You are on page 1of 8

Static and Dynamic Power System Load Emulation in

Converter-based Reconfigurable Power Grid Emulator


Jing Wang, Liu Yang, Yiwei Ma, Jingxin Wang, Leon M. Tolbert, Fred Wang, Kevin Tomsovic
Center for Ultra-Wide-Area Resilient Electric Energy Transmission Networks (CURENT)
The University of Tennessee
Knoxville, TN USA 37996-2250

Abstract — A hardware test-bed platform By customization and optimization, converters could


emulating multiple-area power system scenario be built to satisfy various application environments.
dynamics has been established aiming at multiple
According to different control requirements,
time-scale emulations. In order to mimic real power
loading conditions, and switching capabilities of
flow situation in the system, the load emulators
semiconductor devices, closed-loop controlled
have to behave like real ones in both its static and
behavior bandwidth of the power electronic converter
dynamic characteristics. A constant-impedance,
can be demonstrated up to several kHz.
constant-current, and constant-power (ZIP) model
has been used for static load type, while a three- Because of their fast, accurate, and flexible closed-
phase induction motor model has been built to loop behaviors, it is feasible to control them to mimic
represent dynamic load types. In this paper, ways of static or slower-time scaled electrical dynamics.
modeling ZIP and induction motor loads and the Power system study in general, is observed by the
performance of each load emulator are discussed. A time scale of milliseconds, seconds, minutes or even
comparison between simulation and experimental hours. These dynamic details can be easily captured by
results are shown as well for the validation of the power converters.
emulator behavior.

Index Terms — reconfigurable power grid emulator,


regenerative converter, power hardware-in-the-loop, power
electronic load emulator, ZIP load, three-phase induction
motor load, control algorithms, numerical method, starting up
transient, dynamic performance.

I. INTRODUCTION

There are various ways of analyzing power system


scenarios: simulation software like PSSE, FPGA based
hardware emulation tool like RTDS and etc. However,
as the virtual simulation and emulation methods have
different focus, such as concentration on small time Figure 1. Power electronic converter based
scale in order to adequately capture transients which power system component emulator.
results in insufficient computing resource for long
duration simulations, or they focus on large time scale As illustrated from Figure 1, the three-phase side is
for overall system behavior but then have little the electrical behavior connection interface with the
information about transient dynamics. rest of the system. Voltage/current inputs are sampled
in real time for updating the power system component
Power electronics are known for their versatile model reference calculation. The references from the
control algorithms and different power stage options. model will be given to the hardware converter closed
Sponsored by NSF, DOE and CURENT industrial partners.

978-1-4799-5776-7/14/$31.00 ©2014 IEEE 4008


loop for behavior control. Converters react within its scenario emulation could be demonstrated.
bandwidth capability to the voltage/current reference Figure 2(a) shows a typical two-area system with
changes. four conventional generations, two loading centers and
The power system emulator integrates both the an interconnecting long-distance transmission line
advantages of power electronics high bandwidth between the two areas [7]. The corresponding
performance, and also the slower power system emulation network structure consisting of real power
behavior. In addition, by combining different types of electronic converter hardware is shown in Figure 2 (b),
emulators together according to specific scenario with red highlighted three-phase interface as grid
emulation dynamics analysis platform.
network, an emulation system could be both detailed,
presenting the transient dynamics, and full-pictured The actual power system under analysis has a power
with appropriate recording technique [2][3][4][6][8]. base of 900 MVA. After being converted to per unit
system, it is scaled to the ratings of the three-phase
Under this development principle, conventional power converter base of 15 kVA. Transmission line
synchronous generators [9][16] , renewable energy impedance values are calculated and replaced with
generations [10][11][12], as well as most kinds of passive inductors (for details of corresponding values,
power system loads could be emulated by flexible please see the reference [16]).
power electronic converters [5][6][9].
Synchronous generators (or renewables
By connecting the multiple modular and generations), load (or energy storage) and transmission
reconfigurable units together, flexible network and lines (in progress) are replaced by converters for the

(a)

(b)
Figure 2. Example two-area power system: (a) scenario. (b) corresponding configuration of hardware test-bed.

