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

net/publication/269293014

Determination of acceptable inertia limit for ensuring adequacy under high

levels of wind integration

Conference Paper · May 2014

DOI: 10.1109/EEM.2014.6861300

CITATIONS READS

12 1,677

3 authors:

Nakisa Farrokhseresht Hector Chavez

Delft University of Technology University of Santiago, Chile
5 PUBLICATIONS 27 CITATIONS 78 PUBLICATIONS 513 CITATIONS

SEE PROFILE SEE PROFILE

Mohammad Reza Hesamzadeh

KTH Royal Institute of Technology
131 PUBLICATIONS 1,853 CITATIONS

SEE PROFILE

All content following this page was uploaded by Hector Chavez on 20 December 2018.

The user has requested enhancement of the downloaded file.

 Determination of Acceptable Inertia Limit for
Ensuring Adequacy under High Levels of Wind
Integration
Nakisa Farrokhseresht Héctor Chávez Oróstica Mohammad Reza Hesamzadeh
Student Member, IEEE Member, IEEE Senior Member, IEEE
M.Sc. Student EIT-KIC Energy Eng. Assistant Professor Assistant Professor
Power Systems Dep. - KTH, Sweden Electrical Engineering Dep. EMReG group
Email: [email protected] University of Santiago, Chile Power Systems Dep. - KTH, Sweden

Abstract—The significant penetration of wind generation in Primary Frequency Control (PFC) is defined by ENTSO-E
power grids has raised new challenges in operational and as “the power delivered by the rotating mass of synchronous
planning decisions of power systems. Wind turbine units almost machines in response to frequency drops” and also “governor
always include power converters decoupling the frequency dy-
namics of the wind power generators from that of the grid. This response that acts to arrest frequency decays” [5].
decoupling causes a reduction in total system inertia, affecting The aim of this paper is to determine the penetration level
the system’s ability to overcome frequency disturbances [1]. To of wind generation that leads to insufficient PFC in case of
study the impact of wind power on system inertia, a system the Swedish grid. The paper is organized as follows: first, in
identification approach is followed. This heuristic identification section II, the concepts of frequency control such as inertial
method determines the parameters of governors and prime
movers. The instantaneous minimum frequency requirements response and governor response are discussed. In section III,
and the rate of change of frequency are simulated using the PFC adequacy concerns in NORDEL and Sweden will be
identified power system, and via an extrapolation, the maximum discussed. In Section IV, a case study of the Swedish grid is
wind power penetration in Sweden is found. Finally, a need for described, modelled by the Nordic 32-A system [6] , and an
market design is introduced in order to ensure adequacy. ad-hoc identification method will be presented to identify the
Index Terms—primary frequency control; power system simu-
lation; system identification; wind power integration; electricity
dynamical parameters of the Nordic 32-A case. Then, wind
market farms will be added in Section V, and the impact of these
wind generators on the rate-of-change-of-frequency (ROCOF)
I. I NTRODUCTION and the minimum frequency (NADIR) will be assessed. A
discussion on a need for market design in order to ensure
Among the various renewable resources, wind power is adequacy will be presented in section VI. Finally, Section
assumed to have the most favorable technical and economic VII summarizes the main conclusions and provides some
prospects and offers significant potential for reducing green- recommendations for future research.
house gas (GHG) emissions. Roughly 1.8 % of worldwide
electricity demand has been met by wind power energy and it
is predicted that the share of worldwide wind power will grow II. F REQUENCY C ONTROL C ONCEPT
up to 20 % by 2050 [2]. Wind power is a major new resource
in both Europe and the U.S in 2009, approximately 39 % of all The definition of frequency stability given by CIGRE and
capacity installed in these two parts of the world came from IEEE is the following: Frequency stability refers to the ability
wind power [2]. This trend is also seen in Sweden where the of a power system to maintain a steady frequency following a
share of net electricity production coming from wind power severe system disturbance, resulting in a significant imbalance
has increased significantly from 0.5 % in 2003 to 4.4 % in between generation and load. It depends on the ability to
2012 [3]. Sweden also has a planning framework for wind maintain/restore the equilibrium between system generation
power, projecting a production of 30 TWh by 2020 (20 TWh and load [7].
onshore, 10 TWh offshore) [3]. An imbalance between the mechanical power generated and
A large penetration of wind will reduce the power supplied electrical load leads to a change in the grid frequency. When
by conventional synchronous generators. The variable speed the frequency deviation exceeds a certain limit, controllers
wind machines, which are now dominant in the market do (governors) are activated to change the power output set point
not always have an inertial response [1]. This lack of inertial to the prime movers. This is called the primary frequency
response causes less overall inertia in the grid and the power control and is divided into the phases of inertial response and
frequency control experiences a higher control burden [4]. of governor response.

