optionally, the dual deep learning network model building module specifically includes:
the network architecture forming submodule is used for establishing a dual deep learning network model consisting of a first dual deep learning network and a second dual deep learning network;
the expression relation determining submodule is used for respectively determining the input and output expression relations of the first deep learning network and the second deep learning network; the first heavy deep learning network takes wind power system operation data as input and takes physical attributes, motion rules and operation characteristics of physical entities as output; the second deep learning network takes the physical attributes, the motion rules and the operation characteristics of the physical entities output by the first deep learning network as input and takes the operation behavior model of the physical entities as output;
and the parameter configuration submodule is used for respectively configuring parameters of the first heavy deep learning network and the second heavy deep learning network by adopting a mean square error loss function and utilizing a random gradient descent method.
Optionally, the operation data obtaining module specifically includes:
the information model construction submodule is used for constructing an OPC UA information model for describing the attributes and the relations of the production elements of the whole wind power system;
the heterogeneous data receiving submodule is used for receiving heterogeneous data of the wind power system through an OPC UA server;
the processing submodule is used for fusing and processing the heterogeneous data through an OPC UA server based on an OPC UA information model;
and the subscription submodule is used for subscribing the required type of wind power system operation data processed by the OPC UA server through the OPC UA client by utilizing a subscription mechanism.
Optionally, the long-term and short-term memory neural network rule model building module specifically includes:
the optimization submodule is used for optimizing the long-term and short-term memory neural network by adopting a particle swarm algorithm to obtain an optimized long-term and short-term memory neural network model;
the historical data set forming submodule is used for obtaining historical operating data of the indexes and historical generating power of corresponding physical entities according to the selected indexes to form a historical data set;
the dividing submodule is used for dividing the historical data set into a training set and a test set;
the training submodule is used for training the optimized long-time and short-time memory neural network model by utilizing the training set to obtain the trained long-time and short-time memory neural network model;
the testing submodule is used for carrying out simulation testing on the trained long-time and short-time memory neural network model by utilizing the testing set to obtain a prediction error;
the calling submodule is used for calling the historical data set to form a submodule if the prediction error is larger than the error threshold;
and the output submodule is used for outputting the tested long-time and short-time memory neural network model as a long-time and short-time memory neural network rule model for predicting the generated power of the wind power system if the prediction error is smaller than or equal to the error threshold.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention discloses a data-driven wind power system twinning method and system based on a deep learning network, firstly generating a physical entity static model based on a monocular vision three-dimensional reconstruction method, then a double deep learning network model is constructed, the dynamic operation simulation process of the physical entity in the wind power system is obtained by utilizing the double deep learning network model, and constructing a long-time memory neural network rule model for predicting the power generation power of the wind power system, predicting the power generation power by using the long-time memory neural network rule model, and finally combining the dynamic operation simulation process and the predicted power generation power of the physical entity in the wind power system with the physical entity static model by using the digital twin platform to realize the three-dimensional visualization of the physical entity in the twin wind power system, realize the remote monitoring and management of the wind power system equipment, improve the operation and maintenance level of the wind power system equipment and realize the short-term wind power prediction.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a data-driven wind power system twinning method based on a deep learning network provided by the invention;
FIG. 2 is a schematic diagram of a static model of a physical entity according to the present invention;
FIG. 3 is a diagram of a data transmission structure of a twin wind power system based on OPC UA technology provided in the present invention;
FIG. 4 is a schematic diagram of a dual deep learning network according to the present invention;
fig. 5 is a simulation diagram for predicting the short-term power generation amount of the wind power system based on the PSO-LSTM neural network provided by the embodiment of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a data-driven wind power system twinning method and system based on a deep learning network, so as to realize remote monitoring and management of wind power system equipment, improve the operation and maintenance level of the wind power system equipment and realize short-term wind power prediction.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
The invention provides a data-driven wind power system twinning method based on a deep learning network, which comprises the following steps of:
step S1, acquiring multi-view images of physical entities in the wind power system; the physical entities include wind generators, transformers, overhead lines, etc.
And step S2, generating a static model of the physical entity by adopting a three-dimensional reconstruction method based on monocular vision according to the multi-view image of the physical entity.
