Content uploaded by Amaris Dalton
Author content
All content in this area was uploaded by Amaris Dalton on Sep 19, 2019
Content may be subject to copyright.
52
Wind power variability during the passage of cold fronts
across South Africa
Amaris Dalton1*, Bernard Bekker1, Andries Coenrad Kruger2,3
1Department of Electrical and Electronic Engineering, Stellenbosch University, Private Bag X1, Matieland 7602,
South Africa
2 Climate Services, South African Weather Service, Private Bag X097, Pretoria 001, South Africa
3 Department of Geography, Geoinformatics and Meteorology, University of Pretoria, Pretoria, South Africa
ORCID iDs: A Dalton: https://orcid.org/0000-0002-2893-1961; B Bekker: https://orcid.org/0000-0002-5574-0482
A Kruger: https://orcid.org/0000-0002-9815-570X
Abstract
Wind is a naturally variable resource that fluctuates across timescales and, by the same token, the electricity
generated by wind also fluctuates across timescales. At longer timescales, i.e., hours to days, synoptic-scale
weather systems, notably cold fronts during South African winter months, are important instigators of strong
wind conditions and variability in the wind resource. The variability of wind power production from aggre-
gates of geographically disperse turbines for the passage of individual cold fronts over South Africa was
simulated in this study. When considering wind power variability caused by synoptic-scale weather patterns,
specifically cold fronts, the timescale at which analysis is conducted was found to be of great importance, as
relatively small mean absolute power ramps at a ten-minute temporal resolution, order of 2-4% of simulated
capacity, can result in large variations of total wind power production (at the order of 32–93% of simulated
capacity) over a period of three to four days as a cold front passes. It was found that when the aggregate
consists of a larger and more geographically dispersed set of turbines, as opposed to a smaller set of turbines
specifically located within cold-front dominated high wind areas, variability and the mean absolute ramp rates
decrease (or gets ‘smoothed’) across the timescales considered. It was finally shown that the majority of large
simulated wind power ramp events observed during the winter months, especially at longer timescales, are
caused by the passage of cold fronts.
Keywords: wind power ramps; weather systems; aggregated smoothing;
Highlights:
• Significant wind power variability is caused by the passage of cold fronts.
• Wind power variability becomes larger as longer timescales are considered.
• The smoothing effect becomes greater as geographically dispersed turbines are added to an aggregated
time-series.
Journal of Energy in Southern Africa 30(3): 52–67
DOI: https://dx.doi.org/10.17159/2413-3051/2019/v30i3a6356
Published by the Energy Research Centre, University of Cape Town ISSN: 2413-3051
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International Licence
https://journals.assaf.org.za/jesa
Sponsored by the Department of Science and Technology
Corresponding author: +27 (0)79 408 2122;
email: amaris_dalton@sun.ac.za
Volume 30 Number 3
August 2019
53 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
1. Introduction
Electricity generated by wind turbines fluctuates with
wind speed and wind speed varies at all timescales,
from decades down to sub-seconds (Widén et al.,
2015). At short timescales (less than one hour),
these fluctuations are caused by turbulent eddies in
the boundary layer, brought about by the interaction
of meso-scale weather systems with the local envi-
ronment. Short-term variability may be caused by
local features such as terrain, surface roughness or
micro-climatic phenomena such as sea-breezes. At
longer timescales, of hours to several days, the most
prominent variations are brought about by synoptic-
scale weather systems (Kiviluoma et al., 2016).
Amongst the most frequent of synoptic systems that
affect South Africa’s weather are extra-tropical cy-
clones, and the cold fronts associated with them. Lit-
tle literature could be found where the underlying
causative meteorological phenomena driving wind
ramp events were studied (Couto et al., 2015;
Gallego-castillo et al., 2015; Lacerda et al., 2017).
Furthermore, attempts to find literature in which
wind power ramp events associated with specific
weather systems had been studied for South Africa
were not successful. In the South African context,
cold fronts are deemed highly relevant to the wind
energy industry as they are the main strong wind
producers along the coast and adjacent interior of
South Africa (Kruger et al., 2010). Strong winds
from the passage of cold fronts dominate the south-
ern parts of the country, though these winds often
reach further north, as shown in Figure 1.
The variability and distributed nature of the wind
resource present numerous challenges to the inte-
gration of wind energy into future power networks.
Such challenges include suboptimal unit allocation
of thermal dispatchable units and low capacity credit
of variable renewable energy (VRE) generators
(Albadi et al., 2010; Ueckerdt et al., 2015). These
problems, associated with the incorporation of VRE
into the power network, are expected to become
more pronounced as the level of VRE penetration
increases. Currently, South Africa remains heavily
reliant on coal (Mcewan, 2017), but it is, however,
anticipated that wind energy will play an important
part in the country’s future energy mix, as the coun-
try is rich in wind resources (Knorr et al., 2016) and
has a favourable policy environment for renewable
energy development (Department of Energy, 2018).
