Available online at www.sciencedirect.

com

Solar Energy 84 (2010) 11241136 www.elsevier.com/locate/solener

Design of isolated renewable hybrid power systems

E.S. Sreeraj a, Kishore Chatterjee a, Santanu Bandyopadhyay b,*
b a Department of Electrical Engineering, Indian Institute of Technology Bombay, Mumbai 400 076, India Department of Energy Science and Engineering, Indian Institute of Technology Bombay, Mumbai 400 076, India

Received 7 January 2010; received in revised form 15 March 2010; accepted 17 March 2010 Available online 10 April 2010 Communicated by: Associate Editor Mukund Patel

Abstract Isolated electrical power generating units can be used as an economically viable alternative to electrify remote villages where grid extension is not feasible. One of the options for building isolated power systems is by hybridizing renewable power sources like wind, solar, micro-hydro, etc. along with appropriate energy storage. A method to optimally size and to evaluate the cost of energy produced by a renewable hybrid system is proposed in this paper. The proposed method, which is based on the design space approach, can be used to determine the conditions for which hybridization of the system is cost eective. The simple and novel methodology, proposed in this paper, is based on the principles of process integration. It nds the minimum battery capacity when the availability and ratings of various renewable resources as well as load demand are known. The battery sizing methodology is used to determine the sizing curve and thereby the feasible design space for the entire system. Chance constrained programming approach is used to account for the stochastic nature of the renewable energy resources and to arrive at the design space. The optimal system conguration in the entire design space is selected based on the lowest cost of energy, subject to a specied reliability criterion. The eects of variation of the specied system reliability and the coecient of correlation between renewable sources on the design space, as well as the optimum conguration are also studied in this paper. The proposed method is demonstrated by designing an isolated power system for an Indian village utilizing windsolar photovoltaic-battery system. 2010 Elsevier Ltd. All rights reserved.
Keywords: Design space; Renewable hybrid system; Chance constrained method; Process integration; Battery sizing

1. Introduction Isolated power systems using renewable energy sources like wind, solar, biomass, micro-hydro, etc. can be utilized to provide electricity for remote locations where grid extension is not feasible and/or economical. It was estimated that more than 1500 million people around the world had no access to electricity in 2005 (International Energy Agency, 2006). A vast majority of them are from Sub-Saharan Africa and South Asia, where electrication rates are only 25.8% and 51.8%, respectively (International Energy Agency, 2006). In 2001, about 44% of the households in
*

Corresponding author. Tel.: +91 22 25767894; fax: +91 22 25726875. E-mail address: santanu@me.iitb.ac.in (S. Bandyopadhyay).

India do not have access to electricity ( Govt. of India, 2001). As electricity is important for rapid economic growth and poverty alleviation, Indian government has decided to provide electricity access to all households. Along with rapid expansion in conventional power generation, Indian government has also decided to go for power generation from new and renewable sources. For many remote non-electried rural areas, power generation from stand-alone systems is cheaper than grid extension. The National Electricity policy of India states that wherever it is neither cost eective nor optimal to provide grid connectivity, decentralized distributed generation facilities together with local distribution network would be provided so that every household gets access to electricity (Govt. of India, 2005). Non-conventional sources of energy could be

0038-092X/$ - see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.solener.2010.03.017

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

1125

Nomenclature AP ACC AOM B a C0 COE CRF D d Dactual Dt f fi H Hi IT lP j t n gc gd gi gP Pnet Pdu Ppv total array area (m2) annualised capital cost (Rs) annualised operation and maintenance cost (Rs) battery capacity (kW h) condence level capital cost of the component (Rs) cost of energy (Rs/kW h) capital recovery factor load demand (W) discount rate the deterministic demand to be met taken in the chance constrain (W) time step for the simulation (h) factor representing net charging/discharging eciency factor that represents inverter eciency wind turbine height (m) specied meteorological mast height (m) total radiation incident on the array (W/m2) mean of power available from ith power source (W) life of the component (years) battery charging eciency battery discharging eciency inverter eciency photovoltaic system eciency net power available at the dc bus (W) dumped excess power (W) power generated by the photovoltaic array (W) Pr Pw Pj QB Qmax1 Qmax2 Qmin qj(t) qij rP j t T t tmax1 tmax2 tref vc vf vi vr X z rated electrical power of the wind power generating unit (W) power generated by the wind turbine (W) power generated by the jth power source (W) energy stored in the battery (kW h) maximum stored energy of the battery energy before tref (W h) maximum stored energy of the battery energy after tref (W h) minimum stored energy by the battery (W h) multiplication factor for nding out power produced coecient of correlation between the power available from ith and jth sources standard deviation of power available from ith power source (W) time horizon for the simulation (h) time (h) time at which battery reaches maximum energy before tref (h) time at which battery reaches maximum energy after tref (h) time at which battery reaches minimum energy (h) cut-in wind speed (m/s) cut-o wind speed (m/s) wind speed at reference height Hi (m/s) rated wind speed (m/s) depth of discharge of battery power law exponent

