CN116050072A

CN116050072A - Wind turbine theoretical power curve identification method and device based on random sampling consistency

Info

Publication number: CN116050072A
Application number: CN202211564545.6A
Authority: CN
Inventors: 安鸯; 闫相臣; 钟晓刚; 钱峰
Original assignee: China National Software & Service Co ltd
Current assignee: China National Software & Service Co ltd
Priority date: 2022-12-07
Filing date: 2022-12-07
Publication date: 2023-05-02

Abstract

The invention relates to a wind turbine generator set theoretical power curve identification method and device based on random sampling consistency. The method is based on data driving, firstly, the characteristics related to the theoretical power curve of the wind turbine are extracted from the data of the normal operation of the wind turbine, and then, through data exploration, the corresponding mathematical analysis formula is selected according to the distribution of the data; further, the corresponding relation of wind speed and power of the wind turbine is obtained by self-adaptive fitting of data by using random sampling consistency and combining a nonlinear least square algorithm; and finally, calculating a corresponding theoretical power curve under each wind speed section through the wind speed data sub-buckets, so that the fitting of the theoretical power curve of the wind power generation system is realized from the data driving angle, the application of intelligent operation and maintenance of wind power plant data is completed, and the intelligent transformation in the wind power field can be promoted. The invention does not increase the operation and maintenance cost of the wind field, and has stronger universality and expandability.

Description

Wind turbine theoretical power curve identification method and device based on random sampling consistency

Technical Field

The invention relates to the field of analysis of theoretical power curves of wind turbines, in particular to a method and a device for identifying a theoretical power curve of a wind turbine based on random sampling consistency.

Background

Wind power generation can convert wind energy into electric energy, and is a main means for human to utilize the wind energy as clean energy. At present, the proportion of wind power generation in the field of electric power energy is increased year by year. In a wind power generation system, a wind turbine generator is core equipment and plays an important role in converting wind energy into electric energy. With the increase of service time of the wind turbine, the faults of the wind turbine are increased gradually, and particularly, the faults of major components of the wind turbine, such as impellers, main bearings, gear boxes, generators and the like, not only can the long-time shutdown of the wind turbine be caused, but also the generated energy is influenced, and even safety production accidents, such as wind turbine tower falling, blade breakage and the like, can be caused by serious faults of the major components.

In order to solve the problems, the intelligent and remote wind power production operation mode is realized by actively exploring and establishing a centralized control center for each wind power enterprise in order to further improve the management efficiency, increase the refinement level, increase the operation efficiency, reduce the production cost and increase the economic benefit. With the improvement of industry operation and maintenance level, the gains brought by the traditional fault diagnosis are gradually capped, and the cost reduction and efficiency improvement of wind farm operation are required to further define the sources of losses.

However, due to the technical differences of wind turbines and different geographical environments, the wind speed-theoretical power curve of the wind turbine is difficult to determine an explicit mathematical analysis formula. Furthermore, the fitting of the wind power curve depends on the wind speed-power scatter plot, and the prior art suffers from distortion problems when fitting to the scatter points, resulting in the failure to obtain the desired smooth wind power curve. Even if the variables such as blade angle, temperature, rotation speed and the like in the SCADA (Supervisory Control And Data Acquisition, i.e. monitoring control and data acquisition) system are subjected to working condition filtering in advance, the application range is limited, and the whole working condition cannot be covered. Therefore, reasonably measuring the lost electric quantity of the wind turbine generator set by fitting a theoretical power curve is a key for further improving the operation and maintenance efficiency of the wind farm.

Disclosure of Invention

The invention provides a wind turbine generator theoretical power curve identification method and device based on random sampling consistency, which aims to solve the problems that the existing wind turbine generator theoretical power curve is lost in analytic mode, difficult to fit, large in influence of external factors on fitting results and the like.

According to the invention, the theoretical power curve of the wind turbine generator is identified, and curve fitting is realized on wind speed-power scattered points based on random sampling consistency. The method is based on wind speed and power data acquired by the SCADA system. And eliminating scattered points which seriously do not accord with actual conditions in the original data through a wind speed threshold value and a power threshold value which are known by the wind turbine generator. Then, aiming at the screened wind speed and power data, observing the distribution in a data scatter diagram, respectively selecting a plurality of functions including a quadratic polynomial, a cubic polynomial and a Sigmoid function as analytic functions, applying a nonlinear least square fitting algorithm based on random sampling consistency, iterating for a plurality of times, and calculating to obtain the theoretical power corresponding to the wind speed. And dividing the obtained wind speed-theoretical power data into barrels at equal intervals according to the magnitude of the wind speed value, calculating the average power value in each barrel, and finally obtaining the theoretical power curve of the wind turbine.

