Professional Documents
Culture Documents
6 MW WIND TURBINE-
GENERATORS FOR STABILITY SIMULATIONS
Nicholas W. Miller, Juan J. Sanchez-Gasca, William W. Price Robert W. Delmerico
GE Power Systems GE Research
Schenectady, NY Niskayuna, NY
-
Abstract GE Power Systems has an ongoing effort dedicated GE 3.6 MW WTG the converter is connected as shown
to development of models of GE wind turbine generators through a transformer at a lower voltage than the stator
(WTG) suitable for use in system impact studies. This paper winding. For the GE 1.5 MW WTG the converter is
documents the present recommendations for dynamic connected at the same volage as the stator winding. Machines
modeling of the GE 1.5 and 3.6 MW WTG for use in system of this structure are termed 'double fed', and have
impact studies. The paper includes recommended model significantly different dynamic behavior than either
structure and data, as well the assumptions, capabilities and conventional synchronous or induction machines. Modeling
limitations of the resulting model. of the GE 1.5 and 3.6 machines with conventional dynamic
models for either synchronous or induction machines is, at
1. Introduction best, highly approximate and should be avoided.
The goal of this effort is the development of a simple model
appropriate for bulk power system dynamic studies. It is
valuable to put the model limitations in the context of what
analysis is required. First and most important, this model is
for positive sequence phasor time-domain simulations, e.g.
T
PSLF or PSSle. Second, the analysis is mainly focused on
how the wind turbine-generators (WTGs) react to
disturbances, e.g. faults, on the transmission system. Third,
the model provides for calculation of the effect of wind speed
fluctuation on the electrical output of the WTG. Details of the
device dynamics have been substantially simplified.
Specifically, the very fast dynamics associated with the
control of the generator converter have been modeled as Figure 2-1. GE WTG Major Components.
algebraic (i.e. instantaneous) approximations of their response. The fundamental frequency electrical dynamic performance of
Representation of the turbine mechanical controls has been the GE WTG is completely dominated by the field converter.
simplified as well. The model is not intended for use in short Conventional aspects of generator performance related to
circuit studies. intemal angle, excitation voltage, and synchronism are largely
irrelevant. In practice, the electrical behavior of the generator
2. Model Overview and Philosophy and converter is that of a current-regulated voltage source
inverter. Like other voltage source inverters (e.g. a BESS or a
2.1 Fundamentals
STATCOM), the converter will make the WTG behave like a
A simple schematic of the GE Wind System is shown in voltage behind a reactance that results in the desired active
Figure 2-1. The GE WTG generator is unusual from a system and reactive current being delivered to the device terminals.
simulation perspective. Physically, the machine is a relatively
conventional technology wound rotor induction (WRI) The rotation of the machine means that the ac frequency on
machine. However, the key distinction is that this machine is the rotor winding corresponds to the difference between the
equipped with a solid-state AC excitation system. The AC stator frequency (60Hz) and the rotor speed. This is the slip
excitation is supplied through an ac-dc-ac converter. For the frequency of the machine. In the vicinity of rated power, the
GE 1.5 and 3.6 machines will normally operate at 120%
speed, or -20% slip. Control of the excitation frequency
G E Power Systems Energy Consulting allows the rotor speed to be controlled over a wide range,
IEEE WTG Modeling Panel ?30%. The rotation also means that the active power is
divided between the stator and rotor circuits, roughly in
Session July 2003 proportion to the slip frequency. For rotor speeds above
synchronous, the rotor active power is injected into the
network through the converter. The active power on the rotor
is converted to terminal frequency (60Hz), as shown in Figure
1978
generator models, all of the electrical/flux state variables have selected by the user. The regulator itself is a PI controller with
been reduced to their algebraic equivalents. The flux a low pass time constant, Tv. The time constant reflects the
dynamics of the double fed machine with AC excitation are delays associated with cycle time, communication delays and
too fast to have significant impact on system stability. The net additional high frequency attenuation needed to maintain
result is an algebraic, controlled current source that injects the controller stability. Unlike voltage regulators associated with
active and reactive power specified by the WTG electrical conventional synchronous machines, this regulator produces a
control model into the network. Conceptually, the model is reactive power command (rather than a field voltage
trivial. Details of the implementation of the controlled current command) that is imposed very rapidly by the converter. This
source are unique to each simulation package. provides the potential for very fast and effective voltage
regulation. However, excessively high gains and short time
4.2 WTG Electrical Controller Model constants should he avoided.
The WTG Electrical Controller model dictates the active and
reactive power to he delivered to the system based on input A provision is shown for adding a reactive power modulation
from the turbine model and power system conditions. The signal (Q modulation) to the output of the voltage regulator.
