Yaw control strategy optimization

In recent years, with the rapid development of large-scale wind turbines, the issue of yaw misalignment has attracted increasing attention. This problem not only reduces the power generation capacity of the turbine but also affects the fatigue load on its blades. Therefore, in this paper, the shortcomings of common yaw control strategies were introduced first, and we propose a decoupling approach for wind direction angle and yaw angle to eliminate physical deviations and improve accuracy in aligning with the wind direction.


Background information
In recent years, China has accelerated the construction of a clean and low-carbon energy system, with an increasing proportion of clean energy and non-fossil fuel consumption.Wind power plays an increasingly important role in the adjustment of our country's energy structure as a clean energy source.With the continuous maturity of wind power technology, wind turbines are constantly transitioning to larger sizes.At the same time, starting in 2021, newly approved onshore wind power projects will no longer receive subsidies from the central government but will implement grid parity pricing.This poses higher cost constraints on the production and operation of the wind power industry.
Research has shown that yaw misalignment has a much more severe impact on the horizontal-axis wind turbine than wake, turbulence, and vertical wind shear.While reducing the power output of the wind turbine, it also increases the gyroscopic torque due to yaw misalignment, posing challenges to the safety and reliability of the turbine assembly [1] .
The ability of a wind turbine to maintain a stable yaw direction during operation is crucial for the normal functioning of a wind energy generation system.Firstly, a stable yaw direction ensures that the turbine blades are always at the optimal angle concerning the airflow, thereby improving rotational efficiency and power generation effectiveness [2] .To achieve this goal, engineers typically employ various control strategies to ensure that the wind turbine maintains a stable yaw direction under different operating conditions.

Original yaw control strategy
In the original yaw strategy, wind direction data V was collected through the wind vane, and a relatively stable average wind direction V1 was obtained after low-pass filtering.V1 was used as the basis for yaw drive and stop.The start and stop strategies are summarized below and are not listed because they are only used to illustrate the method.
IOP Publishing doi:10.1088/1742-6596/2771/1/012006 2 1) Starting conditions: the deviation between the average wind direction V1 and 180 degrees is greater than 9 degrees, lasting for 210 seconds, we start yaw; 2) Stop condition: the deviation between the mean wind direction V1 and 180 degrees is less than 5 degrees, lasting for 3 seconds, we stop yaw.
The yaw control block diagram is shown in Figure 1.180°, then Y is 30.It needs to be emphasized here, that the reference frame of the wind direction angle value is the nacelle position.The size of the wind direction angle data value not only depends on the wind direction but also depends on the nacelle position.That is, the wind direction angle is a representation of the wind direction relative to the nacelle position [3] .
Average wind direction angle V1: after low-pass filtering, the high-frequency component of wind direction fluctuation is eliminated, only the DC component and the low-frequency component are Ԣ retained.Y is 0 when the wind direction is not yawed.The average wind direction angle represents the deviation of the wind direction from the nacelle position.When yawing, due to the movement of the reference frame, the value of the wind direction changes linearly (such as the linear change from 150 to 180), but because of the attenuation and hysteresis of the filter, the numerical transmission is not linear, and Ԣ the value of the attenuation lag Yl.

