Bladed modeling method for wind turbine based on scada data extraction

Abstract

This invention relates to a bladed modeling method for wind turbines based on SCADA data extraction. It involves establishing a bladed model equivalent to the operating conditions of a real wind turbine using operational data. The method is characterized by preprocessing the SCADA data collected from the wind farm, using least squares and fuzzy weighted least squares methods to model the wind turbine's operating state and bladed turbine under different conditions, creating wind files for the SCADA wind speed under the studied conditions so that the bladed wind file wind speed can represent the actual wind speed, and finally simulating the wind turbine using different wind files. The simulation results are then compared with the operating parameters of the selected SCADA system after correction. The purpose of this invention is to solve the problem of simulating wind turbine faults to avoid accidents; and to achieve optimized control of wind turbines, enabling safe, stable, and efficient operation.

Description

Technical Field

[0001] This invention belongs to the field of wind turbine modeling, and in particular relates to a method for establishing a Bladed model equivalent to the operating condition using real wind turbine operating data. Background Technology

[0002] China currently possesses relatively mature design and manufacturing technologies for large-scale wind turbines, making it both the world's largest installed wind power capacity country and the largest producer of complete wind turbines. Its wind power equipment manufacturing has reached a leading level, and its equipment industry chain is internationally competitive.

[0003] With the development of the wind power industry, the competitiveness of wind power generation against traditional power generation is constantly improving. The increasing size of wind turbines, the stability of power generation, and the reliability of wind turbine operation are the most effective ways to enhance competitiveness. Currently, many wind turbine manufacturers in China are no longer focusing on low-capacity wind turbines, but instead are producing larger capacity turbines suitable for different environments. The capacity of onshore wind turbines in my country has increased from the initial kilowatt level to the current megawatt level, and many companies are now deploying offshore turbines, with capacities reaching 5MW or 6MW and above.

[0004] As the capacity of a single wind turbine increases, the cost of building wind turbines rises. Furthermore, wind farms are typically located in areas with abundant wind resources, such as coastal or mountainous regions, where the natural environment is often unfavorable for building wind turbines suited to the local conditions, hindering their operation. Secondly, turbine failures can occur, leading to significant losses and complicating maintenance. While the lifespan of a wind turbine is expected to exceed 20 years, in actual operation, many wind turbines experience varying degrees of failure, impacting their normal operation. Summary of the Invention

[0005] Purpose of the invention:

[0006] This invention provides a bladed modeling method for wind turbines based on SCADA data extraction. Its purpose is to solve the problem of how to simulate the faults of wind turbines and avoid many accidents. It can also optimize the control of wind turbines and achieve safe, stable and efficient operation of wind turbines.

[0007] Technical solution:

[0008] The objective of this invention is achieved as follows:

[0009] A bladed modeling method for wind turbines based on SCADA data extraction, characterized by the following steps:

[0010] (1) Perform data preprocessing on the SCADA data of wind turbine units collected from the wind farm, that is, repair the missing and erroneous data of wind speed-power, speed-torque and wind speed-pitch angle;

[0011] (2) A fuzzy weighted least squares method is proposed, and the least squares method and the fuzzy weighted least squares method are used to model the operating status of wind turbines under different working conditions. The advantages and disadvantages of the two models are analyzed, and the model with higher accuracy is selected.

[0012] (3) Based on the unit parameters of the wind turbine, bladed unit modeling is performed, and the blades, impellers, hubs, towers, nacelles, transmission chains and generators are digitally modeled.

[0013] (4) Create a wind file for the SCADA wind speed under the studied working conditions, and perform a deviation analysis of the Euclidean distance between the wind file and the original wind speed so that the wind speed in the Bladed wind file can replace the real wind speed.

[0014] (5) Finally, SCADA data is used to identify the system and tune the parameters of the controller. The obtained PI parameters are input into the Bladed controller and simulated using different wind files. The simulation results are compared with the operating parameters of the selected SCADA system after repair. The operating status and the model established by the fuzzy weighted least squares method are compared and analyzed.

[0015] Step (1) processes the collected data on wind speed, power, rotational speed, torque, and pitch angle, and repairs the data using the following regression function:

[0016]

[0017]

[0018] In the formula, Let y be the value of the i-th repaired value. i x is the i-th value that does not need to be repaired. i For y i The corresponding value of the i-th element. The mean of values ​​that do not require repair. It is the propeller pitch angle. is the mean of x, and i is a natural number from 1 to n.