model is greatly simplified, hut maintains those aspects that This feature could be used for an explicitly modeled
are crucial to capturing the dynamic performance of interest to supervisory control. Other signals could be injected here as
the system. It is within this model that the power system well.
planner may select parameter settings in order to meet the
particular requirements of a system application. The control Each individual machine is presently equipped with power
functions are segregated into two major functional groupings, factor control. The initial power factor of the WTG is set by
executed in sequence: (1) Closed Loop Controls, and ( 2 ) Open the initial conditions from the loadflow. Subsequent
Loop Controls. fluctuations in active power command will result in a
corresponding change in reactive power output, thereby
maintaining the specified power factor. The power factor
Closed Loop Controls.
control logic is in the box on the right-hand side of Figure 4-1,
The closed loop controller shown in Figure 4-1 includes
When constant power factor control is enabled, as is normally
voltage regulation and power factor control functions. The
the case, this simplified model forces the reactive power to
voltage regulation function represents an equivalent of the
follow changes in active power command thereby maintaining
supervisory control for the entire farm. It is appropriate to use
a nominally constant power factor. Modifications to the
this function when using a single WTG model to represent an reactive power request from the voltage regulator effectively
aggregation of machines under supervision. The function
change the power factor setting.
monitors the collector bus voltage and compares it against the
reference voltage. Normally this reference voltage is the
Table 4-1 below includes recommended settings for the WTG
initial voltage of the regulated bus (from the load flow). The
electrical controller model. All settings are given in terms of
reach of the regulator can be modified by inclusion of a
compensating impedance, as shown in Figure 4-1. The
compensating impedance (Rc and Xc) can he omitted or
Open Loop Controls functions in the box of Figure 4-1 are shown in detail in
The open loop controls respond to large disturbances and Figure 4-2. These control functions are primarily
include the important protective functions relating to the responsive to large variations in system voltage, and are
electrical aspects of the WTG. The open loop control
1979
usually inactive whenever the terminal voltage is within a Three other voltage thresholds are shown in this function
normal range (defined by the limits VLIand VH,). box. These represent the type of optional open loop
controls than can be implemented to improve system
A principle feature of the converter is that the maximum performance for large voltage deviations resulting from
current limit must be observed. Under conditions when systems events. These controls force the reactive power to
both the active and reactive power cannot both he satisfied pre-specified levels as the voltage deviations persist. Table
without violation of the current limit, the converter control 4-2 gives a summary of the control chronology. For
will give priority to the reactive current. voltages helow VU, the reactive power command is current
limited. As with all open loop controllers of this type,
The terminal voltage and the reactive power order from the hysteresis is needed to avoid hunting. Once the voltage
closed loop control are shown entering a function box in the thresholds are crossed and the open loop reactive power
upper left hand portion of Figure 4-2. When the terminal command is issued, the threshold voltage is shifted up (or
voltage falls helow the low voltage trip threshold or above down for high voltage events) by a specified amount, Vhyn.
the high voltage trip threshold, the machine trips following
a delay. The trip sequence is shown on the right-hand side Following the function box, the reactive power command is
of the figure. Once the trip flag is set, it will not reset, even subjected to MVAr limits and a current magnitude limit.
if the voltage recovers. (Implementation of this These transient MVAr limits should he set to match the
instantaneous hip flag requires a minimum of two steady-state MVAr limits used in the loadflow. The
successive integration steps in simulations so as to avoid controller responds to the turbine power order, Pd, as a
false trips from numerical aberrations.) current order. The active current is limited such that the
reactive current has priority, as shown in the figure.
1980
shaft power that is dependent on wind velocity, rotor speed speed trip. Details of the turbine control actions at these
and blade pitch. This algebraic relationship is provided in a extremes have been neglected in this model.
The model does not include high and low wind speed cut-
1.0 , /
out for the turbine. In situations where system performance
questions hinge on this behavior, the user can simply trip
the machine.
The rotational model of the machine is approximated as a
single rotational inertia, representing the total inertia of the
machine, gearbox and turbine. This is a similar
0.0 : approximation to that used for conventional thermal and
0.7 1.0 1.2 hydro generation. However, the relatively low natural
torsional frequencies typical of wind systems make this
Gen. Speed (P.u.)
approximation somewhat less valid here. Further
Figure 4-3 Open Loop Electrical Controls investigation of this approximation is planned.
separate model, Wind Power Model, described helow. The Table 4-1 WTG Electrical Controller Parameters
wind turbine model represents all of the relevant controls I Parameter I Recommended I UserAdiustoble? 1 Reouired? I
and mechanical dynamics of the wind turbine. The model Nome I Value I I
is based on power calculations, rather than torque, for T" I 0.5 Yes I Yes, i / K p or Ki
simplicity. It directly accepts the machine terminal active
power from the WTG Electrical Control Model and the
mechanical power calculated by the Wind Power Model.
