CN107124003A - Wind power plant wind energy Forecasting Methodology and equipment - Google Patents
Wind power plant wind energy Forecasting Methodology and equipment Download PDFInfo
- Publication number
- CN107124003A CN107124003A CN201710296557.8A CN201710296557A CN107124003A CN 107124003 A CN107124003 A CN 107124003A CN 201710296557 A CN201710296557 A CN 201710296557A CN 107124003 A CN107124003 A CN 107124003A
- Authority
- CN
- China
- Prior art keywords
- data
- wind energy
- neuron
- wind
- layer
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000013528 artificial neural network Methods 0.000 claims abstract description 28
- 210000002569 neuron Anatomy 0.000 claims description 43
- 210000004205 output neuron Anatomy 0.000 claims description 31
- 241001269238 Data Species 0.000 claims description 26
- 210000002364 input neuron Anatomy 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 19
- 238000005259 measurement Methods 0.000 claims description 15
- 238000012544 monitoring process Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 5
- 210000005036 nerve Anatomy 0.000 claims description 3
- 238000002834 transmittance Methods 0.000 claims 2
- 238000010586 diagram Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- 238000013213 extrapolation Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000007935 neutral effect Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000000712 assembly Effects 0.000 description 1
- 238000000429 assembly Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000739 chaotic effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005684 electric field Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Classifications
-
- H02J3/386—
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- 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
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A30/00—Adapting or protecting infrastructure or their operation
-
- 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
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/70—Wind energy
- Y02E10/76—Power conversion electric or electronic aspects
Landscapes
- Wind Motors (AREA)
Abstract
A kind of wind power plant wind energy Forecasting Methodology and equipment are provided.Methods described includes：The first data and the second data are obtained, wherein, the first data include historical wind speed data, and the second data include history wind energy data corresponding with the first data；Using the first data as the artificial neural network for predicting wind energy data input variable and using the second data as the artificial neural network target variable, to train the artificial neural network；The 3rd data are obtained, wherein, the 3rd data include the air speed data of prediction；By by the 3rd data input to the trained artificial neural network, to obtain the wind energy data of prediction.The wind power plant wind energy Forecasting Methodology and equipment provided by using the present invention, can be predicted the wind energy data of longer time scope, and the result predicted can have higher robustness and accuracy.
Description
Technical field
The present invention relates to a kind of wind power plant wind energy Forecasting Methodology and equipment.More particularly, it is related to a kind of user
The short-term wind energy Forecasting Methodology of wind power plant and equipment of artificial neural networks.
Background technology
Although the environmental benefit of wind-power electricity generation is significant, the continuous and chaotic fluctuation of wind speed causes wind power plant
Power output completely random.Therefore, the assessment of the power output of this generator is always relevant with some uncertainties.Due to this
Plant uncertain, a large amount of wind energies, which are connected in power system, may bring huge challenge.However, this challenge is not
It can not overcome.In order to increase the economic benefit and acceptability of wind energy, and reduce due to over-evaluating or underestimating and leading to wind energy
The punishment from spot market caused to wind energy and wind speed, it is necessary to carry out Accurate Prediction.
Nowadays, several prediction wind energies and the method for wind speed have been developed.Based on used forecast model, existing method
It can be concluded as statistic law, Physical and time series modeling method.These conventional forecast models have the model between 3-6 hours
The predictive ability enclosed.
However, a large amount of wind energies are connected to the challenge that may be brought in power system in order to overcome, it is only small to following 3-6
When wind energy carry out Accurate Prediction be far from being enough.Therefore have what following at least 24 hours wind energy was accurately predicted
The demand of short-term wind energy Predicting Technique.
The content of the invention
According to the one side of exemplary embodiment, there is provided a kind of wind power plant wind energy Forecasting Methodology, the wind-force hair
Electric field wind energy Forecasting Methodology includes：The first data and the second data are obtained, wherein, the first data include historical wind speed data, the
Two data include history wind energy data corresponding with the first data；First data are used as to the artificial god for predicting wind energy data
Input variable through network and using the second data as the artificial neural network target variable, to train the artificial neuron
Network；The 3rd data are obtained, wherein, the 3rd data include the air speed data of prediction；By the way that the 3rd data input is instructed to process
The experienced artificial neural network, to obtain the wind energy data of prediction.
The artificial neural network is baek-propagetion network, and it has the feedforward net of the layer of multiple interconnections
Network, the layer includes input layer, multiple hidden layers and output layer；Wherein, input layer includes defeated for describing one or more
Enter the input neuron of variable, hidden layer includes multiple hidden neurons, and output layer includes the output for being used to describe output variable
Neuron.
The baek-propagetion network includes forward path and amendment path；The forward path is from input layer
Input neuron starts, and sequentially passes through the hidden neuron of each hidden layer, reaches the output neuron of output layer；The forward direction
Path is used to set each input neuron according to the first data that Online Monitoring Control and data collecting system are monitored, and counts successively
The numerical value of the hidden neuron of each hidden layer is calculated, and calculates output neuron；The output neuron represents history wind energy number
According to；The amendment path sequentially passes through the hidden neuron of each hidden layer since the output neuron of output layer, reversely；Institute
The connection weight weight values that amendment path is used to adjust each layer neuron are stated, to build new forward path.