4009
behavior emulations. The power flow is shown with a II. LOAD EMULATOR CONTROL SCHEME
green arrow: from generator to load and also
transmitted from Area 1 to Area 2 (bus 7 to bus 9). The The load emulator was first developed by
whole structure is reconfigurable and could be changed researchers to substitute the passive loads and provide
to emulate a different power system structure. flexible loading conditions during testing power
Figure 3 shows the current architecture and real electronic converter applications such as
hardware of a power electronic converter based scaled uninterruptable power supply, transformers, switching
reconfigurable grid emulator single area. Each single devices and protection functions, etc. It works as
cluster includes: active rectifier for establishing and current sink with various static or dynamic current
stabilizing dc bus voltage, generators 1 and 2 in each behaviors.
area, combined static and dynamic load, and local
transmission line inductors.
The HTB will allow testing, integration and
demonstration of key technologies on the topics of
power system monitoring, control, and actuation. By
reconfiguring the test-bed, various kinds of scenarios
can be built, and the impact of new technologies, like
renewable energy sources [10][11][12], responsive
loads, and energy storage can also be evaluated [17].
In order to mimic the actual power system loading
behaviors, voltage- and frequency-dependent load
models should be established and programmed for
(a)
quantifying an emulator’s references.
vab vbc vca
This paper specifically shows the power system
load models adopted, including static and dynamic
characteristics. Also, verification for the load emulator Line voltage
behaviors is demonstrated with results comparisons
between simulation and experiments. Phase voltage
va vb vc
integral
θ ω
abc/dq 1/s
vq ‐
0
PI
+
limit lowpass
filter
(b)

(c)
Figure 4. Load emulator basic implementation:
(a) structure of current load emulator.
(b) PLL closed loop control structure details.
Figure 3. Back view of a cluster hardware (c) PI current controller in dq domain.
representing a power system area (curent structure).

4010
Different from conventional passivee loads, a power For solution to the second d technical challenge,
electronic load offers the option of circulating the although the control parameeters of the current
energy, which is especially beneficial for high power controller are modified for bettter stability within the
applications [14]. structure, the trade-off on the current tracking
capability is trivial compared within power system
The load emulator functions as a ccontrolled active
dynamics time scale.
rectifier mode. The input three-phasse currents are
controlled to behave like the references calculated from
the emulated load model. Three-phasse currents are
sampled in real time for closed-loop control; three-
phase voltages are sampled and phasee locked for dq
domain frequency and real-time inpput updates for
emulated model. The current control sysstem is classical
PI controller designed in dq frame wiith decoupled d
and q axes. The calculation for each sw witching interval
current reference and actual duty cycle are done within
the Texas Instruments DSP28335. Botth the emulated
model and controller are established in dq domain, and
all the variables in steady state are connstant dc values,
thus enabling easier modeling and higher control
accuracy.
However, compared with other cuurrent controlled Figure 5. Detailed example experrimental comparison
load emulators, there are several technnical challenges for induction motor emulatorr refs and outputs
(y axis as current in Amps, x axis as time in seconds).
for the load emulator developmennts within the
reconfigurable grid emulator, including:: Figure 5 illustrates an examplle of detailed induction
motor emulator experimentaal startup transients
1) Control stability within paralleeled converter waveform. As an example of dynamic
d power system
structure. load, the induction motor has h the time-variant
2) Detailed power system load perform
mance transients references with frequency of 60 Hz in dq domain. The
tracking capability. recorded waveform from DSP sh hows strong and robust
3) Accurate load model that closelly imitates the tracking capability of the emulator under either the fast
emulated object behavior. changing or constant references for different operating
points.
For the determination of each emulaator unit current
controller parameters, classical thhree-phase dq For solution to the third technical challenge, each
averaging circuit is analyzed for smalll signal stability load object model is establlished with classical
around a steady state operating pooint. The ideal mathematical equations. Underr the assumption of
bandwidth for the current controller forr the emulator is correct classical model, the emulator outputs would be
verified through comparison of simulated and
designed to be 2 kHz.
experimented outputs under samee input conditions.
However, so far no standard or em mpirical research
modeling method for the control desiign of multiple III. ZIP LOAD EMULATOR
M
paralleled ac three-phase converters is applicable yet.
The paralleled converters different behaviors are not In power system analysis, co
onventional static load
considered in the current controller deesign procedure. model is represented by a series of equations (shown
Thus, for the solution to technical chhallenge number below), which are called ZIP load. They stand for the
one, the trial-and-error method for modifying combination of constant impeedance (Z), constant
controller parameters based on the dessigned values is current (I) and constant power (P) loads in both real
used. Nevertheless, experimental resultts show that the and reactive power [7].
control parameters are functioning up to satisfactorily Commonly adopted in power flow calculation, ZIP
within the dynamic operating rangee of the grid model represents the combin nation of the static
emulator. characteristics of power system load. The mathematical