In one example, based on a three-dimensional reconstruction principle of monocular vision, an SFM algorithm is utilized to realize picture feature extraction and matching, camera calibration and pose estimation to construct sparse point cloud of a physical entity of a wind power system; and (3) densifying the sparse point cloud of the wind power system by using a CMVS-PMVS algorithm, and generating a physical entity static model of the wind power system by using the dense point cloud in combination with a Poisson surface reconstruction algorithm and a texture mapping method. The method specifically comprises the following steps:
respectively extracting feature points from the multi-view images by adopting an SIFT feature extraction algorithm;
matching the extracted characteristic points of the multi-view images by using a fast approximate nearest neighbor algorithm;
calculating a transformation matrix between the successfully matched feature points;
calculating rotation and translation information between images according to the transformation matrix;
converting the pixel point coordinates of the multi-view images in the world coordinate system into the pixel point coordinates of the multi-view images in the camera coordinate system according to the rotation and translation information among the images, so as to realize camera calibration; the principle of conversion to points in the camera coordinate system is as follows:
Figure 100122DEST_PATH_IMAGE001
wherein subscript C represents coordinates in the camera coordinate system; subscript W represents coordinates under the world coordinate system; r represents a rotation matrix; t represents a translation matrix;
projecting pixel points of multi-view images calibrated by a camera into the same three-dimensional coordinate system, and constructing sparse point clouds of physical entities in a wind power system;
using a multi-view Clustering (CMVS) algorithm and a patch-based multi-view stereo-stereo algorithm to densify the sparse point cloud of the physical entity in the wind power system to obtain a dense point cloud;
performing surface reconstruction on the dense point cloud by adopting a Poisson surface reconstruction algorithm;
and performing texture mapping on the dense point cloud after surface reconstruction to generate a physical entity static model.
The SIFT feature extraction and matching algorithm comprises the following steps:
1. a Gaussian scale space is constructed, and the realization method of the scale space comprises the following steps:
Figure 115876DEST_PATH_IMAGE002
whereinI(x,y) A representation of the target image is shown,
Figure 277736DEST_PATH_IMAGE003
is a Gaussian kernel with a variable scale,
Figure 415325DEST_PATH_IMAGE004
is a target imageI(x,y) With a variable scale 2-dimensional Gaussian function
Figure 487710DEST_PATH_IMAGE003
And (5) convolution operation results.
2. Determining the position of a feature point
3. And giving directions to the characteristic points, calculating the module length and the direction of the gradient of the scale image in the region by taking the key point as the center and 3 x 1.5 as the radius on the scale image where the key point is located by the key point obtained by the steps, wherein the formula is as follows:
Figure 761566DEST_PATH_IMAGE005
Figure 625486DEST_PATH_IMAGE006
wherein
Figure 987721DEST_PATH_IMAGE007
，
Figure 798551DEST_PATH_IMAGE008
Is a model length and direction representing the gradient at (x, y) of the scaled image.
Figure 333438DEST_PATH_IMAGE009
Representing the pixel value of the scaled image at (x +1, y),
Figure 305942DEST_PATH_IMAGE010
representing the pixel value of the scaled image at (x-1, y),
Figure 636034DEST_PATH_IMAGE011
represents the pixel value of the scale image at (x, y + 1),
Figure 188238DEST_PATH_IMAGE012
representing the pixel value of the scaled image at (x, y-1).
A schematic diagram of the generated static model of the physical entity is shown in fig. 2.
Step S3, constructing a dual deep learning network model; the dual deep learning network model comprises a first dual deep learning network and a second dual deep learning network; the first heavy deep learning network is used for determining physical attributes, motion rules and operation characteristics of the physical entity according to the operation data of the wind power system; and the second deep learning network is used for obtaining the operation behavior model of the physical entity according to the physical attributes, the motion rules and the operation characteristics of the physical entity determined by the first deep learning network.
And constructing a dual deep learning network model, and determining an expression relation between input and output of the dual network model. Namely, the first deep learning network is used for expressing the operation data of the wind power system and the physical attribute, the motion rule and the operation characteristic information of the entity equipment. The second expression is the expression of physical attributes, motion rules, operation characteristic information and an entity operation behavior model. And respective parameters of the double deep learning networks are configured, and a random gradient descent method and a mean square error loss function are adopted in the aspects of a solution algorithm and a loss function. A schematic diagram of the dual deep learning network model is shown in fig. 4.
In one example, the method specifically comprises the following steps:
establishing a dual deep learning network model consisting of a first dual deep learning network and a second dual deep learning network;
respectively determining the expression relations of input and output of the first heavy deep learning network and the second heavy deep learning network; the first heavy deep learning network takes wind power system operation data as input and takes physical attributes, motion rules and operation characteristics of physical entities as output; the second deep learning network takes the physical attributes, the motion rules and the operation characteristics of the physical entities output by the first deep learning network as input and takes the operation behavior model of the physical entities as output;
and respectively configuring parameters of the first heavy deep learning network and the second heavy deep learning network by adopting a mean square error loss function and utilizing a random gradient descent method.
Step S4, acquiring the operation data of the physical entity in the wind power system in real time by using an OPC UA (Unified Architecture) technology.
