Renewable and Sustainable Energy Reviews 30 (2014) 452460

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

A comprehensive review on wind turbine power curve

modeling techniques
M. Lydia a,n, S. Suresh Kumar b,1, A. Immanuel Selvakumar a,2, G. Edwin Prem Kumar c,3
a
Department of Electrical & Electronics Engineering, Karunya University, Coimbatore 641 114, Tamil Nadu, India
b
Department of Electronics & Communication Engineering, Dr. NGP Institute of Technology, Coimbatore 641 048, Tamil Nadu, India
c
Department of Information Technology, Karunya University, Coimbatore 641 114, Tamil Nadu, India

art ic l e i nf o a b s t r a c t

Article history: The wind turbine power curve shows the relationship between the wind turbine power and hub height
Received 25 May 2013 wind speed. It essentially captures the wind turbine performance. Hence it plays an important role in
Received in revised form condition monitoring and control of wind turbines. Power curves made available by the manufacturers
27 August 2013
help in estimating the wind energy potential in a candidate site. Accurate models of power curve serve as
Accepted 21 October 2013
an important tool in wind power forecasting and aid in wind farm expansion. This paper presents an
Available online 14 November 2013
exhaustive overview on the need for modeling of wind turbine power curves and the different
Keywords: methodologies employed for the same. It also reviews in detail the parametric and non-parametric
Modeling accuracy modeling techniques and critically evaluates them. The areas of further research have also been
Non-parametric modeling
presented.
Parametric modeling
& 2013 Elsevier Ltd. All rights reserved.
Wind turbine power curve

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
2. IEC power curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
3. Power curve modeling requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
3.1. Modeling objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
3.1.1. Wind energy assessment and prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
3.1.2. Choice of wind turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
3.1.3. Monitoring and troubleshooting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
3.1.4. Predictive control and optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
3.2. Modeling data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
3.2.1. Statistical analysis of wind data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
3.2.2. Factors affecting power curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
3.3. Modeling accuracy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
4. Power curve modeling methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
4.1. Parametric techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
4.1.1. Linearized segmented model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
4.1.2. Polynomial power curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
4.1.3. Maximum principle method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
4.1.4. Dynamical power curve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
4.1.5. Probabilistic model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
4.1.6. Ideal power curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
4.1.7. Four parameter logistic function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458

n
Corresponding author. Tel.: 91 9443445047.
E-mail addresses: [email protected] (M. Lydia), [email protected] (S.S. Kumar), [email protected] (A.I. Selvakumar), [email protected]
(G.E. Prem Kumar).
1
Tel.: 91 9442514130.
2
Tel.: 91 9994534647.
3
Tel.: 91 9443929655.

1364-0321/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.rser.2013.10.030
 M. Lydia et al. / Renewable and Sustainable Energy Reviews 30 (2014) 452460 453

4.1.8. Five parameter logistic function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458

4.2. Non-parametric techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
4.2.1. Copula power curve model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
4.2.2. Cubic spline interpolation technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
4.2.3. Neural networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
4.2.4. Fuzzy methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
4.2.5. Data mining algorithms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
4.3. Analysis of wind turbine power curve modeling techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
5. Inferences and future scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460

1. Introduction components. The conversion of power in the wind into actual

power varies non-linearly because of the transfer functions of
The signicance of alternate energy sources like solar, wind, available generators [4].
biomass etc. has exponentially increased in the recent times due to The theoretical power captured by the rotor of a wind turbine
the ever increasing demand for clean energy. Harvesting of wind (P) is given by
energy has hence gathered sufcient momentum in the recent days.
P 0:5R2 C p u3 1
Estimation of the technical and economic wind energy potential in
various regions have gathered momentum [1]. However, the large where is the air density, R is the radius of the rotor, Cp is the power
scale integration and penetration of wind energy into the power grid coefcient and u is the wind speed [5]. As the air density remains
can result in signicant social, environmental, economical and almost constant at hub height, the power captured signicantly
technical impacts [2]. In order to develop a sustainable power system depends on the power coefcient and wind speed. The power
for the future, these impacts need to investigated and mitigated. coefcient, which denotes the percentage of power captured by the
Wind energy denitely holds out a promising respite, but for turbine, essentially depends on the tip speed ratio () and the
the uncertainty involved in power produced due to the stochastic blade-pitch angle.
nature of wind. A signicant penetration of wind in the present
day power system can be realized only if accurate and reliable
forecasting models are made available. A wind turbine power 2. IEC power curve
curve can go a long way in fullling this. Fig. 1 shows a wind
turbine power curve (WTPC). The International Standard IEC 61400-12-1 has been prepared
The output power of a wind turbine signicantly varies with by the International Electrotechnical Commission (IEC) technical
wind speed and hence every wind turbine has a very unique committee 88: Wind turbines. The standard methodology for
power performance curve. A power curve aids in wind energy measuring the power performance characteristics of a single wind
prediction without the technical details of the components of the turbine has been specied here. It is also applicable for testing the
wind turbine generating system [3]. The electrical power output as performance of wind turbines of varied sizes and types. It can be
a function of the hub height wind speed is captured by the power used to evaluate the performance of specic turbines at specic
curve. The minimum speed at which the turbine delivers useful locations and also aid in comparing the performance of different
power is known as the cut-in speed (uc). Rated speed (ur) is the turbine models or settings [6].
wind speed at which the rated power, which is the maximum The power performance characteristics of wind turbines are
output power of the electrical generator, is obtained. The cut-out ascertained by the measured power curve and the estimated annual
speed (us) is usually limited by engineering design and safety energy production. Simultaneous measurements of wind speed and
constraints. It is the maximum wind speed at which the turbine is power output is made at a test site for sufciently long duration to
allowed to produce power. Power curves for existing machines, create a signicant database under varying atmospheric conditions.
derived using eld tests, can be obtained from the wind turbine The measured power curve is determined from this database. The
manufacturers. The approximate shape of the power curve for a annual energy production is calculated, assuming 100% availability,
given machine can also be estimated using the power character- by applying the measured power curve to reference wind speed
istics of rotor, generator, gearbox ratio and efciencies of various frequency distributions.
The measured power curve is determined by applying the
method of bins, for the normalized datasets using the following
equations:
1 Ni
ui u 2
Ni j 1 n;i;j

