Burton T. (et. al.) Wind energy Handbook

Подождите немного. Документ загружается.

turbine itself (stall hysteresis might be one example), in the driving disturbances

(the wind characteristics are not stationary), or in the controller objectives. For

example, the controller objectives might change around rated wind speed, or non-

linear effects such as fatigue damage might be included in the cost function.

8.5 Pitch Actuators (See also Section 6.7.2)

An important part of the control system of a pitch-controlled turbine is the pitch

actuation system. Both hydraulic and electric actuators are commonly used, each

type having its own particular advantages and disadvantages which should be

considered at the design stage.

Smaller machines often have a single pitch actuator to control all the blades

simultaneously, although there is an increasing trend to use individual pitch

actuators for each blade on larger turbines. This has the advantage that it is then

possible to dispense with the large and ex pensive shaft brake which would other-

wise be needed. This is because of the requ irement for a turbine to have at least two

independent braking systems capable of bringing the turbine from full load to a

safe state in the event of a failure . Provided the individual pitch actuators can be

made independently fail-safe, and as long as the aerodynamic braking torque is

always sufﬁcient to slow the rotor down to a safe speed even if one pitch actu ator

has failed at the working pitch angle, then multiple actuators may be considered to

be independent braking systems for this purpose. There may still be a need for a

parking brake, at least for the use of maintenance crews, but this may then be fairly

small. It must at least be capable of bringing the rotor to a complete stop from a low

speed, not necessarily in high or extreme wind speeds , for long enough to allow a

rotor lock to be inserted.

A collective pitch actuation system commonly consists of an electric or hydraulic

actuator in the nacelle, driving a push-rod which passes through the centre of the

gearbox and hollow main shaft. The push -rod is attac hed to the pitchable blade

roots through mechanical linkages in the hub. The actuator in the nacelle is often a

simple hydraulic cylinder and piston. A charged hydraulic accumulator ensures

that the blades can always be feathered even if the hydraulic pump loses power. An

alternative arrangement is to use an electric servo motor to drive a ballnut which

engages with a ballscrew on the push -rod. Since the push-rod turns with the rotor,

loss of power to the motor causes the ballscrew to wind the pitch to feather, giving

fail-safe pitch action. This requires a fail-safe brake on servo motor to ensure that

the ballnut stops turning if power is lost.

Individual pitch control requires separ ate actuators in the hub for each blade.

Therefore there must be some means of transmitting power to the rotating hub to

drive the actuators. This can be achieved by means of slip rings in the case of

electric actuators, or a rotary hydraulic joint for hydraulic actuators. A rotary

transformer can also be used to transmit electrical power to the hub without the

inconvenience of slip rings, which require maintenance.

The need to ensure a backup power supply on the hub to enable the blades to

pitch even in the event of power loss can be a source of problems. A hydraulic

PITCH ACTUATORS 505

system needs an accumulator for each blade, while electric actuators usually have

battery packs in the hub for this purpose. Such battery packs are large, heavy and

expensive, and alternative methods such as the use of hub-mounted generators,

which can always generate power as long as the hub is turning, have been

proposed. If a battery is used, the actuator motors must either be DC motors or

(more commonly) AC motors with a frequency converter, with the batteries on the

DC link. Since this will form part of the safety system, the reliability of the inverter

between the DC link and the pitch motor is important. A hub-mounted generator

would produce either DC or variable frequency AC, and once again the reliability

of the connection to the pitch motor is important. Since the pitch actuator s have to

be independently fail-safe, separate battery packs or generators and frequency

converters etc. must be provided for each blade.

The friction in the pitch bearing is often a signiﬁcant factor in the design of the

pitch actuation system. The bearing friction depends on the loading applied to the

bearing, and the large overturning moment acting on the bearing can lead to very

high levels of friction.

A hydraulic actuator would usually be controlled by means of a proportional

valve controlling the ﬂow of oil to the actuator cylinder. The valve openin g, and

hence the oil ﬂow rate, would be set in proportion to the required pitch rate. The

demanded pitch rate may come directly from the turbine controller, or it might

come from a pitch-position feedback loop. In this case the turbine controller gener-

ates a pitch-position demand. This is co mpared to the measured pitch position, and

the pitch-position error is turned into a pitch-rate demand through a fast PI or PID

control loop, implemented either digitally or by means of a simple analogue circuit.