The technical scheme adopted for solving the technical problems is as follows:

a wind turbine theoretical power curve identification method based on random sampling consistency comprises the following steps:

selecting two variables of wind speed and active power from an original SCADA data set to form an original data set;

for the original data set, carrying out numerical screening according to a threshold value set by a wind turbine generator to remove data which seriously do not accord with actual conditions;

drawing a wind speed-active power scatter diagram for the screened data set, and observing the data form and distribution of the data set;

selecting a mathematical analysis formula of a wind speed-theoretical power curve, wherein the wind speed is an independent variable, and the power is a dependent variable;

fitting data in a wind speed-active power scatter diagram based on a selected mathematical analysis algorithm by adopting a random sampling consistency algorithm;

determining a definite wind speed-theoretical power relation according to the optimal parameters output by the random sampling consistency algorithm and the selected mathematical analysis formula;

substituting the wind speed into a mathematical analysis formula to obtain a theoretical power value corresponding to each wind speed point;

and grouping the wind speeds according to the grouping result, calculating the average value of the wind speeds and the theoretical power under each group, and taking a curve obtained according to the average value of the wind speeds and the theoretical power as a wind speed-theoretical power curve.

Further, the threshold values include cut-in wind speed, cut-out wind speed, rated power.

Further, the mathematical analysis formula for selecting the wind speed-theoretical power curve is a mathematical analysis formula for selecting a quadratic polynomial, a cubic polynomial and a Sigmoid function as the wind speed-theoretical power curve respectively.

Further, a random sampling consistency algorithm (RANSAC algorithm) is adopted, based on the different mathematical analysis formulas, the data in the wind speed-active power scatter diagram are fitted according to the following steps:

(1) Setting initial parameters, wherein the quantity of the initial parameters is related to a mathematical analysis formula, if a quadratic polynomial comprises three parameters, and if the quadratic polynomial comprises four parameters, a Sigmoid function comprises five parameters;

(2) Randomly dividing the data set into an inner point data set and an outer point data set;

(3) Setting super parameters of a random sampling consistency algorithm, comprising: the threshold value of the number of the inner points, the duty ratio of the number of the inner points, the minimum allowable error and the iteration times;

(4) Selecting an interior point data set as input of a nonlinear least square method, fitting any selected mathematical analysis formula, judging whether the selected data set is an interior point according to whether fitting errors are smaller than minimum allowable errors, and recording a corresponding parameter result as an optimal parameter, wherein the current interior point number is used as an optimal interior point number;

(5) After each iteration, comparing the current parameters with the optimal parameters, and if the number of the interior points of the current new round is larger than the number of the optimal interior points, updating the interior point data set and the optimal parameters;

(6) After each iteration is finished, if the condition (the number of the inner points is larger than the threshold value or the number of the inner points is larger than the threshold value) is met or the upper limit of the iteration times is reached, finishing data fitting, and outputting the optimal parameters and the inner point data set. Otherwise, the step (4) and the step (5) are circularly repeated until the condition is met or the upper limit of the iteration times is reached.

Further, the wind speed sub-barrels are characterized in that the lower limit of wind speed is set to be 0m/s, the upper limit of wind speed is set to be 30m/s, and the interval is set to be 0.5m/s.

A wind turbine theoretical power curve identification device based on random sampling consistency comprises:

the original data set acquisition module is used for selecting two variables of wind speed and active power from an original SCADA data set to form an original data set;

the numerical screening module is used for carrying out numerical screening on the original data set according to a threshold value set by a wind turbine generator factory so as to remove data which seriously do not accord with actual conditions;

the wind speed-active power scatter diagram drawing module is used for drawing a wind speed-active power scatter diagram for the screened data set and observing the data form and distribution of the wind speed-active power scatter diagram;

the mathematical analysis formula selection module is used for selecting a mathematical analysis formula of a wind speed-theoretical power curve, wherein wind speed is an independent variable, and power is a dependent variable;

the data fitting module is used for fitting the data in the wind speed-active power scatter diagram based on the selected mathematical analysis formula by adopting a random sampling consistency algorithm;

the wind speed-theoretical power relation determining module is used for determining a definite wind speed-theoretical power relation according to the optimal parameters output by the random sampling consistency algorithm and the selected mathematical analytic expression;

the theoretical power value acquisition module is used for substituting the wind speed into a mathematical analysis formula to obtain a theoretical power value corresponding to each wind speed point;

the wind speed-theoretical power curve acquisition module is used for grouping wind speeds into barrels according to barrel grouping results, calculating average values of wind speeds and theoretical powers under each group, and taking a curve obtained according to the average values of the wind speeds and the theoretical powers as a wind speed-theoretical power curve.

