Energy Policy 39 (2011) 1265–1276

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

A methodology for analysis of impacts of grid integration of

renewable energy
Mel George, Rangan Banerjee n
Department of Energy Science and Engineering, Indian Institute of Technology, Bombay, Powai, Mumbai 400 076, India

a r t i c l e i n f o abstract

Article history: Present electricity grids are predominantly thermal (coal, gas) and hydro based. Conventional power
Received 8 April 2010 planning involves hydro-thermal scheduling and merit order dispatch. In the future, modern renew-
Accepted 29 November 2010 ables (hydro, solar and biomass) are likely to have a significant share in the power sector. This paper
presents a method to analyse the impacts of renewables in the electricity grid. A load duration curve
Keywords: based approach has been developed. Renewable energy sources have been treated as negative loads to
Renewable energy obtain a modified load duration curve from which capacity savings in terms of base and peak load
Grid integration generation can be computed. The methodology is illustrated for solar, wind and biomass power for
Load duration curve Tamil Nadu (a state in India). The trade-offs and interaction between renewable sources are analysed.
The impacts on capacity savings by varying the wind regime have also been shown. Scenarios for
2021–22 have been constructed to illustrate the methodology proposed. This technique can be useful
for power planners for an analysis of renewables in future electricity grids.
& 2010 Elsevier Ltd. All rights reserved.

1. Introduction generation, the methods of analysis for making optimal decisions

The present share of renewables in most electric grids is
marginal. Electricity generation from renewables is assuming
increasing importance in the context of rapid and accelerated
growth in energy demand and negative environmental external-
ities due to predominance of fossil fuels in the generation mix.
The share of renewables in electricity generation is around 18%,
with 15% of global electricity coming from hydroelectricity and
3.4% from new renewables (Global Status Report, 2007). In
particular, the wind and solar photovoltaic markets show yearly
growth rates between 20% and 30% in the last few years (Waldau
and Ossenbrink, 2004). At the end of 2008, worldwide wind farm
capacity was 120,791 megawatts (MW), representing an increase
of 28.8% during the year, and wind power produced 1.3% of global
electricity consumption. Wind power accounts for approximately
19% of electricity use in Denmark, 9% in Spain and Portugal, and
6% in Germany and the Republic of Ireland (World Wind Energy
Association, 2009). The National Solar Mission is a major initiative
of the Government of India and targets the deployment of
20,000 MW of solar power by 2022 (MNRE, 2010).
However, regardless of the type of renewable energy source,
there are obstacles to their implementation in the current
centralised conventional distribution system. For conventional
n
Corresponding author. Tel.: +91 22 2576 7883.
E-mail address: rangan@iitb.ac.in (R. Banerjee).
are well known, but intermittent non-conventional sources pose
special problems. Conventional power planning offers techniques
for fossil fuel power and hydro power plants. These techniques
deal with optimal hydro-thermal scheduling with varying loads,
using mathematical approaches. A conventional generation
system can be operated to follow the load, which is not possible
with non-schedulable renewable power generation technologies.
The output of a wind/solar power plant has daily and seasonal
variations, depending on the site wind/solar insolation regime
and machine/device characteristics. As the share of variable
renewable energy increases and becomes significant, issues
related to capacity expansion planning and dispatch become
important.
The impact of wind energy in the grid has been generally
represented in literature in terms of the capacity credit. The
capacity credit of wind power has been defined as fraction of
installed renewable capacity by which conventional capacity can
be reduced without a loss in security of supply (Giebel, 2006). It
has been shown that calculating the capacity factor of the
resource over a relevant time period usually provides a good
estimate of capacity credit (Milligan and Parsons, 1997; Milligan
and Porter, 2005). Another method is to compute the wind power
generation for time periods corresponding to high systems risk
hours or approximate the system Loss of Load Probability (LOLP)
curve so that high risk hours receive more weight than other
hours. Most methods to assess the capacity credit of a wind
power plant are based on a related reliability measure called the

0301-4215/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enpol.2010.11.054
1266 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276

Loss of Load Expectation (LOLE) (Milligan and Parsons, 1997; Wijk
et al., 1992). However, the total amount of energy generated by a
specific technology does not by itself indicate the full value of that
technology. The contribution is intermittent and its value is
dependent on the correspondence with the major part of the
demand. For seasonal wind/solar regimes like India, neither the
capacity factor, nor the capacity credit reflects the monthly
variation in the renewable energy generation.
This paper illustrates a new approach based on the annual load
duration curve for generation expansion planning with higher
penetration of renewable power. The method, initially proposed
for wind power in Tamil Nadu state (India) in our earlier work
(George and Banerjee, 2009), has been extended here to show the
impact of variation of wind regimes. The capacity savings
obtained for the Indian wind regime are compared with a
different wind regime (from the UK). The suggested methodology
has also been extended to other renewables (solar and biomass
power). The potential savings in base and peak capacity with
increasing renewable energy have been quantified. The interplay
between intermittent sources like solar and wind power has also
been discussed. Future scenarios for Tamil Nadu state in India for
2021–22 have been illustrated using this methodology.
2. Assessment framework
4. The calculated output power is corrected with suitable adjust-
ments for height, inclination and temperature. The details for
each source are described in the following sections.
5. The output powers for each site are summed up and weighted
based on installed capacity at each site to obtain the total
hourly renewable energy generation for the state.
6. The hourly generation data for renewables is compared with the
actual hourly load demand. The annual load duration curve with
and without renewable energy is drawn and savings obtained are
calculated.
7. The energy replaced is estimated from the area between the
two load duration curves. The capacity savings are calculated
considering the load savings from renewables at 6000 h
(considered as base load savings) and at 15% of peak load
hours (considered as peak load savings). The exercise is
repeated for different penetrations of renewables.
8. The method is extended to future scenario modelling by
including a capacity expansion model, which is calculated
from the Gross Domestic Product (GDP) growth rate and
elasticity. New load duration curves are drawn by extrapolat-
ing the existing LDC with the energy growth rate.
Different hypothetical renewable energy capacities may be
used for the simulations and savings and investments required for
each can be obtained.