4011
polynomials are written as (1) and (2) under · ·
_ _ (3)
presumption of minor fluctuations around nominal grid
voltage [7]. Also, frequency dependency effect on the
· ·
load behavior is considered with the coefficients on _ _ (4)
real and reactive power.
When the grid voltage falls below a specified value
(1)
during a fault–caused sag, the polynomials described
above no longer represent the real power flow
(2)
condition, because of the relay protection and stall
effect of motors in the grid. Switching static load
In the equations, and stand for real and models to constant impedance expressions becomes
reactive power base value. and are coefficients necessary under that circumstance. , and , will
for constant impedance portions for real and reactive be set to zero and , will be adjusted in order to
power, while and are for constant current accommodate the change.
portions, and and are for constant power
portions. and are coefficients for the load
frequency dependency characteristics.

Figure 6. Emulated ZIP model inputs and


(a)
outputs relationship.
ZIP model could be used to fit the power
consumption curves for most kinds of power system
loads: higher order polynomials are neglected for
simplifying the model. In the meantime, the percentage
of Z, I and P for system load could be determined by
aggregating different load types and their usage within
the grid. By determining the values for , , and
, , , the static load operating state could be (b)
specified. It is always set such that the sum of , ,
is unity and the sum of , , , is unity.
As presented in Figure 6, the emulated model uses
voltages in dq domain and emulated grid frequency as
inputs for the ZIP model. External base real and
reactive power commands are given to the model for
specifying the actual power desired by customers.
Output currents are calculated using the equations as (3)
(c)
and (4) based on power equivalence relationship in dq Figure 7. ZIP load bus 9 emulator performance:
domain. The load emulator current behavior is (a) output P and Q for ZIP load emulator under voltage and
controlled accurately according to the references so as frequency variations experimental and simulation results
to ensure the power consumptions at the load bus to be comparison in broad time scale.
the ones described in (1) and (2). (b)(c) three phase currents and line to line voltage experimental
results during transients in micro-seconds time scale.