The present study provided a first iteration anal-
ysis that quantifies anticipated induced variability
and wind power ramps during the passage of typical
South African frontal systems. Understanding this
variability and power ramps could assist network op-
erators in unit commitment, load scheduling and
short-term maintenance planning. There is a signifi-
cant body of literature considering the nature and
effects of wind power variability (Kiviluoma et al.,
Figure 1: Map of South Africa defining the dominant strong wind mechanisms responsible for
maximum annual hourly wind speeds (Kruger 2011).
54
2016; Choukri et al., 2017; Kalverla et al., 2017;
Monforti et al., 2016; Sørensen et al., 2018; Prasad
et al., 2009; Thapar et al., 2011). This literature may
be divided into studies that were conducted using
real wind power data (Kiviluoma et al., 2016;
Choukri et al., 2017) and those using meteorological
data. The latter may in turn be divided between
studies using actual observational data (Kalverla et
al., 2017) and studies using outputs from numerical
weather prediction (Monforti et al., 2014; Sørensen,
Heunis, et al., 2018) as proxies for generation data.
Observed or modelled wind speeds are frequently
used to estimate wind power potential through the
use of wind resource assessments (Prasad et al.,
2009a; Prasad et al., 2009b) and transforming wind
speeds to power through turbine power curves
(Thapar et al., 2011). There is significant evidence
supporting the idea of the ‘smoothing effect’, i.e., re-
ductions in the gradient of electricity feed-in,
brought about by aggregation effects from geo-
graphically dispersed wind generators (Widén et al.,
2015). The smoothing effect is attributed to wind
speed that varies with geographical distance, where
the greater the distance the more dissimilar the wind
resource. From the literature considered (Albadi et
al., 2010), it was, however, found that this smooth-
ing effect is more pronounced over short rather than
long time intervals. It may be contended that power
ramps and variability at time periods of smaller than
ten minutes are of greatest importance from a tur-
bine design perspective, rather than a power system
perspective, as it is assumed that fluctuations at
these timescales would be effectively smoothed
(Sørensen et al., 2008; Widén et al., 2015). How-
ever, variability over longer time intervals (greater
than or equal to an hour) across relatively large spa-
tial extents is considered to be of greater importance
to the security and adequacy of power systems
(Kiviluoma et al., 2016). Variability at longer time-
scales is also considered to be of greater importance
to this study, as it has been found that longer-time-
scale, synoptic-scale weather patterns (such as cold
fronts) drive variability (Kiviluoma et al., 2016).
It has been noted that it is difficult to make gen-
eralisations on resource variability for a specific re-
gion based on studies done for other regions
(Kiviluoma et al., 2016; Gallego-Castillo et al.,
2015). This is especially true for studies considering
the prevailing meteorological processes inherent to
a specific region, highlighting the importance of con-
ducting region-specific studies. In studies conducted
by Knorr et al. (2016) and Sørensen et al. (2018),
the variability of the wind energy resource in South
Africa was assessed using the Weather Research and
Forecasting Model (Skamarock et al., 2008) reanal-
ysis data. Both studies found that power ramps will
be significantly reduced by aggregated smoothing,
as the number of geographically dispersed genera-
tors feeding into the grid increases over time. Both
Knorr et al. (2016) and Sørensen et al. (2018) fo-
cused on wind power variability and ramps at high
temporal resolutions of ten and 15 minutes respec-
tively, but not on timescales of hours to days. Aggre-
gated smoothing is most evident at short timescales;
however, larger power ramps become evident at
longer timescales due to the impact of slower-mov-
ing, synoptic-scale weather systems (Widén et al.,
2015; Kiviluoma et al., 2016). Indeed, in a study
considering various countries (Mararakanye et al.,
2019), it was found that variations from 15 minutes
to eight hours could be from 10–40% of installed ca-
pacity, which highlights the importance of conduct-
ing variability studies focused specifically on longer
timescales. Therefore, in addition to considering
variability in wind power production introduced by
synoptic-scale weather systems at short timescales,
the present study also analysed such variability and
power ramps at hourly and daily timescales.
2. Methodology
Simulations were conducted representing wind
power generation during the passage of several cold
fronts over South Africa. A cold front may somewhat
simply be defined as the baroclinic boundary be-
tween a mass of cold air and warm air, where tem-
perature decreases by a minimum of 3 °C
(Eumetrain, 2012). There are no strict empirical def-
initions of a cold front, especially in terms of its sig-
nificance, therefore, noteworthy historic cold fronts
that had made landfall in South Africa were identi-
fied from news media reporting or from increases in
observed wind speeds from Wind Atlas for South Af-
rica (WASA) observational mast data (Council for
Scientific and Industrial Research (CSIR), 2010;
ECR Newswatch, 2016).