utilized even where grid connectivity exists, provided it is found to be cost eective (Govt. of India, 2003). Isolated systems using renewables can be powered by a single or a combination of renewable power sources. The power available from the renewable sources is stochastic in nature. However, some of the renewable resources like solar and wind are complementary in nature. This means that during seasons of low insolation, the wind speed is typically higher and the wind speed is generally low for seasons of high insolation. In Fig. 1, monthly average value of wind speed, monthly average total rainfall, and monthly average daily global insolation for an Indian town, Ratnagiri during each month is plotted (Mani and Rangarajan, 1982). From Fig. 1, it can be observed that the amount of rainfall and wind speed is negatively correlated to solar insolation. During monsoon months, when the rainfall is high, the wind speed is also high and solar insolation is low. The wind speed and rainfall are low for summer months when the insolation is high. Thus, it is apparent from Fig. 1 that it may be advantageous to make a wind solar or a micro-hydro-solar hybrid power system. Due

to complimentary nature of wind and solar power and the cost eectiveness of hybridizing these two systems, a vast literature deals with windsolar hybrid systems and is recently reviewed by Deshmukh and Deshmukh (2008). The performance of a hybrid system depends upon proper sizing of the system. Design and simulation followed by optimization are main steps involved in sizing an isolated hybrid system. The size of a system, that can supply the required power demand, can be determined by simulating the entire system using the resource and the demand data. Optimization of the entire system may be performed to arrive at a sizing which satises certain cost and reliability criteria. This is typically achieved by minimizing the net present cost of the system or the levelized cost of generated energy. The reliability of the power produced by the hybrid system is also generally included in the optimization process either in the form of constrains or as another variable to be maximized. In the latter case, a multi-objective optimization routine has to be revoked and the solution set generally consists of a set of Paretooptimal congurations, out of which a suitable one has

1126
1000 900

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

10

25

700 600 500 400 300 200 100 0

monthly average windspeed(km/h)

800

20

15

10

05

6 month

10

12

Fig. 1. Monthly average rainfall, wind speed and solar insolation at Ratnagiri, India.

to be selected (Gavanidou and Bakirtzis, 1992). The indices that are generally used to evaluate system reliability are: (a) number of days of autonomy (Natarajan and Rajendra Prasada, 2006), (b) loss of load probability (Gavanidou and Bakirtzis, 1992; Beyer and Langer, 1996), (c) loss of power supply probability (Ai et al., 2003; Yang et al., 2003), and (d) unmet load (Chedid and Rahman, 1997). There are primarily two dierent approaches to design and simulate a hybrid system: deterministic and probabilistic (Deshmukh and Deshmukh, 2008). In deterministic approaches, the renewable energy resources and the demand are considered as deterministic quantities and their variation with respect to time is assumed to be known. Usually, time for which the system has to be analyzed (time horizon), is divided into smaller time periods, during which the resources availability and load are assumed to be constant. In deterministic methods, the chronological sequence of the data is extremely important. Sometimes the calculation based on the worst case scenario, say worst month, can also be used in designing of the system (Beyer and Langer, 1996; Protogeropoulos et al., 1997; Celik, 2002a). Typical weather year for a particular location can also be used to design a renewable power system (Yang et al., 2003). However, the typical year data for more than one renewable resource are dicult to obtain. Celik (2002b) has employed synthetically simulated weather data to simulate a hybrid system. These approaches are either computationally intensive (if it uses data for a long time period, say many years) or produce sub-optimal results (if it uses the worst month scenario for calculation). In probability-based approaches, energy generated by power sources, and load demand in some cases, are considered as random variables. Some methods may not consider the chronological sequence of the data, which makes them less accurate. Karaki et al. (1999) proposed a probabilistic model for an autonomous photovoltaicwind system with several wind machines, accounting uncertainties related to uctuations in primary energy output, and outages of