In the case of an electric actuator, the motor controller usually requires a torque

demand signal. This may be derived from a speed controller, which uses a fast PI or

PID controller acting on speed error to generate a torque demand. Once again the

speed demand may come directly from the turbine controller or from a position

feedback loop.

Simpler actuators are sometimes used, particularl y when a fast pitch response is

not so important, as in the case of a turbine which is controlled by pitching to stall

rather than to feather. In this case an actuator which merely pitches at a ﬁxed rate in

either direction may be adequate.

8.6 Control System Implementation

Previous sections have explained some of the techniques whereby control algo-

rithms can be designed. The system and controller dynamics have been described

in continuous time in terms of the Laplace operator, s. While it is possible to

implement a continuous-time controller, for example using analogue circuitry, the

use of digital contr ollers is now almost universal. The greater ﬂexibility of digital

systems is a factor here: simply by making software changes, the control logi c can

be changed completely.

A consequence of using digital control is that the control actions are calculated

and updated on a discrete time step, rather than in continuous time. Control

506 THE CONTROLLER

algorithms designed in continuous time must therefore be converted to discrete

time for implementation in a digital controller. It is also possible to design

controllers in discrete time, if the linearized model of the turbine is ﬁrst discretized.

The following sections brieﬂy describe some of the practical issues involved in

implementing a control algor ithm in a real digital controller. Once again, the reader

is referred to standard contro l theory texts for more detailed treatments.

8.6.1 Discretization

Supposing a control algorithm has been designed in continuous time as a transfer

function (such as Equation (8.1) for a PID controller, for example), it must be

discretized before it can be implemented in a digital controller. Discretized transfer

functions are usually represented in terms of the delay operator, z, where z

 k

represents the value of x sampled k timesteps ago. As a simple example, a moving

average or lag ﬁlter from x to y is often implemented as

¼ Fy

k1

þ (1  F)x

This is a difference equation which can readily be implemented in code in a discrete

controller. In terms of the delay operator, it can be written as

(1  Fz

1

)y ¼ (1  F)x

or alternatively as a transfer function consisting of a ratio of polynomials in z

1

y ¼

(1  F)

(1  Fz

1

)

Now the Laplace operator can be considered as a differentia tion operator, and so

as a simple appr oximation, it might be possi ble to convert a continuous transfer

function into discrete form by replacing s by (1  z

1

)=T, where T is the timestep.

In fact by simple algebraic manipulation, it is straight forward to show that with this

substitution, the above discrete transfer function is in fact equivalent to the contin-

uous tran sfer function representation of a ﬁrst-order lag with time constant , namely

y ¼

1 þ s

with the factor F being given by =(T þ ).

Clearly any discretized equation can only be an approximation to the continuous-

time behaviour. There are other discretization methods, and the so-called bilinear

or ‘Tustin’ approximation often works better in practice. In this case the Laplac e

operator is replaced by

(1  z

1

)

(1 þ z

1

)

CONTROL SYSTEM IMPLEMENTATION 507

Discretization results in a phase shift compared to the continuous-time process.

This phase shift increases with frequency. If the algorithm performance is particu-

larly sensitive to the phase shift at a certain frequency, then the discretization can

be ‘pre-warped’ to this frequency. Pre-warping modiﬁes the phase shift so that the

phase of the discrete transfer function is correct at the chosen frequency, but

deviates at lower and higher frequencies. An example of a situation where pre-

warping may be important is in the case of a drive train resonance damper in a

variable-speed turbine (Section 8.3.6). The resonant frequency which is being tar-

geted is usually fairly high, typically around 4 Hz, and unless the controller

timestep is very short the phase lag caused by discretization may signiﬁcantly affect

the performance of the damping algorithm.