1 Ni
Pi P 3
N i j 1 n;i;j
where ui is the normalized and averaged wind speed in bin i, un,i,j
is the normalized wind speed of dataset j in bin i, Pi is the
normalized and averaged power output in bin i, Pn,i,j is the
normalized power output of dataset j in bin i and Ni is the number
of 10 min data sets in bin i. The accuracy of WTPC models have
improved by using the prole information available using remote
Fig. 1. Typical wind turbine power curve. sensing instruments [7]. However, it has been stated in [8], that
454 M. Lydia et al. / Renewable and Sustainable Energy Reviews 30 (2014) 452460

the IEC-based power curve gives the behavior of the wind turbine
with the inuence of site turbulence. Though the current site data
is rendered with reliable accuracy, the IEC power curve contains
the hidden effect of current site turbulence, in such a way that
its blind application to other sites is not very correct. The IEC
procedure also ignores the fast wind uctuations through the
10 min averaging and the results in obtaining the behavior of the
machine independent of wind uctuations. Hence the need for
modeling site-specic WTPC has gained great signicance.

3. Power curve modeling requirement

The manufacturer provided power curve for any turbine gives the
relationship between wind speed and power at a particular air
density. But this curve is neither site-specic nor does it take into
account the wear and tear of the turbine. It was observed that there
were notable discrepancies between the small wind turbine manu-
facturer power curves and the test results carried out at high wind
speeds [9]. The WTPC is not an adequate model for estimation of
power of variable speed wind turbines as it ignores the dynamic
behavior of wind [14]. Hence it is necessary that we model the wind
Fig. 2. WTPC modeling objectives.
turbine power curve, taking into account all these varying para-
meters. The main objective of modeling of a wind turbine power
curve, the statistical analysis of data that forms the basis of modeling wind dynamics and has resulted in more accurate power predic-
techniques and performance metrics that validates the modeling tion [14].
procedures are discussed in detail in this section.
3.1.2. Choice of wind turbines
3.1. Modeling objective WTPC models aid the wind farm developers to choose the
generators of their choice, which would provide optimum ef-
A WTPC built from the measured data in a particular site using ciency and improved performance. The impact of WTPC on the
better modeling techniques will denitely overcome the draw- cost of energy and optimal system conguration in a small wind
backs posed by the manufacturer provided power curve and the off-grid power system has been presented in [15]. Judicious choice
IEC power curve. A power curve built from the measured data of a wind turbine generator that yields higher energy at higher
deviates when some power outputs are negative implying wind capacity factor can be done by using the normalized power curves
turbine is consuming energy due to low wind speed and some proposed in [16]. These generalized curves, obtained from a new
power outputs vary even when the wind speed is constant. Hence ranking parameter known as wind turbine performance index, can
it is necessary that a power curve is modeled with minimum error. be used at the planning and development stages of wind power
The objective for modeling a WTPC is four fold: wind energy stations. The wind turbine capacity factor was modeled using the
assessment and prediction, choice of wind turbines, monitoring site wind speed and turbine power curve parameters in [17,18]. An
and troubleshooting and nally predictive control and optimiza- increase in energy yield up to 5% was obtained, when the
tion of wind turbine performance (Fig. 2). proposed model was used for optimum turbine-site matching.

3.1.1. Wind energy assessment and prediction 3.1.3. Monitoring and troubleshooting
Wind resource assessment is the process by which wind farm A WTPC model can serve as a very effective performance
developers estimate the future energy production of a wind farm. monitoring tool [19]. The model developed can be used as a reference
Accurate assessments are crucial to the successful development of for monitoring the performance of wind turbines. An equivalent
wind farms. The meteorological potential of any candidate site is steady state model of a wind farm under normal operating conditions
equivalent to the available wind resource [3]. If the wind speed has been built using data-driven approach and has been utilized for
data of the site is available, a WTPC can facilitate the estimation of creation of quality control charts, with the aim of detecting anom-
wind energy that can be produced over a period of time. Accurate alous functioning conditions of the wind farms [20]. Monitoring the
WTPC models also help in the planned expansion of wind farms performance of a wind farm using three different operational curves
[10]. An analytical method to estimate the output power variation has been presented in [21]. The WTPC has been used to identify
in a wind farm has been devised using dynamic power curves in various faults and its severity in [22]. The wind turbine power output
[11]. Estimating and controlling the variability of wind farm power has been evaluated and deviations that may result in nancial losses
output aids in providing stable wind power to the utility/grid are calculated using online monitoring of power curves [23]. The
and improves loss of load expectation (LOLE). Olaofe and Folly performance of four different data mining approaches has been
have concluded that the analysis of the energy outputs of the compared for this purpose.
wind turbines based on the developed site power curves is more
accurate than the turbine power curves [12]. The WTPC models 3.1.4. Predictive control and optimization
can very well be used for wind power forecasting at varying time Uluyol et al., showed that the WTPC can be very useful for
horizons [13]. Accurate forecasting of wind power in intra-day and performance assessment and for generation of robust indicators
day-ahead electricity markets are the need of the day. The power for component diagnostics and prognostics. Higher reliability and
curve of a variable speed wind turbine has been modied using a lower maintenance costs can be incurred by employing condition-
new curve called the controllers power curve to account for the based rather than hour-based monitoring [24]. A copula model
 M. Lydia et al. / Renewable and Sustainable Energy Reviews 30 (2014) 452460 455