[0019] The relevant parameters of the wind turbine in step (3) are as follows:

[0020] The cut-in wind speed is 3 m / s, the rated wind speed is 11 m / s, the gear transmission speed ratio N is set to 104, the impeller diameter is 82.56 m, and the engine rotational inertia Jg is set to kg·m.2 The moment of inertia of the wind turbine, Jω, is set to 6.25 × 10⁶ kg·m. 2 The pitch angle variation range θ / (°) is taken as (0~20). Based on the above, the Bladed model is set.

[0021] The specific steps for creating the wind document in step (4) are as follows:

[0022] ① First, the wind speed data in the SCADA system is processed into three columns of wind speed data, namely the data in the X, Y and Z directions. Since it is difficult to obtain wind speed data in the Y and Z directions in reality, these two directions are set to 0. Then, the real wind speed is saved in .txt format to ensure that the frequency of use, i.e. the time interval, remains consistent. The X, Y and Z columns of data are separated by tabs to process multiple variable data at the same time.

[0023] ② Set the turbulent wind parameters;

[0024] ③ Configure the parameters for the Kaimal model. When configuring Kaimal, the General option is usually selected.

[0025] ④ In the turbulent wind parameter setting window, add the path to the processed SCADA system data txt file and set the average wind speed for the SCADA data.

[0026] Advantages and effects:

[0027] 1. By establishing a 1:1 Bladed model of a real wind turbine, it can be used for the design of wind turbines in the region, allowing the design of wind turbines to better reflect reality.

[0028] 2. Once the bladed model is built, it can simulate wind turbine failures, thus preventing many accidents from occurring.

[0029] 3. This Bladed model can also optimize the control of wind turbines, thereby improving the actual performance of wind turbines.

[0030] 4. Establishing a bladed model of a real wind turbine is beneficial to the safe, stable and efficient operation of the wind turbine. This bladed model has strong economic performance and engineering application value for the research of real wind turbines. Attached Figure Description

[0031] Figure 1 The Bladed model built for this invention is equivalent to a structural schematic diagram of a real wind turbine.

[0032] Figure 2 This is a flowchart of the preprocessing process for SCADA data from wind turbine generators.

[0033] Figure 3 This is a flowchart of the steps of the BP algorithm in the BP neural network of the present invention.

[0034] Figure 4 This is a schematic diagram of the initial PI value predicted based on the BP neural network algorithm. Detailed Implementation

[0035] This invention primarily utilizes the parameters of a wind turbine in a wind farm and SCADA system data to establish a wind turbine model. The modeling software used is Bladed, with the aim of ensuring that the Bladed model's simulation parameters are identical to those of a wind turbine operating in its natural environment. To guarantee the Bladed model's equivalence to the real wind turbine, the SCADA system data needs to be preprocessed before modeling, i.e., anomalies and missing data need to be corrected. Statistical analysis of typical operating states of the wind turbine under different conditions is also performed. Next, Bladed modeling is conducted based on the inherent parameters of the wind turbine in the wind farm, and a Bladed wind file is created based on the natural wind speed where the wind turbine is located. Then, the controller for different operating conditions of the wind turbine is designed using SCADA data. Finally, the simulation results of the Bladed model are identical to the operating state of the wind turbine in its natural environment. By establishing a 1:1 bladed model of a real wind turbine, it can be used for the design of wind turbines in the region, allowing the design to better reflect reality. Once the model is built, it can simulate wind turbine failures, preventing many accidents. The model can also optimize the control of the wind turbine, improving its performance. In conclusion, establishing a bladed model of a real wind turbine is beneficial for its safe, stable, and efficient operation. This bladed model has strong economic performance and engineering application value for the research of real wind turbines.

[0036] The present invention will now be described in detail with reference to the accompanying drawings.

[0037] This invention involves creating a bladed model of a real 1.5MW wind turbine and designing controllers for different operating conditions using the turbine's operational data. This results in a bladed model that closely resembles the actual wind turbine in terms of operating parameters and states. The equivalence between the bladed model and the real wind turbine is measured by the various operating parameters at the same wind speed and the operating parameters and states under different conditions. The process flow of this invention is as follows: Figure 1 As shown.

[0038] Based on the SCADA data, a Bladed model equivalent to a real wind turbine is established. The collected data on wind speed, power, rotational speed, torque, and pitch angle are processed, and the data is repaired using a regression function, as shown below:

[0039]

[0040]

[0041] In the formula, The corrected value is y. i x is a value that does not need to be repaired. i For y i The corresponding value, The mean of values ​​that do not require repair. is the mean of x, and i is a natural number from 1 to n. Regression function data repair is a refinement of the mean-based repair function, eliminating the influence of extreme values.

[0042] Based on the solution process of the least squares method, the solution method of the fuzzy weighted least squares method can be derived, and the model results can be obtained by analogy. Preprocessing and data repair are as follows. Figure 2 As shown.