The approximation for s used for discretization with pre-warping about a

frequency ø is

tan(øT=2)

(1  z

1

)

(1 þ z

1

)

8.6.2 Integrator desaturation

Controllers containing integral terms, such as PI or PID controllers, experience a

particular problem known as integrator wind-up when the control action saturates

at a limiting value. A common example is in pitch control, where the pitch angle is

limited to the ﬁne pitch position when the wind is below rated. For example, a PI

power controller for a ﬁxed-speed turbine can be represented as:

Power error

Pitch

demand

Limits

Above rated, the power error will be zero on average because of the integral term.

Below rated, the pitch saturates at the ﬁne pitch position, and the power error will

remain negative. The integral of the error will therefore grow more and more

negative, and only the application of the limits prevents the actual pitch demand

from doing the same. However, when the wind speed reaches rated again and the

power error becomes positive, it will take a long time before the integrated power

error climbs back up to zero and start s to dem and a positive pitch angle. To prevent

this problem of integrator wind-up, the integral term must be prevented from

integrating when the pitch is at the limit. This is easily implemented by separating

out the integrator, I(z), from the rest of the controller, R(z). R(z) generates a change

in demanded pitch angle, and I(z) then integrates this by adding it to the previous

pitch demand after the limits have been applied.

As an example, a PI co ntroller (Equation 8.20) discretized using the bilinear

approximation would be:

508 THE CONTROLLER

y ¼ K

[(T=2T

þ 1) þ (T=2T

 1)z

1

]

[1  z

1

]

x ¼ R(z)I(z)x

To avoid integrator wind-up, this can be implemented as follows:

˜ y

¼ K

[(T=2T

þ 1)x

þ (T= 2T

 1)x

k1

] (implementation of R(z))



¼ y

k1

þ ˜ y

(integrator I (z) using previous limited

output)

¼ lim(y



) (application of limits)

Higher-order controllers may need more sophisticated desaturation techniques.

References

Anderson, B. D. O. and More, J. B., (1979). Optimal ﬁltering. Prentice Hall, New York, USA.

Astrom, K. J. and Wittenmark, B., (1990). Computer controlled systems. Prentice-Hall, New

York, USA.

Bossanyi, E. A., (1987). ‘Adaptive pitch control for a 250 kW wind-turbine’. Proceedings of the

Ninth BWEA Conference. Mechanical Engineering Publications, Bury St Edmunds, UK.

Bossanyi, E. A. and Gamble, C. R., (1991). ‘Investigation of torque control using a variable

slip induction generator’. ETSU WN-6018, Energy Technology Support Unit, Harwell, UK.

Bossanyi, E. A., (1994). ‘Electrical aspects of variable-speed operation of horizontal axis wind

turbine generators’. ETSU W/33/00221/REP, Energy Technology Support Unit, Harwell,

UK.

Bossanyi, E. A., (2000). ‘Developments in closed-loop controller design for wind-turbines’.

Proceedings of the ASME Wind Energy Symposium, AIAA/ASME.

Caselitz, P. et al., (1997). ‘Reduction of fatigue loads on wind-energy converters by advanced

control methods‘. Proceedings of the European Wind Energy Conference, pp. 555–558.

Clarke. D. and Gawthrop, P., (1975) ‘Self-tuning controller’. Proc. IEE, 122, 9, 929–934.

D’Azzo, J. J. and Houpis, C. H., (1981). Linear control system analysis and design. McGraw-Hill,

New York, USA.

Holley, W., Rock, S. and Chaney, K., (1999). ‘Control of variable speed wind turbines below

rated wind speed’. Proceedings of the Third ASME/JSME Conference (FEDSM99-5295–16).

Knudsen, T., Andersen, P. and To

ffner-Clausen, S., (1997). ‘Comparing PI and robust pitch

controllers on a 400 kW wind turbine by full-scale tests’. Proceedings of the European Wind

Energy Conference, pp. 546–550.

Pedersen, T. K., (1995). Semi-variable speed – a compromise? Proceedings of the Wind Energy

Conversion 17

British Wind Energy Association Conference , pp. 249–260. Mechanical En-

gineering Publications, Bury St Edmunds, UK.

Steinbuch, M., (1989). Dynamic modelling and robust control of a wind energy conversion system.