of WTPC has been used for early identication and detection variation of wind speeds experienced by wind farms across a region.
of incipient faults such as blade degradation, yaw and pitch errors Availability and electrical efciency of offshore sites are generally
[25]. The copula-power curve condition monitoring correlates lower than onshore sites.
faults or anomalies to statistical signatures. Kusiak and Li have The effects of the environmental parameters on wind turbine
shown that power curve models along with data mining based power probability density function curve were studied in [27].
model extraction could be used to predict specic faults with an These parameters included the annual average wind speed,
accuracy of 60 min before they occur [22]. k-factor of Weibull distribution, autocorrelation factor, diurnal
pattern strength, altitude above sea level and variance of monthly
3.2. Modeling data averaged wind speed in one year. It was found out that the altitude
above sea level (which determines the air density indirectly) and
The data required for modeling a power curve is the wind the k-factor of Weibull distribution affected the wind turbine
speed and power output recorded at periodic intervals over a long output power more than all the remaining parameters.
time. The historical data could either be obtained from experi- Site-specic adjustments are required by wind turbine power
mental wind farms or from the Supervisory Control and Data curves in order to address the effects of turbulence, complex
Acquisition (SCADA) system. Once the required data is available, terrain, wind shear, blade fouling and icing, power curve measure-
the energy production of the wind turbine can be analyzed using ment blockage effects and uncertainty in availability of wind
four different approaches namely, direct use of data averaged over farms [28]. It was observed that for a site with 18% turbulence, a
a short time interval, the method of bins, development of velocity 1% reduction of energy took place. Hence for sites, where the
and power curves from data and statistical analysis using sum- predicted turbulence levels and wake effects are more than 15%, a
mary measures [3]. turbulence power curve adjustment factor should be applied. For
complex terrains, it was suggested that an up-ow power curve
3.2.1. Statistical analysis of wind data adjustment factor be applied. To account for the uncertainty in
The wind speed probability distribution describes the like- power curve and wind turbine availability, an allowance equiva-
lihood that certain values of wind speed will occur. The probability lent to 2% of the wind farm energy production has also been
distributions are generally characterized by probability density suggested. As the output power of wind turbine varies as the cube
function (f(u)) or a cumulative density function (F(u)) [3]. The two of the input wind speed, it is the variability in the wind speed that
commonly used probability distributions in wind data analysis affects the power curve most. If the annual mean wind speed
are the Rayleigh and Weibull distribution. Rayleigh distribution varies by 710%, it was observed that the corresponding variation
requires only the knowledge of mean wind speed (u) and hence is in available wind energy was about 725% [29].
the simplest velocity probability distribution. The f(u) and F(u) of
Rayleigh distribution is given below: 3.3. Modeling accuracy

  
u  u2
f u exp 4 The most important criteria to be addressed while formulating
2 u2 4 u
the various techniques for WTPC modeling, is the model accuracy.
 
 u2 Different performance metrics have been used by various
Fu 1  exp 5
4 u researchers. The most common metrics have been listed below
where Pe is the estimated power and Pa is the actual power and
N is the total number of data.
The f(u) and F(u) of Weibull distribution is given in Eqs. (6) and
In [5], absolute error (AE) and relative error (RE) are used to
(7).
evaluate the WTPC models.

 k  1   k 
k u u
f u exp  6 AE jP e i P a ij 8
c c c
 
  k  P e i P a i
u RE    100%
 9
Fu 1  exp  7 P a i
c
where k is the shape factor and c is the scale factor. Higher the
value of k, lesser is the observed wind speed variation. The wind The use of mean absolute error (MAE), symmetric mean
resource of the site under study has been assessed using Weibull's absolute percentage error (sMAPE) and normalized mean absolute
and Rayleigh's distribution in [12]. percentage error (NMAPE) as performance metrics has been
reported in [20].
3.2.2. Factors affecting power curves 1 N
MAE jP e i  P a ij 10
Wind farm power curves are adversely affected by the changing Ni1
environmental and topographical conditions. Equivalent power
curve models incorporating the effect of array efciency, high wind 1 N jP e i  P a ij
speed cut out, topographic effect, spatial averaging, availability and sMAPE  100 11
N i 1 jP e ij jP a ij=2
electrical losses have been built in [26]. The impact of wind speed
reduction due to the wakes created by the wind turbines upstream 1 N jP e i  P a ij
determines the array efciency. The main factors affecting array NMAPE  100 12
i 1 P a i
N i 1 maxN
efciency are wind farm layout, wind regime and the type of terrain.
Offshore wind farms are susceptible to a higher wake loss. The effect
of topography is higher in upland wind farms than the low land The mean absolute error (MAE) and root mean squared error
wind farms, because of the greater variation in wind speed. This can (RMSE) have been used as metrics in [13].
be reduced by averaging the power from a range of power curves at s
different wind speeds. An equivalent regional power curve is 1 N
RMSE P e i  P a i2 13
produced for each wind farm by averaging, in order to reduce the Ni1
456 M. Lydia et al. / Renewable and Sustainable Energy Reviews 30 (2014) 452460

The coefcient of determination R2 is used in [30] to ascertain

the accuracy of the developed models.
2
N
i 1 P e i  P a i
R2 1  2
14
N
i 1 P a i P a i

where P a i is the mean value of actual power.