In one example, as shown in fig. 3, the method specifically includes:
constructing an OPC UA information model for describing attributes and relations of production elements of the whole wind power system;
receiving heterogeneous data of a wind power system through an OPC UA server;
based on an OPC UA information model, fusing and processing heterogeneous data through an OPC UA server;
and subscribing the required type of wind power system operation data processed by the OPC UA server through the OPC UA client by utilizing a subscription mechanism.
And step S5, inputting the real-time acquired operation data of the physical entity into the dual deep learning network model, and outputting the dynamic operation simulation process of the physical entity in the wind power system to realize data driving.
And step S6, selecting indexes influencing the generated power of the physical entity in the wind power system by using a grey correlation analysis method.
Mainly comprises the steps of solving the grey correlation coefficient of the reference sequence and the comparison sequenceε(X i ) And degree of associationτ(i) The degree of correlation is substantially the degree of difference in geometry between curves. Therefore, the magnitude of the difference between the curves can be used as a measure of the degree of correlation. For a reference sequence X0With several comparison arrays X1,X2,…,Xn，ξ(Xi) The correlation coefficient for each comparison series with the reference series at each time instant (i.e., each point in the curve). Since the correlation coefficient is the degree of correlation value of the comparison series with the reference series at each time (i.e., each point in the curve), its number is more than one. Therefore, it is necessary to concentrate the correlation coefficient at each time (i.e., each point in the curve) into one value, i.e., to obtain the average value thereof, and to express the degree of correlation, i.e., the degree of correlation, as the number of degrees of correlation between the comparison sequence and the reference sequenceτ(i). The formula used is as follows:
Figure 436686DEST_PATH_IMAGE013
Figure 111250DEST_PATH_IMAGE014
whereinρThe resolution factor is generally 0 to 1, and usually 0.5. Δ (min) is the second-order minimum difference, Δ (max) is the two-order maximum difference, Δ oi (k) Is a number sequence XiEach point on the curve and the reference number series X0Absolute difference for each point on the curve.ε i0Representing a comparison series XiWith reference number series X0The correlation coefficient of (a) is calculated,τ i representing a comparison series XiWith reference number series X0The degree of association of (a) is,ε i (k) Representing a reference sequence XiIn thatkTime of day (reference series curve numberkPoints) of the correlation coefficient,Nrepresenting a reference sequence XiThe total number of points is counted,kdenotes the second in the reference sequencekAnd (4) points.
And step S7, constructing a long-term and short-term memory neural network rule model for predicting the generated power of the wind power system based on a particle swarm algorithm according to the selected indexes.
In one example, the method specifically comprises the following steps:
optimizing the long-term and short-term memory neural network by adopting a particle swarm algorithm to obtain an optimized long-term and short-term memory neural network model;
acquiring historical operating data of the indexes and historical generating power of corresponding physical entities according to the selected indexes to form a historical data set;
dividing a historical data set into a training set and a test set;
training the optimized long-time and short-time memory neural network model by using a training set to obtain a trained long-time and short-time memory neural network model;
carrying out simulation test on the trained long-time and short-time memory neural network model by using a test set to obtain a prediction error;
if the prediction error is larger than the error threshold, returning to the step of obtaining historical operating data of the index and historical generating power of the corresponding physical entity according to the selected index to form a historical data set;
and if the prediction error is smaller than or equal to the error threshold, outputting the tested long-time and short-time memory neural network model as a long-time and short-time memory neural network rule model for predicting the generated power of the wind power system.
The particle in the particle swarm optimization updates the position and the speed of the particle by monitoring two optimal solutions, wherein the first optimal solution is the optimal solution searched by the particle, namely an individual extreme valuepbest i . The other is the optimal solution found by the whole population, namely the global optimal solutiongbest i When searching for the two optimal solutions, the particles update themselves according to the following formulaρSpeed and new position:
Figure 651340DEST_PATH_IMAGE015
Figure 7235DEST_PATH_IMAGE016
Figure 516714DEST_PATH_IMAGE017
where w is the inertial weight, c1And c2Is a learning factor, rand is [0,1 ]]λ is the velocity coefficient, λ = 1.V i t,Denotes the first in the particle groupiParticles oftThe speed of the moment in time is,V i t+,1denotes the first in the particle groupiParticles oftThe speed at the time +1 is,X i t,denotes the first in the particle groupiParticles oftThe position of the moment of time is,X i t+,1denotes the first in the particle groupiParticles oftThe position at time + 1.
And step S8, acquiring the operation data of the index by utilizing the OPC UA technology.
And step S9, inputting the operation data of the index into a long-time memory neural network rule model, and outputting the predicted power generation power of the physical entity in the wind power system.
And step S10, combining the dynamic operation simulation process, the predicted power generation power and the physical entity static model of the physical entity in the wind power system by using the digital twin platform to realize the three-dimensional visualization of the physical entity in the twin wind power system.
The following specific example of digital twinning is carried out by selecting a physical entity as a wind driven generator and taking the wind driven generator as a research object.