The beneficial effects of the invention are as follows:

(1) The method is based on data driving, firstly, the characteristics related to the theoretical power curve of the wind turbine are extracted from the data of the normal operation of the wind turbine, and then, through data exploration, the corresponding mathematical analysis formula is selected according to the distribution of the data. The method comprises the steps of obtaining a wind speed-power corresponding relation of a wind turbine by using random sampling consistency and combining a nonlinear least square algorithm to perform self-adaptive fitting on data, and finally obtaining a corresponding theoretical power curve under each wind speed section through wind speed data sub-buckets, so that fitting of the theoretical power curve of a wind power generation system is realized from a data driving angle, application of intelligent operation and maintenance of wind power plant data is completed, and intelligent transformation is promoted in the wind power field.

(2) The method is based on the SCADA data of the running wind turbine, is easy to be practically applied, has no special requirement on the wind turbine, and the required characteristic sequence is acquired in real time by the existing sensors at present, so that the operation and maintenance cost of a wind field is not increased. Meanwhile, the system has strong universality and expandability, can be directly applied to wind turbine sets of different brands, and has high theoretical research value and practical application value for intelligent operation and maintenance technology research of various wind fields.

(3) Aiming at the problem of operation and maintenance problems commonly existing in the current wind power field, the method creatively realizes the work of identifying the theoretical power curve of the wind turbine by providing a new artificial intelligence technology through a machine learning algorithm model, and has positive promotion effect on the wind turbine efficiency optimization and operation and maintenance management field based on the artificial intelligence technology.

(4) The novel wind turbine theoretical power curve identification method based on the random sampling consistency algorithm can effectively solve the problem that part of data cannot reflect the actual wind speed-power corresponding relation due to artificial regulation and control or communication faults and the like, and the algorithm model can effectively eliminate the influence of abnormal data, avoid false identification, make up the defects in the traditional wind power operation and maintenance process and realize value mining and analysis of the wind turbine operation data.

(5) Due to the flexibility of algorithm design and the universality and expandability of the algorithm model, the method can try to migrate to other technical problems to be solved of the intelligent operation and maintenance system of the wind power plant, and has strong expandability and mobility.

Drawings

FIG. 1 is a flow chart of steps of a method for identifying theoretical power curves of wind turbines based on random sampling consistency.

FIG. 2 is a visual result of a wind speed-active power scatter plot;

FIG. 3 is a graph of the result of fitting a wind speed-theoretical power curve.

Detailed Description

For a modified understanding of the present invention, the technical solution of the present invention will be further described below with reference to practical use cases and accompanying drawings.

The method is verified in technical effectiveness based on SCADA operation data of 80 wind turbines in the whole farm from 8 months in 2020 to 12 months in 2020 in Bancheng wind farm in Xinjiang. The data set used in this case was sampled at 10 minute intervals with a data time span of 5 months, and each wind turbine group included 22032 pieces of data. Table 1 shows the actual data information, including time, wind speed, active power, equipment number, etc., used in part to fit the theoretical power curve, with each row representing a sample point. Table 2 shows detailed information about the dataset.

TABLE 1 theoretical Power Curve identifies partial actual data in the dataset

Time	Wind speed	Power of	Device numbering
				...	...	...	...
2020/9/2 0:00	6.72	841.57	#1-D1
				2020/9/2 0:10	6.02	642.76	#1-D1
2020/9/2 0:20	6.22	669.67	#1-D1
				2020/9/2 0:30	5.68	523.98	#1-D1
2020/9/2 0:40	6.01	535.14	#1-D1
				...	...	...	...