4012
As observed from Figure 7, the multi-time scale accurate load emulator representing its dynamic
observations of the dynamics could be illustrated in the behavior is crucial to the system overall loading
results. The test conditions are all in per unit values: behavior estimation.
keep the external command P0 and Q0 of 0.6 and -0.2
while vary the power flow of the testing network to
cause the load emulator terminal voltage (top right)
and frequency (top left) to fluctuate.
Figure 7 (a) shows the seconds-scale results
obtained from the central visualized operation recorder,
illustrating the power system dynamics slow-changing
trend and overall power flow amplitude variations. The
accurate correspondence between simulation and
experimental results (bottom left and right) proves the
successful emulation of the ZIP load. (b)(c) present the
transients of load increase and decrease in micro-
seconds time scale. Line to line voltage is in PWM
shape due to no capacitor filtering in the power circuit.
Obvious changes in the current amplitude could be Figure 8. Emulated induction model inputs
observed, corresponding to the power fluctuations and outputs relationship.
shown in (a).
The induction motor per unit model is also
established in dq domain as shown in (5) - (9).
IV. THREE-PHASE INDUCTION MOTOR Electrical inputs for the model are voltage in dq domain
LOAD EMULATOR and synchronous speed angle from the three-phase
voltage. The mathematical matrix equations for the
Three-phase induction motor is a very common model is described with relationships among voltage,
dynamic power system load. Consisting up to 95% of current, flux, torque, inertia and rotating speed. The
the industrial load, a motor is the major element real-time update of electro-magnetic torque and rotor
contributing to the dynamics and transients for load speed in mechanical model are necessary in calculating
variations. The high-current start-up process could accurate current references. In a nutshell, both electrical
induce visible grid voltage and frequency fluctuations, and mechanical variables are working together
while on the other hand, rich dynamic behavior details determining the system’s behavior.
could be observed under grid variations. Thus, an

⎡vqs pu ⎤ ⎡ rs
pu
ω pu ⋅ X ss
pu
0 0 − ω pu ⋅ X M
pu
0 ⎤ ⎡iqs ⎤
pu

⎢ pu ⎥ ⎢ ⎥ ⎢ pu ⎥
− ω pu ⋅ X ss − ω ⋅X M
pu pu pu pu
⎢vds ⎥ ⎢ rs 0 0 0 ⎥ ⎢ids ⎥
⎢v ⎥ ⎢pu
0 0
pu
rs 0 0 0 ⎥ ⎢i0 s ⎥
pu

⎢ 0 s pu ⎥ = ⎢ ⎥ ⋅ ⎢ pu ⎥
(ω pu − ωr ) ⋅ X M (ω pu − ωr ) ⋅ X rr
pu pu pu pu pu
v
⎢ qr ⎥ ⎢ 0 0 rr 0 ⎥ ⎢iqr ⎥
⎢v pu ⎥ ⎢− (ω pu − ω pu ) ⋅ X M pu 0 0 − (ω pu − ωr ) ⋅ X rr
pu pu
rr
pu
0 ⎥ ⎢idr ⎥
pu

⎢ dr pu ⎥ ⎢ r
⎥ ⎢ ⎥
⎢⎣v0 r ⎥⎦ ⎢⎣ rr ⎥⎦ ⎢⎣i0 r ⎥⎦
pu pu
0 0 0 0 0
⎡ X ss pu 0 0 XM
pu
0 0 ⎤ ⎡iqs pu ⎤
⎢ pu pu ⎥ ⎢ pu ⎥
⎢ 0 X ss 0 0 XM 0 ⎥ ⎢ids ⎥
⎢ 0 0 ⎥ 1 d ⎢i0 s ⎥
pu pu
0 X ls 0 0
+⎢ ⎥⋅ ⋅ ⎢ pu ⎥
0 ⎥ ωb dt ⎢iqr ⎥
pu pu
⎢X M 0 0 X rr 0
⎢ 0 XM
pu
0 0 X rr
pu
0 ⎥ ⎢i pu ⎥
⎢ ⎥ ⎢ dr pu ⎥
⎢⎣ 0 X lr ⎥⎦ ⎢⎣i0 r ⎥⎦
pu
0 0 0 0
(5)

(6) (7)

(8) (9)

4013
performance is not according to expectations. The
electro-magnetic torque keeps fluctuating and rotor
speed does not increase. The calculation results are not
accurate or stable while compared with Figure 9.