the individual components due to hardware failures. For stand-alone photovoltaicwind hybrid system, the stochastic nature of the resources is modelled as a three event Markov process by Bagul et al. (1996). This is an extension of the two state Markov process, proposed by Bucciarelli (1984). A transformation theorem based method for sizing of a stand-alone photovoltaicwind system has been proposed by Abousdar and Ramkumar (1990, 1991) where solar insolation, wind speed and load are considered as random variables. The set of feasible congurations that can meet the given load demand forms the design space. The concept of design space is introduced for optimized sizing of solar hot water systems by Kulkarni et al. (2007, 2008, 2009). For isolated power systems, the representation of the design space is used for diesel generatorbattery systems (Arun et al., 2008), photovoltaic-battery systems (Arun et al., 2007), wind-battery systems (Roy et al., 2009) and windphotovoltaic hybrid systems (Roy et al., 2007). The design space representation for a photovoltaic battery system incorporating uncertainty through chance constrained programming has been presented by Arun et al. (2009). Chance constrained programming approach (Charnes and Cooper, 1959), dealing with stochastic programming, has been applied in various elds of engineering to deal with uncertainty (Rao, 1980; Changchit and Terrell, 1993; Azaiez et al., 2005; Li et al., 2008). As mentioned earlier, deterministic and probabilistic approaches for system design and simulation have their own merits and demerits. Design methodology utilizing deterministic methods are either computationally intensive or produce sub-optimal results depending upon the type and amount of data used. Whereas, design methodologies based on stochastic methods are simple. However, deterministic methods take into account the chronological sequence of data, and hence, can produce accurate results compared to that of stochastic methods. In order to overcome the limitation of inaccuracy in stochastic methods

monthly average of daily global insolation(kWh/m 2)

insolation wind speed rainfall

monthly average rainfall(cm)

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

1127

reported in the literature (e.g., Bucciarelli, 1984), a new stochastic method to optimally size a renewable hybrid system with more than one power source and to determine the cost of energy produced is proposed in this paper. The method combines the advantages of both the deterministic and probabilistic approaches while retaining inherent simplicity of stochastic method. The system chosen is constrained to supply power to the load satisfying certain reliability criterion. The probabilistic approach is used to incorporate uncertainty in available energy from each power source. The variability associated with power available from each power source along with coecient of correlation between them is used in the analysis as these quantities aect the performance and reliability of the system. The emphasis is on the system design part and the optimum conguration that is found by doing a search in the design space. The proposed sizing method incorporates a simple and novel methodology to nd the battery capacity when renewable resource availability, ratings of renewable power sources and load demand are known. This methodology also helps to plan further load growth, as it gives the durations for which load can be increased without additional capacity growth. When the random nature of the resources is considered, chance constrained programming is used to arrive at the design space. The eectiveness of the proposed method is demonstrated by designing an isolated power system for an Indian village with wind and solar power as potential options.

system with power sources connected to either ac bus or dc bus and load connected to the ac bus is shown in Fig. 2. During some instants the battery may be fully charged and the net power produced by the sources is greater than the demand. The excess power is dumped using the dump load and it is connected to the dc bus. The system considered has n power sources, with the rst m sources (1st to mth) connected to the ac bus and remaining n m sources (m + 1st to nth) connected to the dc bus. Let Pj(t) represents the power generated by the jth source. The net power available at the dc bus that can be used to charge the battery (Pnet(t)) is the sum of the power generated by the sources connected to the dc bus and the power available from the ac bus. ! n m X X P j t P j t Dt fi t 1 P net t
jm1 j1

The term inside the bracket is the net power generated at the ac bus. D is the power required by the load and fi represents the eciency associated with the inverter (gi). Depending on the sign of the net power generated in the ac bus, that is the dierence of power generated by the various power sources and the load demand, the power ow will take place from ac bus to dc bus or vice versa. The power loss in inverter is accounted by the factor, fi which represents the inverter eciency and is given as follows: fi t gi when
m P j1 1 g i

P j t P Dt

2. Design space generation with deterministic approach System sizing methodology for a renewable hybrid system following a deterministic approach is discussed in this section. Renewable power sources can produce ac or dc electric power depending on the type of generator and power electronic interfaces connected to the system. The hybrid system considered here has both ac and dc power sources. The average power available from each power source during every time step is assumed to be known from the resource data and the capacity of the power source. The time series simulation, based on the energy balance of the overall system is performed to arrive at the required battery capacity. The schematic of an isolated renewable hybrid

otherwise

During some time periods, the net energy that can be produced by the system may be greater than the sum of the energy consumed in the load and that can be stored in the battery. This excess energy has to be dumped or lower amount of energy has to be produced by the power sources. The hybrid system considered in this paper has a dump load connected to the dc bus and the excess energy is dumped using that load. The energy transfer across the battery bank dQB t=dt is proportional to the net power available at the dc bus (Pdc(t)) minus the dumped power (Pdu(t)).