This paper outlines a method which can be used to quantify

the savings achieved under higher penetration of renewables in
the grid, along with the investments required (land and water
requirements). Here, we propose a methodology based on the
annual load duration curve (LDC) to assess impacts of renewables
on the power system. The LDC method also involves detailed
models for wind and solar PV based generation systems.
The basic framework for our analysis has been shown in Fig. 1
and a detailed model is shown in Fig. 2. The model is developed in
a sequential manner, as described below:
1. Major sites for renewable energy generation are considered.
2. Long period (over 10 year) average hourly wind speed and
solar insolation data for each representative site is obtained.
3. Using the existing installed capacity values and the equipment
characteristics, the hourly output from each type of generation
is computed.
2.1. Micro-level models
2.1.1. Wind power
The wind resource model combined with the performance
characteristics of the wind turbine, gives the wind energy gener-
ated. Major wind power sites in the area are considered with their
installed wind power capacities. The hourly wind speeds at each
site were obtained and extrapolated to the hub height using the
Power Law Index (PLI) for each site. The equation used here is
 a
v2 z2
¼ ð1Þ
v1 z1
where v2 and v1 are wind velocities at heights z2 and z1,
respectively. a is PLI for the given site.
From the characteristics of a large wind turbine, the hourly
wind power output at each site was obtained. Then a weighted

Wind resource model Solar resource model

WECS performance model Solar PV performance model Biomass power generation

Micro level model

Macro level model Utility generation model Load model

Capacity expansion model Economic scenario

Output: Savings and

investments

Fig. 1. Framework showing sub-models involved in the analysis.

 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276 1267

Hourly wind speed data at

Select major sites Select major sites
sensor height Hourly average insolation

Hub height, power law index Hourly solar power Solar module characteristics
Hourly wind speed at hub
generation at each site
height
Cos ( β)
Wind turbine characteristics
Corrected power for panel
Hourly wind power at hub inclination
height
Power coefficient (W/ °C)
Installed wind power at Corrected power for temperature
each site effect
Extrapolated hourly wind
power generated in the Installed solar power
state Extrapolated hourly solar capacity at each site
power generated in the state

Hourly total renewable

Hourly biomass energy Installed biomass
energy generated
generated power capacity

Hourly load Effective load curve

curve
Divide load curve into equal-sized bins

Record the no. of hrs. in each bin

Frequency distribution of load over the year

Sum up to obtain annual LDC

Input n and n
discrete
Evaluate for n discrete renewable energy capacities
capacities of
Wind, Solar and
Compute base and peak load savings from diff. LDCs
Biomass power

Fig. 2. Detailed flowchart showing the proposed methodology.

addition (based on installed capacity at each site) of these values Finally, the hourly solar power generation at each site was
was performed to get the hourly wind power output for the state. added based on the installed capacities to obtain the hourly solar
PV generation curve for the state.

2.1.2. Solar photovoltaic power

2.1.3. Biomass power model
The micro-model for solar PV generation uses hourly average
Since biomass power can be controlled (unlike intermittent
insolation values. Major sites are considered based on the solar
sources like solar and wind power) and is dispatchable, no separate
energy survey for the state. The efficiency–insolation characteristics
model has been developed. The monthly biomass energy genera-
of a solar module are used to get the hourly solar power generation
tion has been equally divided during the day to get the hourly
at each site. The correlation is given by (Durisch et al., 1996)
biomass power output for the state.
q2 Is
Z25 ¼ q1 Is þ ð2Þ
2.2. Load duration curve based analysis
q3 þIs