Figure 9. Induction motor model direct line free acceleration


startup process simulation results.
The simulation results based on the motor model
above is shown in Figure 9. The motor model is started
up with ideal voltage and frequency input conditions,
no motor drive adjustment and free acceleration. The
top left waveform shows the startup torque, while the (a)
top right one represents the rotor speed. The bottom
pictures illustrate the d and q axis stator currents. The
whole startup process takes around 4 s, which is
determined by both the induction motor electrical
parameters and also the mechanical inertia time
constant.

(b)
Figure 11. Induction motor startup process experimental
results with implicit trapezoidal method:
(a) dsp recorded results in dq comparison with simulation.
(b) oscilloscope results overall and detailed shape in abc
(constant current bases are added on top of motor references).
In order to improve the model’s calculation
Figure 10. Induction motor startup process experimental convergence capability within limited time step,
results using explicit Euler method(failed). explicit 4th order Runge Kutta was applied to the model.
Being the computing resource for the time-variant The convergence indeed was improved by the upgrade
coefficient ordinary differential equations containing of the numerical method, however, judging from the
vectors of 6x1 variables, a digital signal processor fluctuations and spikes on the current reference curves,
could not provide good calculation accuracy and the startup performances were not very robust/up to
stability based on simple discrete numerical method, satisfaction.
because of its limitation on the minimum time step size, Further upgrade of the numerical method to implicit
real-time fluctuations on the inputs and digital trapezoidal method is adopted. With less processing
processor data precision. Figure 10 shows the DSP time, same iteration time interval and high complexity
recorded real-time startup experimental results with the ODEs, the model’s performance is greatly improved.
explicit Euler method: the induction motor’s