Source-1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Source-m+1
Converter
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Source-m
AC Load

Source-n
Battery bank Dump load AC Bus DC Bus

Fig. 2. Schematic of renewable energy based isolated power system.

1128

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

dQB t P net t P du tf t dt

3
Qmax2

where f(t) represents the eciencies associated with the charging (gc) and discharging processes (gd) of the battery. 4
Battery energy

Qmax1

f t gc g1d

when when

P net P 0 P net < 0

The change in stored energy for the time period (Dt) may be expressed as follows: Z tDt P net t P du tf tdt 5 QB t Dt QB t
t

B tmax1

C tmin

D tmax2

Dump energy E time

Qmin

For a relatively small time period, the above equation can be approximated as QB t Dt QB t P net t P du tf tDt Combining Eq. (1) and Eq. (6), we get:
QB t Dt QB t
n X j1m m X j 1

No dumping during this period

Fig. 3. Typical variation of stored energy level.

P j t

P j t Dt fi t P du t f tDt 7

Eq. (7) represents the energy balance across the battery and is solved to nd the battery capacity required to meet the load. The following additional constrains are to be satised: QB t P 0 8 t QB t 0 QB t T 8 9

Eq. (8) ensures that the battery charge level is non-negative. Eq. (9) represents the repeatability of the battery state of energy over the time horizon, T. The power availability and the load at each time step are repeated in the next cycles, and if the initial state of charge of the battery is not equal to the nal, the battery charge will either accumulate or deplete over the time. A methodology is proposed in Section 2.1 to nd the capacity of the battery when the ratings of each power source are known. 2.1. Methodology for nding the battery capacity Since the power available from the sources are known, Eq. (7) is solved to nd the stored energy at the end of each time step over the time horizon by assuming zero dump energy (Pdu = 0). A typical variation of stored energy in the battery is shown in Fig. 3. For nding the minimum battery capacity, the stored energy level is determined for a hypothetical battery with an initial zero state, QB(t = 0) = 0 (point A in Fig. 3). If the nal stored energy level is negative at the end of the time horizon (at point E in Fig. 3), i.e. if QB(t = T) < 0, then the power sources are insucient to meet the load. Starting from zero energy level to positive stored energy at the end of the time horizon is necessary for meeting the demand, incorporating various system losses. The energy to be dumped is indi-

cated by the stored energy level at the end of the time horizon, QB(t = T). Fig. 3 shows the variation of stored energy during the entire time horizon, T. Let, Qmin is the minimum stored energy and it occurs at tref (point C in Fig. 3). On the other hand, Qmax1 is the maximum stored energy level before tref and Qmax2 is the maximum stored energy level after tref. They occur at tmax1 (point B in Fig. 3) and tmax2 (point D in Fig. 3), respectively. Since no energy is dumped in this initial process, the extra energy generated has to be dumped at appropriate time so that Eq. (9) is satised. This is done to reduce the required battery capacity to its minimum value. The excess energy, QB(t = T) can be dumped at various instants to reduce the peak of the stored energy of the hypothetical battery keeping the minimum of the stored energy of the battery to occur at reference time itself. The peak of the stored energy level that occurs after tref, i.e. Qmax2, can be reduced by dumping the excess energy between tref and tmax2, and hence the peak of the energy stored by the actual battery after tref can be lowered by the amount equal to the dumped energy. This is because of the excess energy QB(t = T) that is available between tmax2 and tref. On the other hand, the peak of the stored energy level that occurs before tref, i.e. Qmax1, cannot be reduced as there is no excess energy between tmax1 and tref. The energy stored in an actual battery cannot go negative (8), and a real battery should not be discharged beyond a certain limit. Discharging the battery beyond the specied minimum limit will aect the life of the battery. Typically, a battery is allowed to be discharged till its depth of discharge (X). Combining these, the minimum battery capacity (B) required can be determined as:  B max  10

