Professional Documents
Culture Documents
Mohamed Asmine, Jacques Brochu, Jens Fortmann, Richard Gagnon, Yuriy Kazachkov, Charles-Eric Langlois,
Christian Larose, Eduard Muljadi, Jason MacDowell, Pouyan Pourbeik, Slavomir A. Seman, and Kevin Wiens
I. INTRODUCTION
these types of studies, generally what are referred to as positivesequence (or sometimes RMS) models are adequateaside: for
full-converter and doubly-fed asynchronous generators, positive-sequence models may be limited in their ability to properly
capture the units response to unbalanced faults; more on this
latter in the paper. For equipment design, and the assessment
of fast electromagnetic transients, detailed three-phase electromagnetic transient (EMT) level models are typically used. This
paper is primarily focused on generic (that is, non-vendor specific and publicly available) positive-sequence models for stability analysis, and does not address electromagnetic transient
(EMT) models.
For a more thorough discussion on the reasons for testing
and model validation, particularly from a policy perspective, see
some of the documents being developed by reliability entities
such as [2][5].
IE
E
W
E
eb P
r
Ve oo
rs f
ion
Manuscript received August 09, 2010; revised October 08, 2010; accepted
November 08, 2010. Paper no. TPWRS-00632-2010.
Corresponding author: P. Pourbeik, TF Lead (e-mail: pouyan@ieee.org).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPWRS.2010.2092794
1In the case of an entire wind power plant, one also needs the data related
to the collector system and transformer impedances, cable capacitance, reactive
power compensation at the turbine level, and at the plant level, if any.
stability model testing. The data provided were model specifications or simulations from equipment-design level models.
With this in mind, in the development of the first generation of
generic WTG models, a vendor specific model was considered
the only benchmark available for validating the generic models.
The first generation of generic models were developed for
used in grid interconnection studiesfor detailed description of
these models and the model structures, please refer to our sister
TF paper [1]. As such, they should adequately simulate wind
turbine dynamics in terms of terminal voltage, active, and reactive power response to disturbances coming from the grid. They
are not designed for studying the effect of changing wind. Of
all four available generic models, only the WT3 model (DFAGbased wind turbine) has a reference to the initial per unit wind
speed which is used for a simplified calculation of the aerodynamic torque and initial blade pitch angle. All the rest of the
generic models do not use any wind correlation. In the WT1
(conventional induction generator) and the WT2 (variable rotor
resistance induction generator), aerodynamics and pitch control
are replaced by the pseudo-governor action. In the WT4 (full
size converter connected generator), the machine is decoupled
from the grid and the mechanical side is not modeled at all.
Conventional Induction Generator (WT1 Model): The
electrical generator is modeled as a full-order induction machine, as is done for vendor specific models. The remaining
part of the model is the mechanical sidenamely the turbine
aerodynamics and control. In the course of preliminary testing
of some vendor specific models of this type, it was noticed
that the response of the mechanical power to step changes in
the machine active power Pelec or the rotor slip WTRBSP can
be approximated by a first or second order transfer function.
This led to the idea of the pseudo-governor mimicking the
complicated joint action of the aerodynamic conversion and
pitch control by calculating the mechanical torque WAEROT
as a function of two inputs: Pelec and WTRBSP (see Fig. 1).
Therefore, the main concern is the validity of the pseudo-governor approach. The response of the Mitsubishi 1000-kW Type
1 wind turbine to a three-phase fault at the 230-kV point of interconnection in terms of real electrical power (Pelec), reactive
power (Qelec), and terminal voltage (Vterm) is illustrated by
plots of Fig. 2, for the vendor specific model and generic WT1
model. In general, we see a relatively good fit, with some difference in attenuation of electrical power (Pelec).
An outstanding question is the validity of these models
for system frequency disturbances. Again, the models have
been tested against the vendor specific models; however, it is
acknowledged that comparison to actual recorded disturbance
IE
E
W
E
eb P
r
Ve oo
rs f
ion
Thus, we see from the brief outline above that the following
is needed for defining the process of, and the requirements for,
model validation:
1) Model Structures: the actual model structures to be used.
For WTG, these have been defined through the WECC WG
effort and our sister ad hoc TF within the DPWG WG [1].
2) Data Requirements: to define the minimum data that
needs to be recorded in the field or in the factory from the
wind turbine generator(s).