Subsequently, the corresponding synoptic
weather charts, publicly downloadable as open ac-
cess from the South African Weather Service’s web-
site, were used for more concise identification of
fronts (South African Weather Service, 2018). An
example of such a synoptic weather map depicting
an approaching extra-tropical cyclone with its asso-
ciated cold front is shown in Figure 2.
Wind power simulations were conducted for the
passage of four cold fronts during August 2012, May
2013, and two consecutive events in July 2016 ap-
proximately two days apart. Wind speed and direc-
tion time series at four observational heights (20 m,
40 m, 60 m, 62 m), with a ten-minute temporal res-
olution, were obtained for fifteen WASA observa-
tional masts, represented by WM1-15 in Figure 3.
The observational data, freely available from WASA
(CSIR, 2010), was downloaded for periods corre-
sponding to those of selected cold fronts making
55
Figure 2: Surface synoptic weather map showing a cold front approaching South Africa
(South African Weather Service, 2018).
Figure 3: Locations of observational masts (Google Earth Pro, 2019), indicated by WM 1-10.
landfall in South Africa. The WM1-10 were installed
during the WASA 1 project throughout 2010. The
WM11-15 were installed during the WASA 2 project
throughout 2015; therefore, the 2012 and 2013 sim-
ulations comprised data from ten observational
masts. With the addition of WASA Phase 2 and the
decommissioning of WM04 and WM08, the 2016
simulation consisted of data from 13 observational
masts. Wind speeds were extrapolated from 62 m to
the selected hub height of 80 m using Hellman’s law,
defined by Equation 1.
= (
) (1)
56 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
where v2 is the unknown wind speed at height h2,
v1 is the measured wind speed at h1, and α is Hell-
man’s component or the wind shear component. As
wind speed data is available at two heights, α was
calculated through a log linear model, defined by
Equation 2.
=
(
)
(
) (2)
It is unrealistic to assume that the same turbine
would be installed across the study area. Historic
wind speeds observed at the observational masts
(Mortensen et al., 2012) were, therefore, used to de-
termine the likely International Electrotechnical
Commission (IEC) class of turbines to be installed in
a specific region. Subsequently two turbines from
the same manufacturer, Nordex_N90 (IEC II) and
Nordex_N100 (IEC III), with the same nameplate
capacity of 2.5 MW, were selected. The selection
was in part based on the Renewable Energy Inde-
pendent Power Producer Procurement Programme
wind farm projects, including the South African
Dorper (31.478S, 26.438E) and Kouga (34.056S,
24.625E) wind farms, where these respective tur-
bine types are installed. Power curves consisting of
discrete points were obtained from the manufac-
turer, through the XML package in R (Lang, 2018),
as shown in Figure 4.
Piece-wise functions were defined to emulate the
power curves. Accordingly, the power curve was di-
vided into four distinct functions. Three of these
were simple horizontal straight-line functions: a
function through the rated power, and two functions
through the origin. These were for wind speeds of
less than the cut-in speed and greater than the cut-
out and rated wind speeds, respectively (Joubert,
2017). The final function was represented by a
higher (tenth) order polynomial for wind speeds be-
tween the cut-in and the rated wind speeds. The lm
function, which forms part of R base, was used for
polynomial regression (R Core Team, 2018). It was
found that the derived polynomial functions pre-
dicted the power curve nearly perfectly (R2 ≈ 1), ex-
cept for wind speeds approaching the cut-in and
rated wind speeds. To resolve these difficulties, lin-
ear regression was applied to obtain a linear func-
tion to resolve wind speeds between the cut-in wind
speed and the first discrete point on the power
curve, i.e., 3 ≤ x < 4 m/s. A similar procedure was
performed to be able to predict wind speeds be-
tween the final discrete point on the power curve
and the rated wind speed. Once the power curves
had been defined, wind power simulations were per-
formed for the selected cold fronts. In the simula-
tions, each of the observational masts for which
wind speed data was obtained served as a proxy for
a potential wind turbine. Unanimous indicators for
the time of the cold front making landfall had to be
selected for comparison between events. The follow-
ing range of information sources and indicators were
considered:
• a distinct change in wind direction, usually from
the north-westerly sector to the south-westerly
at WM05 (19.69245°E, 34.61192°S) (however,
due to the orientation of some fronts not follow-
ing the typical textbook examples this would not
always be the case);
• the boundary at the surface between the
warmer and colder air masses in the south-west
of the Western Cape province, i.e., a cold front,
indicated by an observable drop in surface tem-
perature;
• stronger wind speeds due to the change in wind
direction and associated turbulence in the
boundary layer;
• the commencement of rain; and
• interrogation of South African Weather Service
daily synoptic charts.