Qmax Qmin Qmax1 Qmin QB t T ; X X

where QB t T P 0

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

1129

In terms of pinch analysis, the variation of stored energy level is equivalent to the grand composite curve in heat exchange networks and the point of minimum stored energy is equivalent to the pinch point. Pinch analysis began as a thermodynamic-based approach to energy conservation (Linnho et al., 1982), and later evolved over the years to become a powerful tool for resource optimization (Linnho, 1993; Shenoy, 1995; Smith, 1995). Pinch analysis has been fruitfully used in analyzing heat exchanger networks (Shenoy, 1995), utility systems (Shenoy et al., 1998), mass exchanger networks (El-Halwagi and Manousiouthakis, 1989), water networks (Wang and Smith, 1994; Bandyopadhyay et al., 2006; Pillai and Bandyopadhyay, 2007), distillation column (Bandyopadhyay, 2002; Bandyopadhyay et al., 2003, 2004), production planning (Singhvi and Shenoy, 2002; Singhvi et al., 2004), etc. Pinch analysis recognizes the importance of setting targets before design. This allows dierent process design objectives to be screened prior to the detailed design of the process. Pinch analysis provides graphical representation tools and full control to the process designer over decision making processes. Along with nding the minimum battery capacity requirement, the proposed methodology can be extended to nd the maximum possible amount of energy that can be dumped at various instants. This helps to plan further load expansion. A load which consumes less power than the dumped power at that time can be added to the system without any addition in system capacity. The time horizon (T) is divided into four parts, (AB, BC, CD and DE in Fig. 3), using the points A (t = 0), B (t = tmax1), C (t = tref), D (t = tmax2) and E (t = T) as shown in Fig. 3. The excess energy QB(t = T) has to be dumped such that the dierence between the maximum and the minimum of the stored energy level should be minimum. The dierence between Qmax1 and Qmin cannot be reduced further because there is no excess energy between tmax1 and tref. The peak of the stored energy curve that occurs after tref (Qmax2) can be reduced by dumping the excess energy between tref and tmax2. A part of the excess energy, (min(QB(t = T),Qmax2 Qmax1)), has to be dumped during time interval CD. Since the reduction of Qmax2 beneath Qmax1 will not further reduce the battery capacity the remaining excess energy can be dumped during other time intervals. The minimum amount of excess energy that has to be dumped during interval CD is QB(t = T), if Qmax2 Qmax1 P QB t T and Qmax2 Qmax1 , if Qmax2 Qmax1 < QB t T . The remaining excess energy, if any, can be dumped at any time interval except BC. If we dump the excess energy during BC, battery capacity increases as the dierence between Qmax1 and Qmin increases. Table 1 summarizes the results discussed in this paragraph and can be used to allocate timing for dumping excess energy or plan further load growth without increasing the size of the system. The amount of excess energy to be dumped at any instant should always be non-negative, and a negative term in the table indicates no dumping during that period.

Table 1 Time interval for dumping excess energy. Condition Qmax2 Qmax1 > QB t T Qmax2 Qmax1 6 QB t T Dumped power Qmax1 Qmax2 QB t T Qmax2 Qmax1 QB t T Time for dumping AB, CD, DE CD CD

2.2. System optimization The set of all congurations, which can meet the load, forms the design space and it can be used to select the optimum conguration based on the specied objective. The objective function, used in this paper, is the minimum cost of energy (COE) produced by the hybrid system. The cost of energy depends upon the capital cost, operation and maintenance cost and the amount of energy delivered. COE is represented as: COE ACC AOM E 11

where ACC is the annualized capital cost, AOM is the annual operating and maintenance cost, and E is the amount of energy delivered to meet the demand. The annualized capital cost is calculated as: X C 0i CRFi 12 ACC
i

where the capital recovery factor (CRF) is a function of life (n) and the discount rate (d). CRFi d 1 d i n 1 d i 1
n

13

C0i is the capital cost of the ith system component corresponding to dierent power sources (photovoltaic module, wind turbine, etc.) battery bank, inverter, and balance of system. Generation of the design space is illustrated for a windsolar hybrid system in the following section. 2.3. Illustrative example Sukhalai in Hoshangabad district of Madhya Pradesh, India is the location considered for this example. The load prole is given in Fig. 4. The hourly variation of wind speed and solar insolation are taken from Roy et al. (2007). Various parameters used for sizing calculations are given in Table 2 (Roy et al., 2007). In this example the wind system is connected to the ac bus and the solar system is connected to the dc bus. The economic parameters considered for system optimization are given in Table 3 (Kolhe et al., 2002). Even though the power generated by the photovoltaic array is a function of the ambient temperature, operating voltage and current, wind speed, etc., a simple constant eciency model is used to

1130
6 5

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

Maximum load = 3.8 kW

Load (kW)

4 3
Average load = 1.8 kW

specied meteorological mast height, Hi. The wind speed at turbine height H is calculated by the following correlation (Justus, 1978):  z H 16 v vi Hi where v is wind speed at turbine hub height H, vi is wind speed at reference height Hi, and z is the power law exponent. Figs. 5 and 6 show the power produced by a 1 kWp solar PV array and the power produced by a 1 kW rated wind turbine respectively kept in the location of case study. 2.3.1. PVbattery system The solar array rating is varied from zero to 1000 kWp and the minimum battery capacity required to meet the load is determined to generate the sizing curve of the
0.8

2 1 0 0 2 4 6 8 10 12 14 16 Time (hour of the day) 18 20 22 24

Minimum load = 0.6 kW

Fig. 4. Load variations on a typical day at Sukhalai in Hoshangabad.