4014
The real-time calculated results exhibit high accuracy [5] Wang, Jing, Yiwei Ma, Liu Yang, Leon M. Tolbert, and Fred
and robustness after repetitive tests of the whole startup Wang. "Power converter-based three-phase induction motor
load emulator." In Applied Power Electronics Conference and
process and the sudden jump of torque.
Exposition (APEC), 2013 Twenty-Eighth Annual IEEE, pp.
As shown in Figure 11, left top part shows the 3270-3274. IEEE, 2013.
comparison between d q currents references and outputs [6] Y. Srinivasa Rao and Mukul Chandorkar, “Electrical load
emulation using power electronic converters,” Proceedings of
while the right top one shows the calculated the IEEE International Conference on Industrial Technology,
electromagnetic torque and rotor speed (in per unit) Gippsland, Australia, 19-21 November 2008, pp. 1-6.
inside DSP. Bottom left and right waveforms present [7] P. Kundur, Power System Stability and Control. New York:
the voltage and frequency inputs (in real value) for the McGrawHill, 1994.
induction motor model during the experiment. [8] H. Slater, D. Atkinson, and A. Jack, “Real-time emulation for
power equipment development. Part II: The virtual machine,”
The stable and robust model references calculation Proc. Inst. Elect. Eng., vol. 145, no. 3, pp. 153–158, May
and good tracking capability of the load emulator’s 1998.
control algorithm are illustrated in Figure 11 with the [9] Yang, Liu, Xiaohu Zhang, Yiwei Ma, Jing Wang, Lijun Hang,
smooth transients and high alignment between the Keman Lin, Leon M. Tolbert, Fred Wang, and Kevin
Tomsovic. "Hardware implementation and control design of
references and the outputs. generator emulator in multi-converter system." In Applied
Power Electronics Conference and Exposition (APEC), 2013
V. CONCLUSION Twenty-Eighth Annual IEEE, pp. 2316-2323. IEEE, 2013.
[10] Y. Ma, L. Yang, J. Wang, F. Wang, and L. M. Tolbert, “Full-
This paper focuses on developing various load converter wind turbine emulator in converter based power grid
emulators with ZIP static and induction motor dynamic emulation system,” IEEE Applied Power Electronics
Conference and Exposition (APEC), Mar. 2013, in press.
models. By flexibly programming the load emulator’s [11] W. Cao, Y. Ma, J. Wang, L. Yang, J. Wang, F. Wang, and L.
models in the digital controller, different load behaviors M. Tolbert, “Two-stage PV inverter system emulator in
could be presented. Their applications in the power converter based power grid emulation system,” IEEE Energy
system emulation system of hardware test-bed could Conversion Congress and Exposition (ECCE), Sep. 2013, pp.
enable various scenarios emulation and dynamics 4518–4525.
forecast. [12] Cao, Wenchao, Yiwei Ma, Jingxin Wang, and Fred Wang.
"Virtual series impedance emulation control for remote PV or
wind farms." In Applied Power Electronics Conference and
ACKNOWLEDGEMENT Exposition (APEC), 2014 Twenty-Ninth Annual IEEE, pp.
411-418. IEEE, 2014.
This work was supported primarily by the [13] Emadi, Ali. "Modeling of power electronic loads in AC
Engineering Research Center Program of the National distribution systems using the generalized state-space
Science Foundation and the Department of Energy averaging method." Industrial Electronics, IEEE Transactions
on 51, no. 5 (2004): 992-1000.
under NSF Award Number EEC-1041877 and the [14] Y. Ma, L. Yang, J. Wang, X. Shi, F. Wang, and L. M. Tolbert,
CURENT Industry Partnership Program. “Circulating current control and reduction in a paralleled
converter test-bed system,” IEEE Energy Conversion
REFERENCES Congress and Exposition (ECCE), Sep. 2013, pp. 5426–5432.
[15] K. L. Shi, T. F. Chan, Y. K. Wong, and S. L. Ho, “Modeling
[1] Wang, Jing, Liu Yang, Yiwei Ma, Xiaojie Shi, Xiaohu Zhang, and simulation of an induction motor,” Int. J. Elect. Eng.
Lijun Hang, Keman Lin, Leon M. Tolbert, Fred Wang, and Educ., vol. 36, no. 2, pp. 163–172, 1999.
Kevin Tomsovic. "Regenerative power converters [16] L. Yang, J. Wang, Y. Ma, J. Wang, X. Zhang, W. Cao, L.
representation of grid control and actuation emulator." In Hang, L. M. Tolbert, F. Wang, and K. Tomsovic,
Energy Conversion Congress and Exposition (ECCE), 2012 "Development of converter based reconfigurable power grid
IEEE, pp. 2460-2465. IEEE, 2012. emulator," IEEE Energy Conversion Congress and Exposition
[2] H. Slater, D. Atkinson, and A. Jack, “Real-time emulation for (ECCE), Sep. 2014.
power equipment development. Part II: The virtual machine,” [17] Yang, Liu, Xiaohu Zhang, Yiwei Ma, Jing Wang, Lijun Hang,
Proc. Inst. Elect. Eng., vol. 145, no. 3, pp. 153–158, May Keman Lin, Leon M. Tolbert, Fred Wang, and Kevin
1998. Tomsovic. "Stability analysis of inverter based generator
[3] M. Armstrong, D. J. Atkinson, A. G. Jack, and S. Turner, emulator in test-bed for power systems." In Energy
“Power system emulation using a real time, 145 kW, virtual Conversion Congress and Exposition (ECCE), 2013 IEEE, pp.
power system,” in Proc. Eur. Conf. Power Electron. Appl., 5410-5417. IEEE, 2013.
2005, 10 pp. - P.10. [18] J. Wang, L. Yang, C. Blalock, and L. M. Tolbert, “Flywheel
[4] A. Emadi and M. Ehsani, “Multi-converter power electronic energy storage emulation using reconfigurable hardware test-
systems: definition and applications,” in Proc. IEEE 32nd bed of power converters,” Electrical Energy Storage
Power Electron. Spec. Conf., Vancouver, BC, Canada, Jun. Applications and Technologies (EESAT), San Diego, CA,
2001, pp. 1230–1236. USA, Oct. 20-23, 2013.

4015

You might also like