Deficiency
It can be seen from Figure 1 that the actual wind deviation input to the yaw system is 180-Vb-Ԣ Yl -Vl.This pair of wind deviations can be divided into two parts, the low-frequency components of direct current deviation (180-Vb-Ԣ Yl) and wind direction fluctuation (Vl).The start command and stop command of yaw are determined by the joint action of these two parts.
Over-yawing due to DC component lag: in the absence of fluctuations, this behavior can be explained as shown in Figure 2. When yawing, in the absence of wind direction fluctuation, the deviation to the wind is 150 degrees.If yawing responds to the deviation from wind, it only needs to yaw from 30 degrees to 180 degrees.However, as the average wind direction angle is adopted as the basis when the actual yaw reaches 180 direction, the angle reflected by the average wind direction angle is less than 180 degrees, so the yaw will continue until the average wind direction angle reaches 180 degrees.At this time, the true deviation from the wind may have been greater than 180 degrees, and there is over-yawing behavior.Different filter coefficients have different effects on the average IOP Publishing doi:10.1088/1742-6596/2771/1/0120063 wind direction output.The larger the filter coefficient is, the more obvious the attenuation and lag effects are, the greater the deviation is, the greater the yaw displacement is, and the greater the deviation is.Yaw deficiency or yaw chasing behavior caused by low-frequency fluctuations of wind direction: the yaw system should essentially be a response to DC components, and the components of fluctuations should be ignored unless the fluctuation period is long and the amplitude of the fluctuation is large (high-frequency fluctuations cannot respond, and small fluctuations do not need to respond).Then the existing filtering coefficient is selected as 1600, the corresponding turning period is 200 s, and the fluctuations within 200 s are attenuated.Under this filtering coefficient, there are a large number of fluctuations within 1 minute or 2 minutes, and these fluctuations themselves are random variables.After superposition to the DC component, the wind direction stability reflected by the average wind direction angle is poor, so the yaw start and stop based on this average wind direction are also affected by the randomness of fluctuations, resulting in insufficient yaw or yaw chase behavior.
To sum up the above two cases, it can be seen that the original yaw control strategy is insufficient.In addition, the above two behaviors are directly related to the selection of wind direction angle filtering coefficient and are contradictory.To obtain stable wind direction, it is necessary to increase the filtering coefficient, but a larger filtering coefficient will lead to serious DC component lag; a smaller filtering coefficient can reduce DC component lag, but the fluctuation component in wind direction angle will increase, and the randomness of yaw stop will be amplified.It can be seen that the contradiction is irreconcilable with the existing methods.

Wind direction data decoupling
The fundamental reason for the irreconcilable selection of filtering parameter size in the original strategy is that the original data of the wind vane is the data of the wind direction relative to the nacelle position, and the value only reflects the wind condition under the relative reference frame [4] .Therefore, it can be said that the wind vane data is coupling data, which is the coupling of wind direction fluctuation and yaw displacement.The volatility of the wind conditions themselves makes it necessary to use filters to obtain a relatively stable wind direction.However, if the wind vane data is filtered directly by the filter, it will lead to the hysteresis of the DC component of the wind deviation.
To avoid wind direction lag caused by filtering parameters, the wind direction must be decoupled from the wind direction in the relative coordinate system to the wind direction in the absolute coordinate system [5] .The wind direction of an absolute coordinate system is then used for low-pass filtering, because the filter only acts on the components of wind fluctuations, the wind direction processing can be decoupled.The average wind direction data of the absolute coordinate system is converted back to the average wind direction data of the relative coordinate system, which is used as the basis of yaw drive.The decoupling process of wind direction data is shown in Figure 3. A) In practical applications, to avoid the low-frequency wind direction interference described in Chapter 2, the default value of the low-pass filter coefficient as an initialization parameter is 4800, which is the average wind direction of 10 minutes.
B) The yaw position is calculated to prevent the mean wind direction angle calculated in the yaw position dead zone from remaining unchanged for a long time, resulting in over-yaw.The calculation process is shown in Chapter 6.

New yaw strategy
Since the concept of absolute wind direction has been occupied, to represent the wind direction in the absolute reference frame, a new concept, relative wind direction, is introduced.It means the deviation angle of the wind direction relative to the yaw position 0°.Since the yaw position 0° is a stationary reference, the fluctuation of the wind direction relative to the yaw position 0° can reflect the fluctuation of the wind direction in the absolute coordinate system.
According to the formula "relative wind direction = yaw position -wind direction angle", the relative wind direction concerning a certain direction can be calculated.This formula is based on the difference between the yaw position and the wind direction angle.
Relative wind direction = yaw position -wind direction angle: where Yb is the yaw position at the yaw start time.As seen in the calculation formula, it can be Ԣ observed that during yawing, there is a change in the numerical value of yaw position ( Y), as well Ԣ as in the numerical value of wind direction angle ( Y).By subtracting these two variables, the Ԣ resulting data eliminates Y.In other words, in terms of numerical representation, the increase in yaw position and the increase in direction angle cancel each other out.Therefore, when filtering data based on relative wind direction, no longer affected by yawing and accurately reflects actual changes in wind direction.In the new strategy, wind fluctuation and yaw operation are decoupled.The new yaw strategy is shown in Figure 4.After the relative wind direction is low-pass filtered, the stable relative wind direction angle is obtained, and then the relative wind deviation is converted back to drive yaw.
The DC component of the wind deviation is 180-Vb .