It should be noted that all the above signals do
not occur at the same time, but can be hours apart.
However, considering the information available,
Figure 4: Nordex N90 & N100 power curves (Lang, 2018).
57
commencement dates of the movement of the cold
fronts over the virtual wind turbines were selected,
indicated as yellow lines in Figure 5. After the wind
power simulation of individual fronts, the simulation
was extended for passage of cold fronts throughout
the months of interest, i.e., August 2012, May 2013
and July 2016.
Two simple methods of aggregation of simulated
wind power time series were used. For the first set of
simulated time series, hereafter referred to as the
‘uniform aggregation’, the ten-minute time-step
wind power values generated by each of the proxies
were aggregated into a single time series using a sim-
ple mean. For the second set of simulation time se-
ries, hereafter referred to as the ‘cold front domi-
nated aggregation’, only the proxy generators lo-
cated in the cold front dominated geographic re-
gions, as presented in Figure 1, were selected. The
‘cold front dominated aggregation’ consisted of a
mean from the WM04, WM05, WM07, WM08 and
WM10 time series, for the 2012 and 2013 simula-
tions. For the 2016 simulation, WM04 and WM08
data were not available, so that data from WM11,
WM12 and WM13, which are also located within
cold front-dominated high wind areas, were added.
The aggregated time series of wind power outputs
for each of the simulations was considered as a per-
centage of the installed capacities considered within
the simulation. The wind power simulations were
then analysed within the context of the differences
in generation between consecutive time intervals,
i.e., the power ramps. Accordingly, the mean-simu-
lated wind power time series from each of the dis-
tinct observational masts were aggregated at ten-mi-
nute, hourly and daily timescales.
3. Results and discussion
When considering the aggregated time series against
that of each individual simulation, the smoothing ef-
fect becomes clearly evident, as shown in Figure 6,
where a less pronounced fluctuation between min-
ima and maxima values is evident.
Figure 5: Wind direction (in degrees clockwise from North) during the month of August 2012 with
the passage of several cold fronts, where the yellow lines indicate the start of the passage of cold
fronts through the region under consideration.
Figure 6: Aggregated versus individual power time series during the passage of a cold front at ten-
minute timescale. The yellow line indicates the start of the passage of the cold front through the
region under consideration.
58
This is fairly similar to what was observed by
Knorr et al. (2016) and it could be inferred that this
line would become even smoother at the ten-minute
timescale should more generators be added to the
aggregate. Despite significant smoothing evident in
the aggregated time series, a considerable degree of
variability remained evident if considering the simu-
lation in its entirety across four days. Figure 7 repre-
sents the two wind power simulations conducted for
the cold front making landfall in the Western Cape
on 22 August 2012 and should be read in conjunc-
tion with Figure 8, showing a synoptic weather map
of this cold front. Figure 7 shows that a sharp and
immediate ramp up in wind power generation, fol-
lowed by a more gradual ramp down, was associ-
ated with the passage of the front.
The variability of these simulations, expressed in
terms of mean absolute ramp rate, coefficient of var-
iation and maximum power ramps (up and down),
is summarised in Table 1 at the ten-minute, hourly
Figure 7: Wind power simulation for (a) ‘uniform aggregation’, and (b) ‘cold front-dominated
aggregation’ methods, during the passage of a cold front making landfall on 24 August 2012. The
yellow line indicates the start of the passage of the cold front (CF) through the region under
consideration.
Figure 8: Synoptic weather map depicting the cold fronts considered for the 24 August 2012 wind
power simulation (South African Weather Service, 2018).
(a)
(b)
59
and daily timescales. Table 1 shows that wind power
variability was greater in the ‘cold front-dominated
aggregation’ than in the ‘uniform aggregation’,
across all measured indices (except from the daily
max ramp down). This can likely be attributed to the
increase in turbine numbers in the simulation and
turbine placements outside of geographic areas
where cold fronts were the dominant strong wind-
producing mechanism. Table 1 also presents an in-
creasing variability across all measured indices as
the measurement period increased.
Figure 9 represents the two wind power simula-
tions conducted for the cold front making landfall in
the Western Cape on 26 May 2013. A synoptic
weather map of this cold front is shown in Figure 10.
The wind power variability in these simulations, ex-
pressed in terms of the mean absolute ramp rate, co-
efficient of variation and maximum ramp ups, is
summarised in Table 2 at the ten-minute, hourly and
daily timescales. Figure 9 shows a ramp up-ramp
down pattern similar to that in Figure 7, although in
this instance these ramps are relatively subdued,
likely because of the differing intensities and tracks
of the respective systems. Table 2 presents a wind
power that is more variable in the ‘cold front aggre-
gation’ than in the ‘uniform aggregation’ methods,
which is also similar to what was observed in the
previous simulation. Although the mean power
ramps remained the largest at the longest timescale
in Table 2, this was different for the coefficient of
variation and max ramp ups, which was greater at
the 10-minute timescale.