TABLE 2 statistical information of data variables in data set

Variable name	Meaning of variable	Variable type
			Time	Time of current SCADA data recording	Datetime
Wind speed	Wind speed of current wind turbine generator	Float
			Power of	Active power of current wind turbine generator	Float
Device numbering	Numbering of current wind turbine generator	String

The process flow of the method for identifying the theoretical power curve of the wind turbine generator set based on random sampling consistency is shown in fig. 1, and the detailed implementation scheme is as follows:

1) Selecting device number d from raw SCADA dataset _i (i=1, 2,3,..80), wind speed V _{Actual practice is that of} Active power P _{Active power} An original dataset is formed. In the invention, active power refers to the actual power value of the wind turbine, is also a key index of the wind turbine, is positively correlated with the actual wind speed, and is generally lower than the theoretical power under the condition of the same wind speed; the theoretical power refers to the power value of the wind turbine generator under the ideal condition and the normal working condition, and is generally higher than the active power under the condition of the same wind speed.

2) And carrying out numerical screening on the original data set according to a threshold value set by a factory of the wind turbine generator to remove data which seriously do not accord with actual conditions.

The threshold value mainly comprises the following contents: cut-in wind speed V _cut-in =2.5 m/s, cut-out wind speed V _cut-out =25m/s, rated wind speed V _rated =13 m/s, rated power P _rated ＝1500KW。

The specific screening mode is as follows:

for any sample point, when V _{Actual practice is that of} <V _cut-in And P is _{Active power} >P _rated Deleting the data point when the data point is deleted;

for any sample point, when V _{Actual practice is that of} >V _cut-out Deleting the data point when the data point is deleted;

for any sample point, when V _{Actual practice is that of} >V _cut-in ，V _{Actual practice is that of} <V _rated And P is _{Active power} >P _rated When that data point is deleted.

3) For the screened dataset, the wind speed V is plotted _{Actual practice is that of} Active power P _{Active power} A scatter plot, the data morphology and distribution was observed, as shown in fig. 2.

4) At wind speed V _{Actual practice is that of} As an independent variable, theoretical power P _{Theoretical power} As a dependent variable, a cubic polynomial was selected as a mathematical analysis of the wind speed-theoretical power curve, as shown in detail below, wherein A, B, C, D is the parameter to be fitted:

5) According to the equipment number d _i The original data sets are grouped to obtain i groups of sub data sets. For each sub-data set, a random sampling consistency algorithm is applied, and based on the mathematical analysis formula, the wind speed and the active power are fitted.

The specific steps for fitting the data are as follows:

(1) Setting initial parameters A=0, B=0, C=0 and D=0, and storing the initial parameters in the form of a one-dimensional array;

(2) Setting super parameters of a random sampling consistency algorithm, comprising: number of inliers threshold num _in-lier Upper ratio of the number of inner points _in-lier Minimum allowable error sigma, upper limit item of iteration times, and optimal number of inner points num _{up_threshold} Optimum parameter para _best ；

(3) Randomly dividing each sub-data set into two parts of an inner point data set and an outer point data set;

(4) The fitting algorithm selects nonlinear least square, takes an interior point data set as input of the nonlinear least square method, fits a cubic polynomial, judges whether a current sample point is an interior point according to whether fitting error is smaller than minimum allowable error sigma, and records a corresponding parameter result para as an optimal parameter para _best The current number of interior points is taken as the optimal number of interior points num _{up_threshold} ；

(5) Ending each iteration, if the number of the interior points of the current new round is larger than the number num of the optimal interior points _{up_threshold} Then the interior point data is updatedAnd uses the current parameter para to update to the optimal parameter para _best ；

(6) Ending each iteration, if the number num of the interior points is larger than the number num of the optimal interior points _{up_threshold} Or the number of inliers is greater than the upper limit ratio _in-lier Or the iteration number reaches the upper limit item of the iteration number, judging the random sampling consistency to finish data fitting, and outputting the optimal parameter para _best An interior point dataset. Otherwise, the step (4) and the step (5) are circularly repeated until one of the above judgment conditions is satisfied.

6) After fitting, outputting optimal parameters para according to a random sampling consistency algorithm _best And determining a definite wind speed V by a cubic polynomial function _{Actual practice is that of} And theoretical power P _{Theoretical power} Is a relationship of (3).