where Z25 is the solar module efficiency at 25 1C, Is the global
insolation (W/m2), and q1, q2 and q3 are constants. Here,
q1 ¼  0.00487, q2 ¼23.01 and q3 ¼188.1 (Durisch et al., 1996).
By using the solar insolation data, efficiency and effective
module area the variation of solar power during the day was
calculated. The major assumption here is the absence of any
storage for the power produced.
This was then adjusted to include the inclination of the solar
panels. The panel inclination angle (b) for different latitudes is
obtained from Farret and Simoes (2006). This was corrected to
include the effect of temperature. The temperature coefficient
method suggested by Evans (1981), Notton et al. (2005)
and Skoplaki and Palyvos (2009) has been used here. The module
efficiency is assumed to fall at a constant rate with increasing
temperature as per the relation:
Zc ¼ ZTref ½1bref ðTc Tref Þ
where Zc is the cell/module efficiency at temperature Tc, Tref is
the reference temperature (25 1C) and bref is the temperature
coefficient (K  1).
A load duration curve (LDC) is used to illustrate the relationship
between generation capacity requirements and capacity utilisation.
It is similar to a load curve but the demand data is ordered in
descending order of magnitude, rather than chronologically. The
LDC shows the capacity utilisation requirements for each incre-
ment of load. The height of each slice is a measure of capacity, and
the width of each slice is a measure of the utilisation rate or
capacity factor. The product of the two is a measure of electrical
energy (e.g. kWh). Load duration curves are constructed by divid-
ing the chronological loads into different time periods. The smaller
the steps, more is the chronological information retained.
The effect of adding intermittent renewable resources into the
system can be modelled as a negative demand (Wijk et al., 1992;
Castro et al., 1996; Fsadni and Mallia, 2006), which accounts for
the inherent fluctuations of the produced power. The LDC meth-
ð3Þ odology identifies the same trends as hour-by-hour simulations.
Here, the power system is characterised as a black box to
which different power stations are connected in order to share
the required energy supply. This system is characterised by the
load duration curve, which accounts for the needs of electricity
1268 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276

100

LDC with wind power = 0 MW

90

80 y - average peak
capacity saved

Percentage of peak load

70 x - base
capacity saved
LDC with wind power = p MW
60

50

40

30

20

10

0
0 876 1752 2628 3504 4380 5256 6132 7008 7884 8760
Time in hours

Fig. 3. Generalised illustration of LDC methodology.

consumers. The consequences of adding non-schedulable genera- Present LDC

tion depends on how it changes the effective load duration curve.
The main impact is a saving in energy and capacity requirements
from existing generation facilities. The energy replaced is readily
estimated by the difference between the two load duration curves.
To encapsulate the load duration characteristics of the
demand, the load was segmented into separate load bins (each
of 100 MW, in this work) and an algorithm was developed to
record the number of hours during the year when the load was
within the range of the bin. The energy to be served in each bin is
found by multiplying the number of hours in the bin by the
median value of the bin. The programming was done using
MATLAB 7.1. Considering the base load to last for 6000 h and
average peak load to be the average of 15% of the highest load
hours, the base and average peak capacity values were obtained
for different penetrations of wind, solar PV and biomass power in
the Tamil Nadu grid.
The use of the method is shown using two general load
duration curves in Fig. 3.
The figure shows two general load duration curves, without
and with wind power (wind power considered as negative load on
the system). The shaded region shows total energy supplied by
wind power. ‘x’ denotes the base capacity saved due to wind
power. It is calculated from the difference of the two LDCs at the
6000 h. point. ‘y’ is the average peak capacity saved and is
computed by averaging the difference between the two LDCs in
the region up to 15% of the peak load hours (about 1300 h).
2.3. Projections for the future
The LDC methodology proposed above can also be used to plan
capacity expansions in future. Different scenarios can be consid-
ered and each option’s technical potential and inputs evaluated.
The load duration profile for the state system in future is
developed using projections for the Gross Domestic Product
(GDP) growth rate and elasticity. Here, it is assumed that the
load shape will remain the same, but the method can incorporate
a change in the future load profile. In general, the methodology
for scenario development is shown in Fig. 4.
GDP growth rate
Projected year
Elasticity
Projected LDC
Projected LDC with Projected renewables
renewables installed capacity
Evaluate of discrete
renewable energy capacities
Compute base and peak
capacity saved
Fig. 4. Extension of LDC methodology for future scenario assessments.
3. Illustrations of methodology: wind power
3.1. Wind power in Tamil Nadu
Tamil Nadu, lying between 8150 N to 131350 N and 761150 E to
801200 E, is one of the 28 states in India. It has a size
(130,058 km2), almost equal to Greece and a population of about
66.4 million in 2008 (GoTN, 2009; Census of India, 2009). The
location of Tamil Nadu in India is shown in Fig. 5, along with the
major wind and solar power sites.
The power situation in the state has been summarised in
Table 1. It can be seen that wind power constitutes about 27% of
the installed capacity in the state and about 10% of the generation.
For Tamil Nadu, hourly wind power generation profiles were
not available. These were modelled considering the major wind
sites and using hourly average wind speed data. The hourly wind
power generation curves were developed using the wind turbine
characteristics and ten year average wind velocities from four
sites in the state (Muppandal, Kethanur, Ayikudy and Kayattar).
These four sites and the location of Tamil Nadu in India are shown
 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276 1269

9000
Present
LDC

8000 peak 3600 MW

wind 5500 9000 MW
capacity base capacity
MW wind saved
saved
7000

MW
6000

5000

4000

3000
0 876 1752 2628 3504 4380 5256 6132 7008 7884 8760
time in hrs

Fig. 6. Annual load duration curves for Tamil Nadu under different wind power
scenarios.

Fig. 5. Map of Tamil Nadu, showing location in India and wind and solar power
sites considered in the study.

