The content of the invention
The technical problems to be solved by the invention be optimize Wind turbines power there is provided a kind of Wind turbines power optimization
Method and system.
The technical scheme that the present invention solves above-mentioned technical problem is as follows:A kind of method of Wind turbines power optimization, including
Following steps,
Step 1, running of wind generating set data are gathered;
Step 2, the data collected are handled, that is, chooses the data of second level temporal resolution, reject failure
Data;
Step 3, generated output model P (t) and vibration amplitude model A (t) are built according to the data after processing;
Step 4, the minimum generated output model of Select Error mean-square value and vibration amplitude model are optimal power generation power mould
Type and optimal vibration amplitude model;
Step 5, it is excellent according to optimal power generation power module and optimal vibration amplitude model buildings power optimization function and vibration
Change function.
On the basis of above-mentioned technical proposal, the present invention can also do following improvement.
Further, the data gathered in step 1 include the following operational parameter data collection of Wind turbines, wind speed v (t),
Go off course y (t), wind direction d (t), becomes vane angle p (t), temperature T (t).
Further, the fault data in step 2 is included because of the data and obvious exception that Wind turbines fault collection is arrived
Data.
Further, P (t)=f (v (t), y (t), d (t), p (t), T (t))+error, A (t)=g in step 3 (v (t),
Y (t), d (t), p (t), T (t))+error, wherein f (), g () is the regression class algorithm of data mining, and t is the fortune of Wind turbines
Row time, error is the error amount of the generated output model and vibration amplitude model.
Further, also include determining the optimal power generation power module and the optimal vibration amplitude model in step 4
Corresponding wind speed v (t), driftage y (t), wind direction d (t), the parameter value for becoming vane angle p (t) and temperature T (t).
Further, treated data are randomly divided into two parts in step 4, wherein 70% is used for generated output mould
The training data of type and vibration amplitude model training, 30% is used for the predictive data set of MSE error mean squares in addition.
Further, before step 5 is carried out, resampling can be used to the non-equilibrium data of training data part, makes institute
State parameter wind speed v (t), driftage y (t), wind direction d (t), the change vane angle in generated output model P (t) and vibration amplitude model A (t)
P (t) and temperature T (t) each interval quantity in five dimension state spaces represents interval censored data quantity no less than preset value n1, n1
Preset value.
Further, in step 5 before majorized function is built, it is necessary to by parameter wind speed v (t), driftage y (t), wind direction d
(t) vane angle p (t) and the controlled optimizations of temperature T (t), are become.
Further, power optimization function is Tp=max { P (t) } in step 5, and Tp refers to the performance number after power optimization, shaken
Dynamic majorized function is TA=max { A (t) }, and wherein TA refers to the vibration after optimization, and C=max (Tp | TA<=C), C represents vibration
Limit value, i.e., vibration can with acceptable conditionses under power it is optimal.
Another technical scheme that the present invention solves above-mentioned technical problem is as follows:A kind of Wind turbines power optimization is
System, including data acquisition module, data processing module, model construction module, optimal models build module and majorized function is taken
Model block;
The data acquisition module is used to gather running of wind generating set data;
The data processing module is used to be handled the data collected, that is, chooses second level temporal resolution
Data, reject fault data;
The model construction module is used to build generated output model P (t) and vibration amplitude mould according to the data after processing
Type A (t);
The optimal models, which builds module, is used for the minimum generated output model and vibration amplitude mould of Select Error mean-square value
Type is optimal power generation power module and optimal vibration amplitude model;
The majorized function, which builds module, to be used for according to optimal power generation power module and optimal vibration amplitude model buildings work(
Rate majorized function and vibration majorized function.
The beneficial effects of the invention are as follows:From the power optimization of the angle research Wind turbines of Controlling model, examine simultaneously
Consider the optimization of driving-chain and tower vibration.
Embodiment
The principle and feature of the present invention are described below in conjunction with accompanying drawing, the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the present invention.
Consider to pass simultaneously in the power optimization of angle research Wind turbines of the invention from Controlling model, optimization aim
Dynamic chain and the optimization of tower vibration.The present invention has inquired into data digging method to optimize the basic ideas frame of blower fan production capacity and vibration
Frame, and realized with an Engineering Projects by pitch control power optimization.