4. Power curve modeling methodology

Fig. 4. Linearized segmented model.
A critical analysis of the various methods used for mathema-
tical modeling of wind turbines has been presented in [31]. The
two different kinds of models developed by them are models 4.1.1. Linearized segmented model
based on fundamental equation of power available in the wind and This is the simplest parametric model where piecewise approx-
models based on the concept of power curve of the turbine. It was imation of the WTPC has been carried out using the equation of a
concluded that models based on the equation of power were very straight line [13,31,38].
cumbersome. Models based on the power curve of the turbines
P mu c 16
gave fairly accurate results. The different techniques available in
literature for WTPC modeling have been classied into parametric where P is the output power and u is the wind speed, m is the
techniques and non-parametric techniques as shown in Fig. 3. slope of the segment and c is any constant (Fig. 4). The data is
tted on to the linear segments using the method of least squares,
which estimates the coefcients by minimizing the summed
4.1. Parametric techniques square of residuals. The residual of the ith data point ri is dened
as the difference between the actual power output Pa(i) and the
Parametric techniques are based on solving mathematical tted response value Pe(i), and is identied as the error associated
modeling expressions. The actual wind turbine generator power with the data. The summed square of residuals (S) is given by
output (Pa) can be expressed as given below:
N N
8 S r 2i P a i  P e i2 17
<0
> u o uc ; u 4 us
i1 i1
P a u pu uc r u r ur 15
>
:P
r ur r u r us
The least squares criterion assumes that the wind measures or
where u is the wind speed, uc is the cut-in speed, ur is the rated forecasts are error-free, which is never true in practice. This
speed and us is the cut-out speed, p(u) is the linear variable region problem could be overcome by using the total least square (TLS)
between the cut-in speed and rated speed and Pr is the rated power. criterion, in which the contribution of the noise components in
both power and meteorological variables are accounted for in the
model parameter estimation [32].

4.1.2. Polynomial power curve

A WTPC has been modeled using polynomial expressions of
varied orders in different literatures [33]. Seven different models
were used to model the linear region of the wind turbine power
curve in [34] and their energy output yields were calculated.
A review of four commonly used equations for representation of
power curves of variable speed wind turbines namely polynomial
power curve, exponential power curve, cubic power curve and
approximate cubic power curve has been done in [30]. All these
four equations have been used to model the linear region of the
WTPC.

1. Quadratic power curve

A second degree polynomial expression has been used for
modeling the WTPC in [30].

pu c1 c2 u c3 u2 18

where c1, c2 and c3 are constants determined from uc, us and Pr.
A WTPC based on the method of least squares, using quadratic
expressions for the linear region has been presented in [31].
Three different quadratic expressions have been used to
approximate the linear region guaranteeing better accuracy.
8
< c11 u c12 u c13 f or uc r u ru1
2
>
pu c21 u2 c22 u c23 f or u1 r u r u2 19
>
: c u2 c u c
31 32 33 f or u2 r u r us
Fig. 3. WTPC modeling techniques.
 M. Lydia et al. / Renewable and Sustainable Energy Reviews 30 (2014) 452460 457

where c11, c12, c13, c21, c22, c23, c31, c32, c33 are coefcients of the speed bin and k(j) denotes the power bin with
quadratic equation and u1 and u2 and wind speeds at heights h1 Nk : P i  P k ui  uj
(m) and h2 (m) respectively. i
2. Cubic power curve
and
A WTPC has been modeled using a cubic power expression in
[30]. Nkj Z N k 25
1 where x is a Heaviside function dened by
pu AC p;eq u3 20
2