Fig. 7. Mean monthly wind speeds at Valley, Wales in the UK.

Table 1
Break-up of installed capacity in Tamil Nadu state as on 31 March 2009
3.2. Illustration II: UK wind profile
(TNEB, 2009).

Source Installed Annual energy Annual The Indian wind regime is seasonal in nature. To illustrate the
capacity generated average methodology for a different wind regime, the United Kingdom
(MW) (MU) capacity factor was chosen. It must be mentioned here that the results are
(%) illustrative and for comparison purposes only (they do not exactly
Coal 2970 21,230 81.6
represent the wind power impact in the UK since only one site has
Gas 424 1945 52.4 been considered). The method can be extended to demonstrate
Hydro 2187 6290 32.8 the impacts on the UK load curves by including more sites. Here,
Firm central sharea 2825 17,785 71.9 we study the impact of wind power and load correlation on
Wind (state +private) 3875 5270 18.6
capacity savings.
Other renewables (solar PV, 556 1220 25.1
biomass and bagasse based The UK has a total installed wind power capacity of about
cogeneration) 3200 MW (GWEC, 2010) and a peak load of about 45,000 MW (UK
Independent power projects 1180 6360 61.5 National Grid, 2009). Wind power in UK has a high capacity factor
(coal, lignite, diesel or gas ranging from 0.35 (on shore) to 0.42 (off shore).
based)
Assistance from other regional 519 2280 50.1
Data was available for one site, Valley in Hollyhead, near
grids Anglessey in Wales. The wind speed data was obtained from
Total 14,536 63,370 49.8 the European Wind Atlas (2009) and load data was taken from
the UK National Grid (2009). The chosen site has a very flat wind
a
The allotted generation capacity for the state electricity utility from the speed profile during the year (Fig. 7), typical of European
power plants operated by Central Government owned generation companies.
wind regimes. There is only a small increase in the mean monthly
wind speed during November to December.
It is seen that wind power in UK has a much higher correlation
in Fig. 5. The load curves for the state were obtained from the with the load curve (than the Tamil Nadu example), as evident from
Southern Regional Power Committee (SRPC, 2010) while wind the graphs shown in Figs. 8 and 9. These illustrate the normalised
speed data was taken from Mani (1992). wind and load profiles in UK during March and July, respectively.
Different wind profiles ranging from 0 MW installed capacity The correlation factors during these months are 0.45 and 0.88,
to 100% of the peak demand (9000 MW) were used and the effect respectively. The annual average correlation factor is about 0.43.
on the annual load duration curve was studied. The impacts of The impacts of two wind power scenarios on the annual LDC
some of these wind profiles on the net load duration curve are are shown here in Fig. 10. The base and peak capacity savings are
shown in Fig. 6. Detailed results have been demonstrated given in Table 2. It can be seen that the base and peak capacity
in George and Banerjee (2009). savings are much higher than that for an Indian wind regime.
1270 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276

3.3. Comparison of UK and Tamil Nadu results for wind power It is seen that wind power installation in UK saves about four
times as much base capacity and three times the peak capacity
The capacity savings due to 1000 MW of wind power installed in than an Indian wind regime. This is due to the predominantly flat
Tamil Nadu, India and at Valley, UK are compared here in Table 3. wind profile in UK and the high wind-load correlation (0.43) as
compared to Tamil Nadu (0.32).
1.2
Load
Wind power 4. Extension of methodology to other renewables
Normalised generation w.r.t. peak

1
We now extend the proposed methodology to other renewables
0.8 (solar PV and biomass), considering Tamil Nadu as a case study. We
also consider the interaction between different intermittent sources

0.6

Table 2
0.4 Base and peak capacity savings in UK for different wind power scenarios.

Installed Base Peak Base capacity saved Peak capacity

0.2 wind capacity capacity (% installed wind saved (% installed
power saved saved power capacity) wind power
(MW) (MW) (MW) capacity)
0
0 5 10 15 20 25 3200 2740 420 85.6 13.2
Hrs. 6000 4800 2790 80 46.5

Fig. 8. Wind and load profiles (normalised) for UK during the month of March.

Table 3
1.2 Base and peak capacity saved by 1000 MW wind power installation.
UK load
wind power Location Base capacity Peak capacity
1 saved (MW) saved (MW)
normalised power generated

Tamil Nadu 200 65

0.8 United Kingdom 788 180

0.6

Table 4
0.4 Assumed installed capacities of solar PV at sites in Tamil Nadu.

Site Location Installed SPV (MWp) Location co-ordinates

0.2
1 Chennai 40 13.01N, 80.21E
2 Kodaikanal 35 10.21N, 77.51E
0 3 Tuticorin 45 8.81N, 78.21E
0 5 10 15 20 25 4 Coimbatore 30 11.01N, 77.11E
Hrs.

Fig. 9. Wind and load profiles (normalised) for UK during the month of July.