[0043] In summary, the steps of the fuzzy weighted least squares algorithm are as follows (taking the speed and torque under the maximum wind energy capture condition as an example):

[0044] (1) Obtain the functions of the speed and torque model using the least squares method;

[0045] (2) Substitute the torque into the equation In the calculation, the deviation between the torque and the function obtained in step 1 is calculated;

[0046] (3) Determine the number of thresholds and calculate them based on the membership function;

[0047] (4) The torque data is assigned to the corresponding fuzzy subsets according to the maximum membership principle, and weights are added to the torque according to the fuzzy subset assignment;

[0048] (5) Substitute the weighted torque into the equation The calculation yields the fuzzy weighted speed and torque models as functions.

[0049] When simulating a 1.5MW wind turbine model built with Bladed, ensuring consistency with real wind turbine conditions is crucial for better comparison and analysis. Real-world winds vary over time, so the wind speed in the Bladed simulation needs to be set as a wind file to match real-world wind speeds as closely as possible. Creating a wind file first requires defining three-dimensional turbulent wind and calculating the wind file. The following are the steps for creating a wind file based on real-world SCADA wind speeds:

[0050] ① First, the wind speed data in the SCADA system is processed into three columns of wind speed data, namely the data in the X, Y and Z directions. Since it is difficult to obtain wind speed data in the Y and Z directions in reality, these two directions are set to 0. Then, the real wind speed is saved in .txt format to ensure that the frequency of use, i.e. the time interval, remains consistent. The X, Y and Z columns of data are separated by tabs to process multiple variable data at the same time.

[0051] ② Set the turbulent wind parameters; the comprehensive calculation capability in the direction of the wind turbine should be less than 5m at nodes; the comprehensive calculation capability in the direction of the turbine height should be less than 5m at nodes; the value should be greater than the diameter of the wind turbine; the value should be greater than the sum of the hub height and the radius of the wind turbine; the simulation duration for wind speed should be 600s; the frequency of wind speed should be greater than 10Hz; the average wind speed of the sampled SCADA data should be used; different random numbers should generate different wind files; ③ Set the parameters for the Kaimal model, selecting General for Kaimal; the model parameters are: Longitudinal = 8.1L; Lateral = 2.7L; Vertical = 0.66L; Coherency scale parameter = 8.1L; Coherency delay constant = 12; where L is the scale parameter, determined based on the hub height H;

[0052] ④ In the turbulent wind parameter setting window, add the path to the processed SCADA system data txt file and set the average wind speed for the SCADA data.

[0053] After creating the wind file, it needs to be compared and analyzed with the actual wind speed. Euclidean distance deviation analysis is performed on it. To simplify, the deviation between wind speeds is assumed to follow a normal distribution. According to the 3σ principle of normal distribution, the number of data exceeding the threshold range is counted. The proportion of data exceeding the threshold range is calculated by comparing it with the total number of data. If the proportion of data exceeding the threshold range is within 5%, it means that more than 95% of the data is covered within the threshold range. This indicates that the wind speed in the wind file can represent the actual wind speed.

[0054] The specific steps for establishing an optimal model equivalent to the actual system of a real wind turbine based on SCADA data are as follows:

[0055] (1) Input and output data: In system identification, the data in the SCADA system is usually used as the input and output sequence as the data set for identification.

[0056] (2) Model category: Before system identification, the structure of the model needs to be selected and the parameters adjusted.

[0057] (3) Optimality Criterion: The optimality criterion is the basis for evaluating the system model. The optimality criterion is generally the error between the model and the actual output, as shown in the following formula:

[0058]

[0059] In the formula, y(i) is the output of the actual system. n (i)——The output of the system identification model, f(k)——The difference function between the actual model and the identification model.

[0060] The principle of a PI controller using a BP neural network algorithm is to use the algorithm to predict the PI controller parameters, making them infinitely close to the actual output value. If the deviation from the actual system output is large during the calculation process, the error needs to be calculated and fed back into the system for re-prediction. This process is repeated until the error is minimized, thus obtaining the initial K. p With K i Value; Schematic diagram of the initial PI value predicted based on the BP neural network algorithm is shown below. Figure 4 As shown.

[0061] The above settings are used to perform a Bladed simulation. Finally, the operating parameters and operating status of the Bladed model are compared with those of the real wind turbine.