1 if  =2 r x o =2 with the particular bin width

where Cp,eq is a constant equivalent to the power coefcient. x
0 else
Model for WTPC based on cubic law has also been presented in
26
[31]. Since the fraction of the wind power that gets converted
to electrical power depends on several parameters like wind
speed, rotational speed of the turbine, angle of attack, pitch However, it was proved that Rauh's method of maximum
angle, mechanical and electrical efciencies, the accuracy principle overestimated the points in the region of transition to
decreases. the rated power in the WTPC and the accuracy of the method was
3. Approximate cubic power curve also not good [35].
An approximate cubic power curve model has been derived by
assigning maximum value to the power coefcient (Cp,max) in [30]. 4.1.4. Dynamical power curve
1 Determination of WTPC through a dynamical approach has been
pu AC p; max u3 21 presented using the Langevin Model in [35]. The main objective of
2
this method is to separate the dynamics of the wind turbine power
output into two parts: a deterministic and a stochastic part. The
4. Exponential power curve
deterministic part corresponds to the actual behavior of the wind
The variable speed WTPC can be modeled using an exponential
turbine and the stochastic part corresponds to other external factors
equation as given below [30]:
such as the wind turbulence. The wind turbine power output is
1 described as a stochastic process that satises the Markovian
pu AK p u  uc 22
2 property and hence can be separated into a drift and a diffusion part.
where Kp and are constants. Pt P stat u pt 27
5. Ninth degree polynomial
The performance of polynomial models of fourth degree, where P(t) is the time series power data, Pstat is the stationary power
seventh degree and ninth degree has been compared using value dependent on the wind speed u and p(t) corresponds to short-
curve tting toolbox of MATLAB in [39]. It was observed that time uctuations around this value caused by wind turbulence. The
the ninth degree polynomial (Eq. 23) performed best. performance of the dynamical power curve was compared with the
IEC power curve and the maximum principle method according to
P a u c1 u9 c2 u8 c3 u7 c4 u6 c5 u5 c6 u4 c7 u3 c8 u2 c9 u c10 Rauh and was found to be much accurate.
23 The advantages of the dynamical power are that it could extract
where c1,,c10 are constants. the dynamical behaviour of any wind turbine with better accuracy
The shapes of the WTPC of different turbines with varied and produce machine-specic and site-independent results. Mea-
design ratings are different. Hence, the biggest disadvantage surements taken for a short-time is enough for this approach,
of the polynomial models is that there can never be a unique where as the IEC power curve procedure requires long-term data
set of generalized characteristic equations that can be used for and also averages out all the dynamics [36].
all types of turbines.
6. Model based on Weibull's parameters 4.1.5. Probabilistic model
A WTPC based on Weibull's parameters has been used in [31] A WTPC modeled using polynomial expressions is deterministic
but the accuracy of modeling was very poor. in nature, since the relationship with the output power and the
input wind speed is pronounced by the modeling expressions. Jin
pu a buk 24
and Tian [11] proposed a probabilistic model for WTPC as follows:
where a P r ukc =ukc  ukr and b P r =ukr  ukc
pu C p u3 28
It was observed that the model based on Weibull's parameters
lacked accuracy in the range of cut-in to rated speed since Eq. (24) In this model, the wind turbine output power is a random
did not accurately represent the wind turbine power curve shape number whose value is determined by u, the wind speed and , the
in that range [31]. variation of the power output. This model characterizes the
Among the polynomial based power curves, the quadratic dynamics of wind energy production and estimates the uncer-
power curve showed the worst results due to its sensitivity to tainty in wind power when the wind turbine generators operate in
the data given by the manufacturer and the approximate cubic the region between cut-in and rated wind speed. The wind turbine
power curve recorded a better performance [30]. However a power is assumed to follow the normal distribution with a varying
polynomial of higher degree recorded better performance in [39]. mean and constant standard deviation.

4.1.3. Maximum principle method 4.1.6. Ideal power curve

The maximum principle method proposed by Rauh denes an The ideal power curve, as proposed in [8], describes the intrinsic
empirical power curve using a very simple method [35]. The performance of the turbine, eliminating the hidden effect of the site
power curve is dened by the location, where, in a given wind turbulence. The ideal power curve refers to ideal conditions such as
speed bin, the maximal density of points Pi is found i.e. the power steady and laminar ow of wind, absence of yaw error and steady
curve is given by the points {uj, Pk(j)}, where j is the number of the state power output. Assessment of wind energy available in a test
458 M. Lydia et al. / Renewable and Sustainable Energy Reviews 30 (2014) 452460

site and extension of power curve to sites with different turbulence 4.2.1. Copula power curve model
levels are the main applications of the ideal power curve. It is Copula is a distribution function in statistics and is used to
analytically derived by a Taylor's expansion and uses an accurate describe the dependence between random variables. A copula
assumption of the ideal power coefcient. The convergence of the model of wind turbine performance has been developed in [25,37].
Taylor's expansion has been improved by applying the Shanks' This method includes the measures of uncertainty while estimat-
transformation. The proposed ideal power curve was successfully ing the performance and also allows comparison of inter-plant
compared with the IEC power curve. The calculation of annual performance. A copula representation of a WTPC is constructed by
energy estimated using the ideal power curve was well within the considering the power curve to be a bivariate joint distribution. To
inherent experimental error. make sure that the transformed variables have uniform distribu-
tion, accurate estimation of wind speed and power marginals are
essential. An estimated power curve copula is shown as a non-
4.1.7. Four parameter logistic function
parametric probability density estimate in [25]. But this approach
The shape of the power curve is similar to the four parameter
can be made fully useful, only if a more advanced method of
logistic function and hence WTPC models have been built based
parametric estimation of marginals and dependency is in place
on this in [5,19].
which may take the form of a mixture density estimate of the
P a1 me  u= =1 ne  u= 29 marginals and cubic spline estimate of the copula. This would aid
in capturing and identifying changes in the operating regime also.

The vector parameters of the logistic function, a, m, n and

4.2.2. Cubic spline interpolation technique
determine its shape. The parameters of the logistic function have
Interpolants and smoothing spline are the non-parametric
been estimated using the least squares method, maximum like-
tting techniques used to draw a simple, smooth curve through
lihood method and evolutionary programming (EP) method in [5].
the data [42]. Interpolation is the process of estimating values that
The parameters have been obtained using genetic algorithms (GA),
lie between two known data points. The different kinds of
particle swarm optimization (PSO) and differential evolution (DE)
interpolant methods include linear interpolation, nearest neighbor
in [13]. Power curve models developed based on these showed
interpolation, cubic-spline and Piece-wise Cubic Hermite Inter-
more accuracy than several non-parametric techniques like neural
polation (PCHIP). The WTPC model has been approximated using
networks etc.
the cubic-spline interpolation technique in [31]. This method ts a
different cubic polynomial between each pair of data points. The
4.1.8. Five parameter logistic function method of least squares and cubic spline interpolation performed
The ve parameter logistic function was originally used in extremely well for wind turbines with smooth power curve.
biological applications and was rst applied in wind turbine power
curve modeling in [13]. 4.2.3. Neural networks