978-1-4799-6095-8/14/$31.00 ©2014 IEEE

A. Inertial Response 1 e 1 + T rs c 1 g
VAR(L) +
(nref) 1 + T fs rTrs 1 + T gs
If the frequency of the power system changes, the mechan- –. Velocity and
Position Limits
ical speed of the synchronized machines within the system Speed +
SPEED
(SPEED)
changes as well. Since the kinetic energy stored in the ro- +
tational parts of the generators is proportional to the square R Dturb
of the mechanical speed, the kinetic energy changes as the
frequency does. If frequency drops (increases) kinetic energy
X
is liberated (absorbed), and the change in kinetic energy tends –
to regulate the frequency change. g 1 q +
PMECH
X – X At
Tws +
For multi-machine systems, consider that a load suddenly q h + –
increases by ΔPL at time t = 0, at bus k. Then, at t = 0+ , 1. q
NL

each generator i will react according to its closeness to the

change, as captured by the synchronizing power coefficient Fig. 1. Hydraulic and Governor model of HYGOV [9]
PSiK between units i and k, according to [8] :
(−PSik ) (−ΔPL ) PSik Now substitute eq. (8) into eq. (5) and obtain:
ΔPei =  =  ΔPL (1)
n n  n  
PSkj PSkj 2  dΔω
j=1 j=1 Hi = −ΔPL , (9)
ωRe i=1 dt
∂Pik
where PSik = |δ (2) and finally:
∂δik ik0 dΔf −ΔPL fRe
= 
n ≡ mf , (10)
The swing equation for machine i is: dt
2 Hi
2Hi d2 Δδi i=1
= −ΔPei (3) where mf , the initial slope of the frequency deviation versus
ωRe dt2
time, is called the rate-of-change-of-frequency (ROCOF) [1].
With δi the rotor angle of generator i and ωRe the nominal It can be seen from eq. (10) that the ROCOF depends on
pulsation. The inertia constant H has the dimension of time the total inertia for the entire interconnected system. The more
and indicates the time that the system can provide nominal inertia in the system, the smaller the ROCOF, and a less severe
power by using only the energy stored in its rotating masses. frequency drop will take place.
Then:  
2 d2 Δδi PSik ΔPL B. Primary Control
2
=− 
n . (4)
ωRe dt Hi A governor is the feedback controller that senses the system
PSkj
j=1 frequency and acts on generator’s prime movers to regulate
frequency deviations. An example of one of the standard
Use Δωi instead of Δδi for all generators 1,..., n,: models that has been applied in the Nordic 32-A model [6],
n known as the HYGOV governor, and used for the hydro plants,
n PSik is shown in Fig.1. R is called the permanent droop. Other
2 dHi Δωi
= − i=1
n ΔPL = −ΔPL . (5) important parameters are Tr , the governor time constant and
ωRe i=1 dt
PSkj Tf , the filter time constant and Tw , the water time constant.
j=1
III. PFC A DEQUACY C ONCERN IN NORDEL
Now define the “Center Of Inertia (COI)” of the system:
NORDEL was established in 1963 and is a body for