Table 2 Input parameters used in the system sizing and optimization. Photovoltaic system eciency, gP (%) Mast height, Hi (m) Turbine hub height, H (m) Power law index, z Cut-in wind speed, vc (m/s) Rated wind speed, vr (m/s) Cut-o wind speed, vf (m/s) Net charging eciency, gc (%) Net discharging eciency, gd (%) Depth of discharge, gD (%) Inverter eciency, gi (%) 10 10 20 0.14 3 8 15 90 90 60 90

Output power from 1kWp PV array (kW)

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Table 3 Economic parameters considered for system optimization. Discount rate (d%) Wind generator life (years) Photovoltaic system life (years) Battery bank life (years) Converter life (years) Cost of photovoltaic system ($/kWp) Cost of wind generator ($/kW) Cost of battery bank ($/kW h) Cost of converter ($/kW) Operation and maintenance cost as a % of total capital cost 10 20 20 5 10 3213 3427 86 386 1

10

15

20

25

time (hour of the day)

Fig. 5. Hourly average power available from a 1 kWp PV array at Sukhalai in Hoshangabad.

model the PV array. The power generated by the photovoltaic array (Ppv) is given as: P pv gp AP I T 14

output power from a 1kW wind turbine(kW)

0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

where gP is the photovoltaic system eciency, IT is the total radiation incident on the array (W/m2) at that time step, and AP is the total array area (m2). The power generated by the wind turbine (Pw) is obtained from the power curve of the wind turbine (Powell, 1981). 8 v2 v2 c > for vc < v < vr 2 < P r v2 r vc 15 Pw Pr for vr < v < vf > : 0 otherwise Pr is the rated electrical power (W), vc is cut-in wind speed (m/s), vf is cut-o wind speed, (m/s) and vr is the rated wind speed (m/s). The wind speed data is generally available at a

10

15

20

time(hour of the day)

Fig. 6. Hourly average power produced by 1 kW rated wind turbine at Sukhalai in Hoshangabad.

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

1131

system. The design space for the solar alone system is the portion above the sizing curve and is shown in Fig. 7. As the rating of photovoltaic array is increased, the minimum battery capacity required to meet the load decreases till a certain point, beyond which the battery capacity remains the same. This is because of the non-availability of solar insolation during night time and a minimum amount of storage is required for any amount of PV array to supply the load at night. The optimum conguration corresponds to a system with PV array of 10.2 kWp, battery capacity of 57.5 kW h, and the cost of energy is $0.38/kW h. It is seen from Fig. 7 that the optimum conguration corresponds to the one with the maximum amount of storage, with the minimum PV array rating. 2.3.2. Windbattery system The wind turbine rating is varied from 0 to 80 kW to determine the sizing curve and the design space for the system (Fig. 8). Optimum system conguration calls for a wind turbine capacity of 8 kW and battery capacity of 19.7 kW h. The minimum cost of energy is determined to be $0.24/kW h. Comparing the sizing curves (Figs. 7 and 8), it is seen that there is always a requirement of storage for any large solar array capacity. This is primarily due to non-availability of power during night. As wind is always available, for large values of wind turbine capacity, the storage capacity falls to zero and the system can be designed with no storage (Fig. 8). However, the optimum conguration based on lowest cost of energy contains a large amount of storage for both the cases. The amount of storage required for the minimum PV array is larger than the amount of storage required for the minimum wind turbine capacity and this is also due to the non-availability of insolation for the entire night. 2.3.3. Windsolar hybrid system The wind turbine rating is varied from 0 to 20 kW and PV array peak rating from 0 to 100 kWp to obtain the min58 56 54

25

20

Optimum configuration

Battery rating(kWp)

15

Design space
Feasible region

10

Sizing curve
Infeasible region
5

0 0

10

20

30

40

50

60

70

80

Wind trbine rating(kW)

Fig. 8. Design space for wind-battery system.

imum battery size required to meet the load. The sizingcurve, shown in Fig. 9, is a surface in three-dimensional space formed by the PV array rating, wind turbine rating and the minimum battery storage capacity required to meet the load. Every point above the sizing curve in the gure is a feasible design point and collection of all of them constitutes the design space. The sizing curve is also shown on the battery capacity vs. total generator rating diagram for various values of solar PV and wind turbine ratings in Fig. 10. For a typical conguration with solar and wind turbine rating of 5 kWp and 4 kW, respectively, the minimum battery capacity required is determined to be 36.6 kW h and the corresponding cost of energy is $0.31/kW h. From storage capacity vs. total generation graph for various xed values of wind turbine/PV array rating, it is seen that all optimum points lie towards the maximum storage side. As the cost of generators is comparatively more than storage cost and the conguration with largest battery storage is the optimum one. The overall optimum conguration corresponds to