The method comprises the steps of obtaining different multi-angle images of the wind driven generator through a camera, and searching a wind power system containing historical weather data and a wind power generation power data set.
According to the method of the invention:
1) and the physical entity of the wind power system is constructed based on a monocular vision method.
2) And constructing a data interaction system architecture of the wind power system.
3) And constructing a dual deep learning network, extracting physical attributes, motion rules and operation behavior characteristics of the wind power system equipment, and constructing an equipment operation behavior model.
4) A traditional LSTM neural network is improved based on a particle swarm algorithm, and a short-term generated power prediction model of the wind power system is constructed in a data-driven mode. FIG. 5 is a simulation diagram for predicting the short-term power generation capacity of a wind power system based on a PSO-LSTM neural network.
The digital twin technology can map various attributes of the physical equipment into a virtual space to form a digital mirror image which can be disassembled, copied, transferable, modifiable, deleteable and repeatedly operated, so that the remote monitoring of the object is realized, and the development of the object is more comprehensively analyzed and predicted.
The method realizes three-dimensional visualization and application service of the twin wind power system based on the digital twin platform, can effectively monitor the running condition of the wind power system, improves the safety and reliability of the wind power system, and realizes low-cost maintenance of the wind power system.
The invention also provides a data-driven wind power system twin system based on the deep learning network, which comprises:
the multi-view image acquisition module is used for acquiring multi-view images of physical entities in the wind power system;
the three-dimensional reconstruction module is used for generating a physical entity static model by adopting a monocular vision-based three-dimensional reconstruction method according to the multi-view image of the physical entity;
the dual deep learning network model building module is used for building a dual deep learning network model; the dual deep learning network model comprises a first dual deep learning network and a second dual deep learning network; the first heavy deep learning network is used for determining physical attributes, motion rules and operation characteristics of the physical entity according to the operation data of the wind power system; the second deep learning network is used for obtaining an operation behavior model of the physical entity according to the physical attribute, the motion rule and the operation characteristic of the physical entity determined by the first deep learning network;
the operation data acquisition module is used for acquiring the operation data of the physical entity in the wind power system in real time by utilizing an OPC UA technology;
the dynamic operation simulation process output module is used for inputting the operation data of the physical entity acquired in real time into the dual deep learning network model and outputting the dynamic operation simulation process of the physical entity in the wind power system to realize data driving;
the index selection module is used for selecting indexes which influence the power generation power of the physical entity in the wind power system by utilizing a grey correlation analysis method;
the long-time memory neural network rule model building module is used for building a long-time memory neural network rule model for predicting the power generation power of the wind power system based on a particle swarm algorithm according to the selected index;
the index operation data acquisition module is used for acquiring the operation data of the index by utilizing an OPC UA technology;
the generating power prediction module is used for inputting the operation data of the index into the long-time memory neural network rule model and outputting the predicted generating power of the physical entity in the wind power system;
and the three-dimensional visualization module is used for combining the dynamic operation simulation process, the predicted power generation power and the physical entity static model of the physical entity in the wind power system by using the digital twin platform to realize the three-dimensional visualization of the physical entity in the twin wind power system.
The three-dimensional reconstruction module specifically comprises:
the characteristic point extraction submodule is used for respectively extracting characteristic points from the multi-view images by adopting an SIFT characteristic extraction algorithm;
the matching submodule is used for matching the extracted characteristic points of the multi-view images by utilizing a fast approximate nearest neighbor algorithm;
the transformation matrix calculation submodule is used for calculating a transformation matrix between the successfully matched characteristic points;
the transformation submodule is used for calculating rotation and translation information between images according to the transformation matrix;
the camera calibration submodule is used for converting the pixel point coordinates of the multi-view images in the world coordinate system into the pixel point coordinates of the multi-view images in the camera coordinate system according to the rotation and translation information among the images so as to realize camera calibration;
the sparse point cloud construction sub-module is used for projecting pixel points of the multi-view images after the camera calibration to the same three-dimensional coordinate system and constructing sparse point cloud of a physical entity in the wind power system;
the denseness submodule is used for using a multi-view clustering and a dense matching algorithm based on a patch model to denseness the sparse point cloud of the physical entity in the wind power system to obtain the dense point cloud;
the surface reconstruction submodule is used for performing surface reconstruction on the dense point cloud by adopting a Poisson surface reconstruction algorithm;
and the texture mapping submodule is used for performing texture mapping on the dense point cloud after the surface reconstruction to generate a physical entity static model.