ub g An articial neural network (ANN) is an information-processing
P d a  d= 1 30 model simulating the operation of the biological nervous system. It has
c
a signicant capacity to derive meaning from complicated or imprecise
data and nds application in extraction of patterns and detection of
Parametric models of WTPC has been developed using ve trends that are too complex to be identied by humans [43].
parameter logistic expression and the parameters were solved using Under normal conditions, the equivalent steady state model of
GA, EP, PSO and DE [13]. This paper has used the merits offered by the wind farm has been built using three different neural network
modern nontraditional optimization techniques for modeling of wind models namely, generalized mapping regressor (GMR), a feed-
turbine power curve. It is indeed a highly complicated process to nd forward multi layer perceptron (MLP) and a general regression
solution for parametric expression using traditional techniques of neural network (GRNN) in [20]. GMR is a novel incremental self-
optimization for huge datasets. The application of nontraditional organizing competitive network. NN models like radial basis
techniques improves the accuracy and is very easy to implement. network and generalized regression network was used for estima-
A review on the current state of the art techniques and the latest tion of annual energy in [33]. The WTPC was modeled using a
research advances in computational optimization methods with multilayer feed forward back propagation network in [13].
application to renewable and sustainable energy technologies has
been presented in [40]. A review of application of multi-objective
4.2.4. Fuzzy methods
optimization methods using evolutionary algorithms for renewable
Fuzzy logic is basically a multi-valued logic which deals with
energy technologies has been presented in [41].
approximate reasoning. Fuzzy logic based on TakagiSugeno model
The power curve models developed based on the ve para-
was used to model the annual wind energy produced in [33].
meter logistic equation gave the best results when compared to all
Modeling of WTPC using fuzzy based methods includes fuzzy cluster
other parametric and non-parametric models.
center method, fuzzy c-means clustering and subtractive clustering.

4.2. Non-parametric techniques (1) Fuzzy cluster center method

Ustunas and Sahin proposed the application of fuzzy model
Non-parametric techniques are used to solve the following based on cluster center estimation to WTPC modeling [44].
assumption: The wind turbine power generation data are clustered and the
cluster centers are determined using the model algorithm. The
P f u 31
more the number of clusters, higher is the accuracy of the
technique. The performance of the fuzzy cluster center method
Several non-parametric methods have been used to nd the is better than the least squares method.
relationship between the input wind speed data and output (2) Fuzzy c-means clustering
power. A brief description of such techniques used to model the A WTPC model has been done using fuzzy c-means (FCM)
WTPC has been given below: clustering algorithm in [39]. Unlike K-means clustering, FCM
 M. Lydia et al. / Renewable and Sustainable Energy Reviews 30 (2014) 452460 459

eliminates the effect of hard membership. It employs fuzzy established wind farms, there is a signicant need for monitoring
measures as the basis for calculation of membership matrix and troubleshooting, predictive control and optimized operation of
and identication of cluster centers, permitting data points to the wind turbines. This can be realized only if the power curve is
have different degrees of membership to each of the clusters modeled based on the historic wind speedpower data of a wind
[45]. Fuzzy clustering and similarity theory have been applied turbine or a wind farm using suitable curve-tting techniques. The
in [46] to classify the measured wind speed data from different huge amount of data available from the wind farm gives a sizeable
time. A xed output value is chosen to represent the wind number of training and testing data. Non-parametric techniques
turbine output power in that category. The non-parametric based on data mining techniques and neural networks perform well
technique FCM, has been used to model the wind turbine but the parametric techniques involving four and ve parameter
power curve in [13] and its performance has been compared logistic expressions, whose parameters are solved by DE give the
with many other models. best results [13]. The performance of the wind turbine power curve
(3) Subtractive clustering modeled using ve parameter logistic expression, with the para-
Subtractive clustering algorithm has been used for modeling meters optimized using DE has been reported to outperform
WTPC in [39]. This algorithm is very similar to mountain the linearized segmented model and the models based on neural
clustering, but the density function is calculated only at every network, fuzzy logic and data mining algorithms.
data point, instead of at every grid point. The number of
computations is reduced signicantly, since the data points
themselves become cluster centers [45]. 5. Inferences and future scope