For the July 2016 simulations, two consecutive
cold fronts were considered: the first making landfall
on 5 July, and the second on 8 July. As with the
previous simulations, a steep ramp up in simulated
wind power generation was followed by a ramp
down. As shown from the synoptic weather maps in
Figure 12, the cold front making landfall on 05 July
was located further to the north and, consequently,
as shown in Figure 11, had a larger impact on the
wind power generated than the cold front making
landfall on 08 July. The variability of these simula-
tions, expressed in terms of mean absolute ramp
rate, coefficient of variation and maximum ramp
ups, is summarised in Table 3. Table 3 reveals simi-
lar trends to Tables 1 and 2, where variability tended
to increase in proportion to the timescale and to de-
crease in inverse proportion to the number of tur-
bines.
Table 1: Summary of the mean absolute power ramp rate, coefficient of variation and maximum
power ramps at ten-minute, hourly and daily timescales for a cold front making landfall on
24 August 2012.
No. of turbines
in aggregate
time series
Period Mean power
ramp rate
(%)
Coefficient
of variation
Max
ramp up
(%)
Max ramp
down
(%)
Total variation
in simulated
capacity (%)
5
10 min 3.67 0.46 40.26 40.00
8.09-100.00
Hourly 6.21 0.46 27.64 20.93
Daily 29.09 0.65 63.07 28.10
10
10 min 2.82 0.56 31.04 21.25
5.49-99.05
Hourly 4.31 0.60 20.49 15.04
Daily 27.37 0.63 49.90 44.10
Table 2: Summary of the mean absolute power ramp rate, coefficient of variation and maximum
power ramps at ten-minute, hourly and daily timescales for a cold front making landfall on
26 May 2013.
No. of turbines
in aggregate
time series
Period Mean
power ramp
rate (%)
Coefficient
of variation
Max ramp
up
(%)
Max ramp
down
(%)
Total variation
in simulated
capacity (%)
5
10 min 3.30 0.39 25.59 13.16
6.22- 93.55 Hourly 5.98 0.29 17.65 17.46
Daily 14.74 0.22 13.32 21.11
10
10 min 2.07 0.37 13.78 10.51
5.77-71.01
Hourly 3.65 0.16 10.56 9.80
Daily 14.64 0.35 11.97 26.28
60 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
Figure 9: Wind power simulation for (a) ‘uniform aggregation’, and (b) ‘cold front-dominated
aggregation’ methods, during the passage of a cold front making landfall on 26 May 2013. The
yellow line indicates the start of the passage of the cold front (CF) through the region under
consideration.
Figure 10: Synoptic weather map depicting the cold front considered for the 26 May 2013 wind
power simulation. (South African Weather Service, 2018).
(a)
(b)
61 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
Table 3: Summary of the mean absolute power ramp rate, coefficient of variation and max-
imum power ramps at ten-minute, hourly and daily timescales for the consecutive cold fronts
making landfall on 5 and 8 July 2016.
Date
No. of tur-
bines in ag-
gregate time
series
Period Mean
power
ramp rate
(%)
Coefficient
of variation
Max
ramp up
(%)
Max ramp
down
(%)
Total varia-
tion in simu-
lated capac-
ity (%)
2016-
07-05
6
10 min 2.67 0.52 16.52 12.05
7.99-80.14 Hourly 5.24 0.52 18.28 16.87
Daily 17.05 0.44 39.38 10.89
13
10 min 1.81 0.34 8.99 6.19
12.59-73.16
Hourly 3.09 0.33 10.91 7.73
Daily 11.37 0.21 19.94 10.09
2016-
07-08
6
10 min 2.17 0.47 10.47 10.36
4.78-66.73
Hourly 5.24 0.47 18.28 16.87
Daily 21.04 0.32 28.49 21.04
13
10 min 1.55 0.33 16.52 12.05
7.93-49.84
Hourly 3.38 0.33 9.03 9.03
Daily 18.04 0.28 25.03 11.01
Figure 11: Wind power simulation for (a) ‘uniform aggregation’, and (b) ‘cold-front dominated
aggregation’ methods, during the passage of consecutive cold fronts making landfall on 5 and 8
July 2016. The yellow lines indicate the start of the passage of the cold front (CF) through the region
under consideration.
Figures 13–14 show that the mean absolute
power ramp rate increased significantly as the tem-
poral resolution decreased from ten minutes to
hourly and to daily for all cold fronts considered.