The dual deep learning network model building module specifically comprises:
the network architecture forming submodule is used for establishing a dual deep learning network model consisting of a first dual deep learning network and a second dual deep learning network;
the expression relation determining submodule is used for respectively determining the input and output expression relations of the first deep learning network and the second deep learning network; the first heavy deep learning network takes wind power system operation data as input and takes physical attributes, motion rules and operation characteristics of physical entities as output; the second deep learning network takes the physical attributes, the motion rules and the operation characteristics of the physical entities output by the first deep learning network as input and takes the operation behavior model of the physical entities as output;
and the parameter configuration submodule is used for respectively configuring parameters of the first heavy deep learning network and the second heavy deep learning network by adopting a mean square error loss function and utilizing a random gradient descent method.
The operation data acquisition module specifically comprises:
the information model construction submodule is used for constructing an OPC UA information model for describing the attributes and the relations of the production elements of the whole wind power system;
the heterogeneous data receiving submodule is used for receiving heterogeneous data of the wind power system through an OPC UA server;
the processing submodule is used for fusing and processing the heterogeneous data through an OPC UA server based on an OPC UA information model;
and the subscription submodule is used for subscribing the required type of wind power system operation data processed by the OPC UA server through the OPC UA client by utilizing a subscription mechanism.
The long-time memory neural network rule model building module specifically comprises:
the optimization submodule is used for optimizing the long-term and short-term memory neural network by adopting a particle swarm algorithm to obtain an optimized long-term and short-term memory neural network model;
the historical data set forming submodule is used for obtaining historical operating data of the indexes and historical generating power of the corresponding physical entities according to the selected indexes to form a historical data set;
the dividing submodule is used for dividing the historical data set into a training set and a test set;
the training submodule is used for training the optimized long-time and short-time memory neural network model by utilizing a training set to obtain the trained long-time and short-time memory neural network model;
the testing submodule is used for carrying out simulation testing on the trained long-time and short-time memory neural network model by utilizing a testing set to obtain a prediction error;
the calling submodule is used for calling the historical data set to form a submodule if the prediction error is larger than the error threshold;
and the output submodule is used for outputting the tested long-time and short-time memory neural network model as a long-time and short-time memory neural network rule model for predicting the generated power of the wind power system if the prediction error is smaller than or equal to the error threshold.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims (10)

1. A data-driven wind power system twinning method based on a deep learning network is characterized by comprising the following steps:
acquiring a multi-view image of a physical entity in a wind power system;
generating a static model of the physical entity by adopting a three-dimensional reconstruction method based on monocular vision according to the multi-view image of the physical entity;
constructing a dual deep learning network model; the dual deep learning network model comprises a first dual deep learning network and a second dual deep learning network; the first heavy deep learning network is used for determining physical attributes, motion rules and operation characteristics of the physical entity according to the operation data of the wind power system; the second deep learning network is used for obtaining an operation behavior model of the physical entity according to the physical attribute, the motion rule and the operation characteristic of the physical entity determined by the first deep learning network;
acquiring the running data of a physical entity in the wind power system in real time by utilizing an OPC UA technology;
inputting the real-time acquired operation data of the physical entity into the dual deep learning network model, and outputting the dynamic operation simulation process of the physical entity in the wind power system to realize data driving;
selecting indexes influencing the power generation power of a physical entity in the wind power system by using a grey correlation analysis method;
according to the selected indexes, constructing a long-term memory neural network rule model for predicting the power generation power of the wind power system based on a particle swarm algorithm;
acquiring the running data of the index by utilizing an OPC UA technology;
inputting the operation data of the index into the long-time memory neural network rule model, and outputting the predicted power generation power of a physical entity in the wind power system;
and combining the dynamic operation simulation process and the predicted power generation power of the physical entity in the wind power system with the physical entity static model by using the digital twin platform to realize the three-dimensional visualization of the physical entity in the twin wind power system.
2. The data-driven wind power system twinning method based on the deep learning network as claimed in claim 1, wherein the generating of the static model of the physical entity by using a monocular vision-based three-dimensional reconstruction method according to the multi-view image of the physical entity specifically comprises:
respectively extracting feature points from the multi-view images by adopting an SIFT feature extraction algorithm;
matching the extracted characteristic points of the multi-view images by using a fast approximate nearest neighbor algorithm;
calculating a transformation matrix between the successfully matched feature points;
calculating rotation and translation information between images according to the transformation matrix;
converting the pixel point coordinates of the multi-view images in the world coordinate system into the pixel point coordinates of the multi-view images in the camera coordinate system according to the rotation and translation information among the images, so as to realize camera calibration;
projecting pixel points of multi-view images calibrated by a camera into the same three-dimensional coordinate system, and constructing sparse point clouds of physical entities in a wind power system;
using a multi-view clustering and a patch model-based dense matching algorithm to densify sparse point clouds of physical entities in a wind power system to obtain dense point clouds;
performing surface reconstruction on the dense point cloud by adopting a Poisson surface reconstruction algorithm;
and performing texture mapping on the dense point cloud after surface reconstruction to generate a physical entity static model.