Among all these three methods, the fuzzy cluster center The manufacturer power curve and the IEC power curve are
method gives the best model of the WTPC. invariably affected by the site turbulence. Hence it is essential that
accurate models are developed incorporating all the possible factors
that affect energy conversion in a wind turbine generating system.
4.2.5. Data mining algorithms Further research on using WTPC models should enable them to be
Data mining is all about solving problems and extracting valuable used not only for online monitoring but also for identifying links
information and patterns by analyzing data present in huge data- between interrelated anomalies as well as correlation between them.
bases. The huge volumes of data stored in the SCADA systems of The application of copulapower curve model can also be improved
wind farms present a priceless opportunity for the application of by using sophisticated method of parametric estimation of marginals
data mining algorithms for wind turbine technology. and piecewise application of copula models [25]. Accurate estimation
Non-parametric models of a WTPC have been obtained using ve and control of the variability in the wind turbine output power can
data mining algorithms namely multi-layer perceptron (MLP), ran- greatly aid utility companies in establishing good distributed gen-
dom forest, M5P tree, boosting algorithm and k-nearest neighbor eration systems and for deploying smart grid systems. Models with
(k-NN) in [5]. Among all these, the k-NN algorithms performed best. reduced error can pave way for efcient control, monitoring and
Four data mining algorithms namely bagging, M5P, REP Tree optimization of wind farms.
and M5Rules were used for modeling the WTPC in [13]. Since the high wind speed sites are almost full, identication,
The different parametric and non-parametric methodologies assessment and development of low wind speed sites is the need
employed by researchers for modeling of WTPC ultimately aim at of the hour. Wind turbine power curve models will be of great
capturing the wind turbine performance accurately and thus use use in this regard. Research and development of wind turbines
for energy prediction, monitoring and predictive control of wind that have a very low cut-in speed and that which will reach rated
turbine operation. power at a lower wind speed is required. The rated power of the
wind turbine is reached at approximately 13 m/s in all wind farm
4.3. Analysis of wind turbine power curve modeling techniques classication types at present, but it is more likely to have rated
wind speed for onshore turbines by 2030 as 12 m/s or lesser [26].
According to [31], the models based on the basic concept of power The big increase in wind energy potential for increasing hub
available in the wind (Eq. (1)), like the probabilistic model, cubic height of wind turbines has greatly encouraged the manufacture
power law model etc. do not give accurate results. This is because of and placement of tall turbines as high as 80 m or above.
the fact that the fraction of wind power that is converted to electrical Developing site-specic WTPC models for offshore sites is one
power depends on several other parameters like rotational speed of of the biggest challenges in wind industry. This will enable wind
the turbine, turbine blade parameters like angle of attack, pitch angle resource assessment in these sites, aid in wind farm power
and the efciencies of the mechanical transmission system and forecasting and also facilitate online monitoring and maintenance
generator efciency. The models based on the shape of the power of wind turbines.
curve, like the linearized segmented model, model based on Weibull's Better facilities of calibration, instrumentation and measure-
parameters etc. do not perform best because the performance of the ment can also help build accurate models. Availability of historical
wind turbines with different design parameters and ratings cannot be datasets for various sites can also be of much help in this research.
modeled using a single set of general equations [31].
It has also been stated in [31] that modeling methods in which
characteristic equations are developed based on the actual power 6. Conclusion
curve of the wind turbine is the best. This could be helpful while
using the wind turbine power curve model for wind resource This paper presents a comprehensive overview on the wind
estimation and for identifying potential wind farm sites. This will turbine power curve modeling techniques. The drawbacks posed
also aid the wind farm owners to make the right choice of turbines. by the standard IEC power curve approach and the manufacturer
But in an established wind farm, where turbines of different types provided power curve lay down the necessity for power curve
are installed, this would be impossible. Wind turbines of the same modeling. WTPC models assist the customers in making the appro-
type and make may also output different power for the same wind priate choice of wind turbines, aid in wind energy assessment and
speed due to several reasons [11]. Moreover, the manufacturer prediction, and revolutionize wind turbine performance monitoring,
power curve is unavoidably affected by the site turbulence [8]. In troubleshooting and predictive control. The various parametric and
460 M. Lydia et al. / Renewable and Sustainable Energy Reviews 30 (2014) 452460