3) Adequacy: to define at what point the model performance
is considered adequate. This is perhaps the most difficult
task and can at times be subjective.
The process of model validation does not necessarily mean a
perfect match between the measured and simulated response,
but rather an adequate match that clearly demonstrates:
capturing the relevant dynamics;
proper representation of the plants dynamic response; and
the ability to account for possible discrepancies.
C. Wind Turbine Generator Model Validation Case Studies
In this section, we present a series of examples of model validation efforts for wind turbine generator models as conducted
by a few manufacturers and utilities.
As stated in the previous section, model validation is ideally achieved by comparing model response to actual measured
equipment response either in the field or in factory tests. In the
early development of WTG models, however, such data was rare
and only now is there an emphasis on obtaining such data. As
such, the early development of generic WTG models was based
on comparing simulations between the generic models and the
vendor specific proprietary models. Thus, in this section, we will
start with a brief summary of that approach and results achieved,
and then present more recent and rigorous approaches using
measured unit response. Here the focus is on detailed model validation of dynamic stability models, reference [6] can be consulted for an example of model validation with an emphasis also
on validating the collector system and steady-state conditions.
Some other valuable references are [7][9].
1) Model Comparison Between Detailed Vendor Specific
Models and Generic Models: Numerous vendor specific
models have been developed in the Siemens PTI PSSE platform in direct collaboration with WTG vendors. In this effort,
thus far, none of these models were based on any manufacturer
providing results of field or factory tests as a benchmark for
IE
E
W
E
eb P
r
Ve oo
rs f
ion
Fig. 4. Results of validation of the WT2 model against Vestas V80 60-Hz
vendor specific model.
IE
E
W
E
eb P
r
Ve oo
rs f
ion
IE
E
W
E
eb P
r
Ve oo
rs f
ion
Fig. 9. Setup to measure the effect of voltage dips on a power generation unit,
for example a wind turbine. PGU stands for power generating unit, i.e., in this
case, the wind generation unit or plant.
Fig. 14. Measurement points for typical wind power plant in Hydro-Qubec.
The process of model validation in Germany has been described and illustrated through an example of a 45% voltage dip.
It should be noted that FRT-Tests down to lower voltages and
the validation of active power is significantly more demanding.
Nonetheless, the process has been shown to work adequately. In
[11], an example is shown also with a two-phase fault, used to
validate a model of a WT3 turbine.
4) Hydro-Qubec Experience With Model Validation:
Hydro-Quebec plans to integrate 4000 MW of wind power to
its network by 2015. Consequently, research has been conducted to insure reliable integration of wind generation into the
system. Here a brief account is given on some of the pertinent
issues related to modeling and model validation. See [13][16]
for more details.
Validation of Type-3 Wind Turbine and Wind Power Plant
Models Using Online Disturbance Monitoring: For the purpose
of model validation, online monitoring equipment has been installed on a typical wind power plant (WPP) connected to the
Hydro-Quebec network. This WPP is composed of 73 1.5-MW
type-3 wind turbines. Fig. 14 shows the WPP, with the voltage
and currents monitored and their locations at the turbine, feeder,
and point of interconnection (POI) levels.
From 2007 to 2009, various disturbances (e.g., faults and frequency deviations) were recorded. Those recordings have been
used to build and validate a type-3 WTG model in the EMT domain.
The model validation process used is based on playback techniques, where the model is fed with recorded voltages from the
actual wind turbine, and validation is confirmed when the model
produces the same current as those recorded during the disturbance. Following the same approach of waveform playback, the
entire WPP model has also been validated, using recorded voltages and currents at the POI level. The wind plant management
system was also modeled and validated in the mean time.
Fig. 15 shows an example of the comparison between simulation and field measurement for an event. It can be seen that
the conformity of the model with the field measurements is very
good for this particular event.
Such good correspondence of the model for a number of
different operating conditions and recorded disturbances has
greatly contributed to increase the confidence in the validity
of the model. Fine-tuning the model is a process relatively
straightforward for small disturbance, but it becomes more
complex with large and/or unbalanced disturbance due to various nonlinearities. Regardless of the disturbance severity, this
IE
E
W
E
eb P
r
Ve oo
rs f
ion
Fig. 11. (1) Measured and (2) simulated reactive current at measurement
point 2 for a symmetric voltage dip. The average values are calculated
for each interval (doted lines) for (1) measurement and (2) simulation.
