See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/282348751

Modeling of Deadband Function of Governor Model and its Effect on Frequency

Response Characteristics

Conference Paper · October 2015

DOI: 10.1109/NAPS.2015.7335089

CITATIONS READS
9 1,427

4 authors, including:

Saeed Mohajeryami Iman Moghaddam

University of California, San Diego University of North Carolina at Charlotte
46 PUBLICATIONS   300 CITATIONS    13 PUBLICATIONS   116 CITATIONS   

SEE PROFILE SEE PROFILE

Zia Salami
University of North Carolina at Charlotte
7 PUBLICATIONS   34 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Improving Customer Baseline Load (CBL) calculation for residential customers View project

Lightning Transients Analysis View project

All content following this page was uploaded by Saeed Mohajeryami on 01 October 2015.

The user has requested enhancement of the downloaded file.

 Modeling of Deadband Function of Governor Model
and its Effect on Frequency Response Characteristics
Saeed Mohajeryami, Aanand R. Neelakantan, Iman N. Moghaddam, Zia Salami
Electrical and Computer Engineering,
University of North Carolina at Charlotte
Charlotte, NC 28223, USA
Email: {smohajer,aneelak1,inazirim,zsalami}@uncc.edu

Abstract—Frequency Response has been the focus of increased simulations that assess the reliability. This standard aims to
attention and analysis recently due to its importance to the assure that the parameters accurately represent the real power
reliability of the network. The models used by the utilities proved response of the generator unit to system frequency variation
to be insufficient and the inability to duplicate events using the [12].
existing models and databases has led to several initiatives to
The modeling of the governor/turbine is one of the primary
address these shortfalls. In this work, the effect of the dead-band
in the governor/turbine model is investigated. For this purpose, concerns in several NERC reliability regions, particularly in
IEEEG1 model is used to study the frequency response of the Eastern interconnection. According to Eastern Interconnection
steam generators. Many of the existing industrial software used Frequency Initiative Whitepaper, the majority of the utilities
by the utilities ignore the dead-band function in their governor that responded to the generator survey, have generator
model. Using the incorrect parameters for the dead-band parameter values far from the standard values recommended
function can lead to the significantly incorrect frequency by NERC [13].
response, it is even worth in case of ignoring the dead-band Because of the continuous load variation, the frequency
function. In this paper, the effect of the dead-band setting on the varies constantly in the power system. In the absence of a
frequency response is studied. The response of the generator
major disturbance, these variations are innocuous. But the
during the frequency deviation is analyzed both with and without
the dead-band. ETAP is used as a good candidate for an response to the small frequency variations can cause wear and
industrial software to study the effect of the frequency deviation tear in steam boilers and turbines. Therefore, a dead-band
on the generator response. Ignoring the dead-band is shown to controller in the governor is a necessity. The dead-band
depict a very optimistic picture of the frequency response. controller does not let the system to respond to the frequency
deviations that do not exceed a pre-defined value. So, the
Keywords—Primary frequency response; IEEEG1 governor/
turbine model; modeling; dead-band; MOD-027; ETAP speed governor is inactive within the dead-band limits [14].
Conversely, the dead-band controller can affect the stability of
I. INTRODUCTION the power system. [15] studied the dead-band effect on the
NERC report of August 14th, 2003 about blackout in eastern stability and showed that in South Eastern Europe, the
interconnection shows a huge discrepancy between what governor dead-band can drastically change the stability of the
utilizes had expected based on the simulation of their models European inter-area oscillation mode and create the limitation
and what was observed [1]. There are many speculations about in the response. [16] discussed how to compensate the adverse
the reasons for this discrepancy in the technical literature [2- effect of the dead-band on the stability by introducing a
6]. The modeling of the different components of the power frequency bias parameter in the state space model of the
system is mentioned to be one of the major contributing governor. In [17-18], load dispatch concerns regarding
factors [7-11]. Among others, the frequency response is one of frequency response variation is investigated. Relation between
the main issues. There are many parameters in the faults, transients and frequency response are addressed in [19].
governor/turbine models which lack enough accuracy to model Western Electricity Coordinating Council (WECC)
the physical governor/turbine. recording of its large generation trips showed that it only
In response to the blackout, NERC approved 14 observed 40% of the simulated governor response. Among
recommendations to address the shortcomings that contributed other things, the dead-band is mentioned as one of the
to the blackout. Incidentally, it reports that after-the-fact contributing factors [20-21].
models developed to simulate the blackout indicate that the So far, the reduced-form of the dead-band model is used for
dynamic modeling assumptions, including generator the study of the effect of the dead-band on the stability. The
parameters, were inaccurate [1]. This report initiated several linearized reduced-form arguably cannot capture the
standards to address the modeling issues in different nonlinearity of the dead-band. In this work, the full model of
components of the power system [11]. MOD-027 is a NERC the dead-band is used in the newly introduced user-defined
standard which mandates the generator owners to test their module (UDM) of ETAP. This module enables the users to
generator components and then based on the test results verify create the sophisticated models that do not exist in the ETAP
the turbine/governor model parameters used in dynamic library.
 Fig. 1. IEEEG1 model