First data may include many datas with very first time interval in the range of the very first time, and the second data can be wrapped
Many datas with the second time interval in the second time range are included, the 3rd data may include the tool in the 3rd time range
There are many datas of the 3rd time interval.
The very first time interval, the second time interval and the 3rd time interval can be 10 minutes.
Very first time scope and the second time range can be at least to pass by 1 year, and the 3rd time range can be at least future
24 hours.
First data may also include at least one in historical temperature data, history wind direction data and history humidity data,
3rd data may also include at least one in temperature data, the wind direction data of prediction and the humidity data of prediction of prediction.
First data and the second data can be the historical datas from Online Monitoring Control and data collecting system.
3rd data can be the data by using numerical weather forecast model prediction, wherein, the Numerical Weather is pre-
Model is reported to be used to determine the wind speed at hub of wind power generator height.
According to the another aspect of exemplary embodiment there is provided a kind of pre- measurement equipment of wind power plant wind energy, the wind energy is pre-
Measurement equipment includes：Acquiring unit, it is used to obtain the first data, the second data and the 3rd data, wherein, the first data include going through
History air speed data, the second data include history wind energy data corresponding with the first data, and the 3rd data include the wind speed number of prediction
According to；Training unit, it is used for the input variable as the artificial neural network for predicting wind energy data and general using the first data
Second data as the artificial neural network target variable, to train the artificial neural network；Predicting unit, it is used for
By by the 3rd data input to the trained artificial neural network, to obtain the wind energy data of prediction.
The artificial neural network is baek-propagetion network, and it has the feedforward net of the layer of multiple interconnections
Network, the layer includes input layer, multiple hidden layers and output layer；Wherein, input layer includes defeated for describing one or more
Enter the input neuron of variable, hidden layer includes multiple hidden neurons, and output layer includes the output for being used to describe output variable
Neuron.
The baek-propagetion network includes forward path and amendment path；The forward path is from input layer
Input neuron starts, and sequentially passes through the hidden neuron of each hidden layer, reaches the output neuron of output layer；The forward direction
Path is used to set each input neuron according to the first data that Online Monitoring Control and data collecting system are monitored, and counts successively
The numerical value of the hidden neuron of each hidden layer is calculated, and calculates output neuron；The output neuron represents history wind energy number
According to；The amendment path sequentially passes through the hidden neuron of each hidden layer since the output neuron of output layer, reversely；Institute
The connection weight weight values that amendment path is used to adjust each layer neuron are stated, to build new forward path.
First data may include many datas with very first time interval in the range of the very first time, and the second data can be wrapped
Many datas with the second time interval in the second time range are included, the 3rd data may include the tool in the 3rd time range
There are many datas of the 3rd time interval.
The very first time interval, the second time interval and the 3rd time interval can be 10 minutes.
Very first time scope and the second time range can be at least to pass by 1 year, and the 3rd time range can be at least future
24 hours.
First data may also include at least one in historical temperature data, history wind direction data and history humidity data,
3rd data may also include at least one in temperature data, the wind direction data of prediction and the humidity data of prediction of prediction.
First data and the second data can be the historical datas from Online Monitoring Control and data collecting system.
3rd data can be the data by using numerical weather forecast model prediction, wherein, the Numerical Weather is pre-
Model is reported to be used to determine the wind speed at hub of wind power generator height.
According to the another aspect of exemplary embodiment there is provided a kind of computer-readable recording medium, have program stored therein, it is described
The executable wind power plant wind energy Forecasting Methodology according to the present invention of program.
According to the another aspect of exemplary embodiment there is provided a kind of computer, including the computer program that is stored with is readable
Medium, the executable wind power plant wind energy Forecasting Methodology according to the present invention of the computer program.
The wind power plant wind energy Forecasting Methodology and equipment provided by using the present invention, can be predicted longer time scope
Wind energy data, and the result predicted can have higher robustness and accuracy.
Other features and aspect will be apparent from following embodiment, drawings and claims.
Brief description of the drawings
From the description carried out below in conjunction with accompanying drawing to example embodiment, these and/or other aspect will be apparent and more
It is readily appreciated that, wherein：
Fig. 1 is the block diagram for the pre- measurement equipment of wind energy for showing the exemplary embodiment according to the present invention；
Fig. 2 is the flow chart for the wind power plant wind energy Forecasting Methodology for showing the exemplary embodiment according to the present invention；
Fig. 3 is the schematic diagram for the artificial neural network for showing the exemplary embodiment according to the present invention；
Fig. 4 is the original for showing artificial neural network forward path and amendment path according to the exemplary embodiment of the present invention
Reason figure；
Fig. 5 is set according to the use wind power plant wind energy prediction as in Figure 1 and Figure 2 of the exemplary embodiment of the present invention
The diagram for the test data that standby and method is simulated.