As shown in figure 1, a kind of method of Wind turbines power optimization, comprises the following steps:
Step 1, gather running of wind generating set data, using the running of wind generating set time as variable, collection Wind turbines as
Lower operational parameter data collection, wind speed v (t), driftage y (t), wind direction d (t), becomes vane angle p (t), temperature T (t);
Step 2, data processing, selects the data of right times resolution ratio, according to engineering experience, general selection second rank
Data, reject fault data, including the data and substantially abnormal data arrived by Wind turbines fault collection.The data processing
Process be conducive to improve data the degree of accuracy, it is ensured that the accuracy subsequently calculated.
Step 3, generated output model P (t) and vibration amplitude model A (t), wherein P (t) are built according to the data after processing
=f (v (t), y (t), d (t), p (t), T (t))+error, A (t)=g (v (t), y (t), d (t), p (t), T (t))+error
Wherein f (), g () are the regression class algorithm of data mining, and the effect above can be realized in data mining algorithm has neural
Network, support vector machine, k nearest neighbor and random forest method etc., error are model
Error amount.
Step 4, the minimum generated output model of Select Error mean-square value and vibration amplitude model are optimal power generation power mould
Type and optimal vibration amplitude model, and determine the optimal power generation power module and the corresponding wind speed v of optimal vibration amplitude model
(t), driftage y (t), wind direction d (t), the parameter value for becoming vane angle p (t) and temperature T (t).In this process, will be treated
Data are randomly divided into two parts, wherein 70% is used for the training data of model training, 30% is used for MSE (i.e. error mean square) in addition
Predictive data set, wherein MSE is to prevent overfitting using predictive data set.
In order to reduce calculated load, speed-up computation uses resampling to the non-equilibrium data of training data part, made above-mentioned
Parameter wind speed v (t), driftage y (t), wind direction d (t), change vane angle p in generated output model P (t) and vibration amplitude model A (t)
(t) in n, (correspondence number of parameters, i.e. n is no less than preset value n1 5) to tie up the quantity in each interval in state space with temperature T (t)
(n1 depends on total training sample amount, and under the conditions of the determination of interval number, training sample is bigger, and n1 is bigger, the number after resampling
According to amount no more than former training data), even if quantity of the parameter in some n ties up state space is no less than n1.So both retain
Original training data can cover all possible point, and the volume of data is reduced again.
Step 5, it is excellent according to optimal power generation power module and optimal vibration amplitude model buildings power optimization function and vibration
Change function;Before majorized function is built, parameter should use controlled optimization, have certain true thing with the result for ensuring optimization
Manage meaning;So that parameter is becomes the power optimization of vane angle as an example, optimization aim can be decomposed into, Tp=max { P (t) }, and Tp refers to power
Performance number after optimization, should controlled optimization, p_min within the specific limits wherein becoming vane angle change<=p (t)<=p_max, its model
Enclose for currency+- someValue (10 ° of acquiescence).
It is also possible to build the majorized function of vibration optimization, TA=max { A (t) } should be certain wherein becoming vane angle change
The controlled optimization of scope, p_min<=p (t)<=p_max, in the range of currency+- someValue (10 ° of acquiescence).
Optimize the majorized function of power and vibration simultaneously, variable element is to become propeller angle, and C=max (Tp | TA<=C), C generations
Table vibration limit value, i.e., vibration can with acceptable conditionses under power it is optimal.
As shown in Figure 2, a kind of system of Wind turbines power optimization, including data acquisition module, data processing module, mould
Type builds module, optimal models structure module, training data resampling module and majorized function and builds module;
Data acquisition module is used for using the time as variable, gathers the following operational parameter data collection of Wind turbines, wind speed v
(t), driftage y (t), wind direction d (t), become vane angle p (t), temperature T (t);
Data processing module is used for the data for selecting right times resolution ratio, general to choose second rank according to engineering experience
Data, reject fault data, including the data and substantially abnormal data arrived by Wind turbines fault collection.At the data
The process of reason is conducive to improving the degree of accuracy of data, it is ensured that the accuracy subsequently calculated.
Model construction module is used to build generated output model P (t) and vibration amplitude model A according to the data after processing
(t), wherein P (t)=f (v (t), y (t), d (t), p (t), T (t))+error, A (t)=g (v (t), y (t), d (t), p (t), T
(t))+error, wherein f (), g () are the regression class algorithm of data mining, can realize the effect above in data mining algorithm
Have neural network, support vector machine, k nearest neighbor and random forest method etc.,
Error is the error amount of model.