The paper will continue with the description of the governor

and dead-band model in section II. Section III describes the
case study and the results for the test case and discusses the
results. Section IV closes the paper with drawing conclusion
from the provided discussion and results.

II. MODELING

A. Governor model
The IEEEG1 model is the IEEE recommended model for
steam turbine. There are many other legacy models
recommended by IEEE, but IEEEG1 is more detailed and also
it entails more parameters that can accurately model the
governor/turbine. The model is shown in Fig. 1. The model is
used in ETAP UDM to model the governor/turbine. It’s Fig. 2. Developed Governor/Turbine model in ETAP UDM
illustrated in Fig. 2. IEEEG1 is a general model of the steam
turbine system with both the speed governor and the turbine
stages. In this model, the dead-band block is included. The TABLE 1: IEEEG1 PARAMETERS
dead-band block represents the intentional dead-band that the PARAMETER DESCRIPTION VALUES
plant control system incorporates into its working to prevent
DB Deadband block [Hz] 0.036
the governor from responding to the tolerable changes in the
frequency [22]. Table 1 contains the parameter values used in K Governor Gain (1/droop) [pu] 20.0
the developed IEEEG1 model. These values are the real
parameter values of one of the steam power plants in the T1 Lag Time Constant [s] 0.82
eastern interconnection. T2 Lead Time Constant [s] 0
B. Implementation of Droop characteristics
T3 Valve Position Time Constant [s] 0.97
The droop characteristics is used to control the magnitude
Uo Maximum Valve Opening Rate [pu/s] 1.38
of governor response for a given change in frequency. With the
dead-band involved, there are two ways of implementing the Uc Maximum Valve Closing Rate [pu/s] -1.25
droop characteristics: step implementation and non-step
implementation. Pmax The maximum output of the Unit [pu] 1.03