Transient periods start at = 1, t = 2, and t = 4 s.
IE
E
W
E
eb P
r
Ve oo
rs f
ion
Fig. 15. Comparison of recorded and simulated waveforms at the type-3 WTG
level during a fault on the network.
Fig. 17. Comparison of recorded and simulated waveforms (using the HydroQuebec model) for a remote fault seen from a substation with type-2 WTGs.
IE
E
W
E
eb P
r
Ve oo
rs f
ion
Fig. 18. Comparison of a detailed WPP model with a one-, two-, and
four-WTG equivalent WPP models. More than one equivalent WTG is not
necessary for modeling a WPP when all WTGs are exposed to the same wind
speed.
NREL appears to offer precise results for various types of disturbances and operating conditions, for both EMT and stability
studies. Fig. 18 shows the performance of the NREL method
for WPP modeling. In this figure, a two-phase fault is applied
to four different models of WPP: a detailed 73-WTG model,
and three different aggregated WPPs consisting of one, two,
and four equivalent WTGs. The significance of this plot is to
show that a single aggregated equivalent generator with a single
equivalent impedance representing the entire collector system
can adequately represent plant response for studying grid disturbancesadmittedly this assumes one is studying a transient
phenomena and wind speed is constant for the duration of the
grid event.
Generic Equivalent Collector System Parameters for Large
WPP for Preliminary Studies: The equivalent collector systems
of 17 WPPs rated between 50 and 300 MW were analyzed.
Using this sample, a set of generic equivalent collector system
parameters were calculated to be used for prospective powerflow and stability studies of WPPs for which little or no information is available yet. An exhaustive sensitivity study based
on EMT simulations has confirmed the adequacy of the generic
equivalent collector system parameters proposed in [16].
5) ABB Experience With Model Validation for a Type-4
WTG Using the Generic WT4 Model: Using documentation in
Siemens PTI PSSE for the WECC developed WT4 generic
model, ABB developed the model shown below.
IE
E
W
E
eb P
r
Ve oo
rs f
ion
Fig. 25. Voltage dip for model versus measured response for a three-phase
voltage dip.
IE
E
W
E
eb P
r
Ve oo
rs f
ion
10
Fig. 26. Total converter current for model versus measured response for a threephase voltage dip.
Fig. 27. Active converter current for model versus measured response for a
three-phase voltage dip.
Fig. 28. Reactive converter current for model versus measured response for a
three-phase voltage dip.
11
IE
E
W
E
eb P
r
Ve oo
rs f
ion
Fig. 30. Reactive power and voltage control structure for GE 1.5-MW wind
turbine and windcontrol plant management.
IE
E
W
E
eb P
r
Ve oo
rs f
ion
12
Fig. 33. The 10-MVAr capacitor removal response measured from WindCONTROL.
The need for modeling and modeling validation should be apparent to the reader, as it is a mandated need in many continents
Fig. 35. The 10-MVAr capacitor removal field test versus simulation results.
IE
E
W
E
eb P
r
Ve oo
rs f
ion
13
IV. CONCLUSIONS
4Note: in actuality, three-phase voltage and currents are the measured quantities, real and reactive power and bus frequency are then calculated from the
measured voltage and currents. Most modern digital recording devices are able
to make these calculations internally.
14
IE
E
W
E
eb P
r
Ve oo
rs f
ion
voltage, real, and reactive power at the terminals of a few representative generating units in the plant. A more clear and thorough definition of measurement is needed. This is presently
being pursued by other groups such as the IEC TC88 WG27
and the NERC IVGTF [27].
It should be understood that there are inherent errors in the
measurement process (up to 3%5%). Furthermore, there are
dynamics, particularly associated with fast transients, which
are seen in measurements and are not represented in stability
models. Moreover, in the case of unbalanced faults, measurement data needs to be filtered to extract the positive sequence
response for comparison to stability modelsthis introduces
further complications, for the negative and zero sequence
behavior cannot be faithfully represented in positive-sequence
simulation tools. All these considerations must be fully recognized when embarking on model validation, and they must be
taken into consideration when setting standards for acceptance
of model validation results. That is, when comparing the simulated response to measured response and trying to identify
when a good-enough fit has been achieved, the criteria used
should fully recognize all these issues and potential sources of
error that are unavoidable.
ACKNOWLEDGMENT