Embodiment
Hereinafter, some example embodiments be will be described in detail with reference to the accompanying drawings.Ginseng on distributing to the element in accompanying drawing
Examine label, it should be noted that in the case of any possible, even if identical element is illustrated in different drawings, similar elements
Also it will be denoted by the same reference numerals.In addition, in describing the embodiments of the present, when thinking and known related structure or function
When detailed description will cause the explanation of the disclosure fuzzy, the detailed description to known related structure or function will be omitted.
In addition, can be used the term such as first, second, A, B, (a), (b) to describe component here.In these terms
Each term is not used in essence, the order or sequence for limiting corresponding assembly, and is only used for carrying out corresponding assembly and other components
Distinguish.Term as used herein is only used for describing the purpose of specific embodiment, is not intended to limitation.Unless the context clearly
Indicate, otherwise singulative as used herein is also intended to including plural form.It will also be understood that when used herein, term " bag
Include " and/or specified feature, entirety, step, operation, element and/or the component in the presence of narration of "comprising", but do not exclude the presence of
Or add one or more of the other feature, entirety, step, operation, element, component and/or their group.
It shall yet further be noted that in some selectable embodiments, function/action of proposition can not be by the order shown in accompanying drawing
Occur.For example, two accompanying drawings continuously shown actually can be performed or can performed in reverse order sometimes simultaneously substantially, this
Depending on the function/action included.
Some example embodiments are described above.However, it should be understood that various repair can be made to these example embodiments
Change.If for example, description technology be performed in a different order, and/or if description system, framework, device or circuit
In component be combined in a different manner and/or by other assemblies or they equivalent replace or supplement, then can be closed
Suitable result.Therefore, other embodiment is within the scope of the claims.
Fig. 1 is the block diagram for the pre- measurement equipment 100 of wind energy for showing the exemplary embodiment according to the present invention.
The pre- measurement equipment 100 of wind energy can be used for the generated output of prediction wind power generation field in the next few days.According to exemplary
Embodiment, the pre- measurement equipment 100 of wind energy includes：Acquiring unit 110, training unit 120 and predicting unit 130.
Acquiring unit 110 is the data needed for the pre- measurement equipment 100 of wind energy is obtained.Acquiring unit 110 is used to obtain the first number
According to, the second data and the 3rd data.Wherein, the first data and the second data are historical data, and the 3rd data are the number of prediction
According to.According to exemplary embodiment, the first data may include historical wind speed data, and the second data may include history wind energy data, and
And second data with the first data be corresponding.For example, the actual wind speed number that the first data can sometime be put for history
According to then the second data can be the actual wind energy data of the historical time point.In addition, the 3rd data may include the wind speed number of prediction
According to.For example, the 3rd data can be following air speed data of prediction.
In the exemplary embodiment, the first data and the second data can come from Online Monitoring Control and data acquisition
The historical data of (Supervisory Control And Data Acquisition, SCADA) system.Online SCADA system
Wind speed and the composite record of power output and generator availability are provided, its basis that can be predicted as short-term wind-power electricity generation.
The levels of precision influence foreseeable on wind of 3rd data (that is, the air speed data of prediction) be can not ignore.Obtain
Wind data there is several methods that：Observation, data mining and numerical value weather simulation.Obtain the most directly and reliable of wind data
Method is by field observation.However, they can not generally provide predicted value.Data mining is more flexible, but it reduces day
Gas data scale it is limited in one's ability.Numerical weather forecast (Numerical Weather Prediction, NWP) model uses energy
The physics conservation of equation is measured, so allows more real data scale to reduce.In fact, the high-resolution NWP of wind-force is to wind energy
Prediction plays a key effect.
In the exemplary embodiment, the 3rd data can be the data by using NWP model predictions.NWP refers to basis
Air actual conditions, under certain initial value and boundary condition, numerical computations are made by mainframe computer, are solved description weather and are drilled
The hydrodynamics of change process and thermodynamic (al) equation group, the air motion state of prediction following certain period and the side of weather phenomenon
Method.It is outer with reference to log law and wind shear force method etc. using the meteorological data at 10 meters of ground in present inventive concept
Push away method to pre-process meteorological data, obtain the wind speed (that is, the 3rd data) at hub of wind power generator height.Due to
Height of the wheel hub of the wind power generating set of various models apart from ground is different, and every unit wheel can be predicted using NWP models
The wind speed of hub position (for example, 70 meters of height, 80 meters of height etc.).
First data, the second data and the 3rd data may each comprise many with intervals of certain time scope
Data.Specifically, the first data may include many datas with very first time interval in the range of the very first time, second
Data may include many datas with the second time interval in the second time range, and the 3rd data may include the 3rd time model
Enclose interior many datas with the 3rd time interval.
Specifically, because the second data are corresponding with the first data, the second time range can be equal to first
Time range, the second time interval can be equal to very first time interval.For example, very first time scope and the second time range can be
At least pass by 1 year, the very first time interval and the second time interval can be 10 minutes.