In step implementation of the droop characteristics, the Pmin The Minimum output of the Unit [pu] 0
regulation curve will step into the droop characteristics outside
T4 Time Constant for Steam Inlet [s] 0.80
the dead-band. Through this setting, the governor response
would step into the actual response that one would expect T5 Time Constant for Second Boiler Pass [s] 6.36
without the dead-band in place.
T6 Time Constant for Third Boiler Pass [s] 11.14
Mathematically the step implementation can be represented
as follows: T7 Time Constant for Fourth Boiler Pass [s] 0.86
K1 HP Fraction 0.19
K3 HP Fraction 0.22
K5 HP Fraction 0.21
K7 HP Fraction 0.38
=y 0 − D ≤ x ≤ +D in the main grid. The test case is inspired by a real system in an
 eastern interconnection
 y = − kx x ≥ +D (1)
To analyze the effect of deadband on the frequency
 y = + kx x ≤ −D
 response, the generator was equipped with two governor
models; one without the deadband model and the other one
Where k is the slope of the characteristics = 1/R (R = Droop with the deadband model using the non-step implementation of
Setting. Usually 5%) and D is a dead-band. NERC standards the droop characteristics. The reason for choosing the non-step
require the plant owners to have a dead-band not more than implementation was that the inspiring real model used as a
±0.036 Hz. The step implementation is represented in Fig. 3a.
In the non-step implementation of the droop, the
characteristic curve starts from the dead-band and does not step
into the standard curve. Through this implementation, the
starting point of the curve will be at the dead-band and the
slope will be at the regulation value (usually at 5%).
Mathematically the non-step implementation can be
represented as follows:

=y 0 − D ≤ x ≤ +D

 y = −k ( x − D) x ≥ +D (2)
 y = + k ( x + D) x ≤ −D

Where k is the slope of the characteristics = 1/R (R =
Droop Setting. Usually 5%) and D is a dead-band. The non-
step implementation of the droop characteristics is shown in
Fig. 3b. Fig. 3a. Step implementation of droop characteristics

III. RESAULTS AND DISCUSSION

A. Case study
In this section, a small microgrid consisting of a generator
and 4 loads is selected as a test case. Microgrids are small-
scale form of an electric network and typically include
combination of distributed generators and interconnected loads.
Standby diesel generators, fuel cells, micro turbines, wind
generators, and solar panels are typical types of generators in
microgrids [9]. When a microgrid is connected to the main
electrical network, it can work either in a grid-connected mode
or an island mode. Grid-connected is a normal operation mode
for a microgrid. However, under some circumstances,
microgrids must function as an island or off-grid mode. The Fig. 3b. non-step implementation of droop characteristics
switching of grid connection can either be done automatically
(because of a fault) or manually. They will be disconnected
from traditional network to operate separately for grid
disturbance mitigations and grid resilience improvement.
Microgrids are able to work continuously to mitigate grid
disturbances and increase flexibility of power systems.
Consequently, they increase penetration of renewable energy
resources in the grid. To observe the effect of the deadband in
modeling of governor, a microgrid in an islanding mode has
been picked. In the main grid, the effect of deadband under a
small fault is not big enough. In case of a blackout, there would
be a cascading faults one after another and in that case, the
effect of deadbands are pronounced enough to be observed.
Since it is hard to model the whole grid, a microgrid has been
picked for this study. The effect of a small fault in a microgrid
is as big as combined faults in the moments before a blackout Fig. 4. Test Case Single Line Diagram
model of this test case has implemented its droop in the same output to support the frequency. The response from the
way. The frequency response from the generator with and generator in terms of its mechanical power and corresponding
without the deadband is then compared. speed was recorded for analysis.
A.1. Initial Condition B. Results and Discussion
The generator of the system is rated at 711MW (790MVA B.1. Case 1
at 0.9p.f) and is set to supply a total grid load of 370MVA
represented by four lumped loads rated at 150MVA, 140MVA, The generator is equipped with a governor that has no
50MVA and 30 MVA respectively. Initially the system is set to deadband setting in its system. The response from the
supply 351MW of power (working with %50 of nominal generator is shown in Fig. 5. If observed, since the governor is
power) to the system with the grid loads taking up 342MW not reflecting the deadband setting, the generator responds to
(370 MVA with power factor of 0.95). The grid loads are split even the slightest change in frequency. The response continues
among two buses with each bus connected to the point of till the speed (frequency) settles down at 3600 rpm (60Hz).
interconnection with transmission lines represented by B.2. Case 2
impedances.
The generator is equipped with the governor that has the
A.2. Test Condition deadband setting incorporated into its model. The deadband is
set at ±2 rpm (±0.033Hz) with non-step implementation of
To test the generator for its frequency response, a 3 phase
fault is simulated on one of the buses in the grid side and is droop characteristics. The response of the generator is shown
cleared in 0.1 seconds. This leads to an increase in grid
frequency, for which the governor responds by changing its