This highlights the importance of conducting analy-
sis at longer time scales when studying the impact of
synoptic scale systems on wind power variability.
Mean absolute power ramps across timescales were
larger for the ‘cold front-dominated aggregation’
than for the ‘uniform aggregation’. The mean differ-
ence in percentage in mean absolute power ramps
between the two aggregation methods were:
30.30% at ten minutes, 36.52% at hourly, and
13.52% at daily. At the ten-minute and hourly time-
scales, differences between the aggregation scenar-
ios were larger than at the daily timescale. When
comparing differences in the coefficient of variation
between the aggregation scenarios, it was similarly
(a)
(b)
62 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
found that the cold front dominated aggregate and
usually had a higher coefficient of variation than the
uniform aggregate across timescales. The differ-
ences between aggregation scenarios were: 11.95%
at ten minutes, 20.18% at hourly, and 2.19% at
daily. Similar to comparisons between mean abso-
lute ramp rates, differences in the coefficient of var-
iations between aggregation scenarios were larger at
the ten-minute and hourly timescales than at the
daily timescale.
Figures 15–20 illustrate wind power simulations
for the two geographic aggregation scenarios con-
sidered throughout this study, for the months of Au-
gust 2012, May 2013 and July 2016, at ten-minute,
hourly and daily timescales. Visual inspection shows
that most of large spikes in wind power production,
especially at the daily timescale, were preceded by a
cold front making landfall in the Western Cape.
Figure 12: Synoptic weather maps depicting the respective cold fronts considered for the July 2016
wind power simulation, where (a) shows the front making landfall on the 5 July, and (b) shows the
front making landfall on 8 July (South African Weather Service, 2018).
Figure 13: Mean absolute ramp rates from ‘uniform aggregation’ wind power simulations during the
passage of individual cold fronts (CF).
0
5
10
15
20
25
30
10min Hourly Daily
Mean absolute ramp rate (%)
Time Scale
Uniform Aggregation
CF -
2012/08/25
CF -
2013/05/26
CF -
2016/07/05
CF -
2016/07/07
(a)
(b)
63 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
Figure 14: Mean absolute ramp rates from ‘cold front-dominated aggregation’ simulations during the
passage of individual cold fronts (CF).
Spikes in power production that was not accom-
panied by a cold front could also be identified. An
example of this was 27 August 2012, where mostly
clear sky conditions were observed and the weather
was dominated by a ridging high pressure system
extending east from the quasi-stationary Atlantic
high pressure system onto the sub-continent. The
Atlantic or Indian Ocean high pressure systems ridg-
ing onto the subcontinent were also identified by
Kruger et al. (2013) as a common source of annual
maximum wind gusts. Notwithstanding such less
common instances in winter, Figures 15–20 show
that the majority of large ramp events during the
months considered, especially at longer timescales,
were caused by the passage of cold fronts.
Figure 15: Wind power simulation for the ‘uniform aggregation’ for August 2012. The yellow lines
indicate the start of the passage of the cold fronts (CF).
0
5
10
15
20
25
30
10min Hourly Daily
Mean absolute ramp rate (%)
Time Scale
Cold Front Dominated Aggregation
CF -
2012/08/25
CF -
2013/05/26
CF -
2016/07/05
CF -
2016/07/07
64 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
Figure 16: Wind power simulation for the ‘cold front-dominated aggregation’ for August 2012.
The yellow lines indicate the start of the passage of the cold fronts (CF).
Figure 17: Wind power simulation, consisting of an aggregate of 10 turbines for May 2013. The
yellow lines indicate the start of the passage of the cold fronts (CF).
Figure 18: Wind power simulation for the ‘cold front-dominated aggregation’ for May 2013. The
yellow lines indicate the start of the passage of the cold fronts (CF.
65 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
Figure 19: Wind power simulation for the ‘uniform aggregation’ for July 2016. The yellow lines
indicate the start of the passage of the cold fronts (CF).
Figure 20: Wind power simulation for the ‘cold front dominated aggregation’ for July 2016. The
yellow lines indicate the start of the passage of the cold fronts (CF).
4. Conclusions
Based on the simulations of wind power production
during the passage of cold fronts as conducted
within this study, the following conclusions are
made:
• When aggregating a wind power production
time series from geographically distributed gen-
erators, a reduction in wind power variability
was represented both in terms of mean absolute
power ramps and the coefficient of variation.
• A sharp ramp up in wind power production fol-
lowed by a slightly more gradual, but still signif-
icant, ramp down, occurred with the passage of
individual cold fronts despite the reduction in
wind power variability through aggregation ef-
fects. The severity of this ramp-up ramp-down
is dependent on the severity and location of the
cold front in question.