7) Wind velocity V _{Actual practice is that of} The theoretical power value corresponding to each wind speed point is obtained after the wind speed point is brought into the three-time polynomial analysis type;

8) For the raw dataset, for wind speed V _{Actual practice is that of} The lower limit of the barrel is set to be 0m/s, the upper limit of the barrel is set to be 30m/s, the interval is 0.5m/s, and a barrel dividing result corresponding to each wind speed value, namely the belonging group is obtained;

9) Grouping the original data sets according to the barrel grouping result, and calculating the wind speed V under each group _{Actual practice is that of} Theoretical power P _{Theoretical power} A two-dimensional array is obtained, and a curve (the average value is connected) obtained according to the calculation result is taken as a wind speed-theoretical power curve, as shown in fig. 3.

The above embodiment selects the cubic polynomial as the mathematical analysis of the wind speed-theoretical power curve. In other embodiments of the present invention, a quadratic polynomial, sigmoid function, etc. may be selected as the mathematical analysis formula of the wind speed-theoretical power curve.

When a quadratic polynomial is selected as a mathematical analysis formula of a wind speed-theoretical power curve, theoretical power P _{Theoretical power} The calculation formula of (2) is as follows:

in the above formula, V _{Actual practice is that of} Represents the actual wind speed, P _{Theoretical power} And A, B, C represents the theoretical power corresponding to the actual wind speed, and is a parameter to be fitted.

When the Sigmoid function is selected as a mathematical analysis formula of the wind speed-theoretical power curve, theoretical power P _{Theoretical power} The calculation formula of (2) is as follows:

in the above formula, V _{Actual practice is that of} Represents the actual wind speed, P _{Theoretical power} And e represents natural logarithm, and A, B, C, D is a parameter to be fitted.

Wherein the specific implementation of each module is referred to the previous description of the method of the present invention.

Based on the same inventive concept, another embodiment of the present invention provides a computer device (computer, server, smart phone, etc.) comprising a memory storing a computer program configured to be executed by the processor, and a processor, the computer program comprising instructions for performing the steps in the inventive method.

Based on the same inventive concept, another embodiment of the present invention provides a computer readable storage medium (e.g., ROM/RAM, magnetic disk, optical disk) storing a computer program which, when executed by a computer, implements the steps of the inventive method.

The above-disclosed embodiments of the invention are intended to facilitate an understanding of the principles of the invention and to be practiced otherwise than as specifically described herein. Those skilled in the art will appreciate that various substitutions, changes and modifications are possible without departing from the spirit and scope of the invention. The invention should not be limited to what has been disclosed in the examples of the specification, but rather by the scope of the invention as defined in the claims.

Claims

1. A wind turbine generator set theoretical power curve identification method based on random sampling consistency is characterized by comprising the following steps:

2. The method of claim 1, wherein the threshold value comprises a cut-in wind speed V _cut-in Cut-out wind speed V _cut-out Rated wind speed V _rated Rated power P _rated 。

3. The method of claim 2, wherein the numerical screening according to the threshold value set by the wind turbine generator comprises:

for any sample point, when the actual wind speed V _{Actual practice is that of} <V _cut-in And P is _{Active power} >P _rated Deleting the data point when the data point is deleted;

4. The method according to claim 1, wherein the mathematical analysis formula for selecting the wind speed-theoretical power curve is a mathematical analysis formula for selecting a quadratic polynomial, a cubic polynomial, and a Sigmoid function as the wind speed-theoretical power curve, respectively.

5. The method of claim 1, wherein fitting the data in the wind speed-active power scatter plot based on the selected mathematical resolution using a random sampling consistency algorithm comprises:

(1) Setting initial parameters, wherein the number of the initial parameters is related to a mathematical analysis formula;

(6) After each round of iteration is finished, if the condition is met or the upper limit of the iteration times is reached, finishing data fitting, and outputting optimal parameters and an interior point data set; otherwise, the step (4) and the step (5) are circularly repeated until the condition is met or the upper limit of the iteration times is reached.

6. The method of claim 5, wherein the satisfaction of step (6) is that the number of interior points is greater than a threshold or that the number of interior points is greater than a threshold.

7. The method according to claim 1, wherein the wind speed is divided into barrels, the lower limit of the wind speed is set to be 0m/s, the upper limit of the wind speed is set to be 30m/s, and the interval is set to be 0.5m/s.

8. Wind turbine generator system theoretical power curve recognition device based on random sampling uniformity, characterized by comprising:

9. A computer device comprising a memory and a processor, the memory storing a computer program configured to be executed by the processor, the computer program comprising instructions for performing the method of any of claims 1-7.

10. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program which, when executed by a computer, implements the method of any one of claims 1-7.