风力发电机协调控制：模型预测控制约束处理方法

需积分: 9 88 浏览量更新于2024-09-13 收藏 1.17MB PDF 举报

"这篇论文详细探讨了风力发电机的控制策略，特别关注了在约束条件下的控制方法。文章提出了一个基于模型预测控制（Model Predictive Control, MPC）的风力发电机控制器，该控制器能够处理硬性约束和软性约束，如执行器限制，甚至在极端事件或执行器尺寸不足的情况下也能有效工作。研究通过与传统的比例积分控制器（Proportional-Integral Controller, PI Controller）带防积分饱和功能的性能对比，展示了新控制器在应对极端操作条件，如突然风速变化时的优势。此外，还分析了新控制器在各种工况下的运行能力，证明了其在维持风力发电机稳定性和效率方面的优越性。" 这篇论文发表在IET控制理论与应用期刊上，由来自丹麦技术大学风能部门的L.C.Henriksen、M.H.Hansen和信息与数学建模部门的N.K.Poulsen共同撰写。研究者们提出了一种新的风力发电机控制策略，旨在解决在风力发电系统中常见的约束问题，比如执行器限制和其他部件的性能边界。模型预测控制作为一种适合处理约束的控制方法，被选为控制器的基础。文章的重点在于，当风力发电机遭遇极端天气事件或执行器性能不足时，如何有效地管理和优化控制策略。通过对比模型预测控制与传统的比例积分控制器，特别是在防积分饱和功能下的表现，论文展示了前者在应对突发风速变化时的快速响应和动态性能。这一对比实验进一步证实了模型预测控制在处理约束和避免系统过载方面的优势。此外，论文还详细讨论了新控制器在不同工作状态下的表现，包括正常操作条件和异常工况，以全面评估其在保障风力发电机稳定运行和高效能量转换方面的效果。通过对这些复杂场景的分析，研究结果表明，基于模型预测控制的新型控制器能够提供更优的控制性能，对提升风能利用效率和保护设备安全具有重要意义。这篇研究对于理解和改进风力发电系统的控制策略，尤其是如何在约束条件下优化运行，提供了重要的理论和技术支持。

www.ietdl.org

the tower, rotor and nacelle equivalent mass and K

and D

denotes the tower spring and damping constants.

The actuators are assumed linear under the assumption

that a low-level controller, for example, proportional–

integral–differential (PID) or some type of nonlinear

controller, for example [15], is operating in closed

loop with the actuator mechanics. The pitch actuator

dynamics, including local low-level controller, are described

by a second-order ordinary differential equation, an

approximation which under the proper conditions can be

justiﬁed [16]

θ + 2ζ

θ + ω

θ = ω

ref

(13a )

subject to



min



≤





≤



max



(13b)

where ω

and ζ

are the natural frequency and damping

ratio of the actuator and θ

ref

is the reference signal from

the controller. The generator torque actuator is described by

a ﬁrst-order ordinary differential equation [17]

+ τ

−1

= τ

−1

ref

(14a )

subject to



g,min



≤





≤



g,max



(14b)

where τ

is the time constant of the generator torque actuator

and Q

ref

is the reference signal from the controller.

The components can be gathered into a nonlinear model

composed of a nonlinear ordinary differential equation

f : R

n+n

→ R

, a state noise matrix G ∈ R

n×n

and a

measurement equation g : R

n+n

→ R

˙x(t) = f (x(t), u(t)) + Gw(t) (15a )

y(t) = g(x(t), u(t)) + v(t) (15b)

that describes the relationship between the state vector x ∈

, the input vector u ∈ R

, the output vector y ∈ R

and the state and measurement noise vectors w ∈ R

and

v ∈ R

, respectively, of the complete design model. These

vectors contain the following variables

x =[



φψ

eff

θ Q

]

u =[θ

ref

g,ref

]

w =[e]

y =[



θ Q

]