Optimum configuration
Battery capacity(kWh)

60 40 20 0 20 15 10

Battery rating(kWp)

52 50 48 46 44 42 40 38 0 100 200 300 400 500 600 700 800 900 1000

Feasible region

Design space

Design space
Feasible region

Sizing curve

Sizing curve Infeasible region

100 80 60 5 40 0 20 0

Infeasible region
Wind rating(kW) Solar array rating (kWp)

PV array rating(kWp)

Fig. 7. Design space for solar-battery system.

Fig. 9. Three dimensional plot showing the sizing surface and the design space.

1132
60

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

3. Generation of design space and optimization with probabilistic approach

optimum configurations No wind turbine

50

Battery capacity(kWh)

for each sizing curve

40

Wind turbine=2 kW

30

Wind turbine=4 kW Wind turbine=6 kW

20

10

No PV array
0 0 20 40 60 80 100 120

total generation capacity(kW)

Fig. 10. Sizing curves for dierent options for isolated power generation.

the wind-battery system with no photovoltaic array. It is economical to go for a stand-alone wind-battery system than the hybrid system for this location. Variation of cost of generators and battery bank aect the optimum system conguration and the COE. Fig. 11 shows the type of hybrid system which is cost eective as the cost of solar PV array and wind turbine varies. It shows that for most small variations around the price considered, the optimum conguration is a wind alone system. Hence for these cases the optimum cost is not dependent on the cost of PV array. But as both cost of PV array decreases by a large amount and cost of wind turbine increases by a large amount, the system moves from wind alone to windsolar. Solar alone system is not cost eective, even if the solar cost decreases by 50% and wind turbine cost increases by 50% from their respective base cost.

The methodology used to arrive at the design space in deterministic approach is extended to incorporate the eects of uncertainty in the power available from the resources. The hourly values of power available from the power sources are considered as random variables. The system is sized for a specied reliability level using chance constrained programming approach. This approach is used by Arun et al. (2008) for sizing a photovoltaic battery system. It is assumed that the power available from the ith power source (Pi(t)) during each hour follows a normal distribution with mean lP i t and standard deviation rP i t . qij is the coecient of correlation between the power available from ith and jth power sources. The load is assumed to be deterministic over the time step. The source uncertainty is expressed as a probabilistic constraint. The system has to cater to the specied deterministic demand with probability greater than a specied value. The chance constrain relating the probability of the demand, D(t) being met by the system is: P Dt P Dactual t P a 17

where Dactual(t) is the deterministic demand to be met over the time step and a is the specied reliability of compliance of the constraint or the condence level. Combining (17) with the energy balance (7), the overall chance constraint can be written as: ! n m X X QB t Dt QB t P P j t P j t Dactual t f tDt f tDt j1m j1
fi t P du t P 0 P a 18

Separating stochastic and deterministic variables and modifying the above equation, we get: P
n X j1m

1.5

P j t

m X j1

P j tfi t 6

QB t Dt QB t f t D t f t D t 19

actual cost of PV array/cost considered

1.4 1.3 1.2 1.1

Dactual tfi t P du t 6 1 a

Wind only
1

Since the random variables (Pj(t)) are normally distributed, the sum of the stochastic variables in the above expression is a new random variable Pt(t) with mean lP j t and standard deviation rP j t such that lP t t
Wind-solar
n X j 1 n X n X j 1 k 1

0.9 0.8 0.7 0.6 0.5 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

lP j t qj

20

r2 P t t where

qjk rP j t rP k t qj tqk t

21

actual cost of wind turbine/cost considered

 qj t

Fig. 11. Type of hybrid system which is cost eective as the cost of PV array and wind turbine varies.

fi t 1

if j 6 m otherwise

22

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

1133

where qjk is the coecient of correlation between the power available from jth and kth sources. The deterministic equivalent of (19) can now be expressed as: Q B t D t Q B t Dactual tfit P du t f tDt f tDt lP t t rP t t Z a 23

where za is the inverse of the cumulative normal probability distribution corresponding to the required condence level a with zero mean and unity standard deviation. Expressing the deterministic equivalent in terms of the battery energy values for any time step, we get: QB t Dt QB t lP t t rP t t za Dactual tfi t P du tf tDt 24