The turbine control model simultaneously sends a power
order to the electrical control, requesting that the converter
deliver this power to the grid. The electrical control may or
may not he successful in meeting this power order. The
electric power actually delivered to the grid is returned to
the turbine model, for use in the calculation of rotor speed.
The dynamics of the electrical controller are extremely fast.
Dynamically, the combination of blade pitch control and
electric power order results in the two distinct operating
conditions. For power levels significantly below nominal
(-75%). the power is controlled so as to reduce the turbine
speed according to Figure 4-3. This is approximated by
adjusting the speed reference. The model does not allow
for motoring of the turbine.
The turbine controller modifies the blade pitch. In this
model the blade position actuators are rate limited and there
is a short time constant associated with the translation of
blade angle to mechanical output. The speed controller
does not differentiate between shaft acceleration due to
increase in wind speed or due to system faults. In either
case, the response is appropriate and relatively slow
compared to the electrical control.
The turbine control acts so as to smooth out electrical
power fluctuations due variations in shaft power. By Table 4-2 WTG Open Loop Reactive Controller
allowing the machine speed to vary around nominal rated
Chronology
(120%1,the inertia of the machine functions as a buffer to
mechanical power variations.
I Voltope Condition I For lime durotion I Omn Lo00 Reactive
The turbine control model includes protection against over
and underspeed operation. Normally, the turbine controls
will prevent such trips. However excessively high or low
wind speeds, and sustained load rejection may cause a
1981
The function of the wind power module is to compute the
Parameter values for the wind turbine model are shown in wind turbine mechanical power (shaft power) from the
Table 4-3. None of these values should be modified by the energy contained in the wind. The well-known relationship
user.
Table 4-3 Turbine Parameters
For power system simulations involving grid disturbances, For the GE WTGs, parameters given in Table 4-4 will
it is a reasonable approximation to assume that wind speed result in Pmh in p.u. on the MW base.
remains uniform for the 5 to 30 seconds typical of such
cases. However, the mechanical power delivered to the Table 4-4. U'TG Mechanical Constants
shaft is a complex function of wind speed, blade pitch angle
and shaft speed. Further, with wind generation, the impact
of wind power fluctuations on the output of the machines is
of interest. The turbine model depends on the wind power Kb 56.6 69.5
model to provide this mapping.
OveriUnder
! R /
I
Speed Trip
Logic:
speed > mm
a p e d < mi"
i I / I
j
P
", J
1982
C, is a characteristic of the wind turbine and is usually
provided as a set of curves (C, curves) relating C, to 1,with
8 as a parameter. The C, curves for the GE wind turbine
used in this model shown in Figure 4-5 are fit with a fourth
order polynomial on Band 1.
Initialization of the wind power model for simulations
requires care. There are two distinct states of interest: 1)
initial electrical power (from the loadflow) is less than
rated, or 2) initial electrical power equal to rated. For each
condition, Pmuhis known from the loadflow and w equal to
%r as a dynamic input parameter. From these two
quantities, the C, can be calculated. For the fust condition,
the turbine control will initially he at minimum pitch, and
therefore B is known. From this information the 1
necessary to produce the required C, can he calculated.
Notice from Figure 4-5, that two values of 1will generally
satisfy the required C, for a given 0. The wind speed v,
corresponding to the higher Ais used to calculate the initial
wind speed. When P,I, is initially equal to the rated power,
The wind turbine model presented in this paper is based on
then B is not necessarily at minimum and B and v are not presently available design information, test data and
uniquely defined for the C,. In this case, one of these two
extensive engineering judgment. This model was
quantities, B or Y, must he specified. developed specifically for the GE 1.5 and 3.6 MW WTGs.
This model is not designed for, or intended to he used as, a
general purpose WTG. There are substantial variations
between models and manufacturers.
The modeling of wind turbine generators for hulk power
system stability studies is the focus of intense activity in
many parts of the industry. This model is expected to give
realistic and correct results when used for hulk system
performance studies. It is expected that these model
components will continue to evolve, in terms of parameter
values and structure, as experience and additional test data
are obtained.
1983