The nonlinear model can be linearised using ﬁrst-

order Taylor series approximation where δ denotes small

variations away from the linearisation point, which in this

case is equivalent with equilibrium points for a given

wind speed. The linearised model is time-discretised by the

zero-order-hold method [18] into the form

δx

k+1

= Aδx

+ Bδu

+ Gw

, w ∈ N (0, R

) (16a )

δy

= Cδx

+ Dδu

+ v

, v ∈ N (0, R

) (16b)

where the subscript k refers to the discrete points in time t

The state, input and output vectors are related in a linear way

via the matrices A

∈ R

×n

, B ∈ R

×n

, C ∈ R

×n

and D ∈

×n

. Note the small variations symbol δ has been omitted

from the remainder of this paper to simplify notation.

3 Hybrid controller setup

The term hybrid controller relates to the fact that four

controllers, named K

–K

, are active under different

operating conditions governed by a switching mechanism.

The primary static objective of the wind turbine control

system is to optimise power production for the given wind

speed. Load reduction is a typical secondary objective,

which shall however not be discussed further in this section.

The primary objective for a given wind speed can be

formulated as the constrained minimisation of nonlinear

quadratic cost function concerning generator power. An

additional term concerning generator speed is included in

the cost function, this term is only active when the ﬁrst

term concerning power is zero. The steady state constrained

optimisation problem for a given wind speed is

min



,θ

− P

nom

)

+ w (

− 

g,nom

)

(17a )

where

w =



0, for P

= P

nom

1, for P

= P

nom

subject to

0 = f (x, u) (17b)



∈[

g,low

, 

g,nom

] (17c)

θ ∈[θ

min

, θ

max

] (17d)

where the equality constraint (17b) ensures steady state

operation. The generator speed and pitch angle are limited to

certain ranges (17c) and (17d). The terms min/max indicate

constraints to be respected, whereas low/nom indicate lower

and upper set points that should be tracked. As explained

later in this section the actual set point depends on the given

wind speed: for low wind speed the low set point is tracked,

for mid-range wind speeds the set point is between ‘low’ and

‘nom’ and for high wind speeds the set point is ‘nom’. Set

points are not respected in the same way as constraints and

for example generator speed is allowed to vary around its

given set point, whereas for example, the pitch angle cannot

go beyond its min/max limits. The distinction between

min/max and low/nom is maintained throughout this paper.

There are also other constraints such as pitch rate and

acceleration limits etc. but these velocity constraints are not

active during the present steady state optimisation sweep,

as velocities and accelerations are zero at steady state.

Above rated wind speeds, where the generator power is

at its nominal value, there is no unique solution for the

optimisation problem. The pitch angle and generator speed

could have any number of solutions as long as they are on

the appropriate contour line of the C

curve. The generator

speed weight w ensures that for above rated wind speed

there is a unique solution with generator speed at its nominal

value.

The optimisation gives the characteristic diagrams for

generator power, generator speed and pitch angle against

wind speed as seen in Fig. 1. Four regions of operation

O ∈ (I, ...,IV) also denoted (O

, ..., O

) respectively, are

derived from (17). The regions and their respective primary

control objectives are as follows:

• O

: Optimise power. Generator speed at lower level. Pitch

angle ﬁxed at its minimum value.

1724 IET Control Theory Appl., 2012, Vol. 6, Iss. 11, pp. 1722–1734

剩余12页未读，继续阅读

zhw0704

粉丝: 0
资源: 1

风力发电机协调控制：模型预测控制约束处理方法

Wind Turbine Control Systems_turbine_control_windturbine_matlab_

Wind Turbine Control Algorithms (DOWEC)

Wind Turbine Control Systems

Wind Turbine Control Toolbox for Matlab

Control analysis of the wind turbine with transmission to split power

wind turbine control algorithms

WIND_TURBINE2.rar_control_wind_wind turbine

Wind_Turbine.rar_matlab wind turbine_wind turbine模块_wind_turbine

Wind_Turbine_modeling.rar_TURBINE CONTROL_mid trubine_wind_wind

WT.rar_wind_wind turbine_wind turbine matlab_wind_turbine_wt

最新资源