n 
n
co-operation between the transmission system operators in
Hi ω i Hi Δωi
i=1 i=1 Denmark, Finland, Iceland, Norway and Sweden, whose ob-
ω≡ 
n or Δω ≡ n . (6) jective is to establish a Nordic electricity market. The installed
Hi Hi capacity of the NORDEL is about 100.8 GW, of which about
i=1 i=1
8.9 GW is wind power [10]. Integration of renewable energy
Differentiating Δω with respect to time: leads different concerns for adequate PFC in power systems

n [11], of which two will be discussed in this paper:
d(Hi Δωi )
dt • ROCOF adequacy: to prevent the system having a RO-
dΔω i=1
≡ 
n , (7) COF larger than the ROCOF Islanding Detection relay
dt
Hi setting after a sudden losses of generation.
 ni=1    • NADIR adequacy: to prevent the minimum frequency

n
d (Hi Δωi )  dΔω (NADIR) from dropping below the setting of the gen-
or = Hi . (8)
i=1
dt i=1
dt erator’s Under Frequency Load Shedding (UFLS) relay.
 50.2 0
Measured Frequency Response
10-04-16 (530MW)
50.1
0.002
09-07-02 (800MW)

0.004
50

Per unit frequency deviation [pu]

Frequency(Hz)

Original Nordic32 Model

0.006
49.9 09-08-10 (1100MW) Measured Frequency Response

0.008
49.8

0.01
49.7 10-04-16 Original Nordic 32 Model
09-08-10
0.012
49.6 09-07-02

0.014
49.5
0 50 100 150 200 250 300
0.016
Time(seconds)