Fig. 7. Governor response with 20 rpm dead-band setting

Fig. 5. Governor response without dead-band setting

Fig. 8. Comparison of mechanical power

Fig. 6. Governor response with 2 rpm dead-band setting Fig. 9. Comparison of generator speed
in Fig. 6. It can be observed that the response is much smaller REFERENCES
than the response of the case without deadband and the speed [1] North American Electric Reliability Corporation (NERC), "August 14,
settles down at a value outside the deadband value of 2rpm. 2003 Blackout: NERC Actions to Prevent and Mitigate the Impacts of Future
Having 2rpm deadband helps the generator not change its Cascading Blackouts,” available online: http://goo.gl/x8HbnN
mechanical power frequently. It saves a lot of unnecessary [2] J.E Chadwick, "How a smarter grid could have prevented the 2003 U.S.
wear and tear. But the downside is that the speed does not cascading blackout," IEEE Power and Energy Conference at Illinois (PECI),
UIUC, Feb. 2013
settle where it was before the fault. But arguably there are [3] U.S. – Canada Power System Outage Task Force, “Final Report on the
secondary systems that helps the governor speed to settle to August 14, 2003 Blackout in the United States and Canada: Causes and
3600 rpm eventually after a few minutes. So, as far as the Recommendations”, April 2004, available online: http://goo.gl/hxKB2a
system is capable to bring the speed into an acceptable range, [4] S.M. Moghadasi, A. Kazemi, M. Fotuhi-Firuzabad, A.A. Edris,
the system is safe. "Composite System Reliability Assessment Incorporating an Interline Power-
Flow Controller," IEEE Trans. on Power Delivery, vol.23, no.2, pp.1191-
B.3. Case 3 1199, April 2008
[5] Y.V. Makarov, V.I. Reshetov, A. Stroev, N.I. Voropai, "Blackout
Without the dead-band in place, the speed deviation was Prevention in the United States, Europe, and Russia," Proceedings of the
around 10 rpm due to the fault condition. To analyze the IEEE, vol.93, Nov. 2005
response from the generator during conditions where the [6] J.W. Bialek, "Why has it happened again? Comparison between the UCTE
blackout in 2006 and the blackouts of 2003," IEEE conference of Power Tech
deviation is within the dead-band, the setting in the governor at Lausanne, July 2007
was changed to a ±20 rpm (±0.33Hz) dead-band and the [7] I.N. Moghaddam, Z. Salami, L. Easter, "Sensitivity Analysis of an
system was tested for the same fault condition. The response is Excitation System in order to Simplify and Validate Dynamic Model Utilizing
shown in Fig. 7. Comparing the mechanical power with case 1 Plant Test Data," IEEE Trans. on Ind. Appl., 2015
and 2, there is hardly any response from the generator and the [8] S. Mohajeryami, Z. Salami, I.N. Moghaddam, "Study of effectiveness of
under-excitation limiter in dynamic modeling of Diesel Generators," IEEE
speed does not settle down at the rated value of 3600 rpm. Power and Energy Conference at Illinois (PECI), UIUC, March 2014
Arguably it can be dangerous under some circumstance. So, [9] I.N. Moghaddam, Z. Salami, S. Mohajeryami, "Generator excitation
there is always a tradeoff between the sensitivity of the systems sensitivity analysis and their model parameter's reduction," Power
governor and the amount of wear and tear that the mechanical Systems Conference at Clemson university (PSC), 2014, March 2014
system can tolerate in the long run. [10] R. Yousefian, S. Kamalasadan, "Design and real-time implementation of
optimal power system wide area system-centric controller based on temporal
Figs 8 and 9 compare the results of all three cases. As it’s difference learning," IEEE Industry Applications Society Annual Meeting,
shown, if the deadband is not modeled properly, the system pp.1-7, Oct. 2014
[11] NERC, “IEEE Power Engineering Society 2008 General Meeting Control