Therefore, embodiments in accordance with the present invention, the first data can be at least pass by 1 year with 10 minutes for many of interval
Bar air speed data, the second data can be at least pass by 1 year with 10 minutes a plurality of wind energy data for interval.
In addition, the 3rd time range is smaller than very first time scope and/or the second time range.For example, in exemplary reality
Apply in example, very first time scope and the second time range are at least to pass by 1 year, the 3rd time range is that at least future 24 is small
When.3rd time interval can be equal to very first time interval and/or the second time interval, for example, in the exemplary embodiment, first
Time interval, the second time interval and the 3rd time interval are 10 minutes.Therefore, according to the exemplary embodiment of the present invention,
3rd data can be at least following 24 hours with 10 minutes air speed datas for a plurality of prediction at interval.But according to another
Exemplary embodiment, the 3rd time interval can be not equal to very first time interval and/or the second time interval.For example, the very first time
Interval and the second time interval are 10 minutes, and the 3rd time interval is 5 minutes.However, it will be understood by those skilled in the art that
One time interval, the second time interval and the 3rd time interval, and very first time scope, the second time range and the 3rd time
Scope can be not limited to these examples.
In addition to air speed data, the first data and the 3rd data may also include in temperature data, wind direction data and humidity data
At least one.The type of 3rd data should be corresponding to the type that the first data are used.For example, when the first data include going through
During history air speed data, the 3rd data may include the air speed data of prediction.When the first data include historical wind speed data and history temperature
During degrees of data, the 3rd data may include the air speed data of prediction and the temperature data of prediction.When the first data include historical wind speed
When data, historical temperature data and history wind direction data, the 3rd data may include the air speed data of prediction, the temperature data of prediction
With the wind direction data of prediction.When the first data are wet including historical wind speed data, historical temperature data, history wind direction data and history
During degrees of data, the 3rd data may include the wet of air speed data, the temperature data of prediction, the wind direction data of prediction and the prediction of prediction
Degrees of data.
Training unit 120 trains artificial neural network (the Artificial Neural for predicting wind energy data
Networks, ANN).Specifically, training unit 120 can be by the way that the first data to be used as to the ANN's for predicting wind energy data
Input variable and using the second data as ANN target variable, to train ANN, to obtain reflecting for the first data and the second data
Penetrate relation.
For example, the first data are a plurality of actual history air speed data that is obtained from online SCADA system, the second data be from
The a plurality of actual history wind energy data that online SCADA system is obtained, by the way that actual history air speed data as ANN input are become
Amount, using corresponding actual history wind energy data, as ANN target variable, are come on time point with the actual history air speed data
Air speed data and the relation of wind energy data are trained, and by being iterated training to many datas, can finally obtain air speed data
With the mapping relations between wind energy data.
Here, ANN is a kind of powerful data modeling tool, can capture and represent complicated input/output relation.
ANN is made up of the neuron of a large amount of height interconnections, and these neurons unanimously work to solve particular problem.
As shown in figure 3, ANN is the feedforward network of the layer with multiple interconnections, these layers include input layer, Duo Geyin
Hide layer and output layer.Wherein, input layer includes the input neuron X for being used to describe one or more input variable, hidden layer
Comprising multiple hidden neuron H, output layer includes the output neuron Y for being used to describe output variable.
ANN input neuron XjWith output neuron YiBetween relation provided by equation 1：
【Equation 1】
Wherein, XjIt is j-th of input neuron, YiIt is i-th of output neuron, WijIt is input neuron XiConnect with it
The output neuron Y connectiBetween connection weight, n for input neuron quantity, biIt is the deviation of i-th of output neuron,
fiIt is to determine the activation primitive of the property of neutral net.
ANN mean square error (Mean Square Error, MSE) is defined by equation 2：
【Equation 2】
Wherein, TpIt is the desired value of pth group training sample, YpIt is the network output predicted value in pth group training sample, N is
The quantity of training sample.
In the exemplary embodiment, the ANN used is baek-propagetion network.
As shown in figure 4, the baek-propagetion network in this example includes forward path and amendment path.
Wherein, forward path sequentially passes through the hidden neuron of each hidden layer since the input neuron of input layer,
Reach the output neuron of output layer.
In forward path, the historical wind speed data that are monitored according to online SCADA system (can also be temperature, humidity,
The meteorological datas such as wind direction) each input neuron is set, the numerical value of the hidden neuron of each hidden layer is calculated successively, and it is defeated
Go out neuron.The output neuron represents history wind energy data.
Specifically, the output of all neurons in neutral net is calculated in forward path.Hidden from first layer
Layer starts, using the independent variable from the training dataset history monitoring data of online SCADA system (this example be) as
Input value.By performing, related summation and activation primitive are assessed, and nerve is calculated to all neurons in the first hidden layer
Member output.Using these outputs as the input of the neuron in the second hidden layer, summation and the activation letter of correlation are performed again
Number is calculated, to calculate the output of second layer neuron.
