You are on page 1of 5

Fault Induced Delayed Voltage Recovery (FIDVR):

Modeling and Guidelines


Sarina Adhikari * Niroj Gurung Alex Mogilevsky
Jens Schoene Commonwealth Edison Company CEATI International
EnerNex, A CESI Company Chicago, IL USA Quebec, Canada
Knoxville, TN USA alex.mogilevsky@ceati.com
*Contact: sadhikari@enernex.com

Abstract—Voltage sensitive loads are broadly classified as static The concept of composite load model is considered a state-
or steady state, and dynamic load types. After several years of of-art dynamic load model [1]. It is important to replicate the
research and development, Bonneville Power Administration behavior of the increasing amount of single phase and three
(BPA) and Western Electricity Coordinating Council (WECC) phase motor loads along with other types of static and
developed the Composite Load Model structure comprising of electronic loads in industrial, commercial and residential area.
different types of static loads, electronic loads, three phase and Bonneville Power Administration (BPA) and Western
single phase compressor motors. Single phase compressor Electricity Coordinating Council (WECC) developed
motors are the main drivers of the phenomenon termed as Fault Composite load model that has been implemented in PTI
Induced Delayed Voltage Recovery (FIDVR). In the present
PSS/e as well as GE PSLF software platforms. This load
work, we developed a computer simulation model in GE PSLF
that could replicate the FIDVR event in the Valley substation of
model comprises of ZIP loads, electronic loads, three different
Southern California Edison (SCE) that occurred in 2004. We types of three phase motor loads, and single phase air
also performed some sensitivity analysis of various parameters conditioner (A/C) compressor motor loads [1].
of three-phase and single phase compressor A/C motor loads The phenomenon in which the system voltage remains at
and produced a guideline showing some typical values of excessively low levels for the duration of several seconds,
different parameters to replicate the real world FIDVR events. even after the clearing of a transmission, subtransmission, or
At last, we have presented a broad guideline for steady state
distribution fault are classified as FIDVR [1]. FIDVR events
cascading simulations. This paper is based on a project report
are believed to be caused by a large percentage of single phase
prepared for the CEATI International Power System Planning
and Operations Program, which consists of North American induction motor loads with constant torque characteristics that
electrical utilities focused on Transmission Systems. stall when the system voltage dips to some lower values
during the fault conditions. These types of motors draw a large
Index Terms—Fault Induced Delayed Voltage Recovery amount of reactive power from the grid. In this stalled or
(FIDVR), Voltage Sensitive load, Single phase A/C compressor locked-rotor condition, these motor draw 5-6 times their
motors, sensitivity study, general guideline steady state current. This might lead to voltage collapse and
several cascading events in the system. In [2], various impacts
I. INTRODUCTION of FIDVR events to the grid and the distribution customers are
discussed in detail. The effects could be related to power
Based on the behavior of voltage sensitive loads during quality issues, opening of automatic reclosers due to high
normal and disturbance conditions, voltage sensitive loads are currents or load imbalance or, even tripping off of the
classified into static or steady state, and dynamic load types. distribution circuits due to current imbalance caused by
The static or steady state loads such as resistive and lighting uneven distribution of stalled air conditioner (A/C) units.
loads are represented as the ZIP models in which Z represents
constant impedance loads, I represents constant current, and P A short description on the FIDVR phenomenon is
represents constant power loads. Dynamic models are capable provided in Section II. In Section III, a brief review on the
of capturing the dynamic behavior of loads during disturbance characteristics of different types of voltage sensitive loads that
conditions. Composite load models are models that capture the impact the recovery of voltage following any types of system
behavior of static, electronic, and dynamic loads. The disturbance is presented. Building upon several literature in
composite load models accurately aggregate different types of this area, this paper presents some results of modeling and
loads and help in understanding the behavior of different loads replicating the real world FIDVR event in Section IV. The
under various system contingency conditions which, as a main contributions of the paper are to provide a detailed
result, help in understanding the system impacts due to voltage guideline for modeling FIDVR types of events and a guideline
sensitive load loss. for the simulation of cascading events caused due to FIDVR
events. The parametric sensitivity study and guidelines are

Centre for Energy Advancement through Technological Innovation


(CEATI International)