[0062] In summary, this invention first preprocesses SCADA data to obtain data for normal operation of the wind turbine. Then, it uses the least squares method and fuzzy weighted least squares method to model the operating states under maximum wind energy capture and constant power conditions, and analyzes the merits of the two models to select the optimal model. A bladed turbine model is established using the inherent parameters of a 1.5MW wind turbine, and a wind file is created based on the SCADA wind speed to ensure that the simulated wind speed in the bladed model is the same as the wind speed of the real wind turbine operating in the natural environment. Then, the initial PI value of the wind turbine controller is obtained using SCADA data. Finally, bladed simulation is performed, and the simulation results are compared and analyzed with the SCADA operating results and operating states. The results show that the bladed model is equivalent to the real wind turbine.

Claims

1. A bladed modeling method for wind turbines based on SCADA data extraction, characterized in that: This method is performed according to the following steps: (1) Perform data preprocessing on the SCADA data of wind turbine units collected from the wind farm, that is, repair the missing and erroneous data of wind speed-power, speed-torque and wind speed-pitch angle; (2) A fuzzy weighted least squares method is proposed, and the least squares method and the fuzzy weighted least squares method are used to model the operating status of wind turbines under different working conditions. The merits of the two models are analyzed, and the model with higher accuracy is selected. (3) Based on the unit parameters of the wind turbine, bladed unit modeling is performed, and the blades, impellers, hubs, towers, nacelles, transmission chains and generators are digitally modeled; (4) Create a wind file for the SCADA wind speed under the studied working conditions, and perform a deviation analysis of the Euclidean distance between the wind file and the original wind speed so that the wind speed in the Bladed wind file can replace the real wind speed. (5) Finally, SCADA data is used to identify the system and tune the parameters of the controller. The obtained PI parameters are input into the Bladed controller and simulated using different wind files. The simulation results are compared with the operating parameters of the selected SCADA system after repair. The operating status and the model established by the fuzzy weighted least squares method are compared and analyzed. The steps of the fuzzy weighted least squares algorithm are as follows: (1) Obtain the functions of the speed and torque model using the least squares method; (2) Substitute the torque into the equation In the formula, the deviation between the torque and the function obtained in step 1 is calculated; where, For the first A value that does not need to be repaired. for The corresponding number The value of , It is a natural number from 1 to n; (3) Determine the number of thresholds and calculate them based on the membership function; (4) The torque data is assigned to the corresponding fuzzy subsets according to the maximum membership principle, and weights are added to the torque according to the fuzzy subset assignment; (5) Substitute the weighted torque into the equation In this process, the fuzzy weighted speed and torque models can be obtained through calculation; The specific steps for creating the wind document in step (4) are as follows: ① First, the wind speed data in the SCADA system is processed into three columns of wind speed data, namely the data in the X, Y and Z directions. Since it is difficult to obtain wind speed data in the Y and Z directions in reality, these two directions are set to 0. Then, the real wind speed is saved in .txt format to ensure that the frequency of use, i.e. the time interval, remains consistent. The X, Y and Z columns of data are separated by tabs to process multiple variable data at the same time. ② Set the turbulent wind parameters; the comprehensive calculation capability in the direction of the wind turbine should be less than 5m at nodes; the comprehensive calculation capability in the direction of the turbine height should be less than 5m at nodes; the value should be greater than the diameter of the wind turbine; the value should be greater than the sum of the hub height and the radius of the wind turbine; the simulation duration for wind speed should be 600s; the frequency of wind speed should be greater than 10Hz; the average wind speed of the sampled SCADA data should be used; different random numbers should generate different wind files; ③ Set the parameters for the Kaimal model, selecting General for Kaimal; the model parameters are: Longitudinal=8.1L; Lateral=2.7L; Vertical=0.66L; Coherency scale parameter=8.1L; Coherency delay constant=12; where L is the scale parameter, determined based on the hub height H; ④ In the turbulent wind parameter setting window, add the path to the processed SCADA system data txt file and set the average wind speed for the SCADA data.

2. The blended modeling method for wind turbines based on SCADA data extraction according to claim 1, characterized in that: Step (1) involves processing the collected data on wind speed, power, rotational speed, torque, and pitch angle, and repairing the data using the following regression function: ； ； In the formula, For the first The repaired value, For the first A value that does not need to be repaired. for The corresponding number The value of , The mean of values ​​that do not require repair. It is the propeller pitch angle. It corresponds The mean of , where i is a natural number from 1 to n.

3. The blended modeling method for wind turbines based on SCADA data extraction according to claim 1, characterized in that: The relevant parameters of the wind turbine in step (3) are as follows: cut-in wind speed Rated wind speed The gear transmission speed ratio N is set to 104, the impeller diameter is 82.56 m, and the engine rotational inertia is... Set as Wind turbine rotational inertia Set to 6.25×106 Pitch angle variation range The range is (0~20). Based on the above, the Bladed model is set.

Images