• A substantial increase in mean absolute ramp
rate occurred when increasing the period under
consideration from timescales of minutes to
hours to days. This highlighted the importance
of considering longer periods when studying
and attempting to quantify the variability of
wind power production with the passage of syn-
optic-scale weather systems. This is also of im-
portance to electricity network operators in
terms of load scheduling and planning of short-
term maintenance in thermal power plants with
the approach and onset of weather systems
such as cold fronts.
• The relatively small mean absolute ramp rate at
the ten-minute timescale could result in a mas-
sive variation of total simulated capacity over
the four-day simulation period (e.g. 5.49–
99.05% for the 25 August 2012 ‘uniform aggre-
66 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
gation’ simulation) with the passage of a cold
front. This highlighted the importance of analys-
ing wind power variability at longer timescales.
• Most large ramp events during the winter
months considered, especially at longer time-
scales, are caused by the passage of cold fronts
across South Africa.
Acknowledgements
The authors would like to acknowledge the support pro-
vided by the Centre for Renewable and Sustainable En-
ergy Studies at Stellenbosch University, South Africa, and
the feedback received from the WindAC Conference,
2018.
Author roles
Amaris Dalton: Lead author, corresponding author, per-
formed central experiments and wrote draft article.
Bernard Bekker: Co-author; revised draft article prior to
submission, gave detailed feedback and provided super-
vision throughout the project.
Andries Kruger: Co-author; significant contributions to
sections of the article dealing specifically with meteorology
including identification of the time of passage of cold
fronts and the strong wind mechanisms in South Africa.
References
Albadi, M.H. & El-Saadany, E.F. 2010. Overview of wind power intermittency impacts on power systems. Electric
Power Systems Research 80(6): 627–632. Available at: http://dx.doi.org/10.1016/j.epsr.2009.10.035.
Choukri, K., Naddami, A. & Hayani, S.M. 2017. Deep analysis of wind variability and smoothing effect in Moroccan
wind farms. Wind Engineering 41(4): 272–281.
https://doi.org/10.1177/0309524x17709731
Council for Scientific and Industrial Research [CSIR]. 2010. No title. Available at:
http://wasadata.csir.co.za/wasa1/WASAData [Accessed July 10, 2018].
Couto, A., Costa, P., Rodrigues, L., Lopes, V. V. & Estanqueiro, A. 2015. Impact of weather regimes on the wind
power ramp forecast in Portugal. IEEE Transactions on Sustainable Energy 6(3): 934–942.
https://doi.org/10.1109/tste.2014.2334062
Department of Energy. 2018. Draft Integrated Resource Plan. Pretoria.
ECR Newswatch. 2016. PICS: Cold weekend brings snow in parts of KZN. Available at:
https://www.ecr.co.za/news/news/pics-cold-weekend-brings-snow-kzn/ [Accessed July 10, 2018].
Eumetrain. 2012. South African Cold Fronts. Available at: http://www.eumetrain.org/satmanu/CMs/SACF/print.htm.
Gallego-Castillo, C., Cuerva-Tejero, A. & Lopez-Garcia, O. 2015. A review on the recent history of wind power ramp
forecasting. Renewable and Sustainable Energy Reviews 52: 1148–1157.
https://doi.org/10.1016/j.rser.2015.07.154
Gallego-castillo, C., Garcia-bustamante, E., Cuerva, A. & Navarro, J. 2015. Identifying wind power ramp causes from
multivariate datasets: A methodological proposal and its application to reanalysis data. IET Renewable Power
Generation 9(8): 867-875
Google Earth Pro, US Department of State Geographer, DATA SIO, US Navy, NOAA, & Image Landsat/ Copernicus.
2019. South Africa.
Joubert, C.J. 2017. Geographical location optimisation of wind and solar photovoltaic power capacity in South Africa
using mean- variance portfolio theory and time series clustering. Master of Engineering dissertation, Stellenbosch
University, Stellenbosch South Africa.
https://doi.org/10.29252/jafm.12.06.29789
Kalverla, P.C., Steeneveld, G.J., Ronda, R.J. & Holtslag, A.A.M. 2017. An observational climatology of anomalous
wind events at offshore meteomast IJmuiden (North Sea). Journal of Wind Engineering and Industrial
Aerodynamics 165 (2017): 86–99. Available at: http://dx.doi.org/10.1016/j.jweia.2017.03.008.
https://doi.org/10.1016/j.jweia.2017.03.008
Kiviluoma, J., Holttinen, H., Weir, D., Scharff, R., Söder, L., Menemenlis, N., Cutululis, N.A., Lopez, I.D. & Lannoye,
E. 2016. Variability in large-scale wind power generation. Wind Energy 19:1649–1665.
https://doi.org/10.1002/we.1942
Knorr, K., Zimmermann, B., Bofinger, S., Gerlach, A.K., Bischof-Niemz, T. & Mushwana, C. 2016. Wind and Solar PV
Resource Aggregation Study for South Africa.