3. The data-driven wind power system twinning method based on the deep learning network as claimed in claim 1, wherein the constructing of the dual deep learning network model specifically comprises:
establishing a dual deep learning network model consisting of a first dual deep learning network and a second dual deep learning network;
respectively determining the expression relations of input and output of the first heavy deep learning network and the second heavy deep learning network; the first heavy deep learning network takes wind power system operation data as input and takes physical attributes, motion rules and operation characteristics of physical entities as output; the second deep learning network takes the physical attributes, the motion rules and the operation characteristics of the physical entities output by the first deep learning network as input and takes the operation behavior model of the physical entities as output;
and respectively configuring parameters of the first heavy deep learning network and the second heavy deep learning network by adopting a mean square error loss function and utilizing a random gradient descent method.
4. The data-driven wind power system twinning method based on the deep learning network of claim 1, wherein the obtaining of the operation data of the physical entity in the wind power system in real time by using the OPC UA technology specifically comprises:
constructing an OPC UA information model for describing attributes and relations of production elements of the whole wind power system;
receiving heterogeneous data of a wind power system through an OPC UA server;
based on an OPC UA information model, fusing and processing the heterogeneous data through an OPC UA server;
and subscribing the required type of wind power system operation data processed by the OPC UA server through the OPC UA client by utilizing a subscription mechanism.
5. The data-driven wind power system twinning method based on the deep learning network of claim 1, wherein the establishing of the long-term and short-term memory neural network rule model for predicting the generated power of the wind power system based on the particle swarm algorithm according to the selected index specifically comprises:
optimizing the long-term and short-term memory neural network by adopting a particle swarm algorithm to obtain an optimized long-term and short-term memory neural network model;
acquiring historical operating data of the indexes and historical generating power of corresponding physical entities according to the selected indexes to form a historical data set;
dividing the historical data set into a training set and a test set;
training the optimized long-time and short-time memory neural network model by using the training set to obtain the trained long-time and short-time memory neural network model;
carrying out simulation test on the trained long-time and short-time memory neural network model by using the test set to obtain a prediction error;
if the prediction error is larger than the error threshold, returning to the step of obtaining historical operating data of the index and historical generating power of the corresponding physical entity according to the selected index to form a historical data set;
and if the prediction error is smaller than or equal to the error threshold, outputting the tested long-time and short-time memory neural network model as a long-time and short-time memory neural network rule model for predicting the generated power of the wind power system.
6. A data-driven wind power system twinning system based on a deep learning network is characterized by comprising:
the multi-view image acquisition module is used for acquiring multi-view images of physical entities in the wind power system;
the three-dimensional reconstruction module is used for generating a physical entity static model by adopting a monocular vision-based three-dimensional reconstruction method according to the multi-view image of the physical entity;
the dual deep learning network model building module is used for building a dual deep learning network model; the dual deep learning network model comprises a first dual deep learning network and a second dual deep learning network; the first heavy deep learning network is used for determining physical attributes, motion rules and operation characteristics of the physical entity according to the operation data of the wind power system; the second deep learning network is used for obtaining an operation behavior model of the physical entity according to the physical attribute, the motion rule and the operation characteristic of the physical entity determined by the first deep learning network;
the operation data acquisition module is used for acquiring the operation data of the physical entity in the wind power system in real time by utilizing an OPC UA technology;
the dynamic operation simulation process output module is used for inputting the operation data of the physical entity acquired in real time into the dual deep learning network model and outputting the dynamic operation simulation process of the physical entity in the wind power system to realize data driving;
the index selection module is used for selecting indexes which influence the power generation power of the physical entity in the wind power system by utilizing a grey correlation analysis method;
the long-time memory neural network rule model building module is used for building a long-time memory neural network rule model for predicting the power generation power of the wind power system based on a particle swarm algorithm according to the selected index;
the index operation data acquisition module is used for acquiring the operation data of the index by utilizing an OPC UA technology;
the generating power prediction module is used for inputting the operation data of the index into the long-time memory neural network rule model and outputting the predicted generating power of a physical entity in the wind power system;
and the three-dimensional visualization module is used for combining the dynamic operation simulation process and the predicted power generation power of the physical entity in the wind power system with the physical entity static model by utilizing the digital twin platform to realize the three-dimensional visualization of the physical entity in the twin wind power system.