978-1-7281-1981-6/19/$31.00 ©2019 IEEE


Authorized licensed use limited to: University of Exeter. Downloaded on July 16,2020 at 18:31:55 UTC from IEEE Xplore. Restrictions apply.
discussed in Section V and VI respectively. Lastly, Section voltage sensitive loads that play important role in FIDVR
VII provides concluding remarks and some recommendations events are described below [1].
based on the outcome of this work.
A. Three phase Induction Motors
II. PHENOMENON OF FAULT INDUCED DELAYED VOLTAGE Three-phase induction motors are usually larger motors
RECOVERY (FIDVR) supplied with three-phase voltages. The key elements that
Figure 1 shows a typical delayed voltage recovery profile determine their dynamic response during events such as
on a 230 KV circuit located in Southern California following a voltage sags and swell are inertia, speed-torque
fault [3]. As observed in the figure, the voltage decreased to characteristics, load-torque characteristics, and their
around 79 percent of the nominal voltage (point 1) following protection scheme. The inertia time constant of three phase
the fault. This reduction in voltage caused the air conditioning motors can vary from as low as 0.1 second to as high as 1
units to stall. These stalled A/C units limited the system second. Motors with higher inertia are more likely to ride
voltage from the nominal voltage level (point 2). As the through a voltage dip caused by a system disturbance.
thermal overload protection switches of the A/C units operated Three-phase induction motors have a local control board
by tripping the units off of the connected system, the voltage and motor contactors that drop out between 55-65% of the
recovered to the value higher than the nominal by around 6 nominal voltage [1]. These motors generally stall at around
percent in this case (point 3). 50% voltage. Therefore, it is very likely that the contactors
drop out before the motor stalling phenomenon. Some of the
4. LTC/CAPS Off due to Overvoltage
three-phase motors might also have undervoltage protection
5. Load coming
which operates around 2 seconds for 80% voltage dip. Large
+6%

back On
Nominal Voltage
three-phase motors such as the chiller motors have an
3. Overshoot due to
LTC/CAPS advanced control board with different types of protection
‐21%

6. Undervoltage due to caps still schemes such as undervoltage, overcurrent, and phase
off line
unbalance.
2. Delayed Recovery as Stalled A/
Cs Disconnect Via Thermal
Overload Protection
B. Single-phase Residential A/C Motors
1. Fault & Fault Motors driving air compressors for residential cooling are
Clears
typically supplied by single-phase voltage. There are
differences in dynamic characteristics of single-phase
residential A/C motors compared to the large three-phase
Figure 1. Typical delayed voltage recovery – this event was recorded induction motors used in commercial and industrial scale
following a 230 kV transmission fault in Southern California [3]. These single-phase loads make up a significant portion of the
overall load during peak-summer demand and, consequently,
The overshooting of the voltage response happened as they play a critical role in selecting proper cable sizes, switch
some of the motor loads were taken out of the system whereas ratings, and other system design decisions. Single-phase
the capacitor banks are still online. The overvoltage condition residential A/C motors are believed to be the main driver for
resulted in the tripping of the capacitor banks and the FIDVR events and it is therefore, very important to
operations of the load side transformer taps as shown in point understand the dynamics of these motors during grid faults.
4. After some interval of time, the loads, including the A/C The inertia of single-phase motors used in residential air
loads, began to reconnect back to the system. Since some of conditioning is very small (0.03-0.05 seconds). This leads to
the capacitor banks had already tripped, the system voltage
rapid deceleration of the A/C motors during the grid events in
went below the nominal voltage as shown by points 5 and 6 in
the figure. These types of conditions make the system which the voltage is suppressed. Therefore, these motors are
vulnerable to other cascading events. The light inertia highly susceptible to stalling even for the faults that are
induction motors, such as air conditioner motors, lose their cleared in as low as 3 cycles, according to the laboratory tests
speed rapidly following the voltage dip conditions. Such conducted by SCE and BPA [5,6].
motors will stall for all faults with the duration more than five There are two types of protection schemes equipped in
cycles and the voltage dip magnitude less than 60% of the single-phase induction motors: (1) contactors, and (2) thermal
nominal voltage [4]. protection relay. The contactors trip the motors off the grid if
the voltage drops below 40%. In practice, the stalled voltage
III. CHARACTERIZATION OF VOLTAGE SENSITIVE LOAD is typically higher than 40% and thus, the contactors do not
Voltage sensitive loads like motor loads with undervoltage operate. Even in situations when the contactor trips the
protection scheme disconnect from the system if the supply compressor, it was found that it delays the stalling of the
voltage deviation exceeds pre-set limits. Various types of motors but cannot completely prevent the stalling
voltage sensitive loads that exist in the present system are dc phenomenon; the compressor motor stalls as the contactor
drives, A/C compressor motors, adjustable speed motor drives recloses. The thermal protection relay is an inverse-time
(ASDs), motor contractors, electromechanical relays, and overcurrent relay that trips off the motor during stalling after
personal computers. These loads are highly sensitive to a time-delay determined by the protection curve [1].
voltage sags and swells. The characteristics of some types of