160
50000 March
July
140 September
No wind power
45000 December
3200 MW wind (present 120
installed capacity)
40000 100
6000 MW wind (2011
estimate)
MW

80
MW

35000

60
30000
40

25000 20

20000
0
0 876 1752 2628 3504 4380 5256 6132 7008 7884 8760 0 4 8 12 16 20 24
time in hrs hrs

Fig. 10. Wind power impacts on annual LDC for UK. Fig. 11. Hourly solar PV generation in Tamil Nadu during different months.
 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276 1271

250

Monthly generation for 1000 MW solar PV (MU)

225

200

175

150

125

100
jan feb mar apr may jun jul aug sep oct nov dec
months

Fig. 12. Variation of monthly solar PV generation in Tamil Nadu.

Fig. 13. Monthly variation of biomass based power generation for Tamil Nadu.

like wind and solar and their impacts on the capacity savings 100
obtained.
90
4.1. Impacts of solar photovoltaic power WS S W
percentage of peak load

80
The micro-model for solar PV generation uses hourly average
insolation values (W/m2) obtained from Mani and Rangarajan (1982).
70
Four major sites in Tamil Nadu have been considered based on the
solar energy survey for the state by Mani. Since the present installed
solar photovoltaic power capacity in the state is negligible, a total of 60
150 MWp capacity was assumed to be distributed across the four
sites.1 Table 4 gives the details for these sites. 50
The hourly solar power generation at each site was added
based on the installed capacities to obtain the hourly solar PV 40
generation curve for the state (Fig. 11), using the methodology
described in Section 2.1. The monthly solar energy generation has 30
been shown in Fig. 12. It can be seen that the energy generated is 0 876 1752 2628 3504 4380 5256 6132 7008 7884 8760
time in hrs
1
India has plans to significantly augment its solar power generation with the Fig. 14. Combined impacts of wind and solar PV generation on the load
National Solar Mission that targets 10 GW of solar power by 2020. duration curve.
1272 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276

Table 5 highest during the months February to April (about 220 MU) and
Capacity savings with wind and solar power. lowest (around 150 MU) during two monsoon periods for the
state, i.e. June to August and October to November. The average
Scenario Wind Solar Base capacity Peak capacity saved (% of
power power saved installed renewable energy generated is about 180 MU and average annual capacity
(MW) (MW) (% of installed power) factor is about 0.21.
renewable
power)
4.2. Biomass based power in Tamil Nadu
W 1000 0 20 6.4
S 0 1000 42 7.2
WS 1000 1000 31 10.4 Tamil Nadu has an installed capacity of 415 MW, which
includes 90 MW from biomass based combustion and 325 MW
from bagasse based cogeneration (TEDA, 2009). For the biomass
power generation model, the monthly generation is assumed
to be equally distributed across all hours during the month. This
is justifiable since biomass based power plants can be
scheduled to meet the demand. Biomass power may be operated
to reduce the peak demand. The monthly generation is as shown
in Fig. 13.
Due to the higher amount of bagasse based power in Tamil
Nadu, there is a wide variation in biomass power generation
during the year. It varies from about 160 MU during the
sugarcane crushing season (March to June) and reduces to about
20 MU just before the harvesting season (December to January).
This seasonal variation has been accounted in the model by
considering the monthly energy generated and distributing it
equally over the hours in the month. The average capacity
factor is 0.33, which is higher than other renewable energy
sources.

Fig. 15. Normalised wind power, solar power and load variation for Tamil Nadu. 5. Interplay between wind and solar power

The usefulness of renewable energy sources cannot be

Table 6
assessed without considering their operation within the context
Correlation factors between wind, solar power and load in Tamil Nadu.
of the connected electric system. The value of wind and solar
Correlation factor w.r.t. load power is strongly dependent on the relationship of the times at
which power is available to the times of peak load. The con-
Wind 0.32 straints imposed by the nature of renewable sources, such as
Solar power 0.24
Wind+ solar power 0.29
variability, stress the need for assessing the potential of hybrid
systems.

Fig. 16. Extrapolated surface showing base capacity savings for different penetrations of wind and solar power in Tamil Nadu.
 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276 1273

2000

Peak capacity savings in MW

1600

1200

800

400

0
9000
7500
9000
6000 7500
4500 6000
3000 4500
3000
Solar PV in MW 1500 1500 Wind in MW
0 0

Fig. 17. Extrapolated surface showing peak capacity savings for different penetrations of wind and solar power in Tamil Nadu.

Table 7
Savings in base and peak capacity under different hybrid scenarios.

Scenario Wind Solar Biomass Base capacity saved Base capacity saved Peak capacity saved Peak capacity saved
(MW) (MW) (MW) (MW) (% of installed renewable (MW) (% of installed renewable
capacity) capacity)

A 4000 1000 1000 1568 30.31 1252 27.22

B 5000 1000 1500 2137 28.66 1423 26.79
C 3000 3000 3000 2984 33.15 2276 25.29

The methodology proposed in Fig. 2 was used to study the 9000

impacts of three scenarios:
8000
 W—with 1000 MW wind power in the grid,
 S—with 1000 MW solar PV in the grid and
 WS—with 1000 MW each of wind and solar PV in the grid. 7000