7. The data-driven wind power system twin system based on the deep learning network as claimed in claim 6, wherein the three-dimensional reconstruction module specifically comprises:
the characteristic point extraction submodule is used for respectively extracting characteristic points from the multi-view images by adopting an SIFT characteristic extraction algorithm;
the matching submodule is used for matching the extracted characteristic points of the multi-view images by utilizing a fast approximate nearest neighbor algorithm;
the transformation matrix calculation submodule is used for calculating a transformation matrix between the successfully matched characteristic points;
the transformation submodule is used for calculating rotation and translation information between images according to the transformation matrix;
the camera calibration submodule is used for converting the pixel point coordinates of the multi-view images in the world coordinate system into the pixel point coordinates of the multi-view images in the camera coordinate system according to the rotation and translation information among the images so as to realize camera calibration;
the sparse point cloud construction sub-module is used for projecting pixel points of the multi-view images after the camera calibration to the same three-dimensional coordinate system and constructing sparse point cloud of a physical entity in the wind power system;
the denseness submodule is used for using a multi-view clustering and a dense matching algorithm based on a patch model to denseness the sparse point cloud of the physical entity in the wind power system to obtain the dense point cloud;
the surface reconstruction submodule is used for performing surface reconstruction on the dense point cloud by adopting a Poisson surface reconstruction algorithm;
and the texture mapping submodule is used for performing texture mapping on the dense point cloud after the surface reconstruction to generate a physical entity static model.
8. The data-driven wind power system twin system based on the deep learning network as claimed in claim 6, wherein the dual deep learning network model building module specifically comprises:
the network architecture forming submodule is used for establishing a dual deep learning network model consisting of a first dual deep learning network and a second dual deep learning network;
the expression relation determining submodule is used for respectively determining the input and output expression relations of the first deep learning network and the second deep learning network; the first heavy deep learning network takes wind power system operation data as input and takes physical attributes, motion rules and operation characteristics of physical entities as output; the second deep learning network takes the physical attributes, the motion rules and the operation characteristics of the physical entities output by the first deep learning network as input and takes the operation behavior model of the physical entities as output;
and the parameter configuration submodule is used for respectively configuring parameters of the first heavy deep learning network and the second heavy deep learning network by adopting a mean square error loss function and utilizing a random gradient descent method.
9. The data-driven wind power system twin system based on the deep learning network as claimed in claim 6, wherein the operation data obtaining module specifically comprises:
the information model construction submodule is used for constructing an OPC UA information model for describing the attributes and the relations of the production elements of the whole wind power system;
the heterogeneous data receiving submodule is used for receiving heterogeneous data of the wind power system through an OPC UA server;
the processing submodule is used for fusing and processing the heterogeneous data through an OPC UA server based on an OPC UA information model;
and the subscription submodule is used for subscribing the required type of wind power system operation data processed by the OPC UA server through the OPC UA client by utilizing a subscription mechanism.
10. The data-driven wind power system twin system based on the deep learning network of claim 6, wherein the long-term and short-term memory neural network rule model building module specifically comprises:
the optimization submodule is used for optimizing the long-term and short-term memory neural network by adopting a particle swarm algorithm to obtain an optimized long-term and short-term memory neural network model;
the historical data set forming submodule is used for obtaining historical operating data of the indexes and historical generating power of corresponding physical entities according to the selected indexes to form a historical data set;
the dividing submodule is used for dividing the historical data set into a training set and a test set;
the training submodule is used for training the optimized long-time and short-time memory neural network model by utilizing the training set to obtain the trained long-time and short-time memory neural network model;
the testing submodule is used for carrying out simulation testing on the trained long-time and short-time memory neural network model by utilizing the testing set to obtain a prediction error;
the calling submodule is used for calling the historical data set to form a submodule if the prediction error is larger than the error threshold;
and the output submodule is used for outputting the tested long-time and short-time memory neural network model as a long-time and short-time memory neural network rule model for predicting the generated power of the wind power system if the prediction error is smaller than or equal to the error threshold.