As shown in figure 4, amendment path sequentially passes through the hidden of each hidden layer since the output neuron of output layer, reversely
Hide neuron.In the positive path of repairing, the propagation of calculation error and the adjustment of weight are carried out.From the error of the neuron in output layer
Calculating starts.Conventional error function is the output neuron Y in certain group training samplepWith the desired value T of the neuronpBetween
The difference of two squares.In the present embodiment, the desired value T of the neuronpThe history wind energy data monitored for online SCADA system.
To each its error amount of data point Continuous plus in training data, the weight W of the connection of each layer neuron is adjustedijNew value,
Until error amount reaches zero or less than specific threshold.Now, one group of new weighted value is obtained, is built newly based on new weighted value
Forward path.
During training, forward path and amendment path are circulated through.
In a preferred embodiment, the foreseeable time interval of wind is 10 minutes, and prediction timeliness is 24 hours, can
To be used as one group of training sample (totally 144 groups of training samples) using the history meteorological data every 10 minutes.
When training unit 120 trains the ANN for predicting wind energy data, predicting unit 130 can be by the way that the 3rd be counted
According to trained ANN is input to, to obtain the wind energy data of prediction.
In addition, when the 3rd data is the meteorological datas at 10 meters of ground from NWP, wind energy prediction meanss 100
It may also include preprocessor (not shown).Preprocessor is used for using the Extrapolation method such as log law and wind shear force method to the
Three data are pre-processed, and obtain the appropriate wind-force information the (that is, the pretreated 3rd at hub of wind power generator height
Data).
According to the exemplary embodiment of the present invention, when training unit 120 using at least pass by the actual wind speed data of 1 year with
At least pass by the actual wind energy data of 1 year to train during ANN, the wind speed in 24 hours futures of prediction can be used in predicting unit 130
Data input is to the ANN trained with the wind energy data in 24 hours futures for obtaining prediction.
Therefore, the result of prediction can have higher robustness and accuracy, and can realize to wind energy it is short-term (for example,
24 hours) prediction.
Fig. 2 is the flow chart for the wind power plant wind energy Forecasting Methodology for showing the exemplary embodiment according to the present invention.Portion
Divide term to be explained and illustrate with reference to Fig. 1, therefore repeated description will be omitted.
In step slo, acquiring unit 110 obtains the first data and the second data.First data include historical wind speed number
According to the second data include history wind energy data corresponding with the first data.In response to receiving the first data and the second data, obtain
Take unit 110 that first data and the second data are sent into training unit 120.
Here, the first data may include many datas with very first time interval in the range of the very first time, the second number
According to many datas with the second time interval that may include in the second time range.In addition to historical wind speed data, the first data
Also include at least one in historical temperature data, history wind direction data and history humidity data.
In addition, the first data and the second data can be the historical datas from online SCADA system.
In step S20, training unit 120 using the first data as the ANN for predicting wind energy data input variable
And using the second data as ANN target variable, to train ANN.
Here, artificial neural network is baek-propagetion network, and it has the feedforward of the layer of multiple interconnections
Network, the layer includes input layer, multiple hidden layers and output layer；Wherein, input layer includes being used to describe one or more
The input neuron of input variable, hidden layer includes multiple hidden neurons, and output layer includes being used to describe the defeated of output variable
Go out neuron.
In addition, baek-propagetion network includes forward path and amendment path；Forward path is from the defeated of input layer
Enter neuron to start, sequentially pass through the hidden neuron of each hidden layer, reach the output neuron of output layer；Forward path is used
Each input neuron is set in the first data monitored according to online SCADA system, hiding for each hidden layer is calculated successively
The numerical value of neuron, and calculate output neuron；The output neuron represents history wind energy data；Path is corrected from output layer
Output neuron start, reversely sequentially pass through the hidden neuron of each hidden layer；The amendment path is used to adjust each layer
The connection weight weight values of neuron, to build new forward path.
In step s 30, acquiring unit 110 obtains the 3rd data, wherein, the 3rd data include the air speed data of prediction.
In response to receiving the 3rd data, the 3rd data are sent to predicting unit 130 by acquiring unit 110.
In addition, when the 3rd data are the meteorological datas at 10 meters of ground from NWP, step S30 may also include
The 3rd data are pre-processed using the Extrapolation method such as log law and wind shear force method, obtained in hub of wind power generator
The highly appropriate wind-force information (that is, pretreated 3rd data) at place, and the 3rd data by pretreatment are sent to
Predicting unit 130.
Here, the 3rd data may include many datas with the 3rd time interval in the 3rd time range.In example
In property embodiment, the 3rd data can be the data by using NWP model predictions.NWP refers to according to air actual conditions,
Under certain initial value and boundary condition, numerical computations are made by mainframe computer, the fluid force for describing weather modification process is solved
Learn and thermodynamic (al) equation group, the method for the air motion state and weather phenomenon of prediction following certain period.In structure of the present invention
In think of, using the meteorological data at 10 meters of ground, with reference to the Extrapolation method such as log law and wind shear force method to meteorology
Data are pre-processed, and obtain the wind speed (that is, the 3rd data) at hub of wind power generator height.Due to the wind of various models
Height of the wheel hub of power generator group apart from ground is different, using NWP models can predict every machine unit hub position (for example,
70 meters of height, 80 meters of height etc.) wind speed.
