Yaw control system plays an important role in helping large-scale horizontal wind turbines capture the wind energy. To track the stochastic and fast-changing wind direction, the nacelle is rotated by the yaw control system. Therein, a difficulty consists in the variation speed of the wind direction much faster than the rotation speed of the nacelle. To deal with this difficulty, model predictive control has been recently proposed in the literature, in which the previewed wind direction is employed into the predictive model, and the estimated captured energy and yaw actuator usage are two contradictive objectives. Since the performance of the model predictive control strategy relies largely on the weighting factor that is designed to balance the two objectives, the weighting factor should be carefully selected. In this study, a fuzzy-deduced scheme is proposed to derive the weighting factor of the model predictive yaw control. For the proposed fuzzy-deduced strategy, the variation degree and the increment of the wind direction during the predictive horizon are used as the inputs, and the weighting factor is the output, which is dynamically adjusted. The proposed model predictive yaw control is demonstrated by some simulations using real wind data and its performance is compared with the conventional model predictive control with the fixed weighting factor. Comparison results confirm the outweighing performance of the proposed control strategy over the conventional one.
Since this new century, renewable energies have attracted great attention due to the global warming and the gradual depletion of fossil energy. Among various renewable energies, the wind energy has been globally utilized and rapidly developed for its environmental-friendly feature and low production cost. According to the global wind report 2019 [
To support the accelerated level of growth and compete with other types of energy resources, it is an urgent demand to further decrease the cost of energy through developing advanced technologies for wind turbines (WTs). Among various technologies, WT control technology could directly affect the active power output and component load of WTs, and it has been regarded as one of the key points of WTs research [
In terms of the actuator type, the control technology of horizontal-axis WT can be divided into three parts [
Recently, the Light Detection and Ranging technology (LIDAR) has been mature and is able to acquire the wind information in front of the WT [
To date there have been only several researchers that studied the model predictive control for the yaw control system [
Since the control performance of a WT is simultaneously affected by the wind condition and the control parameters, one potential way for improving the control performance is to adapt the control parameters to the wind condition. For instance, the gains of the proportional-integral (PI) controller are generally scheduled according to the varying operating points determined by the wind speed [
The remaining sections are arranged as follows: material and method are elaborated in Section 2. Method validation and result analysis are described in Section 3. Finally, the conclusion is presented in Section 4.
The baseline MPYC was introduced in [
In the baseline MPYC, there are two objectives, namely, minimizing the energy loss and the yaw actuator usage. To do that, the two objectives are integrated into the objective function by employing the weighting factor. Since the two objectives are related to the yaw error, the yaw error is selected and taken as the predictive variable, which is estimated by a predictive model. In the predictive model, the yaw rate sequence is the control input, and thus it will be predefined during the prediction horizon. Finally, the optimal algorithm is introduced to search out the yaw rate sequence.
The MPYC system is with two objectives: minimizing the energy loss due to the yaw error while having the acceptable yaw actuator usage. The yaw actuator usage is measured using the usage ratio, and the production reduction factor is used to derive the dimensionless factor of evaluating the power production performance. On this basis, the two objectives are combined by the weighting factor
where
where
In
In
where
The nacelle position
where
With
From
In
With
When more than one control objectives are considered, the weighting factors must be appropriately tuned to obtain the satisfactory performance. Due to the lack of the systematic design method, the determination of the weighting factor is still an open research topic in general for the model predictive control area. In the literature, there are a great number of works considering the determination of weighting factors [
where
where
In the fuzzy logic control technique, the Mamdani-type minimum inferential method cooperating with the center-of-area defuzzification procedure is utilized to produce the crisp evaluator output. The center-of-area defuzzification calculates the center of gravity of the reasoned results to obtain the crisp output and the calculation is indicated by:
where
After determining the input variables and inferential method, the membership functions corresponding to the
The scheme of the fuzzy-deduced weighting factor emulates the expert experience to achieve appropriate weighting dispatch, making it easy to incorporate heuristic rules that reflect specialist experience into the calculating processes. Consequently, the weighting distribution is driven by a set of control rules rather than a fixed value.
In
VS | S | M | L | VL | |
---|---|---|---|---|---|
S | L | M | M | VS | VS |
M | VL | L | M | S | VS |
L | VL | L | M | S | S |
This section aims to demonstrate the capabilities of the proposed MPYC strategy using fuzzy-deduced weighting factor and to investigate its performance with different settings of the yaw control system. Therefore, the real wind direction data collected by the wind wanes mounted on the nacelle of an operating WT are used, of which the time series curve is drawn in
The proposed MPYC strategy is validated by simulation tests, and compared with the baseline MPYC with the fixed weighting factor. Since the yaw rate plays a key role in the yaw control system, three simulation cases with different yaw rate are defined as follows: In simulation Case 1: In simulation Case 2: In simulation Case 3:
In Case 2 for the yaw system with the medium yaw rate, the simulation results are shown in
To further compare the results, the average actuator usage, energy loss and values of objective function in the three cases are calculated. Under the three cases, the results are drawn in Bar plot, which are shown in
Case | Average actuator usage (%) | Average energy loss (%) | Average objective function (-) |
---|---|---|---|
1 | 17.01 | 1.117 | 0.010909 |
2 | 12.96 | 1.007 | 0.009643 |
3 | 11.40 | 0.929 | 0.008611 |
Case | Average actuator usage (%) | Average energy loss (%) | Average objective function (-) |
---|---|---|---|
1 | 16.76 | 1.116 | 0.009734 |
2 | 12.64 | 1.012 | 0.008365 |
3 | 11.25 | 0.926 | 0.007665 |
From
Thus, it is shown that, on one hand, the overall performance of the MPYC is enhanced when the yaw rate is increasing. The reason is that a larger yaw rate is beneficial to tracking the fast-varying wind direction for the yaw control system. On the other hand, the performance of the yaw system is enhanced by the proposed MPYC using dynamic weighting factor. Especially, the proposed MPYC reduces the average actuator usage by 0.25%, 0.32%, and 0.15% for the three cases, in comparison with the baseline controller.
These results show that the MPYC using the dynamic weighting factor has better comprehensive performance than the baseline MPYC using the fixed one under different yaw rate.
This paper has proposed an adaptive model predictive control method using the dynamic weighting factor for the yaw system of the horizontal variable-speed wind turbines, in which the weighting factor is dynamically adjusted by the fuzzy-deduced strategy. Since the performance of model predictive yaw control is influenced by the coming wind direction, the variation degree and the increment of the wind direction during the predictive horizon have been utilized as two inputs of the fuzzy-deduced strategy. In this way, the weighting factor is varying and driven by the two inputs and the fuzzy-deduced strategy. Through some simulations using real wind data, it has been shown that the proposed method has improved the overall performance of the predictive control-based yaw system under different yaw rate. In the future study, the proposed method could be potentially applied to the model predictive controls for the pitch control and torque control systems of the wind turbines.