CN-117212050-B - Wind turbine generator system headroom prediction control method

CN117212050BCN 117212050 BCN117212050 BCN 117212050BCN-117212050-B

Abstract

The invention provides a method for predicting and controlling the clearance of a wind turbine, which comprises the steps of firstly collecting historical clearance data in a recent period of time, including clearance data cl h , a pitch angle alpha and an impeller rotating speed omega, then obtaining a clearance predicting result through a machine learning method or a compartment query predicting method, and finally adopting unit protection action control according to the clearance predicting result. The wind turbine generator system clearance prediction control method provided by the invention can predict the clearance value based on the short-term clearance data and the working condition data, thereby realizing timely control intervention under extremely severe working conditions, preventing the risk of tower sweeping caused by continuous descending of the clearance, realizing early interference of the wind turbine generator system, and ensuring safe and stable operation of the wind turbine generator system.

Inventors

GUO XIAOLIANG
WANG DONGLI
Shu Jianyuan
AO RUI
SUN KAI
WANG HAICHAO

Assignees

陕西中科启航科技有限公司

Dates

Publication Date: 20260508
Application Date: 20230918

Claims (5)

1. The wind turbine generator system headroom prediction control method is characterized by comprising the following steps of: S1, acquiring historical clearance data in a recent period of time, wherein the historical clearance data comprises clearance data cl h , a pitch angle alpha and an impeller rotating speed omega; s2, obtaining a headroom prediction result through a machine learning method or a binning inquiry prediction method; if the machine learning method is adopted, the method specifically comprises the following steps: A1, loading collected historical clearance data, and preprocessing the historical clearance data, wherein the preprocessing comprises detection and revision of missing values, abnormal values and repeated values; A2, grouping historical clearance data according to time sequence, wherein each group of data comprises historical information of 3 blades in the process of continuously rotating the blades for two circles And a first clearance value cl A3 of the first blade after two rotations, wherein a first clearance value of the first blade and a pitch angle and an impeller rotation speed at the same time in history information are respectively represented by cl A1 、α 1 and ω 1 , a first clearance value of the second blade and a pitch angle and an impeller rotation speed at the same time are respectively represented by cl B1 、α 2 and ω 2 , a first clearance value of the third blade and a pitch angle and an impeller rotation speed at the same time are respectively represented by cl C1 、α 3 and ω 3 , a second clearance value of the first blade and a pitch angle and an impeller rotation speed at the same time are respectively represented by cl A2 、α 4 and ω 4 , a second clearance value of the second blade and a pitch angle and an impeller rotation speed at the same time are respectively represented by cl B2 、α 5 and ω 5 , and a second clearance value of the third blade and a pitch angle and an impeller rotation speed at the same time are respectively represented by cl C2 、α 6 and ω 6 ; then for a group of data A representation; A3, combining according to priori knowledge and expert advice, deriving new features according to the existing features, and normalizing feature data; A4, randomly splitting each group of data into a training set and a testing set according to the proportion of 7:3; a5, training the multiple linear regression model by a training set, and optimizing parameters of the linear regression model by adopting a gradient descent algorithm until the loss rate of the model is reduced to the minimum, so as to obtain an initial headroom prediction model; A6, evaluating an initial headroom prediction model by using test set data, and if the performance is insufficient, performing a characteristic engineering and parameter optimization method to improve the performance until the headroom prediction model with the best performance is obtained; A7, predicting a headroom value by using a headroom prediction model with the best performance to obtain a headroom prediction result; if the method for predicting the bin query is adopted, the method specifically comprises the following steps: B1, dividing historical clearance data into bins according to rotating speed and pitch angle, taking the average value of the clearance data in each bin as a reference value, and obtaining a rotating speed pitch angle-clearance reference table; b2, grouping historical clearance data according to time sequence, wherein each group of data comprises historical information of 3 blades in the process of continuously rotating the blades for two circles Wherein the first clearance value and the co-time pitch angle and the impeller speed of the first blade are denoted by cl A1 、α 1 and ω 1 , respectively, the first clearance value and the co-time pitch angle and the impeller speed of the second blade are denoted by cl B1 、α 2 and ω 2 , respectively, the first clearance value and the co-time pitch angle and the impeller speed of the third blade are denoted by cl C1 、α 3 and ω 3 , respectively, the second clearance value and the co-time pitch angle and the impeller speed of the first blade are denoted by cl A2 、α 4 and ω 4 , respectively, the second clearance value and the co-time pitch angle and the impeller speed of the second blade are denoted by cl B2 、α 5 and ω 5 , respectively, the second clearance value and the co-time pitch angle and the impeller speed of the third blade are denoted by cl C2 、α 6 and ω 6 , Then any one group of data is used A representation; B3, for any group of data d, inquiring the corresponding pitch angle and impeller rotating speed at each moment in a rotating speed pitch angle-clearance reference table to obtain a corresponding reference clearance value, namely, an ith reference clearance value cl bi is obtained by inquiring the ith pitch angle alpha i and impeller rotating speed omega i in the rotating speed pitch angle-clearance reference table, and then the data d is converted into data ; B4, according to the formula Further converting the data d into data ; B5, calculating to obtain a predicted headroom wind shearing component cl q7 according to the following empirical formula: , wherein k 1 、k 2 and k 3 are a first drop coefficient, a second drop coefficient and a third drop coefficient respectively, and a preset empirical value is adopted; b6, calculating a predicted rotation speed omega 7 and a predicted pitch angle alpha 7 according to the following formula: , , B7, inquiring in a rotating speed pitch angle-clearance reference table by using the predicted rotating speed omega 7 and the predicted pitch angle alpha 7 to obtain a predicted reference clearance value cl b7 ; B8, calculating to obtain a predicted empty value cl A3 by the following formula: ; and S3, adopting unit protection action control according to the headroom prediction result.
2. The method for predicting and controlling the clearance of the wind turbine generator set according to claim 1, wherein the step S1 clearance data cl h is acquired by a clearance measuring system, and the pitch angle alpha and the impeller rotating speed omega are acquired from a fan main control system through an industrial bus.
3. The method for controlling the headroom prediction of the wind turbine generator system according to claim 1, wherein in the step A6, an initial headroom prediction model is evaluated, specifically, a mean square error, an absolute average error and a decision coefficient are adopted for calculation and evaluation, a chart of a predicted value and a true value is drawn, and the predicted effect of the model is observed.
4. The method for predicting and controlling the headroom of a wind turbine generator according to claim 1, wherein step S3 is characterized in that the headroom drop condition is judged according to the predicted headroom value, a soft control strategy is adopted if the predicted headroom value is higher than the upper limit th1 of a dangerous area, an emergency control strategy is adopted if the predicted headroom value is higher than 70% of the upper limit th2 of the dangerous area, and a soft control strategy is adopted if 60% of the actual headroom value is in an excessive area th2, th 1.
5. The method of claim 4, wherein the soft control strategy is to retract the pitch to a safe pitch angle α e at a rate of θ1, and to cut the pitch if the trigger condition is not satisfied for 30 seconds, and the emergency control strategy is to retract the pitch to a safe pitch angle α e at a rate of θ2, and to cut the pitch if the trigger condition is not satisfied for 30 seconds.

