Disclosure of Invention
The purpose of the invention is as follows: the invention provides a wind turbine power characteristic evaluation method based on a BP (back propagation) neural network, which aims to solve the defects of the conventional method.
The technical scheme is as follows: the invention is realized by the following technical scheme:
a wind turbine generator power characteristic evaluation method based on a BP neural network is characterized by comprising the following steps: the method comprises the following steps:
step 1: site evaluation is carried out on each unit in the wind power plant according to the IEC61400-12-1 standard, and one unit is randomly selected as a reference unit for the wind power plant in the general terrain; for a wind power plant with complex terrain, calculating the Roughness Index (RIX) of each unit, and selecting a unit close to the average value of all RIXs as a reference unit;
and 2, step: carrying out power characteristic test on the reference unit according to IEC61400-12-1 standard, and meanwhile, installing a voltage transformer and a current transformer on the low-voltage side of the box transformer of the rest units to obtain the net power output value of each unit;
and 3, step 3: synchronously recording the data of the wind measuring tower, the data of the power characteristic test wind measuring rod, the SCADA data of each unit and the power output data of each unit, and establishing an evaluation database. The wind speed of an engine room in the evaluation database is acquired in an SCADA system of the unit, the free wind speed is acquired in a wind measuring tower and a power characteristic test wind measuring rod in a wind power plant, the database stores data of not less than 180h of continuous measurement time, the data acquisition frequency is within 7 s-10 min, and the wind speed range and the wind condition in a certain range can be covered;
and 4, step 4: evaluating the time synchronization of the data such as the cabin wind speed, the free wind speed, the power and the like in the database, screening and checking the data in the database, and unifying the frequency for 10min, wherein the free wind speed is the wind speed measured by the anemometer tower and the anemometer rod;
and 5: performing data regression, determining the air density of the tested wind power plant according to the wind measuring tower data, wherein the average air density measured during the effective data acquisition period of the tested wind power plant is 1.225 +/-0.05 kg/m 3 In range, there is no need to normalize the air density to the actual average air density; if not, converting according to the related content of IEC 61400-12-1;
step 6: and correcting the cabin wind speed measured by wind measuring equipment at the tail part of the wind turbine generator cabin by utilizing the nonlinear fitting capacity of the BP neural network, so as to obtain the free wind speed, and thus, evaluating the power characteristics of the wind turbine generator. The method adopts a three-layer BP neural network, wherein an input layer is the wind speed of a wind measuring tower and the wind speed of an engine room, an output layer is the free wind speed, reference unit data is selected as training data, and the mapping relation between the free wind speed and the wind speed of the engine room is established;
and 7: taking the cabin wind speed and anemometer tower data of each unit in the wind power plant as input values of the trained BP neural network to obtain the free wind speed of each unit;
and 8: and after the wind speed is corrected, the power characteristic test of the wind turbine is carried out according to the power characteristic test standard requirement of the IEC61400-12-1 wind turbine, a power curve and a power coefficient are drawn, the annual energy production is extrapolated, and finally an evaluation report is issued.
When the method is applied, at least one unit in the wind power plant carries out power characteristic test (the unit becomes a reference unit) according to the IEC61400-12-1 standard, and at least one wind measuring tower is installed in the wind power plant.