In addition, in addition to the air speed data of prediction, the 3rd data also include the temperature data of prediction, the wind direction data of prediction and
At least one in the humidity data of prediction.
In the exemplary embodiment, very first time interval, the second time interval and the 3rd time interval can be 10 minutes.
Very first time scope and the second time range can be at least to pass by 1 year, and the 3rd time range is at least following 24 hours.
In step s 40, predicting unit 130 is by by the 3rd data input to trained ANN, to obtain prediction
Wind energy data.
The simulated test data of the method for the prediction wind energy of the application present invention are described in detail hereinafter with reference to Fig. 5.
Fig. 5 is the use wind power plant wind energy prediction side as in Figure 1 and Figure 2 of the exemplary embodiment according to the present invention
The diagram for the simulated test data that method is simulated.
In order to assess the performance of wind power plant wind energy Forecasting Methodology of the invention, using from online SCADA system
The wind speed and wind energy being recorded from March 25,26 days to 2015 March in 2014 are counted respectively as the first data and second
According to being trained to the ANN (for example, baek-propagetion network) for predicting wind energy.Utilize the every of NWP model predictions
The wind speed of the prediction in ensuing 4 days (on the March 29,26 days to 2015 March in 2015) at 10 minutes intervals is used as the 3rd number
According to by the 3rd data input to trained ANN.Fig. 5 shows the wind energy of prediction on March 26th, 2015.
Predict the outcome and show, the wind energy of prediction is almost similar to the actual wind energy recorded by online SCADA, shows the present invention
Wind power plant wind energy Forecasting Methodology effective performance.
【Table 1】
Table 1 lists the predicated error of the wind energy of daily prediction in 29, in 26 days to 2015 on March March in 2015.In order to
Assess in the accuracy of the wind power plant wind energy Forecasting Methodology of the present invention, table 1 and consider mean absolute error (Mean
Absolute Error, MAE), root-mean-square error (Root Mean Square Error, RMSE) and average absolute percent miss
Poor (Mean Absolute Error Percentage, MAPE).
On March 26th, 2015, error amount of the invention be 12.1620%, on March 27th, 2015 be 15.5106%, 2015
On March 28, be 12.0655%, on March 29th, 2015 be 15.6206%, far below above-mentioned numerical value disclosed in the prior art.
The wind power plant wind energy Forecasting Methodology according to the present invention can be performed according to computer program instructions.These are calculated
Machine programmed instruction may be recorded on computer-readable recording medium.The computer readable recording medium storing program for performing stores it to be any
The data storage device for the data that can be read afterwards by computer system.Programmed instruction and medium can be for the purpose of the present invention
The special programmed instruction and medium for designing and manufacturing, or it is known to the technical staff in terms of computer software that they, which can be,
And available type.
In addition, can also according to can run above-mentioned instruction and/or computer with above computer readable storage medium storing program for executing or
Hardware come perform according to the present invention wind power plant wind energy Forecasting Methodology.Computer or hardware can be for the purpose of the present invention
And the computer or hardware for specially designing and manufacturing, or can be to the technical staff of computer or hardware aspect it is known simultaneously
Available type.
As by it is above-mentioned predict the outcome proved with forecast of the measure error, the method for prediction wind energy of the invention has micro-
Small error, and with acceptable consistent compared with being derived from the data of actual wind-power electricity generation power of online SCADA system
Property, indicate the robustness and accuracy of the ANN wind energy forecast models of the present invention.The model has generated desired prediction essence
Degree, therefore, the wind energy of prediction may be used as the input quantity with one day scheduling of resource in advance in the permeable micro-capacitance sensor of high wind
One of.
Although describing the wind power plant wind energy Forecasting Methodology and equipment according to the present invention with reference to specific embodiment,
But it should be appreciated by those skilled in the art without departing from the spirit or scope of the present invention, can do to the present invention
Go out various variants and modifications, the scope of the present invention is limited by claim and its equivalent.
Claims (16)
1. a kind of wind power plant wind energy Forecasting Methodology, it is characterised in that the wind power plant wind energy Forecasting Methodology includes：
The first data and the second data are obtained, wherein, the first data include historical wind speed data, and the second data include and the first number
According to corresponding history wind energy data；
Using the first data as the artificial neural network for predicting wind energy data input variable and regard the second data as institute
The target variable of artificial neural network is stated, to train the artificial neural network；
The 3rd data are obtained, wherein, the 3rd data include the air speed data of prediction；
By by the 3rd data input to the trained artificial neural network, to obtain the wind energy data of prediction.