0.018
0 10 20 30 40 50 60 70 80
Fig. 2. Measured frequency in the Nordic system from sudden tripping of Time [s]
530, 800 and 1100 MW generation[15]
Fig. 3. Frequency response of original Nordic 32-A and measured frequency
contingency from [15]
Although the Nordic Grid Code has no requirement on the
maximum value for the magnitude of the ROCOF, [11] defines
-0.5 Hz/s as the minimum acceptable ROCOF. The Nordic there is a sudden load-generation unbalance. Fig. 2 shows
Grid Code specifies an automatic load shedding for Sweden the frequency during three generation outage events in the
at 49.4 [email protected] s [12]. Thus, PFC adequacy for Sweden as Nordic system [15]. The contingency on 09-08-10 (1100 MW)
part of NORDEL will be considered in this paper as: is considered in the identification.
• Magnitude of ROCOFs must be less than 0.5 Hz/s, and
Since the installed capacity of the Nordic 32-A model is
• Frequency NADIRs must be larger than 49.4 Hz.
half of that of the real Swedish grid, the machine on bus 1014
IV. C ASE STUDY with 1100/2 =550 MW generation is considered to model
the 1100 MW contingency and the center of inertia of the
A. Developing Nordic 32-A in PSS/E
per unit bus frequency deviations is plotted. It can be seen
The CIGRE Nordic 32-A test system is used as a repre- in Fig. 3 that there is a significant difference between the
sentation of the Swedish transmission grid in this paper. The measured frequency response to contingency of 1100 MW
total installed capacity of Nordic 32-A is 17.5 GW, and the and the center of inertia of the original Nordic 32-A test
load is 10.9 GW. Then, the Nordic 32-A test system is a system simulated in PSS/E. The dynamic behavior obtained
50 % scaled-down version of the Swedish system that has with a transient simulation of the grid modeled according the
36.4 GW installed capacity [13] [14]. Nordic 32-A consists of original Nordic 32-A document does not correspond with the
32 main buses and 9 loads. It is divided in four main areas. measurement results obtained from the real Nordic grid and
The northern area consists of hydro generation and low load an ad-hoc identification process is used to adjust the dynamic
consumption. The central area has a high demand and mainly parameters in the Nordic 32-A case to identify the actual
thermal power generation. The south west area has several data. The inertia constants H of the generators are adjusted
thermal units and low load consumption. Finally the external in the range [0.3, 10.5] s until the simulated center of inertia
area has a mixture of different kinds of generation and levels matches with the measurement data of the frequency inside
of loads. The following are the dynamic models of the thermal the time range [0,2] s. To compare the goodness of fit, a
and hydro generators, exciters and stabilizers that are used in least-square criterion is used. This time range is used because
the Nordic 32-A test system [6]. it is assumed that the governors have time constants of such
• GENSAL: represents a salient pole generator and is used an order that they are not yet active in the first 2 seconds after
for all hydro power generating units. a contingency. Then, the droop constants of the governors
• GENROU: is a cylindrical round rotor type of generator are sought by adjusting them until the simulated frequency
and represents the generators of the thermal power units. matches the measured data in steady state. Finally, various
• SEXS: represents the excitation system’s dynamic model dynamic parameters of the HYGOV governor belonging to
and is used for all types of synchronous generators. the machines in hydraulic plants are sought by matching
• SATB2A: is the stabilizer model and dampens the oscil- the transient in the simulated frequency with the measured
lations in the electrical output power. data. The dynamic parameters in the Nordic 32-A model
• HYGOV: represents the hydro-turbine governor for hydro before and after identification are shown in Table I. It should
plant (thermal units have no any governor). be noticed that the model identification is time dependent,
because different load condition leads to different frequency
B. Model Identification control dynamics.
A starting point for analyzing the frequency stability in the