The effective LDCs obtained in each case are shown in Fig. 14.
MW

6000
The capacity savings obtained in each scenario are given in
Table 5. A
It can be seen that the savings obtained are not the same 5000
B
(savings from 1000 MW wind alone with 1000 MW solar alone is
C
not the same as the savings from 1000 MW each of wind and solar, 4000
or W+SaWS). This can be attributed to the differences between
the two intermittent renewable sources, which are correlated with
3000
the load. The normalised graph showing correlation between wind 0 876 1752 2628 3504 4380 5256 6132 7008 7884 8760
power, solar power and the load for Tamil Nadu is presented in time in hrs
Fig. 15. The Pearson product-moment coefficient or sample corre-
lation coefficient has been calculated using Eq. (4). For a series of n Fig. 18. Three illustrative hybrid renewable scenarios for Tamil Nadu—impacts on
measurements of X and Y, written as xi and yi where i¼ 1, 2, y, n, the annual LDC.

the Pearson correlation coefficient is written as

Pn
ðx xÞðyi yÞ
rxy ¼ i ¼ 1 i
ðn1Þsx sy
where x and y are the sample means of X and Y, and sx and sy are
the sample standard deviations of X and Y. In general, the
correlation coefficient indicates the strength and direction of a
linear relationship between two sets of random variables. The
computed results have been shown in Table 6.
The base capacity savings achieved by varying penetrations of
wind and solar power can be characterised as a surface and this
ð4Þ has been shown in Fig. 16. The illustration shows the possible
base load capacity saved in Tamil Nadu for any combination of
wind and solar power installed capacities.
An extrapolated surface was also obtained for the peak
capacity savings under different wind-solar PV scenarios and it
has been plotted in Fig. 17. Comparing Figs. 16 and 17, we can see
that there is a linear variation of base capacity saved with
installed wind or solar power, but the peak capacity savings
become constant beyond a certain penetration level.
1274 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276

Table 8
Capacity and emission savings for different hybrid renewable scenarios in 2021–22.

Scenario Wind Solar Biomass Base capacity Peak capacity % of total Emission
(MW) (MW) (MW) saved (MW) saved (MW) energy savings
demand (MtCO2/yr)

A 5500 1000 720 2080 750 10 12.1

B 5500 2000 510 2260 840 10 12.1
C 5500 1000 2640 2460 1120 20 24.2
D 5500 2000 2430 2970 1140 20 24.2
E 11,000 1000 1650 3710 1310 20 24.2
F 11,000 2000 1440 3800 1330 20 24.2

Table 9
Inputs required under different hybrid scenarios for 2021–22.

Scenario Wind (including spacing) Solar PV Biomass Total land area Land area per 100 MW of installed Water
(km2) (km2) (low yield) (km2) required (km2) renewable capacity (km2) (cu.m./day)

A 583 26 3154 3763 52.1 6048

B 583 52 2234 2869 35.8 4284
C 583 26 11,563 12,172 133.2 22,176
D 583 52 10,643 11,278 113.6 20,412
E 1166 26 7227 8419 61.7 13,860
F 1166 52 6307 7525 52.1 12,096

6. Hybrid scenarios for Tamil Nadu
At higher penetrations of renewable power, each additional
unit of any particular intermittent renewable energy system has a
smaller benefit than the one before it, as demonstrated using
wind power scenarios in George and Banerjee (2009). On the
other hand, if these systems are used together, then their
complementary characteristics may show increased effectiveness
in the new overall system, and their introduction may be
facilitated, so that a mix of systems develop in the long term.
We have assessed the impacts on hybrid wind-solar PV-biomass
scenarios on the Tamil Nadu grid and the results showing capacity
savings under different scenarios is presented in Table 7. The
impacts of these scenarios on the LDC are illustrated in Fig. 18.
7. Illustrative future scenarios and inputs required
The impacts of a few hybrid scenarios on the future load
duration curve (projected for 2021–22), using the methodology
in Fig. 3 have been simulated. The year 2021 has been chosen
because it is the end of the 14th Five Year Plan for India. A GDP
growth rate of 9% p.a. and an energy-GDP elasticity of 0.8 have
been assumed here.2 The savings achieved and investments
required for these hypothetical scenarios are presented in the
next few sub-sections.
7.1. Capacity and emission savings
Impacts of different scenarios in terms of base and peak
capacity savings in 2021–22 are shown in Table 8. The total
renewable energy generation is obtained by calculating the area
between the original LDC and the effective LDC with renewables.
We have developed scenarios wherein renewables contribute 10%
and 20% of the total energy requirement in 2021. The emission
savings have been computed assuming that no renewable energy
2
The growth rate assumed here is for illustration only. It may be corrected to
include the impacts of the current economic recession.
generated is wasted and it displaces an equivalent amount of
conventional generation. The emission factor considered here is
0.72 tCO2/MWh for the Southern Grid (CEA, 2009).
7.2. Land and water requirements
To assess the inputs required for each of the scenarios to
develop, we have considered land area and water requirements.
The assumptions involved are discussed here.
For wind turbine of 1 MW rating, the area required is about
0.8 acres (NREL, 2009a), while excluding spacing between wind
turbines in a wind farm. When considering the spacing between
the wind turbines (9  diameter between adjacent turbines and
3  diameter between two rows of turbines), the area required for a
1 MW wind farm is about 29 acres in India (MNRE, 2009). According
to Bansal et al. (2002), the land area required is about 25 acres/MW
and we have considered this value in our calculations.
In case of solar PV plants, it has been estimated that a land
area of 6.4 acres is required for 1 MW plant, assuming 20% of the
installations are roof-top based ones and that solar panels will
cover 50% of the land area provided (NREL, 2009b).
For biomass power, 1 Mt of biomass is assumed to produce
1 TWh of output electricity (Ravindranath, 2007). The land area
requirements depend on the productivity (yield). We have
assumed 6.6 t/ha/yr yield of biomass without genetic modifica-
tions or fertilizers and 12 t/ha/yr with genetic improvements and
fertilizers. The water requirements for a 7.5 MW biomass power
plant is about 63 cu.m./day (TEDA, 2009).
Thus, the inputs required for each of the scenarios presented
in Table 8 are computed and shown in Table 9. It may be noted
that the total land area of Tamil Nadu state is about 1,30,000 km2.
The land area required for a 1000 MW coal power plant in
India, including the main power house, ash handling and town-
ship is about 0.6 million km2 (1420 acres) (CEA, 2007). It is seen
that 100 MW from renewables takes up about 60–80 times of this
land requirement. However, the total land area required for these
renewable scenarios is 2–4% of the land area in the state, which
may be available as waste/fallow land.
 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276 1275