2. wind power plant wind energy Forecasting Methodology as claimed in claim 1, it is characterised in that before the artificial neural network is
Reverse transmittance nerve network is presented, it has a feedforward network of the layer of multiple interconnections, the layer includes input layer, multiple hidden
Layer and output layer；
Wherein, input layer includes the input neuron for being used to describe one or more input variable, and hidden layer is comprising multiple hidden
Neuron is hidden, output layer includes the output neuron for being used to describe output variable.
3. wind power plant wind energy Forecasting Methodology as claimed in claim 2, it is characterised in that the feedforward Back propagation neural
Network includes forward path and amendment path；
The forward path sequentially passes through the hidden neuron of each hidden layer since the input neuron of input layer, reaches
The output neuron of output layer；The forward path is used for the first number monitored according to Online Monitoring Control and data collecting system
According to each input neuron is set, the numerical value of the hidden neuron of each hidden layer is calculated successively, and calculate output neuron；
The output neuron represents history wind energy data；
The amendment path sequentially passes through the hidden neuron of each hidden layer since the output neuron of output layer, reversely；
The amendment path is used for the connection weight weight values for adjusting each layer neuron, to build new forward path.
4. wind power plant wind energy Forecasting Methodology as claimed in claim 1, it is characterised in that the first data include the very first time
In the range of have the very first time interval many datas, the second data include the second time range in have the second time between
Every many datas, the 3rd data include the 3rd time range in many datas with the 3rd time interval.
5. wind power plant wind energy Forecasting Methodology as claimed in claim 1, it is characterised in that the first data also include history temperature
At least one in degrees of data, history wind direction data and history humidity data, the 3rd data also include the temperature data, pre- of prediction
The wind direction data of survey and at least one in the humidity data of prediction.
6. wind power plant wind energy Forecasting Methodology as claimed in claim 1, it is characterised in that the first data and the second data are
Historical data from Online Monitoring Control and data collecting system.
7. wind power plant wind energy Forecasting Methodology as claimed in claim 1, it is characterised in that the 3rd data are by using number
It is worth the data of Forecast Model For Weather prediction, wherein, the numerical weather forecast model is used to determine in hub of wind power generator height
Wind speed at degree.
8. a kind of pre- measurement equipment of wind power plant wind energy, it is characterised in that the pre- measurement equipment of wind energy includes：
Acquiring unit, it is used to obtain the first data, the second data and the 3rd data, wherein, the first data include historical wind speed
Data, the second data include history wind energy data corresponding with the first data, and the 3rd data include the air speed data of prediction；
Training unit, it is used for the input variable as the artificial neural network for predicting wind energy data and general using the first data
Second data as the artificial neural network target variable, to train the artificial neural network；
Predicting unit, it is used for by by the 3rd data input to the trained artificial neural network, to be predicted
Wind energy data.
9. the pre- measurement equipment of wind power plant wind energy as claimed in claim 8, it is characterised in that before the artificial neural network is
Reverse transmittance nerve network is presented, it has a feedforward network of the layer of multiple interconnections, the layer includes input layer, multiple hidden
Layer and output layer；
Wherein, input layer includes the input neuron for being used to describe one or more input variable, and hidden layer is comprising multiple hidden
Neuron is hidden, output layer includes the output neuron for being used to describe output variable.
10. the pre- measurement equipment of wind power plant wind energy as claimed in claim 9, it is characterised in that the feedforward backpropagation god
Include forward path and amendment path through network；
The forward path sequentially passes through the hidden neuron of each hidden layer since the input neuron of input layer, reaches
The output neuron of output layer；The forward path is used for the first number monitored according to Online Monitoring Control and data collecting system
According to each input neuron is set, the numerical value of the hidden neuron of each hidden layer is calculated successively, and calculate output neuron；
The output neuron represents history wind energy data；
The amendment path sequentially passes through the hidden neuron of each hidden layer since the output neuron of output layer, reversely；
The amendment path is used for the connection weight weight values for adjusting each layer neuron, to build new forward path.
11. the pre- measurement equipment of wind power plant wind energy as claimed in claim 8, it is characterised in that when the first data include first
Between in the range of have the very first time interval many datas, the second data include the second time range in have the second time
Many datas at interval, the 3rd data include many datas with the 3rd time interval in the 3rd time range.
12. the pre- measurement equipment of wind power plant wind energy as claimed in claim 8, it is characterised in that the first data also include history
At least one in temperature data, history wind direction data and history humidity data, the 3rd data also include prediction temperature data,
The wind direction data of prediction and at least one in the humidity data of prediction.
13. the pre- measurement equipment of wind power plant wind energy as claimed in claim 8, it is characterised in that the first data and the second data
It is the historical data from Online Monitoring Control and data collecting system.
14. the pre- measurement equipment of wind power plant wind energy as claimed in claim 8, wherein, the 3rd data are by using numerical value day
The data of gas forecasting model prediction, wherein, the numerical weather forecast model is used to determine at hub of wind power generator height
Wind speed.
15. a kind of computer-readable recording medium, has program stored therein, it is characterised in that described program perform as claim 1 to
Wind power plant wind energy Forecasting Methodology described in any one of 7.