CN202210213632.0A 2022-03-07 2022-03-07 Data-driven wind power system twinning method and system based on deep learning network Active CN114297947B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210213632.0A CN114297947B (en) 2022-03-07 2022-03-07 Data-driven wind power system twinning method and system based on deep learning network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210213632.0A CN114297947B (en) 2022-03-07 2022-03-07 Data-driven wind power system twinning method and system based on deep learning network

Publications (2)

Publication Number Publication Date
CN114297947A true CN114297947A (en) 2022-04-08
CN114297947B CN114297947B (en) 2022-05-27

Family

ID=80978565

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210213632.0A Active CN114297947B (en) 2022-03-07 2022-03-07 Data-driven wind power system twinning method and system based on deep learning network

Country Status (1)

Country Link
CN (1) CN114297947B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115270491A (en) * 2022-08-06 2022-11-01 福建华电福瑞能源发展有限公司福建分公司 Offshore wind power operation and maintenance platform design method based on multivariate information fusion
CN115983494A (en) * 2023-02-10 2023-04-18 广东工业大学 Short-term wind power prediction method and system for newly-built small-sample wind power plant
CN117477648A (en) * 2023-11-09 2024-01-30 商运(江苏)科创发展有限公司 Wind power equipment modeling and performance evaluation method based on digital twin technology
CN117593702A (en) * 2024-01-18 2024-02-23 深圳市光明顶技术有限公司 Remote monitoring method, device, equipment and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112434848A (en) * 2020-11-19 2021-03-02 西安理工大学 Nonlinear weighted combination wind power prediction method based on deep belief network
US20210110262A1 (en) * 2019-10-14 2021-04-15 Honda Research Institute Europe Gmbh Method and system for semi-supervised deep anomaly detection for large-scale industrial monitoring systems based on time-series data utilizing digital twin simulation data
CN113236491A (en) * 2021-05-27 2021-08-10 华北电力大学 Wind power generation digital twin system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210110262A1 (en) * 2019-10-14 2021-04-15 Honda Research Institute Europe Gmbh Method and system for semi-supervised deep anomaly detection for large-scale industrial monitoring systems based on time-series data utilizing digital twin simulation data
CN112434848A (en) * 2020-11-19 2021-03-02 西安理工大学 Nonlinear weighted combination wind power prediction method based on deep belief network
CN113236491A (en) * 2021-05-27 2021-08-10 华北电力大学 Wind power generation digital twin system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
殷豪等: "基于二次模式分解和级联式深度学习的超短期风电功率预测", 《电网技术》 *
王兴志 等: "基于数字孪生和深度学习的新一代调控系统预调度方法", 《上海交通大学学报》 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115270491A (en) * 2022-08-06 2022-11-01 福建华电福瑞能源发展有限公司福建分公司 Offshore wind power operation and maintenance platform design method based on multivariate information fusion
CN115983494A (en) * 2023-02-10 2023-04-18 广东工业大学 Short-term wind power prediction method and system for newly-built small-sample wind power plant
CN115983494B (en) * 2023-02-10 2023-09-12 广东工业大学 Short-term wind power prediction method and system for newly-built small-sample wind power plant
CN117477648A (en) * 2023-11-09 2024-01-30 商运(江苏)科创发展有限公司 Wind power equipment modeling and performance evaluation method based on digital twin technology
CN117477648B (en) * 2023-11-09 2024-03-08 商运(江苏)科创发展有限公司 Wind power equipment modeling and performance evaluation method based on digital twin technology
CN117593702A (en) * 2024-01-18 2024-02-23 深圳市光明顶技术有限公司 Remote monitoring method, device, equipment and storage medium
CN117593702B (en) * 2024-01-18 2024-04-09 深圳市光明顶技术有限公司 Remote monitoring method, device, equipment and storage medium

Also Published As

Publication number Publication date
CN114297947B (en) 2022-05-27

Similar Documents

Publication Publication Date Title
CN114297947B (en) Data-driven wind power system twinning method and system based on deep learning network
Wilson et al. Deep learning-aided cyber-attack detection in power transmission systems
CN110533631A (en) SAR image change detection based on the twin network of pyramid pondization
CN111462324B (en) Online spatiotemporal semantic fusion method and system
CN114972213A (en) Two-stage mainboard image defect detection and positioning method based on machine vision
CN114359283A (en) Defect detection method based on Transformer and electronic equipment
Moayyed et al. A Cyber-Secure generalized supermodel for wind power forecasting based on deep federated learning and image processing
CN115223049B (en) Knowledge distillation and quantification method for large model compression of electric power scene edge calculation
CN111178585A (en) Fault reporting amount prediction method based on multi-algorithm model fusion
CN111680875A (en) Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model
CN115187154A (en) Neural network-based regional power grid oscillation source risk prediction method and system
CN113569928B (en) Train running state detection data missing processing model and reconstruction method
CN116384020A (en) Digital twin substation space layout method integrating multidimensional semantic information
CN116169675A (en) Power system dynamic stability online evaluation method considering operation mode change
CN113052103A (en) Electrical equipment defect detection method and device based on neural network
CN117371949A (en) Three-dimensional visual model-based power transmission line construction safety monitoring method and system
CN116029618B (en) Dynamic safety partition assessment method and system for power system
CN117076928A (en) Bridge health state monitoring method, device and system and electronic equipment
Lin et al. Post-disturbance dynamic frequency features prediction based on convolutional neural network
JP6950647B2 (en) Data determination device, method, and program
Ma et al. A fault prediction framework for Doubly‐fed induction generator under time‐varying operating conditions driven by digital twin
CN117437230B (en) Photovoltaic power station power prediction method and system based on image restoration
Guo et al. Transmission line galloping detection based on SURF feature and FLANN matching algorithm
CN111354028A (en) Binocular vision-based power transmission channel hidden danger identification and tracking method
Dabou et al. Supervised learning of overcomplete dictionaries for rapid response-based dynamic stability prediction

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