7.3. Wind and solar power for equal peak and base capacity savings
Using the load duration curve approach, an analysis was done
to obtain the wind and solar power installation required to save
1000 MW of peak and base load capacity. The combinations for
saving 1000 MW peak are shown in Fig. 19. It is seen that a very
high level of solar power installation is necessary to achieve this,
presumably because of the very low peak coincidence of solar
photovoltaic power (without storage).
The analysis for base capacity saved is more insightful. It is
quite evident that solar power saves more base capacity,
from Table 5. The wind and solar installed capacity required for
1000 MW of base load savings is shown in Fig. 20. A fairly linear
relationship can be seen. It is interesting to note that the slope of
the line is almost equal to the ratio of the wind and solar power
capacity factors (0.181/0.208¼0.866, which is equal to the slope
of 0.866).
This analysis was repeated with the UK wind regime and the
wind and solar power capacity required for 1000 MW base
capacity replacement is also shown. Again, it was found that the
ratio of the annual average capacity factors is almost equal to the
slope of the line (0.32/0.21 ¼1.52).

8. Conclusions

Renewables are gradually entering the mainstream of power

generation. With their increasing penetration in the grid, the
impacts on the power system need to be explored. The conven-
Fig. 19. Installed wind and solar photovoltaic power for 1000 MW of peak
capacity saving. tional power techniques focus mainly on hydro-thermal coordi-
nation. In this paper, a simple load duration curve methodology
with detailed micro-level modelling has been proposed to study
3500 the impacts of renewable energy systems in electric utility
for Indian wind regime, y1 = -0.8666x + 2798.1 systems. This methodology has been applied to Tamil Nadu state
R2 = 98.3%
in India. The savings potential has been calculated in terms of
3000
for UK wind regime, y2 = -1.5261x + 2690.5 base and peak capacity replaced and emissions. Future hybrid
R2 = 97.7% scenarios using wind, solar PV and biomass power for 2021–22
Installed solar power (MW)

2500 have been developed using the methodology and inputs required
(land and water) and emission savings have been computed. The
2000 Indian wind regime
interplay between wind and solar power in Tamil Nadu has been
studied in detail and their value in terms of capacity savings was
UK wind regime compared. The results for Tamil Nadu have been compared with
1500
Linear (Indian that of a UK wind regime and the impact of wind speed—load
wind regime)
curve correlation has been demonstrated.
Linear (UK wind
1000 regime) A number of alternative energy sources and their combina-
tions can be analysed using this simple LDC methodology, to
500 show their operational and economic impacts. The method can be
used to vary any of the inputs viz. the load duration curve
(representing the hourly demand profile for the site) or the
0
0 500 1000 1500 2000 2500 3000 3500 4000 renewable energy generation (by changing the wind speed or
solar insolation regime) and then obtain the possible base and
Installed wind power (MW)
peak capacity savings, energy and emission savings and future
Fig. 20. Installed wind and solar power required for 1000 MW of base capacity investments required in terms of land and water. The input–
saving (Indian and UK wind regimes). output framework for the model is shown in Fig. 21.

Fig. 21. Input–output framework for the proposed model.