16. a kind of computer, includes the computer-readable recording medium for the computer program that is stored with, it is characterised in that the computer program is held
Wind power plant wind energy Forecasting Methodology of the row as described in any one of claim 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710296557.8A CN107124003B (en) | 2017-04-28 | 2017-04-28 | Wind power plant wind energy prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710296557.8A CN107124003B (en) | 2017-04-28 | 2017-04-28 | Wind power plant wind energy prediction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107124003A true CN107124003A (en) | 2017-09-01 |
| CN107124003B CN107124003B (en) | 2020-09-15 |
Family
ID=59726442
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710296557.8A Active CN107124003B (en) | 2017-04-28 | 2017-04-28 | Wind power plant wind energy prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107124003B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111967689A (en) * | 2020-09-04 | 2020-11-20 | 上海积成能源科技有限公司 | Model and method for wind power generation prediction by combining multivariate and stepwise linear regression and artificial neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101741085A (en) * | 2009-12-11 | 2010-06-16 | 西北电网有限公司 | Method for forecasting short-term wind-electricity power |
| CN103904682A (en) * | 2014-03-18 | 2014-07-02 | 许继电气股份有限公司 | Power prediction method based on wind-solar hybrid model |
| CN106229972A (en) * | 2016-08-16 | 2016-12-14 | 北京国能日新系统控制技术有限公司 | A kind of integrated based on many meteorological sources and the wind power forecasting method of segmentation modeling |
| CN106529719A (en) * | 2016-11-08 | 2017-03-22 | 中国地质大学（武汉） | Method of predicting wind power of wind speed fusion based on particle swarm optimization algorithm |
-
2017
- 2017-04-28 CN CN201710296557.8A patent/CN107124003B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101741085A (en) * | 2009-12-11 | 2010-06-16 | 西北电网有限公司 | Method for forecasting short-term wind-electricity power |
| CN103904682A (en) * | 2014-03-18 | 2014-07-02 | 许继电气股份有限公司 | Power prediction method based on wind-solar hybrid model |
| CN106229972A (en) * | 2016-08-16 | 2016-12-14 | 北京国能日新系统控制技术有限公司 | A kind of integrated based on many meteorological sources and the wind power forecasting method of segmentation modeling |
| CN106529719A (en) * | 2016-11-08 | 2017-03-22 | 中国地质大学（武汉） | Method of predicting wind power of wind speed fusion based on particle swarm optimization algorithm |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111967689A (en) * | 2020-09-04 | 2020-11-20 | 上海积成能源科技有限公司 | Model and method for wind power generation prediction by combining multivariate and stepwise linear regression and artificial neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107124003B (en) | 2020-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kalogirou et al. | Artificial neural networks for the performance prediction of large solar systems | |
| CN105973594B (en) | A kind of rolling bearing fault Forecasting Methodology based on continuous depth confidence network | |
| Jain et al. | Application of ANN for reservoir inflow prediction and operation | |
| Catalao et al. | An artificial neural network approach for short-term wind power forecasting in Portugal | |
| Gomes et al. | Wind speed and wind power forecasting using statistical models: autoregressive moving average (ARMA) and artificial neural networks (ANN) | |
| CN109444740B (en) | Intelligent fault state monitoring and diagnosing method for wind turbine generator | |
| Tesfaye et al. | Short-term wind power forecasting using artificial neural networks for resource scheduling in microgrids | |
| Ghadi et al. | A novel heuristic method for wind farm power prediction: A case study | |
| Castellani et al. | Wind energy forecast in complex sites with a hybrid neural network and CFD based method | |
| Zhang et al. | Research on AGC performance during wind power ramping based on deep reinforcement learning | |
| Ruslan et al. | Flood modelling using artificial neural network | |
| Brahimi et al. | Prediction of wind speed distribution using artificial neural network: The case of Saudi Arabia | |
| Bhatt et al. | Statistical and ANN based prediction of wind power with uncertainty | |
| CN107124003A (en) | Wind power plant wind energy Forecasting Methodology and equipment | |
| CN112465136A (en) | Hydroelectric generating set degradation prediction method and system | |
| CN111680823A (en) | Wind direction information prediction method and system | |
| CN106529713A (en) | Grey GMDH network combination model-based wind speed prediction method and system | |
| CN115965177A (en) | Improved autoregressive error compensation wind power prediction method based on attention mechanism | |
| Natarajan et al. | Determination of wind farm life consumption in complex terrain using ten-minute SCADA measurements | |
| Vallejo et al. | Genetic algorithm applied to the specialization of neural networks for the forecast of wind and solar generation | |
| Jovanović et al. | Application of аrtificial neural networks for prediction of heating energy consumption in university buildings | |
| Sanayha et al. | Remaining useful life prediction using enhanced convolutional neural network on multivariate time series sensor data | |
| De Giorgi et al. | Short-term wind forecasting using artificial neural networks (ANNs) | |
| Gomes et al. | Comparison of statistical wind speed forecasting models | |
| Fonte et al. | ANN approach to WECS power forecast |
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 |