The method comprises the following steps:
the first step is as follows: the method comprises the following steps of performing site evaluation on each unit in the wind power plant, wherein in an evaluation test site, the plane gradient, namely the included angle between a plane passing through a tower foundation of the wind power unit and a sector formed by taking the position of the wind power unit as a vertex and the terrain change on a ground plane meet the requirements of IEC61400-12-1 standard on the terrain of the test site, and the specific requirements of the terrain change around the test site are given in Table 1, wherein L is the distance between the wind power generator unit and a meteorological mast, and D is the diameter of a wind wheel of the wind power generator unit;
table 1 test field requirements: terrain variation
(1) For a wind power plant on a general terrain, randomly selecting one unit as a reference unit;
(2) For a wind power plant with complex terrain, calculating the Roughness Index (RIX) of each unit, and selecting a unit close to the average value of all RIXs as a reference unit;
the second step: carrying out power characteristic test on the reference unit according to IEC61400-12-1 standard, and meanwhile, installing a voltage transformer and a current transformer on the low-voltage side of the box transformer of the rest units to obtain the net power output value of each unit;
the third step: synchronously recording anemometer tower data, power characteristic test anemometer pole data, SCADA data of each unit and power output data of each unit, establishing an evaluation database, wherein cabin wind speed in the evaluation database is acquired in an SCADA system of the unit, free wind speed is acquired in the anemometer tower and the power characteristic test anemometer pole in a wind power plant, the database stores data of continuous measurement time not less than 180h, the data acquisition frequency is within 7 s-10 min, and the anemometer tower, the SCADA system and the SCADA system can cover a certain range of wind speed and wind condition conditions;
the fourth step: the method comprises the steps of (1) checking and screening data, performing range checking and trend checking according to GB/T18710-2002 and by combining with the actual situation of a wind power plant, and referring to tables 2 and 3, and then performing data elimination and correction to ensure that the data can truly and objectively reflect the operation situation of a wind turbine generator, wherein the data elimination is to eliminate the data of the wind turbine generator which does not work or a test system fails;
TABLE 2 reasonable Range reference values for the main parameters
TABLE 3 reasonable trend of Change reference values of the main parameters
Synchronizing the time of the data such as the cabin wind speed, the free wind speed (measured by a wind measuring tower and a wind measuring rod), the power and the like in the evaluation database, wherein the unified frequency is 10min;
the fifth step: performing data regression, determining the air density of the tested wind power plant according to the wind measuring tower data, wherein the average air density measured during the effective data acquisition period of the tested wind power plant is 1.225 +/-0.05 kg/m 3 In the range, air is not required to be sealedNormalizing to actual average air density; if not, converting according to the related content of IEC 61400-12-1;
the air density can be derived from air temperature and air pressure measurements according to equation (1):
wherein: rho 10min An average air density of 10min; t is 10min The measured average absolute air temperature is 10min; b is 10min The measured average air pressure is 10min; r is a gas constant 287.05J/(kg.K);
for the wind turbine generator with automatic power control, the wind speed can be converted to the standard atmospheric pressure by the formula (2):
wherein, V n The wind speed value is converted; v 10min The measured average wind speed value is 10min; rho 10min The obtained average air density for 10min; rho 0 Is a standard air density of 1.225kg/m 3 ;
For a wind turbine with stall control, constant pitch and speed, the measured power output data can be converted using equation (3):
wherein: p n The wind speed value is converted; p 10min The measured average wind speed value is 10min; rho 10min To obtain an average air density of 10min; rho 0 Is a standard air density of 1.225kg/m 3 ;
And a sixth step: and correcting the cabin wind speed measured by wind measuring equipment at the tail part of the wind turbine generator cabin by utilizing the nonlinear fitting capacity of the BP neural network, so as to obtain the free wind speed, and thus, evaluating the power characteristics of the wind turbine generator. The method adopts three layers of BP neural networks as shown in figure 1, wherein the input layer is the wind speed of a wind measuring tower and the wind speed of an engine room, and the output layer is the free wind speed. Selecting reference unit data as training data, and establishing a mapping relation between the free wind speed and the cabin wind speed;
the seventh step: taking the cabin wind speed and anemometer tower data of each unit in the wind power plant as input values of the trained BP neural network to obtain the free wind speed of each unit;
eighth step: and drawing a power curve, after finishing data correction, sorting the selected test data according to a bin method, wherein the selected data group should cover a wind speed range from 1m/s lower than a cut-in wind speed to 1.5 times of the wind speed when the wind turbine generator outputs 85% rated power. The wind speed range should be continuously divided into 0.5m/sbin, with a central value that is an integer multiple of 0.5 m/s. Drawing a power curve by using the power value corresponding to each normalized wind speed bin:
wherein: v i The calculated average wind speed value of the ith bin is obtained; v n,i,j The measured wind speed value of the j data group of the ith bin; n is a radical of i The number of data in the 10min data set for the ith bin; p i The average power value of the ith bin after the conversion is obtained; p n,i,j The measured power value of the j data group of the ith bin;
the ninth step: and (3) drawing a power coefficient curve, wherein the power coefficient can be obtained by calculating the formula (6) according to the measured power curve:
wherein: c p,i Is the power coefficient in bin; v i Average wind speed in bini obtained for the conversion; p i For the resulting power output in the bin; a is the swept area of the wind wheel of the wind turbine set; rho 0 Is the standard air density;
the tenth step: the annual energy generation is calculated by using the power curve obtained by measurement to calculate estimated values of different reference wind speed frequency distributions, and the reference wind speed frequency distribution can be performed by using a Rayleigh distribution which is equivalent to a Weibull distribution with a shape coefficient of 2. The Annual Energy Production (AEP) for an annual average wind speed of 4,5,6,7,8,9,10,l lm/s can be calculated according to the formula (7):
wherein: AEP is annual energy production; n is a radical of h The hour number in one year is about 8760; n is the number of bins; v i The converted average wind speed value in the ith bin is obtained; p is i The average power value of the ith bin after the conversion is obtained;
the function of the rayleigh distribution is:
wherein: f (V) is a Rayleigh distribution function of wind speed; v ave The annual average wind speed value at the central height of the hub of the wind turbine is obtained; v is a wind speed value, and V is set i-1 =V i -0.5m/s and P i-1 Starting superposition when the power is not less than 0.0 kW;
the eleventh step: and (6) generating a report.