model might show a response while in practice, there is not any Center Issues NERC Reliability Standards Development," IEEE Power and
response. The effect of missing a deadband model in the Energy Society General Meeting - Conversion and Delivery of Electrical
governor model is very small, but if the model of all the Energy in the 21st Century, July 2008
generators in a big region miss the deadband model, then the [12] North American Electric Reliability Corporation (NERC), Modeling,
effect is noticeable. Data, and Analysis (MOD) standards, available online: http://goo.gl/dcRmDK
[13] NERC standard MOD-027-1, “Verification of Models and Data for
Turbine/Governor and Load Control or Active Power/Frequency Control
IV. CONCLUSION Functions”
[14] NERC Resource Subcommittee, “Eastern Interconnection Frequency
The deadband setting in a governor/turbine model can Initiative Whitepaper”, Oct. 2013, available online: http://goo.gl/xq9d0U
either be implemented as a step implementation or a non-step [15] J.C. Mantzaris, C.D. Vournas, "Governor-deadband-induced sliding
implementation, along with the droop characteristics. The bifurcations in South-East European Interconnection model," Power Systems
effect of the deadband setting on the frequency response Computation Conference (PSCC), Aug. 2014
characteristics of a generator is analyzed by observing the [16] J. Mantzaris, A. Metsiou and C. Vournas, “Analysis of Interarea
Oscillations including Governor Effects and Stabilizer Design in South-
response of the generator during a fault in the system. Eastern Europe”, IEEE Trans. on Power Syst. , vol.28, Nov. 2013
The governor without the deadband setting is more [17] S.C. Tripathy, G.S. Hope, O.P. Malik, "Optimisation of load-frequency
control parameters for power systems with reheat steam turbines and governor
responsive to frequency deviations when compared to its deadband nonlinearity," IEE Proceedings on Generation, Transmission and
response with the deadband. Also, it is observed that as long Distribution, January 1982
as the frequency deviation is within the deadband, there is [18] S.B. Nejad, S.H. Elyas, A. Khamseh, I.N. Moghaddam, M. Karrari,
hardly any response from the generator. These observations "Hybrid CLONAL selection algorithm with PSO for valve-point Economic
have led to the suggestion that it is important that the deadband load Dispatch," IEEE Electrotechnical Conference (MELECON), March
2012.
setting is modelled as part of the steam unit governor models to [19] A. Mohseni, S. Mohajeryami, A.A.S Akmal, "Sensitivity analysis and
obtain a response that is closer to the actual response. stochastic approach in study of transient recovery voltage with presence of
superconducting FCL," Electrical Power and Energy Conference (EPEC),
Without modeling of deadband, the simulated response 2011 IEEE, Oct. 2011
from the system may predict a more responsive state compared [20] L. Pereira, J. Undrill, D.N. Kosterev, D. Davies, S. Patterson, "A new
to the real observations. As discussed, NERC report showed thermal governor modeling approach in the WECC," IEEE Trans. on Power
that the utilities in eastern interconnection had expected their Syst., May 2003
system to handle the fault in the system. But the system had not [21] L. Pereira, "New thermal governor model development: its impact on
operation and planning studies on the Western Interconnection," IEEE Power
responded as they expected. According to NERC, the models and Energy Magazine, May-June 2005
used by the utilities had shown a very responsive system, but [22] Power System Dynamic Performance Committee, Power System
observation showed that the system is not responsive as fast as Stability Subcommittee and Task force on Turbine-Governor Modelling,
they had expected it to be. "Dynamic Models for Turbine-Governors in Power System Studies," PES-
TR1, Jan 2013.

View publication stats