non-parametric modeling techniques that have been employed for [21] Kusiak A, Verma A. Monitoring wind farms with performance curves. IEEE
WTPC modeling have been presented in detail. The performance Trans Sustainable Energy 2013;4:1929.
[22] Kusiak A, Li W. The prediction and diagnosis of wind turbine faults. Renewable
metrics that measures the accuracy of the models have also been Energy 2011;36:1623.
included. The future areas of research, if properly addressed will [23] Schlechtingen M, Santos IF, Achiche S. Using data-mining approaches for wind
denitely be a major stride in making the stochastic wind resource turbine power curve monitoring: a comparative study. IEEE Trans Sustainable
Energy 2013;4:6719.
into a reliable source of energy, thus transforming the wind farm [24] Uluyol O, Parthasarathy G, Foslien W, Kim K. Power curve analytic for wind
into a wind power plant. turbine performance monitoring and prognostics. In: Annual conference of
the prognostics and health management society 2011.
[25] Gill S, Stephen B, Galloway S. Wind turbine condition assessment through
References power curve copula modeling. IEEE Trans Sustainable Energy 2012;3:94101.
[26] McLean JR. WP2.6- Equivalent Wind Power Curves. EIE/06/022/SI2 2008;442659:
[1] Schallenberg-Rodriguez J. A methodological review to estimate techno- 114.
economical wind energy production. Renewable Sustainable Energy Rev [27] Jafarian M, Soroudi A, Ehsan M. The effects of environmental parameters on
2013;21:27287. wind turbine power pdf curve. CCECE 2008:0011938.
[2] Shaullah GM, Oo AMT, Shawkat Ali ABM, Wolfs P. Potential challenges of [28] Tindal A, Johnson C, LeBlanc M, Harman K, Rareshide E, Graves A M. Site-
integrating large-scale wind energy into the power grida review. Renewable specic adjustments to wind turbine power curve. In: AWEA windpower
Sustainable Energy Rev 2013;20:30621. conference; 2008. p. 111.
[3] Manwell JF, McGowan JG, Rogers AL. Wind energy explained: theory, design [29] Khalfallah MG, Koliub AM. Wind turbines power curve variability. Desalina-
and application. UK: John Wiley & Sons; 2009. tion 2007;209:2307.
[4] Monteiro C, Bessa R, Miranda V, Botterud A, Wang J Conzelmann G. Wind [30] Carrillo C, Obando Montao AF, Cidrs J, Daz-Dorado E. Review of power
power forecasting: state-of-the-art 2009. Decision and Information Sciences curve modelling for wind turbines. Renewable Sustainable Energy Rev
Division, Argonne National Laboratory, ANL/DIS-10-1. 2009. 2013;21:57281.
[5] Kusiak A, Zheng H, Song Z. On-line monitoring of power curves. Renewable [31] Thapar V, Agnihotri G, Sethi VK. Critical analysis of methods for mathematical
Energy 2009;34:148793. modeling of wind turbines. Renewable Energy 2011;36:316677.
[6] IEC 61400-12-1 Ed.1: Wind turbines - Part 12-1: Power performance mea- [32] Pinson P, Nielsen HA, Madsen H, Nielsen TS. Statistical power curve model-
surements of electricity producing wind turbines, 88/244/FDIS, 2005. ling: which power curve for what applicaton? EWEC 2008.
[7] Wagner R, Courtney M. Multi-MW wind turbine power curve measurements [33] Jafarian M, Ranjbar AM. Fuzzy modeling techniques and articial neural
using remote sensing instrumentsthe rst Hovsore campaign 2009. networks to estimate annual energy output of a wind turbine. Renewable
[8] Trivellato F, Battisti L, Miori G. The ideal power curve of small wind turbines Energy 2010;35:200814.
from eld data. Wind Eng. Ind. Aerodyn 2012;107-108:26373. [34] Akdag SA, Guler O. Comparison of wind turbine models. IREC 2010:2159.
[9] Whale J, McHenry MP, Malla A. Scheduling and conducting power perfor- [35] Gottschall J, Peinke J. How to improve the estimation of power curves for wind
mance testing of a small turbine. Renewable Energy 2013;55:5561. turbines. . Environ Res Lett 2008;3:17.
[10] Norgaard P, Holttinen H. A multi-turbine power curve approach. In: Proceed- [36] Milan P. The stochastic power curve analysis of wind turbines. Hydrodynamics
ings of the Nordic wind power conference; 2004. p. 15. & Wind Energy group, University of Oldenburg, Germany: 2008.
[11] Jin T, Tian Z. Uncertainty analysis for wind energy production with dynamic [37] Stephen B, Galloway SJ, McMillan D, Hill DC, Ineld DG. A copula model of
power curves. In: Proceedings of the International conference probabilistic wind turbine performance. IEEE Trans Power Syst 2011;26:9656.
methods applied to power systems; 2010. p. 74550. [38] Khalfallah MG, Koliub AM. Suggestions for improving wind turbine power
[12] Olaofe OZ, Folly KA. Wind energy analysis based on turbine and developed site curves. Desalination 2007;209:2219.
power curves: a case-study of Darling city. Renewable Energy 2013;53:30618. [39] Raj MSM, Alexander M, Lydia M. Modeling of wind turbine power curve. ISGT-
[13] Lydia M, Selvakumar AI, Kumar SS, Kumar GEP. Advanced algorithms for wind India IEEE PES 2011:1448.
turbine power curve modeling. IEEE Trans Sustainable Energy 2013;4:82735. [40] Baos R, Manzano-Agugliaro F, Montoya FG, Gil C, Alcayde A, Gmez J.
[14] Zamani MH, Riahy GH, Ardakani AJ. Modifying power curve of variable speed Optimization methods applied to renewable and sustainable energy: a review.
wind turbines by performance evaluation of pitch angle and rotor speed Renewable Sustainable Energy Rev 2011;15:175366.
controllers 2007. In: IEEE Canada electrical power conference. p. 34752. [41] Fadaee M, Radzi MAM. Multi-objective optimization of a stand-alone hybrid
[15] Simic Z, Mikulicic V. Small wind off-grid system optimization regarding wind renewable energy system by using evolutionary algorithms: a review.
turbine power curve. AFRICON 2007. J Renewable Sustainable Energy Rev 2012;16:33649.
[16] Jangamshetti SH, Rau VG. Normalized power curves as a tool for identication [42] Curve tting toolbox, for use with MATLAB, User's Guide, Version 1. The
of optimum wind turbine generator parameters, IEEE Trans on Energy Mathworks: 2002.
Conversion 2001;16:2838. [43] Sivanandam SN, Deepa SN. Principles of soft computing. New Delhi: Wiley
[17] Albadi MH, El-Saadany EF. Wind turbines capacity factor modelinga novel India (P) Ltd; 2010.
approach. IEEE Trans Power Syst 2009;24:16378. [44] Ustuntas T, Sahin AD. Wind turbine power curve estimation based on cluster
[18] Albadi MH, El-Saadany EF. New method for estimating CF of pitch-regulated center fuzzy logic modeling. Wind Eng. Ind. Aero 2008;96:61120.
wind turbines. Electr Power Syst Res 2010;80:11828. [45] Hammouda K, Karray F. A comparative study of data clustering techniques.
[19] Kusiak A, Zheng H, Song Z. Models for monitoring wind farm power. Renew- Design 625, pp. 121 [Online]. Available: http://citeseerx.ist.psu.edu/ view
able Energy 2009;34:58390. doc/download, Citeseer. Retrieved from 10.1.1.126.3224.
[20] Marvuglia A, Messineo A. Monitoring of wind farms' power curves using [46] Suhua L, Zhiheng L, Yaowu W. Clustering analysis of the wind power output
machine learning techniques. Appl Energy 2012;98:57483. based on similarity theory. DRPT 2008:28159.