Optimal models builds the module generated output model minimum for Select Error mean-square value and vibration amplitude model is
Optimal power generation power module and optimal vibration amplitude model, and determine the optimal power generation power module and optimal vibration amplitude model
Corresponding wind speed v (t), driftage y (t), wind direction d (t), the parameter value for becoming vane angle p (t) and temperature T (t).In this process, will
Treated data are randomly divided into two parts, wherein 70% is used for the training data of model training, 30% is used for MSE in addition
The predictive data set of (i.e. error mean square), wherein MSE is to prevent overfitting using predictive data set.
Training data resampling module is used to use resampling to the non-equilibrium data of training data part, in order to reduce meter
Calculate load, speed-up computation uses resampling to the non-equilibrium data of training data part, make above-mentioned generated output model P (t) and
Parameter wind speed v (t), driftage y (t), wind direction d (t), change vane angle p (t) and temperature T (t) in vibration amplitude model A (t) is (right in n
Answering number of parameters, i.e. n, no less than preset value n1, (n1 depends on total training sample for each interval quantity in 5) dimension state space
This amount, under the conditions of the determination of interval number, training sample is bigger, and n1 is bigger, and the data volume after resampling is trained no more than former
Data), even if quantity of the parameter in some n ties up state space is no less than n1.So both having remained original training data can
All possible point is covered, the volume of data is reduced again.
Majorized function, which builds module, to be used to build majorized function, and before majorized function is built, parameter should use controlled excellent
Change, there is certain actual physical meaning with the result for ensuring optimization;So that parameter is becomes the power optimization of vane angle as an example, optimize mesh
Mark can be decomposed into,
Tp=max { P (t) }, Tp refer to the performance number after power optimization, should be controlled within the specific limits wherein becoming vane angle change
Optimization, p_min<=p (t)<=p_max, in the range of currency+- someValue (10 ° of acquiescence).
It is also possible to build the majorized function of vibration optimization, TA=max { A (t) } should be certain wherein becoming vane angle change
The controlled optimization of scope, p_min<=p (t)<=p_max, in the range of currency+- someValue (10 ° of acquiescence).
Optimize the majorized function of power and vibration simultaneously, variable element is to become propeller angle, and max (Tp | TA<=C), C is represented
Vibration limit value, i.e., vibration can with acceptable conditionses under power it is optimal.
Existing GW1500 machine type data, we will set up simple power optimization function, and variable element is change oar
Angle.
A total of 45 groups of data, in actual analysis, we only study wherein 14 groups variables:WROT.PtAngVal.Bl1、
WROT.PtAngVal.Bl2、WROT.PtAngVal.Bl3、WMAN.State、DateTime、WTUR.PwrAt.InstMag.f、
WNAC.ExlTmp.instMag.f、WNAC.TopBoxTmp、WNAC.IntTmp.instMag.f、WROT.PtCptTmp.Bl1、
WROT.PtCptTmp.Bl2、WROT.PtCptTmp.Bl3、WROT.PtCnvTmp.Bl1、WROT.PtCnvTmp.Bl2、
WROT.PtCnvTmp.Bl3。
Data scrubbing is carried out first.Power curve, change oar curve and speed curves after cleaning is respectively such as Fig. 3, Fig. 4 and figure
Shown in 5.
Wind speed ' NacWS', driftage ' YawPos', wind direction ' Wdir' becomes in oar ' PtAngBl3', temperature ' ExlTmp'
Data are normalized, and ' PwrAct' data are also normalized, using neutral net (settings of attention parameters), built
Formwork erection type, training set data is random 70% initial data, and such as Fig. 6 is error time sequence chart.As can be seen, error time
Sequence is at most not over 15%, and overall MSE is 0.0001, and model can describe the relation of each variable and power very well.
To become vane angle ' PtAngBl3' as variable element, processing is optimized to power, optimum results are obtained, change vane angle
Degree optimization uses controlled optimization, you can the rule for fixing value changes is come for fixed percentage change or currency by currency+-
Adjustment.It is below 95% confidecne curve for becoming oar curve according to side, to control the example of optimization.If Fig. 7 is for standardization
Become in oar curve confidential interval figure, Fig. 7 the bound that the curve 1 and curve 3 marked is confidential interval, curve 2 represents P50, Fig. 8
For the change propeller angle figure before and after optimization, the change propeller angle in Fig. 8 before the representing optimized of curve 1, the change vane angle after the representing optimized of curve 2
Degree, Fig. 9 subtracts the power diagram being not optimised for power after optimization.
The result after being optimized is calculated, result before and after optimization is analyzed, as a result shown, the productivity ratio after optimization is former
To improve 0.6%, the result become before and after the optimization of propeller angle is compared with meeting actual physical change rule.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc. should be included in the scope of the protection.