Nordic system is to study the primary frequency control when
 TABLE I TABLE II
O RIGINAL AND IDENTIFIED DYNAMIC PARAMETERS OF N ORDIC 32-A P OWER FLOW DATA FOR GE 1.5MW AND A WIND FARM
MODEL
DFIG GE 1.5 Wind
Dynamic parameters Original Identified
parameters parameters MW [17] Farm(300
R, Permanent Droop 0.04 pu 0.05 pu lumped
Tg , Servo Time Constant 0.2 s 0.2 s
Tr , Governor Time Con-
turbines)
5s 0.6 s
stant Generator rating,
Tf , Filter Time Constant 0.05 s 0.1 s 1.67 500
Velm , Gate Velocity MVA
0.1 pu/s 0.1 pu/s
Limit Pmax , MW 1.5 450
Gmax , Maximum Gate
0.95 pu 0.9 pu Pmin , MW 0.07 21
Limit
Gmin , Minimum Gate Pgen , MW - 300
0 pu 0 pu
Limit
Tw , Water Time Con- Xsorce , pu 0.8 0.8
1s 0.5 s
stant Unit transformer
Dturb , Turbine Damp-
0 0.4 1.5 525
ing rating, MVA
At , Turbine Gain 1 0.95
H1 , Inertia of Thermal
6s 9.9 s
Generators
H2 , Inertia of Hydro
3s 7.7 s
Units This work concerns the tripping of one unit at bus 1014 with
a rated output of 550 MW which causes a generation loss
of 4.8 % (the total production is 11400 MW). A number of
V. W IND P OWER I NTEGRATION simulations have been performed to further understand the
impact of wind power on frequency stability. Fig. 4 shows the
This paper focuses mainly on the variable-speed type of frequency response of the system for the base case (without
wind turbines which is more commonly used than fixed-speed wind) and with wind power integration. The impact of wind
turbines [1]. There are several studies that detail how the integration on the ROCOF and NADIR after disconnecting
inertia of the blades can be utilized through an implementation machine 1014 is shown in Table III.
of an additional control loop [16], but standard DFIG wind
turbines typically don’t have frequency control capabilities. TABLE III
The wind turbine model for all wind farms in this study is I MPACT OF WIND INTEGRATION ON ROCOF AND NADIR
assumed to be a DFIG type GE 1.5 MW [17]. All simulations
Added Pgen of
are performed in PSS/E. The rated quantities needed for ROCOF [Hz/s] NADIR [Hz]
wind park [MW]
static and dynamic simulation of a typical GE 1.5 MW wind
turbine are taken from [17]. In each wind park, 300 identical Base case (no
-0.11488 49.6113
wind turbines are lumped together. The rated wind turbine wind)
quantities for a static power flow are presented in Table II. The 600 -0.12149 49.6052
dynamic models of a GE 1.5 MW wind turbine consist of an 1200 -0.13818 49.5902
aerodynamics model (GEWTA1), an electrical control model 1800 -0.14868 49.5797
(GEWTE2), a model for the generator and power converter 2400 -0.16236 49.5686
(GEWTG2), a pitch control model (GEWTP1) and a 2-mass 3000 -0.16678 49.5184
model for the turbine shaft (GEWTT1). All the required data
for the dynamic model is described in [17]. The assumptions in
the study when wind farms are added to the identified Nordic
32-A model are:
A. Impact of wind integration on ROCOF
• The load will not change in different scenarios,
• One aggregated wind turbine is used to represent all wind According to Table III, the ROCOF decreases when more
turbines inside a wind farm, wind power is added to the system. As described in Section
• The wind speed remains constant during the simulations, III, this study will consider -0.5 Hz/s as a lower limit for
• The wind power output is 70 % of the installed capacity. ROCOF adequacy. Using the ROCOF data of Table III, and
This captures the fact that the correlation between dif- performing a linear extrapolation, a ROCOF of -0.5 Hz/s is
ferent wind farms within a large geographical area is not predicted when the wind capacity is 20.08 GW. This means
perfect, so it is very unlikely that all wind farms generate that for the Swedish grid, 40.16 GW is the maximum wind
a maximum output at the same time, power that can be integrated into the system. More wind power
• The integration of wind is assumed to replace existing will cause the ROCOF to become inadequate. The expectation
thermal units and the output of the replaced generator is for wind power in Sweden is 26 GW by 2030 [18], and hence,
equal to the output of the integrated wind generation. there is no problem for the immediate future.
 50.2 The ad-hoc identification process may be improved by
Base Case No Wind
Center of Inertia of the bus frequencies [Hz]