Description

Wind turbine generator system headroom prediction control method Technical Field The invention relates to the technical field of wind turbine unit headroom monitoring, in particular to a wind turbine unit headroom prediction control method. Background With the progress of wind power technology, the size of wind turbine generator blades is continuously increased. The larger blades can capture more wind energy, so that the power generation efficiency is improved. From the previous tens to nearly hundreds of meters of large blades, the blades are also moving towards lighter and softer blades. Meanwhile, in order to obtain higher wind speed and more stable wind energy resources, the heights of the wind turbine tower cylinders are gradually increased. The higher tower can promote generating efficiency, makes the unit operation in higher height. All of the above trends present challenges to the design and safety margin of wind turbine clearances. In order to cope with the generation of low clearance or tower sweeping conditions which can be generated, various major complete machine manufacturers develop and apply a clearance monitoring means for early warning the clearance and protecting the unit. The current clearance monitoring means are all real-time measurement, on one hand, due to the measurement calculation and communication of the system, delay of hundreds of milliseconds is commonly existed, on the other hand, extremely severe working conditions are encountered, the clearance is greatly reduced, when a fan main control system receives a dangerous clearance signal, the system is likely to just make countermeasures, the clearance of blades is not restrained timely and effectively, a tower sweeping accident is generated, and the hysteresis is the defect of the unit protection function based on the real-time clearance at present. There are also studies that attempt to predict the headroom value based on environmental data such as wind speed, wind direction, etc., and input historical data into a more mature model such as a neural network model to calculate and predict to obtain the headroom value of the wind turbine generator, and then control and intervention are performed based on the result. However, the method for predicting the wind speed has a plurality of problems that the cabin type wind measuring radar accurately measures the wind speed beyond hundred meters to complete feedforward control of the wind turbine generator, the change rules of the wind speed, the wind direction, the shear and the like are poor, the accurate measurement is difficult, the accurate rule information is difficult to summarize and form, the accurate prediction is difficult to be performed by combining the state information of the wind turbine generator, and even the integrated blade load sensor is difficult to accurately predict. 2. The neural network model often needs to combine too much uncertain information, and coarse control granularity based on average value is adopted, so that the neural network model is used for training with so many variables, and the fitting phenomenon is easy to occur, so that the generalization performance on new data is reduced. 3. In the prediction process, the neural network model needs to perform a series of matrix operation and activation function calculation on input data, and the calculation operations need to consume a large amount of calculation resources, are inconvenient to directly implement in the wind turbine generator PLC, need to independently add hardware, increase the cost of the system and the difficulty of integrating the system with the wind turbine generator PLC. Disclosure of Invention In order to solve the defects of the prior art, the invention provides a wind turbine generator system clearance prediction control method, which can predict a clearance value based on short-term clearance data and working condition data, thereby realizing timely control intervention under extremely severe working conditions, preventing tower sweeping risks caused by continuous descending of clearance, realizing early interference of the wind turbine generator system, and ensuring safe and stable operation of the wind turbine generator system. The invention provides a method for predicting and controlling the clearance of a wind turbine, which aims to solve the technical problems and comprises the following steps: S1, acquiring historical clearance data in a recent period of time, wherein the historical clearance data comprises clearance data cl h, a pitch angle alpha and an impeller rotating speed omega; s2, obtaining a headroom prediction result through a machine learning method or a binning inquiry prediction method; and S3, adopting unit protection action control according to the headroom prediction result. Step S1 clearance data cl h is acquired by a clearance measurement system, and the pitch angle alpha and the impeller rotating speed omega are acquired from a fan main control system thr