Authorized licensed use limited to: University of Exeter. Downloaded on July 16,2020 at 18:31:55 UTC from IEEE Xplore. Restrictions apply.
IV. COMPUTER SIMULATION MODEL DEVELOPMENT
This section provides a detailed description of the
development of a computer simulation model that could
replicate the real world FIDVR event. The composite load
model in GE PSLF platform is utilized to model the dynamic
loads in the system. The composite load model comprise of
four different types of motor loads. Motor A represents small
low-inertia three phase compressor motors, Motor B
represents fan motors used in residential and commercial
Figure 3. Measured vs simulated voltage profile of 115 kV bus (Bus 2) of
buildings such as ventilation fans in buildings or air handler Valley model.
fans. These motors usually have high inertia (0.25-1.0 secs).
Motor C represents direct connected pump motors in
commercial buildings with low inertia in the range of 0.1-0.2
secs. Lastly, Motor D represents single-phase compressor
motors in residential and small commercial cooling and
refrigeration with very low inertia (0.03-0.05 secs).
The summer 2004 FIDVR event at the Valley Station of
the Southern California Edison (SCE) system is considered
the benchmark for the simulation [7,8]. The available data
related to the load composition in the Valley system is also
a) b)
taken from [7]. This paper, however, does not provide the Figure 4. Bus voltage profiles in the valley model: a) without modeling the
detailed system parameters of the employed model. The motor load tripping characteristics; b) all the loads modeled as ZIP loads.
Valley system model as shown in Figure 2, with some typical
system parameter values, was developed in GE PSLF In this case, the voltage cannot recover even after the fault
platform. The total active power and reactive power load is cleared as the stalled motors are not removed from the
served by the Valley station before the fault scenario is system, and there is a possibility of system wide voltage
considered to be 908 MW and 187 MVar respectively collapse, as can be observed from the figure. Therefore, the
according to the data provided in [7]. lack of voltage sensitive load loss will lead to the system
level voltage collapse that shows entirely different system
steady state responses as compared to the system with voltage
sensitive load loss which is close to the behavior of the
practical systems.
Figure 4 b) shows the voltage profiles of the Valley buses
when all the loads in the Valley system are modeled only as
static ZIP loads. In this case, the system voltages show rapid
recovery after the fault. This result is also quite different than
the one observed in the real FIDVR events. This shows that
representation of real world load types with the accurate load
models is very important in order to obtain the real-world
dynamic and steady state responses in the computer
simulation models.
Figure 2. Oneline diagram and data considered to model SCE’s Valley V. PARAMETER SENSITIVITY ANALYSIS
Station. The WECC composite model is very comprehensive and is
A three-phase-to-ground fault is applied at Valley 115 kV comprised of more than 130 parameters. Therefore, it is
Bus 2 at time, t = 0.5 seconds. The fault clears after 4 cycles crucial to understand the sensitivity of some of the important
at t = 0.5666 seconds. parameters of the composite load model in overall load
Figure 3 shows the voltage profile at 115 kV bus (Bus 2) response [9].
of the Valley model from the PSLF simulation as compared For the parametric sensitivity analysis, a three phase to
to the one measured by the Phasor Measurement Unit (PMU) ground fault is applied at the 115 kV bus at t = 0.5 secs and it
during the FIDVR event. It took nearly 30 seconds to recover clears at t = 0.5666 secs which is after 4 cycles.
the voltage back to the pre-fault voltage level. The developed Table 1 shows the fraction of different types of motor loads
PSLF system model with the chosen composite load model in composite load model considered for the sensitivity study.
parameters is able to closely imitate the behavior of this In Scenario 1, fraction of Motor A, B, and C combined is
phenomenon, as seen from this figure. higher that the fraction of Motor D.
Figure 4 a) shows the voltage profiles of different buses in
the Valley model in which the tripping characteristics of the
motor loads are not modeled.