600 MW Wind Farm including more actual data of frequency responses after sudden
50.1 1200 MW Wind Farm losses of generation. It may be also of interest to consider
1800 MW Wind Farm different control strategies for wind frequency response. This
Base Case 2400 MW Wind Farm
50 3000MW
is proposed for future work.
3000 MW Wind Farm
600MW Although wind turbines can provide a frequency response
49.9
1800MW to maintain adequacy, such a frequency response requires the
turbine not to operate at maximum power point tracking.
49.8
If the frequency drops, the turbine cannot increase power
output unless the turbine is operating below the maximum
49.7
1200MW power point tracking, so additional power can be delivered. If
not transformed immediately into electricity, wind energy is
49.6
lost, so any operating condition below the maximum power
2400MW
point represents an opportunity cost. Then, the provision of a
49.5
0 10 20 30 40 50 60 70 80 frequency response from wind turbines cannot be considered
Time [s] “by default” and such an opportunity cost must be analyzed
Fig. 4. Frequency response after contingency of 550 MW for different amount to determine the least cost solution to maintain adequacy. The
of wind integration design of an ancillary service market including an adequacy
assessment is proposed as a future work to represent the op-
portunity cost of wind turbines providing frequency response.
B. Impact of wind integration on NADIR
R EFERENCES
Table III shows the same decreasing trend for the NADIR.
The NADIR requirement is to maintain the system frequency [1] G. Lalor, A. Mullane, and M. O’Malley, “Frequency control and wind
turbine technologies,” IEEE Transactions on Power Systems, vol. 20,
above 49.4 Hz at all times. Based on the data from Table III, no. 4, Nov. 2005.
the wind generation value that causes the system frequency [2] O. Edenhofer, Renewable Energy Sources and Climate change mitiga-
to fail complying with this requirement is 6.2 GW. Therefore tion. New York: Cambridge University Press, May 2009.
[3] “Vindkraftsstatistik,” Energimyndigheten, Tech. Rep., 2012.
a critical wind generation for Sweden would be 12.4 GW. [4] D. Gautam, D. Vittal, and T. Harbour, “Impact of increased penetration
In the other words, when the wind power, integrated into of DFIG-based wind turbine generators on transient and small signal
the Swedish grid, is larger than 12.4 GW, there would be stability of power systems,” IEEE Transactions on Power Systems,
vol. 24, no. 3, pp. 1426–1434, Aug. 2009.
NADIR inadequacy. This is, in principle, of concern as the [5] “Supporting document for the network code on operational security,”
wind capacity is expected to increase up to 26 GW by 2030 ENTSO-E, Tech. Rep., Feb. 2013.
[18]. [6] K. Walve, “Nordic32-A CIGRE test system for simulation of transient
stability and long term dynamics,” Svenska Kraftnät, Tech. Rep., Mar.
1996.
VI. T OWARDS DESIGNING MARKETS FOR ENSURING [7] “Definition and classification of power system stability,” IEEE/CIGRE,
NADIR ADEQUACY Tech. Rep., May 2004.
[8] V. Vittal, “Impact of increased DFIG wind penetration on power systems
According to Section V, there would be a NADIR inade- and markets,” Power Systems Engineering Research Center, Tech. Rep.,
quacy in the immediate future in Sweden; therefore a future Oct. 2009.
study should design a market for ensuring NADIR adequacy [9] “PSSE program application guide volume 2,” Siemens Industry, Tech.
Rep., Mar. 2013.
where the cost of (energy losses + NADIR inadequacy costs [10] “Nordic market report 2013, development in the Nordic electricity
+ the price of producing energy + the price of providing for market,” NordREG, Tech. Rep., Sept. 2013.
NADIR adequacy) is minimized, where the constrain is [11] J. Vikesj, L. Messing, and G. Power, “Wind power and fault clearance,”
Elforsk, Tech. Rep., Apr. 2011.
fN AD ≥ fM IN (11) [12] “Nordic grid code (Nordic collection of rules),” NORDEL, Tech. Rep.,
Jan. 2007.
The NADIR adequacy optimization problem involves various [13] N. Ullah and T. Thiringer, “Effect of operational modes of a wind farm
on the transient stability of nearby generators and on power oscillations:
aspects of frequency dynamics such as inertia and frequency a Nordic grid study,” Wind Energy, vol. 11, no. 1, pp. 63–73, Jan. 2008.
response. [14] “Energy in Sweden 2012,” Swedish Energy Agency, Tech. Rep., Oct.
2012.
VII. C ONCLUSION AND F UTURE W ORK [15] Z. Li, O. Samuelsson, and R. Garcia-Valle, “Frequency deviations and
generation scheduling in the Nordic system,” in IEEE Trondheim Pow-
An ad-hoc identification model of the Nordic 32-A System erTech, the Power of Technology for a Sustainable Society, Trondheim,
has been presented. Wind power was included in the identified Norway, June 2011.
[16] J. Morren, J. Pierik, and H. de Haan, “Inertial response of variable speed
case to simulate the effect of wind integration on ROCOF wind turbines,” Electric Power Systems Research, vol. 77, no. 11, pp.
and NADIR adequacy. There are no major concerns associated 980–987, July 2006.
with ROCOF. But in the case of Nadir adequacy, the system [17] “Modelling of GE wind turbine-generators for grid studies,” GE Energy,
Tech. Rep., Apr. 2009.
is unable to maintain the frequency above 49.4 Hz for a wind [18] B. Lemström, “Wind and grid scenarios for NORDEL,” VTT,
generation of about 12.8 GW or higher. All simulations are TradeWind, Tech. Rep., Jan. 2008.
performed in PSS/E.

View publication stats