Ph.D. Thesis, University of Delft, Holland.

CONTROL SYSTEM IMPLEMENTATION 509

Wind Turbine Installations and

Wind Farms

For any wind turbine installation, there are certain additional activities (e.g.,

construction of foundations and access roads, electrical connections, site erection, as

well as project development and management) that must be undertaken. For ﬂat

onshore sites, which might be found typically in Denmark or North Germany, the

total inve stment cost is approximately 1.3 times the ex-works turbine cost (EUREC

Agency, 1996). In the UK, where sites are often located in more remote, upland

areas the balance-of-plant costs (i.e., all costs other than the wind turbines) tend to

be higher and a more typical breakdown is shown in Table 9.1. Commer cial

developers of wind farms will often prefer larger projects as, in that way, the ﬁxed-

costs, particularly electrical network connection and project development and

management costs, may be spread over a bigger investment. A further encourage-

ment for large projects is that the ﬁxed costs of arranging project ﬁnance are high.

However, there are individuals, community groups and commercial organizations

who develop smaller wind farms and projects of this type are common in Denmark

and Germany. The advantage of community involvement in the project is, of course,

that planning permission is more readily obtained if it is seen that there is tangible

beneﬁt to local people.

9.1 Project Development

The development of a wind farm follows a broadly similar process to that of any

other power generation project, but with the particular requirements that the wind

Table 9.1 Typical Breakdown of Costs for a 10 MW Wind

Farm

Element of wind farm % of total cost

Wind turbines 65

Civil works (including foundations) 13

Wind farm electrical infrastructure 8

Electrical network connection 6

Project development and management costs 8

turbines must be located in high wind speed sites to maximize energy production

and their size makes visual effects a particularly important aspect of the environ-

mental impact. Guidance on the development of wind energy schemes has been

issued by the British Wind Energy Association (BWEA, 1994) in their Best Practice

Guidelines for Wind Energy Development. A similar docum ent is published by the

European Wind Energy Association (EWEA). The three main elements of the

development of a wind-farm project are identiﬁed as: (1) technical and commercial

issues, (2) environmental consid erations, and (3) dialogue and consultation. Per-

haps surprisingly to many engineers and technologists the technical and commer-

cial considerations are often the more straightforward and project success or failure

hinges critically on environmental consid erations and the success of the dialogue

and consultation process with local inhabitants and planning authorities.

Wind-farm development may be divided into a number of phases:

• initial site selection,

• project feasibility assessment (including obtaining a power purchase contract for

the wind farm output)

• preparation and submission of the planning applicati on,

• construction,

• operation, and

• decommissioning and land reinstatement.

9.1.1 Initial site selection

Initially a desk-based study is carried out to locate a suitable site and to co nﬁrm it

as a potential candidate for the location of a wind farm. It may be recalled (see

section 4.5) that the mean power production for a wind turbine (assuming 100

percent availability) is given by:

E ¼ T

P(U) f (U)dU (9:1)

where P(U) is the power curve of the wind turbine, f (U) is the probability density

function (PDF) of the wind speed, and T is the time period.

The power curve is available from the potential turbine supplier(s) while an

initial estimate of the PDF of the wind speed may be obtained from a wind atlas

(European Wind Atlas, 1989). The PDF is generally based on a Weibull distribution

and takes account of regional climatology, roughness of the surrounding terrain,

local obstacles and topology. PDFs are calculated for 12 308 sectors and integrated

with the power curve usually using numerical evaluation techniques. At this stage

of the project development only an approximate indication of the wind-farm output

is required in order to conﬁrm the potential of the site(s). However, it may be useful

to supplement the wind atlas with some form of initial computer modelling.

512 WIND TURBINE INSTALLATIONS AND WIND FARMS

In the UK a wind-speed database exists based on the NOABL airﬂow model

(ETSU, 1999). The model represents the effect of to pography on wind speed and

provides estimates of annual mean wind speeds with a resolution of 1-km square at

either 10 m, 25 m or 45 m above ground level. It does not take account of local

thermally driven wind or local effects. These can have a considerable effect on wind

speed and so on-site measurements are required for an accurate assessment.