Authorized licensed use limited to: University of Exeter. Downloaded on July 16,2020 at 18:31:55 UTC from IEEE Xplore. Restrictions apply.
Table 1 Fraction of motor loads for sensitivity study The thermal time constant of single-phase A/C motors is
another critical parameter while modeling the FIDVR events.
It is clear from Figure 6 b) that, as this time constant
decreases, the system voltage is capable of rapidly recovering
to the normal operational voltage range. This time constant is
highly dependent on the design of the compressor motors.
In Scenario 2, fraction of Motor A, B, C combined is equal Table 2 shows typical values of different important
to the fraction of Motor D, and in Scenario 3, fraction of parameters in the composite load model in order to model the
Motor D is higher than the fraction of Motor A, B, and C real world FIDVR events based on the parametric analysis.
combined. Only some of the selected results from the
Table 2 Typical values of Motor D parameters to model FIDVR events
parametric sensitivity analysis are presented in this paper.

a) b)
Figure 5. Sensitivity of a) first low-voltage trip delay time (secs) in Scenario
1; b) thermal protection trip start and completion level in Scenario 2.

Figure 5 a) shows the sensitivity of first low-voltage trip


delay time of Motors A, B and C for Scenario 1. It can be
observed that the system is capable of rapidly recovering the
voltage when the trip delay time is less which means that the
three phase motors trip sooner than later during the low
voltage condition. Figure 5 b) demonstrates the effects of
thermal protection trip start and completion level on the
system voltage recovery in Scenario 2 which has equal VI. GENERAL GUIDELINES ON MODELING FIDVR EVENTS

fractions of three phase and single phase A/C motor loads. The sensitivity analysis performed in Section V
The sooner the thermal protection trip starts, the sooner the demonstrated that the system voltage response is more
system can rid itself of those stalled motors that consume sensitive to some parameters and less sensitive to others. The
large amounts of reactive power, hence faster voltage most sensitive parameters of the composite load model are
recovery. identified in Table 2. One of the important parameters are
fractions of different types of motor loads. Therefore, the
composition of load in any system is the most important
variable to be determined. Three-phase motor loads are less
susceptible to stall phenomenon during low voltage
conditions because of their high inertia and because the
thermal protection kicks in before reaching the stall voltage.
The choice of Tstall should be carefully determined while
modeling the FIDVR events. It is recommended that the
a) b) chosen Tstall should be lower than 0.06 seconds. A higher
Figure 6. Sensitivity of a) Tstall in Scenario 3; b) Motor D thermal time Tstall value will not be able to model the realistic behavior of
constant in Scenario 3.
the stalled single-phase motor loads.
Figure 6 a) shows the impact of stall time delay, Tstall (in
The contactor voltage at which tripping starts and
seconds) on the system voltage profile. Note that the curves
completes should be chosen such that the stalled motors are
with lower values of Tstall are overlapped with each other. It
removed as soon as the voltage dips below certain threshold.
is clear that if the Tstall is a high value, the single phase A/C
If this voltage threshold for the motor tripping is considered
compressor motors do not stall until this time delay is
as a very low value, the stalled motors continue to consume
reached; the practical FIDVR phenomenon cannot be
huge amount of reactive power that will lead to a system wide
captured in the model. Therefore, it is very important to
voltage collapse.
choose this parameter precisely and appropriately while
The A/C compressor motor thermal time constant is
modeling different load compositions of any system. A
another critical parameter that impacts the voltage recovery
typical range of Tstall is from 0.03 to 0.033 seconds.
during FIDVR events. The voltage recovery is faster for
lower time constants. The thermal time constants vary