In the tenth step, the annual energy production must be calculated on the one hand as "measurement of annual energy production" and on the other hand as extrapolation of annual energy production, and if the measurement does not include the cut-out wind speed value, extrapolation is required to obtain the annual energy production extrapolated from the measured maximum wind speed value to the cut-out wind speed, and the extrapolation of annual energy production partially obtains that all power values of all wind speeds below the lowest wind speed of the tested power curve are assumed to be 0, and all power values in the wind speed range between the highest wind speed on the measured power curve and the cut-out wind speed are assumed to be constant values, and the constant power value used for extrapolation should be the power value of the highest wind speed bin in the measured power curve.
The advantages and effects are as follows:
the invention provides a wind turbine power characteristic evaluation method based on a BP neural network, which only needs to test a reference unit according to IEC61400-12-1 standard and then evaluates the power of all units in the whole field according to test data and SCADA data, so that a large amount of money and time can be saved. Based on the BP neural network, the nonlinear relation from the free wind speed to the cabin wind speed can be realized, and the evaluation result is accurate and reliable;
the method is characterized in that:
and correcting the cabin wind speed measured by wind measuring equipment at the tail part of the wind turbine generator cabin by utilizing the nonlinear fitting capacity of the BP neural network, so as to obtain the free wind speed, and thus, evaluating the power characteristics of the wind turbine generator. The method is suitable for power characteristic evaluation of a single horizontal axis wind generating set connected with or not connected with a grid, and a mapping relation between the free wind speed and the wind speed of an engine room is established by utilizing a BP neural network. When the method is applied, at least one unit in the wind power plant carries out power characteristic test (the unit becomes a reference unit) according to the IEC61400-12-1 standard, and at least one wind measuring tower is installed in the wind power plant.
The method is suitable for power characteristic evaluation of multiple units in the wind power plant. The wind speed of an engine room in a database used for evaluation is acquired from an SCADA system of the unit, the free wind speed is acquired from a wind measuring tower in a wind power plant, the database stores data of not less than 180h of continuous measuring time, the data acquisition frequency is within 7 s-10 min, and the wind speed range and the wind condition within a certain range can be covered. The net power of the wind turbine generator can be obtained by installing a voltage transformer and a current transformer on the low-voltage side of the box transformer of the wind turbine generator. The method adopts three layers of BP neural networks, wherein the input layer is the wind speed of a wind measuring tower and the wind speed of an engine room, and the output layer is the free wind speed. SelectingAnd taking the reference unit data as training data. And evaluating the time synchronization of the data such as the cabin wind speed, the free wind speed (wind speed measured by a wind measuring tower), the power and the like in the database, wherein the unified frequency is 10min. When the method is applied, the average air density measured in the effective data acquisition period of the test site is 1.225 +/-0.05 kg/m 3 In range, there is no need to normalize the air density to the actual average air density; if not, the conversion is carried out according to the related content of IEC 61400-12-1. According to the method, the power curve, the power coefficient curve and the annual energy production of the test unit are estimated according to the measured data.
The specific implementation mode is as follows: the invention is further described with reference to the accompanying drawings in which:
the present invention will be further described with reference to the accompanying drawings and examples, which are described herein for illustrative and explanatory purposes only and are not limiting of the invention.
As shown in fig. 1, fig. 2 and fig. 3, the method for evaluating power characteristics of a wind turbine generator based on a BP neural network provided by the present invention is based on: and correcting the cabin wind speed measured by wind measuring equipment at the tail part of the wind turbine generator cabin by utilizing the nonlinear fitting capability of the BP neural network so as to obtain the free wind speed, thereby evaluating the power characteristic of the wind turbine generator.