If measured site wind speed data is available then the energy yield of a wind

turbine can be estimated as shown in Figure 9.1. by combining the binned wind

speed distribution with the power curve:

Energy ¼

i¼n

i¼1

H(U

)P(U

)(9:2)

where H(U

) is the number of hours in wind speed bin U

, P(U

) is the power output

at that wind speed and there are n wind speed bins.

In addition to assessing the wind resource it is also necessary to conﬁrm that road

access is available, or can be developed at reasonable cost, for transporting the

turbines and other equipment. Blades of large wind turbines can be up to 40 m in

length and so clearly can pose difﬁculties for transport on minor roads. For a large

wind farm, the heaviest piece of equipment is likely to be the main transform er if a

substation is located at the site.

The local electricity utility should be able to provide information on the amount

of generation which the distribution network can accept although for a ﬁrst

approximation it may be useful to consider rules-of-thumb as indicated in Table

10.4. Such rules-of-thumb give approximate information only but do serve to

highlight those sites where the electrical connection will require long extensions of

the power network with associated high cost and environmental impact.

The initial technical assessment will be accompanied by a review of the main

environmental considerations. The most important constraints include special

consideration of National Parks or other areas designated as being of particular

100

200

300

400

500

600

700

800

900

1000

1100

1200

1300

123456789101112131415161718192021

W ind speed (m/s)

Hours per year

100

150

200

250

300

350

400

450

500

550

600

650

700

750

1 2 3 4 5 6 7 8 9 101112131415161718192021

Wind speed (m/s)

Output (kW)

Wind speed

Power Curve

Figure 9.1 Annual Energy Calculation of a Wind Turbine

PROJECT DEVELOPMENT 513

amenity value and ensuring that no turbine is located so close to domestic dwell-

ings that a nui sance will be caused (e.g., by noise, visual domination or light

shadow ﬂicker). A preliminary assessment of visual effects is also required consid-

ering the visibility of the wind farm particularly from important public viewpoints.

If, within the wind-farm perimeter, there are areas of particular ecological va lue,

due to ﬂora or fauna, then these need to be avoided as well as any locations of

particular archaeological or historical interest. Communication systems (e.g., micro-

wave links, TV, radar or radio) may be adversely effected by wind turbines and this

needs to be considered at an early stage.

In parallel with the technical and environmental assessments it is normal to open

discussion with the local civic or planning authorities to identify and ag ree the

major potential issues which will need to be addressed in more detail if the project

development is to continue.

9.1.2 Project feasibility assessment

Once a potential site has been identiﬁed then more detailed, and expensive,

investigations are required in order to conﬁrm the feasibility of the project. The

wind farm energy output, and hence the ﬁnanc ial viability of the scheme, will be

very sensitive to the wind speed seen by the turbines over the life of the project.

Hence it is not generally considered acceptable in complex terrain to rely on the

estimates of wind speed made during the initial site selection but to use the

measure–correlate–predict (MCP) technique to establish a prediction of the long-

term wind resource (Derrick, 1993, Mortimer, 1994).

9.1.3 The Measure–correlate–predict technique

The MCP approach is based on taking a series of measurements of wind speed at

the wind farm site and correlating them with simultaneous wind speed measure-

ments made at a meteorological station. The averaging period of the site-measured

data is chosen to be the same as that of the meteorological station data. In its

simplest implementation, linear regressi on is used to establish a relationship be-

tween the measured site wind speed and the long-term meteorological wind speed

data of the form:

site

¼ a þ bU

long-term

(9:1:3)

Coefﬁcients are calculated for the 12 308 directional sectors and the correction for

the site applied to the long-term data record of the meteorological station. This

allows the long-term wind speed record held by the meteorological station to be

used to estimate what the wind speed at the wind-farm site would have been over

the last, say, 20 years. It is then assumed that the site long-term wind speed is

represented by this estimate which is used as a prediction of the wind speed during

the life of the project. Thus, MCP requires the installation of a mast at the wind farm

site on whic h are mounted anemom eters (usually cup anemometers) and a wind

514 WIND TURBINE INSTALLATIONS AND WIND FARMS