Ԣ
Without considering the influence of fluctuations, when the yaw displacement Y is equal to the DC component of the wind deviation, it can be achieved to face the wind, achieving a complete response without any lag in reflecting the direct current component.The main difference between the Ԣ new and original approaches lies in how Y is transmitted to the yaw input through the yaw feedback channel, bypassing any filtering coefficient processing.This effectively avoids the influence of coefficients on the lagging of the current component [6] .
In the new strategy, components of wind deviation include the DC component, yaw displacement, and wind direction fluctuations present in low-frequency components.The low-frequency components cause interference in both the triggering of yaw initiation and yaw cessation.Therefore, when selecting the coefficients for low-pass filtering, efforts should be made to suppress the influence of fluctuation components.In the original strategy, a filter time factor of 1600 (corresponding to a low-pass filtering transition period of 200 s) was used, referred to as the 60 s average wind direction angle.However, there are still numerous fluctuations in this wind direction data that cannot be responded to by the yaw mechanism.These fluctuations may lead to misjudgments in yaw initiation and termination.Therefore, in the selection of filter coefficients, it is advisable to increase the filter coefficients to obtain a more stable wind direction angle, rather than decreasing them.

Methods for handling dead zones and transitions
Due to the inclusion of a DC component transfer wind deviation through the yaw feedback channel in the new design, there is a higher dependency on the yaw position signal.However, factors such as data discontinuity or dead zones in the cam counter can potentially affect the calculated wind deviation.Therefore, it is necessary to conduct a relevant analysis to assess its potential impact.
Yaw position jump: when there is a jump in the yaw position, it is equivalent to introducing a disturbance R, which is calculated through the flowchart, wind direction=180-Vb-Ԣ Y-Vl-Rg (the high-frequency component of R).If there is a sudden change, the variation period of Rg should be rapid.In yaw initiation conditions, there is a delay of at least 30 seconds, thus it will not affect the yaw initiation.However, in terms of yaw cessation logic, an additional delay of 3 seconds is introduced.If the variation period of Rg is less than 3 seconds, it won't cause any impact.
The impact of dead zone on yaw position: when the yaw position is within the dead zone, it remains unchanged.By calculating wind deviation through flowchart as wind deviation = IOP Publishing doi:10.1088/1742-6596/2771/1/0120066 180-Vb-Ԣ Yl-Vl, it is found that the resulting headwind deviation is consistent with the original plan.Therefore, in the presence of a dead zone, the system will revert to the current plan.However, to suppress fluctuations, a larger filtering coefficient is chosen, which leads to a significant increase in yaw displacement error and exacerbates excessive yawing issues compared to present conditions.
According to the characteristics of being unaffected by dead zones during yaw initiation, it only affects the yaw stop logic and allows for accurate acquisition of yaw speed and time information.Therefore, we can estimate the yaw position signal and convert it back into a headwind-side error angle, triggering the cessation of yaw.Specifically, in the flowchart calculation, when the wind direction reaches 180 degrees, it will trigger a halt in yaw movement.To prevent issues with yaw dead zones, it is necessary to estimate the current yaw position when performing any yaw operations.The following outlines the process for conducting this estimation as shown in Figure 5.

Effect analysis
By increasing the low-pass filter coefficient, relatively stable wind direction data can be obtained.At the same time, when selecting the yaw step size, it can be determined more accurately without waiting for deviation to reach 9 degrees before initiating the yaw.According to the calculation formula of power loss due to wind deviation (1-(cosș)^3), when the wind deviation angle is 5 degrees, the power loss is only 1.11%; however, when the angle increases to 9 degrees, the power loss rises to 3.64%.Despite a 4-degree increase, the power loss more than triples.Therefore, it is advisable to choose an accurate starting yaw angle when initiating yaw.The new yaw control strategy is better than the original as shown in Figure 6.

Conclusion
The yaw system plays a crucial role in the energy capture and health condition of wind turbines.When a turbine is not properly aligned with the wind, it not only reduces the power generation performance but also introduces gyroscopic forces and increases cabin vibrations, posing challenges to the safety and reliability of the turbine.
Firstly, this paper analyzes the drawbacks of the original yaw control strategy.Building upon that analysis, a new yaw control strategy is proposed which decouples wind direction and yaw position.This new strategy reduces the physical deviations during the turbine's yawing process, effectively improving the frequency at which the turbine faces directly into the wind and enhancing the power generation performance of wind turbines.

Figure 1 .
Figure 1.Original yaw control block diagram.Y V V V V g l b '

Figure 5 .
Figure 5. Estimation method of yaw position in dead zone.

Figure 6 .
Figure 6.Comparison of the effects between new and old yaw control strategies.