The basic idea of the invention is as follows: the power characteristic is an important basic attribute of the wind turbine and is directly related to the economic and technical level of the wind turbine. At present, IEC61400-12-1 or IEC61400-12-2 is generally adopted to carry out wind turbine generator power characteristic test. Although the IEC61400-12-1 standard can accurately test the power output condition of a unit, strict requirements on the terrain and the wind measuring tower are required, tests in the mode are probably impossible or difficult to perform on wind farms with complex terrains, even if tests can be performed, the IEC standard wind power characteristic test cannot be performed on dozens or even hundreds of wind farms, and a large amount of manpower, material resources and financial resources are consumed if the test is performed. Although the IEC61400-12-2 standard can evaluate the power of the unit relatively quickly and at a relatively low cost, the application conditions of a cabin Transfer function (NTF) established in the evaluation are strict and are difficult to meet completely, so that a power curve evaluated has a relatively large error, and the judgment of the running state of the unit is influenced.
Wind flows through the rotating impeller, the fluctuation of wind speed is more irregular, and the mapping relation between the cabin wind speed and the free wind speed is complex and nonlinear and is difficult to describe by a linear function. Because the BP neural network can realize any nonlinearity from input to output, the BP neural network is adopted to establish the mapping relation between the free wind speed and the cabin wind speed. And training to obtain the free wind speed of each unit.
A wind turbine generator power characteristic evaluation method based on a BP neural network comprises the following specific implementation steps:
the first step is as follows: and (3) site evaluation is carried out on each unit in the wind power plant, and in the evaluation test site, the terrain change on a plane slope (an included angle between a plane passing through a tower foundation of the wind power unit and a sector formed by taking the position of the wind power unit as a vertex) and a ground plane meets the requirements of IEC61400-12-1 standard on the terrain of the test site. Table 1 gives the specific requirements for the terrain variation around the test site, where L is the distance between the wind generating set and the meteorological mast and D is the wind wheel diameter of the wind generating set. Fig. 1 shows a schematic diagram of terrain variation of a wind turbine power characteristic test site, fig. 2 shows maximum fluctuation of an allowable terrain within 2L, and fig. 3 shows maximum terrain drop from a wind turbine tower foundation horizontal plane allowed by site evaluation in different areas.
Table 1 test field requirements: terrain variation
(1) For a wind power plant with a common terrain, randomly selecting one unit as a reference unit;
(2) For a wind farm with complex terrain, each unit is calculated for the Roughness Index (RIX), and a unit close to the average of all RIXs is selected as a reference unit.
The second step is that: and carrying out power characteristic test on the reference unit according to the IEC61400-12-1 standard. Meanwhile, a voltage transformer and a current transformer are installed on the low-voltage side of the residual unit box transformer so as to obtain the net power output value of each unit.
The third step: synchronously recording the data of the wind measuring tower, the data of the power characteristic test wind measuring rod, the SCADA data of each unit and the power output data of each unit, and establishing an evaluation database. The wind speed of an engine room in the evaluation database is acquired in an SCADA system of the unit, the free wind speed is acquired in a wind measuring tower and a power characteristic test wind measuring rod in a wind power plant, the database stores data of not less than 180h of continuous measurement time, the data acquisition frequency is within 7 s-10 min, and the wind speed range and the wind condition within a certain range can be covered.
The fourth step: and (4) testing and screening the data, and performing range test and trend test according to GB/T18710-2002 in combination with the actual condition of the wind farm (see tables 2 and 3). And then data elimination and correction are carried out, so that the data can truly and objectively reflect the operation condition of the wind turbine generator (the data that the wind turbine generator does not work or the test system breaks down are eliminated).
TABLE 2 reasonable Range of reference values for the principal parameters
TABLE 3 reasonable trend of Change reference values of the main parameters
And (4) synchronizing the time of the data such as the cabin wind speed, the free wind speed (measured by a wind measuring tower and a wind measuring rod), the power and the like in the evaluation database, wherein the unified frequency is 10min.
The fifth step: performing data regression, determining the air density of the tested wind power plant according to the wind measuring tower data, wherein the average air density measured during the effective data acquisition period of the tested wind power plant is 1.225 +/-0.05 kg/m 3 In range, there is no need to normalize the air density to the actual average air density; if not, the conversion is carried out according to the related content of IEC 61400-12-1.