Authorized licensed use limited to: University of Exeter. Downloaded on July 16,2020 at 18:31:55 UTC from IEEE Xplore. Restrictions apply.
between different designs of the A/C motor loads. This is a  Sensitivity analysis of the effects of different
parameter that can be improved with better design. composite load model parameters on the system
In summary, the most sensitive parameters that affect voltage response
voltage response of a system with large percentage of single-  A general guideline about the most sensitive
phase A/C compressor motors fall into the following composite load model parameters, typical values of
categories: those parameters and a guideline on simulations of
1. Thermal contactor trip characteristics steady state cascading
2. Thermal time constants
3. Undervoltage protection characteristics In this paper, we analyzed the behavior of the aggregated
4. Single-phase machine stall characteristics (Vstall, composite load model in simple aggregated test systems. It is
Tstall) also observed that the actual practical system events cannot be
Several series of simulation cases were studied to extract replicated without modeling the loss of voltage-sensitive load
some important requirements in the modeling platform for in the system. While it is understood that the FIDVR
steady state simulations of the FIDVR events. The results of phenomenon occurs due to higher percentage of stalled single-
these cases are not presented in this paper. However, a phase motor loads in a system, the study on the ways to
summary of general guideline for steady-state cascading mitigate such phenomenon is another very interesting and
analysis based on those results are listed below: important area of research and investigation. It is highly
recommended to perform such studies in continuation of this
 In any bulk power systems with a large percentage work to develop a better understanding of the behavior of
of single phase A/C motors, the amount of voltage these loads and to find out the ways to protect the bulk power
sensitive load lost during the FIDVR type events is system from system-wide voltage collapse and cascading
quite high. It can be as high has 43% of the total events. In addition to the impact of voltage-sensitive load loss
load of the system based on the performed into the system, the impact of voltage-sensitive generation loss
simulations. is another important element during FIDVR-type events.
 If the voltage-sensitive load loss is improperly Therefore, this study should be extended to investigate further
modeled in the dynamic simulation environment, the on the generation loss.
results obtained from the computer simulation
models will be inconsistent with the historical event REFERENCES
results obtained in the real world systems. Therefore, [1] NERC Load Modeling Task Force, "Technical Reference Document on
modeling of accurate combination of loads with Dynamic Load Modeling," NERC, December 2016, [Online] Available:
https://www.nerc.com/comm/PC/LoadModelingTaskForceDL/Dynami
appropriate tripping and reconnection characteristics c%20Load%20Modeling%20Tech%20Ref%202016-11-14%20-
is critical in any dynamic simulations. %20FINAL.PDF (Last Accessed: Nov 05, 2018).
 The dynamic simulation should be performed from [2] R. Bravo, R. Yinger and P. Atons, "Fault Induced Delayed Voltage
pre-fault to 30 minutes for the bulk-power systems Recovery (FIDVR) Indicators," in Proc. IEEE PES T&D Conference
and Exposition, 2014
with large percentage of single-phase A/C motor [3] R. Bravo, R. Yinger, D. Chassin, H. Huang, N. Lu, I. Hiskens, and G.
loads prone to stalling as the voltage at the load Venkataramanan, “ Final Project Report - Load Modeling Transmission
interconnection point goes beyond 90%. This Research”, Lawrence Berkeley National Laboratory (LBNL), March
timeframe will capture the operations of all the 2010, [Online] Available:
https://web.eecs.umich.edu/~hiskens/publications/LM_Final_Report.pd
system elements that are expected to be lost, restored f (Last Accessed: Nov 05, 2018)
or changed in status prior to attaining the new steady [4] B. R. Williams, W. R. Schmus and D. C. Dawson, "Transmission
state conditions. voltage recovery delayed by stalled air-conditioner compressors," IEEE
Trans. on Pow. Syst., vol. 7, no. 3, pp. 1173-1181, 1992
 For the scenarios and contingencies where the [5] J. Jayanth and H. Pham, "Residential AC Compressor: Low Line
voltage at the generation interconnection points Voltage Behavior," DOE, 2008.
could fall below 90% of the nominal system bus [6] R. Bravo, J. Wen and B. P. &. Y. R. Kosterev, "WECC Modeling and
voltage, voltage-sensitive generation can be lost. Validation Task Force: Air Conditioner Motor Model Test Report",
[online] Available:
Therefore, the results of the steady-state cascading https://www.wecc.biz/Reliability/WECC%20Air%20Conditioner%20
analysis would be inaccurate if we do not consider Motor%20Model%20Test%20Report--%20Final.pdf (Last Accessed:
the voltage-sensitive generation loss in addition to Nov 05, 2018)
the voltage-sensitive load loss in the simulations. It [7] G. L. Chinn, "Modelling Stalled Induction Motors," in Proc. IEEE PES
Transmission and Distribution Conference and Exhibition, 2006
is recommended and proposed that a future study [8] S. Adhikari, F. Li, H. Li, Y. Xu, J.D. Kueck, and D.T. Rizy,
should consider such generation loss during dynamic “Preventing delayed voltage recovery with voltage-regulating
simulations. distributed energy resources”, in Proc. IEEE Bucharest PowerTech, pp.
1-6, 2009
VII. CONCLUSIONS AND RECOMMENDATIONS [9] Mitsubishi Electric Power Products, Inc. (MEPPI), "Final Report- An
The major outcomes of this work are summarized below: Analysis of the Sensitivity of WECC Grid Planning Models to
Assumptions Regarding the Composition of Loads," MEPPI, 2016
 A validated computer simulation model with the
WECC composite load model developed in GE
PSLF V 21.0_04 platform.

Authorized licensed use limited to: University of Exeter. Downloaded on July 16,2020 at 18:31:55 UTC from IEEE Xplore. Restrictions apply.

You might also like