Kruger, A.C. 2011. Wind climatology of South Africa relevant to the design of the built environment. PhD dissertation,
Stellenbosch Univeristy, Stellenbsoch, South Africa.
Kruger, A.G., Goliger, A.M., Retief, J. V. & Sekele, S. 2010. Strong wind climatic zones in South Africa. Wind and
Structures, An International Journal 13(1): 37–55.
https://doi.org/10.12989/was.2010.13.1.037
67 Journal of Energy in Southern Africa • Vol 30 No 3 • August 2019
Lacerda, M. & Estanqueiro, A. 2017. Wind power ramps driven by windstorms. Energies 10 (10): 1475.
https://doi.org/10.3390/en10101475
Lang, D.T. 2018. Package ‘XML’. Available at: http://www.omegahat.net/RSXML.
Mararakanye, N. & Bekker, B. 2019. Renewable energy integration impacts within the context of generator type,
penetration level and grid characteristics. Renewable and Sustainable Energy Reviews 108(March): 441–451.
https://doi.org/10.1016/j.rser.2019.03.045
Mcewan, C. 2017. Spatial processes and politics of renewable energy transition: Land, zones and frictions in South
Africa. Political Geography 56: 1–12.
https://doi.org/10.1016/j.polgeo.2016.10.001
Monforti, F., Huld, T., Bódis, K., Vitali, L., Isidoro, M.D. & Lacal-arántegui, R. 2014. Assessing complementarity of
wind and solar resources for energy production in Italy. A Monte Carlo approach. Renewable Energy 63: 576–586.
https://doi.org/10.1016/j.rser.2015.12.318
Monforti, F., Gaetani, M. & Vignati, E. 2016. How synchronous is wind energy production among European
countries? Renewable and Sustainable Energy Reviews 59: 1622–1638. Available at:
http://dx.doi.org/10.1016/j.rser.2015.12.318.
Mortensen, N.G., Hansen, J.C., Kelly, M.C., Szewczuk, S., Mabille, E. & Prinsloo, E. 2012. Wind Atlas for South Africa
(WASA) Observational wind atlas for 10 met. stations in Northern, Western and Eastern Cape provinces. Wind
Atlas for South Africa (WASA).
Prasad, R. D., Bansal, R. C. & Sauturaga, M. 2009. Some of the design and methodology considerations in wind
resource assessment. IET Renewable Power Generation 3(1): 53–64.
https://doi.org/10.1049/iet-rpg:20080030
Prasad, R.D., Bansal, R.C. & Sauturaga, M. 2009. Wind energy analysis for Vadravadra Site in Fiji Islands: A case
study. IEEE Transactions on Energy Conversion 24(3): 750–757.
R Core Team. 2018. R: A language and environment for statistical computing. Available at: https://www.r-project.org/.
Skamarock, W. C., Klemp, J. B., Dudhia, J., Gill, D.O., Barker, D. M., Duda, M.G., Huang, X.-Y., Wang, W. &
Powers, J.G. 2008. A Description of the Advanced Research WRF Version 3. NCAR Tech. Note NCAR/TN-
475+STR.
Sørensen, P., Cutululis, N.A., Vigueras-Rodríguez, A., Madsen, H., Pinson, P., Jensen, L.E., Hjerrild, J. & Donovan,
M. 2008. Modelling of power fluctuations from large offshore wind farms. Wind Energy 11(1): 29–43.
https://doi.org/10.1002/we.246
Sørensen, P., Litong-palima, M., Hahman, A.N., Heunis, S., Ntusi, M., and Hansen, J.C. 2018. Wind power variability
and power system reserves in South Africa. Journal of Energy in Southern Africa 29(1): 59–71.
https://doi.org/10.17159/2413-3051/2017/v29i1a2067
South African Weather Service. 2018. No Title. Available at: http://www.weathersa.co.za/climate/publications
[Accessed July 10, 2018].
Thapar, V., Agnihotri, G. & Sethi, V.K. 2011. Critical analysis of methods for mathematical modelling of wind turbines.
Renewable Energy 36(11): 3166–3177.
https://doi.org/10.1016/j.renene.2011.03.016
Ueckerdt, F., Brecha, R. & Luderer, G. 2015. Analyzing major challenges of wind and solar variability in power
systems. Renewable Energy 81(2015): 1-10.
https://doi.org/10.1016/j.renene.2015.03.002
Widén, J. et al. 2015. Variability assessment and forecasting of renewables: A review for solar, wind, wave and tidal
resources. Renewable and Sustainable Energy Reviews 44: 356–375.
https://doi.org/10.1016/j.rser.2014.12.019