The air density can be derived from air temperature and air pressure measurements according to equation (1):
wherein: rho 10min An average air density of 10min; t is 10min The measured average absolute air temperature is 10min; b is 10min The measured average air pressure is 10min; r is the gas constant 287.05J/(kg. K).
For the wind turbine generator with automatic power control, the wind speed can be converted to the standard atmospheric pressure by the formula (2):
wherein, V n The wind speed value is converted; v 10min The measured average wind speed value is 10min; rho 10min The obtained average air density for 10min; rho 0 Is a standard air density of 1.225kg/m 3 。
For a wind turbine with stall control, constant pitch and speed, the measured power output data can be converted using equation (3):
wherein: p n The wind speed value is converted; p 10min The measured average wind speed value is 10min; rho 10min The obtained average air density for 10min; rho 0 Is a standard air density of 1.225kg/m 3 。
And a sixth step: and correcting the cabin wind speed measured by wind measuring equipment at the tail part of the wind turbine generator cabin by utilizing the nonlinear fitting capacity of the BP neural network, so as to obtain the free wind speed, and thus, evaluating the power characteristics of the wind turbine generator. The method adopts three layers of BP neural networks as shown in figure 1, wherein the input layer is the wind speed of a wind measuring tower and the wind speed of an engine room, and the output layer is the free wind speed. And selecting reference unit data as training data, and establishing a mapping relation between the free wind speed and the cabin wind speed (as shown in figure 2).
The seventh step: and taking the cabin wind speed and anemometer tower data of each unit in the wind power plant as input values of the trained BP neural network to obtain the free wind speed of each unit.
Eighth step: and drawing a power curve, after data correction is finished, sorting the selected test data according to a bin method, wherein the selected data group is required to cover a wind speed range from 1m/s below a cut-in wind speed to 1.5 times of the wind speed when the wind turbine generator is output at 85% rated power. The wind speed range should be continuously divided into 0.5m/sbin, with a central value that is an integer multiple of 0.5 m/s. Drawing a power curve by using the power value corresponding to each normalized wind speed bin:
wherein: v i The calculated average wind speed value of the ith bin is obtained; v n,i,j The measured wind speed value of the j data group of the ith bin; n is a radical of i The number of data in the 10min data set for the ith bin; p i The average power value of the ith bin after the conversion is obtained; p is n,i,j The measured power value of the j data set of the ith bin is obtained.
The ninth step: and (3) drawing a power coefficient curve, wherein the power coefficient can be obtained by calculating the formula (6) according to the measured power curve:
wherein: c p,i Is the power coefficient in bin i; v i Average wind speed in bin i obtained for the conversion; p i For the resulting power output in the bin; a is the swept area of the wind wheel of the wind turbine set; rho 0 Is the standard air density.
The tenth step: and (4) calculating annual power generation, wherein the annual power generation is an estimated value calculated by using a power curve obtained by measurement for different reference wind speed frequency distributions. The reference wind speed frequency distribution may be a rayleigh distribution equivalent to a weibull distribution with a shape factor of 2. The Annual Energy Production (AEP) at an annual average wind speed of 4,5,6,7,8,9,10,llm/s can be calculated according to the formula (7):
wherein: AEP is annual energy production; n is a radical of h The hour number in one year is about 8760; n is the number of bins; v i The converted average wind speed value in the ith bin is obtained; p i The calculated average power value at the ith bin is used as the power value.
The function of the rayleigh distribution is:
wherein: f (V) is a Rayleigh distribution function of wind speed; v ave Is the annual average wind speed value at the wind turbine hub center height; v is the wind speed value. Set V i-1 =V i 0.5m/s and P i-1 And the superposition starts when the power is 0.0 kW.
The annual energy production has to be calculated on the one hand as a "measurement of the annual energy production" and on the other hand as an extrapolation of the annual energy production. If the measurements do not include a cut-out wind speed value, extrapolation is required to obtain an annual energy production from the measured maximum wind speed value extrapolated to the cut-out wind speed. The annual energy production extrapolation section is obtained assuming that all power values for all wind speeds below the lowest wind speed of the tested power curve are 0, and assuming that all powers above the wind speed range between the highest wind speed on the measured power curve and the cut-out wind speed are constant. The constant power value used for extrapolation should be the power value of the highest wind speed bin in the measured power curve.
The eleventh step: and (6) generating a report.
The above description is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto. Any changes or substitutions that may be easily made by those skilled in the art within the technical scope of the present disclosure are intended to be included within the scope of the present disclosure. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.