1276 M. George, R. Banerjee / Energy Policy 39 (2011) 1265–1276
The analysis using the proposed method will be useful for
power planners to assess the future impacts of increasing renew-
able energy capacity in different grids and design appropriate
strategies for capacity expansion planning.
References
Bansal, R.C., Bhatti, T.S., Kothari, D.P., 2002. On some of the design aspects of wind
energy conversion systems. Energy Conversion and Management, 2002 43
(16), 2175–2187.
Castro, M.A., Carpio, J., Peire, J., Rodriguez, J.A., 1996. Renewable energy integration
assessment through a dedicated computer program. Solar Energy 57 (6),
471–484.
CEA, 2007. Central Electricity Authority, Ministry of Power, Government of India.
Report on the land requirement of thermal power stations, 2007, see: /www.
cea.nic.inS.
CEA, 2009. Central Electricity Authority, Carbon Dioxide Baseline Emission
Database for the Indian Power Sector, /www.cea.nic.in/planning/
c%20and%20e/user_guide_ver2.pdfS (last accessed January 2, 2010).
Census of India, 2009. Projected population, /http://www.censusindia.gov.inS
(last accessed on January 3, 2010).
Durisch, W., Urban, J., Smestad, G., 1996. Characterisation of solar cells and
modules under actual operating conditions. Renewable Energy 8, 359–366.
European Wind Atlas, 2009. UK wind speed data: GWEFR Cyf hourly-mean wind
speed datasets for sites in the European Wind Atlas, available at /http://www.
gwefr.co.uk/datasets.htmS (last accessed on June 3, 2009).
Evans, D.L., 1981. Simplified method for predicting solar photovoltaic array
output. Solar Energy 1981 27 (6), 555–560.
Farret, F.A., Simoes, M.G., 2006. Integration of Alternative Energy Sources. John
Wiley & Sons, New York, pp. 148.
Fsadni, M., Mallia, E.A., 2006. The integration of wind power generation with
electricity demand in the context of a small grid. Renewable Energy 31, 811–819.
George, M., Banerjee, R., 2009. Analysis of impacts of wind integration in the Tamil
Nadu grid. Energy Policy 37 (9), 3693–3700.
Giebel, G., 2006. Wind power has a capacity credit: a catalogue of 50+ supporting
studies. WindEng EJournal, windeng.net (last accessed May 20, 2009).
Global Status Report, 2007. /http://www.ren21.net/pdf/RE2007_Global_Status_Re
port.pdfS (last accessed on January 18, 2010).
GoTN, 2009. Government of Tamil Nadu, Statistical Handbook, /http://www.tn.
gov.in/deptst/index.htmS (last accessed on January 3, 2010).
GWEC, 2010. Global Wind Energy Council, /http://www.gwec.net/index.
php?id=135S (last accessed on January 2, 2010).
Mani, A., 1992. Wind Energy Resource Survey in India, vol. II. Allied Publishers,
New Delhi.
Mani, A., Rangarajan, S., 1982. Solar Radiation over India. Allied Publishers Pvt.
Ltd., New Delhi.
Milligan M., Parsons B., 1997. A comparison and case study of capacity credit
algorithms for intermittent generators. Technical report, National Renewable
Energy Laboratory, CO, USA.
Milligan, M., Porter, K., 2005. Determining the Capacity Value of Wind: A Survey of
Methods and Implementation, WINDPOWER 2005, Denver, CO, USA.
MNRE, 2009. Ministry of New and Renewable Energy, Notes, /http://mnes.nic.in/S
(last accessed on January 2, 2010).
MNRE, 2010. Ministry of New and Renewable Energy, /http://mnre.gov.in/pdf/
mission-document-JNNSM.pdfS (last accessed on January 3, 2010).
Notton, G., Cristofari, C., Mattei, M., Poggi, P., 2005. Modelling of a double glass
photovoltaic module using finite differences. Applied Thermal Engineering 25,
2854–2877.
NREL, 2009a. National Renewable Energy Laboratory (NREL), Power Technologies
– Energy Data Book, /http://www.nrel.gov/analysis/power_databook/calc_
wind.phpS (last accessed on June 8, 2009).
NREL, 2009b. National Renewable Energy Laboratory (NREL), Power
Technologies—Energy data Book, /http://www.nrel.gov/analysis/power_data
book/calc_pv.phpS (last accessed on January 2, 2010).
Ravindranath, N.H., 2007. Biomass for Energy: Resource Assessment for India, see:
/http://www.me.iitb.ac.in/  rangan/Hewlett/hewlett-presentation/Presenta
tions/S (last accessed January 2, 2010).
Skoplaki, E., Palyvos, J.A., 2009. On the temperature dependence of photovoltaic
module electrical performance: a review of efficiency/power correlations.
Solar Energy 83 (5), 614–624.
SRPC, 2010. Southern Regional Power Committee, Annual and Monthly Reports,
Ministry of Power, Government of India, /http://www.srpc.kar.nic.in/reports.
htmS (last accessed on January 2, 2010).
TEDA, 2009. Tamil Nadu Energy Development Agency. Bioenergy last accessed on
January 2, 2010.
TNEB, 2009. Tamil Nadu Electricity Board, /http://www.tneb.in/S (last accessed
on June 2, 2009).
UK National Grid, 2009. UK Load Curves and Demand Data, /http://www.
nationalgrid.com/uk/Electricity/DataS (last accessed June 8, 2009).
Waldau, A., Ossenbrink, H., 2004. Progress of electricity from biomass, wind and
photovoltaics in the European Union. Renewable and Sustainable Energy
Reviews 8 (2), 157–182.
Wijk van, A., Halberg, N., Turkenburg, W., 1992. Capacity credit of wind power in
the Netherlands. Electric Power System Research 23, 189–200.
World Wind Energy Association, 2009. Wind turbines generate more than 1% of
the global electricity /http://www.wwindea.org/home/images/stories/pr_sta
tistics2007_210208_red.pdfS.