It may be noted that the deterministic energy balance (7) and the probabilistic energy balance (24) are equivalent. As the random variable Pt(t) in the stochastic case is represented by its deterministic equivalent, the methodology developed to nd the design space for the deterministic case can be applied directly. Even though we have assumed that the power available from the sources follows normal distribution, the power sources cannot produce negative power. Eq. (24) is corrected when the deterministic equivalent of the power produced by any power source becomes negative. If the deterministic equivalent of power produced by any source is negative, it is considered that particular source is not producing any power. The mean and standard deviation of the power produced by such sources are zero. This can be incorporated by modifying only the expression of qj(t) in (22). 8 if lP j t rP j t za 6 0 > < 0; 25 qj t fi t; if j 6 m and lP j t rP j t za > 0 > : 1; if j > m and lP j t rP j t za > 0 3.1. Illustrative example Illustrative example, discussed in Section 2.3, is extended to incorporate the uncertainty associated with the power produced by dierent sources. For energy related applications, wind speed is generally described using gamma distribution function, or its special cases like Weibull or Rayleigh distribution functions, or using bivariate normal distribution function (Hennessey, 1977). The solar insolation generally forms a bimodal distribution function and can be approximated by the superposition of two distribution functions. Those two distribution functions can be from normal, beta or Weibull distributions (Khallat and Rahman, 1986). However, if the number of days for which the data taken is suciently large, using central limit theorem (Papoulis and Pillai, 2002), the power produced by the photovoltaic array and the wind turbine tends to follow a bivariate normal distribution with known mean and standard

deviation. The standard deviation of power produced by each generator is assumed to be a xed percentage of the mean value of it. The PV array is considered to be a relatively less variant power source with standard deviation equal to 45% of the mean of the power available. The wind is taken as a high variable source with standard deviation equal to 80% of the mean power available from wind. The reliability requirements for rural applications are considered to be not very high and the specied reliability of compliance is varied from 50% to 90%. The coecient of correlation between the power sources is varied in full possible range to study its eect. The wind turbine rating is varied from 0 to 100 kW and PV array peak rating from 0 to 100 kWp, respectively and the minimum battery size required to meet the load for each case is determined. Similar to Fig. 10, sizing curves for a reliability of 80% are shown in Fig. 12, where the power produced from the two sources are assumed to be independent. Optimum conguration (shown in Fig. 12) with 8 kWp of solar PV, 7 kW of wind turbine, and battery capacity of 44.29 kW h, compliance with a reliability of 80% and the corresponding COE is $0.47/kW h. It may be noted that the COE almost doubled to enhance the reliability of the system from 50% to 80%. It may also be noted that the optimum system is hybrid system, comprising of both the PV and wind generators unlike wind-battery system for deterministic case. Eect of reliability and coecient of correlation of sizing curve of the overall system are shown in Figs. 13 and 14, respectively. From the sizing curves shown in Fig. 13, the increase in storage capacity required with the increase in the system reliability requirement can be observed for a hybrid windsolar system with 6 kW wind turbine. The power available from the two sources is considered as independent. It can be noted from Fig. 14 that as the coecient of correlation between the power sources varies from 1 to +1, the system size increases. The variation of sizing curve for 80% reliable

60
No wind turbine

50
wind turbine=2.5kW

Battery capacity

40
Optimum configuration (8kWp,7kW,44.29kWh)

wind turbine=5kW wind turbine=7.5kW

30

wind turbine=10kW

20
No solar array

10

10

20

30

40

50

60

70

80

90

Total generation capacity(kW)

Fig. 12. Sizing curves for a reliability of 80%.

1134
60 55 50
alpha=0.90

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

1 0.8 0.6

Battery capacity (kWh)

coefficient of correlation

45 40 35
alpha=0.70 alpha=0.80

0.4 0.2 0
Wind solar wind alone solar alone

-0.2 -0.4 -0.6

30 25 20 15
alpha=0.60 alpha=.50,deterministic

-0.8 100 120 -1 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95

20

40

60

80

Total generation (kW)

alpha

Fig. 13. Variation of design space with the condence level (wind turbine rating of 6 kW and zero coecient of correlation).

Fig. 15. Type of hybrid system which is costeective for dierent reliability of compliance and coecient of correlation.

48 46
Optimum configuration for each sizing curve

correlated power sources. So hybridisation is more eective in such cases. Hybridisation is most eective for (a) Moderate values of reliability of compliance. (b) High negative values of coecient of correlation.

Battery capacity (kWh)

44
Coefficient of correlation increasing from-1 to 1 steps of 0.333, 6kW wind turbine, alpha=80%

42

40

38

36

34 0 20 40 60 80 100 120

Total generation capacity (kW)

Constant alpha curves are plotted between PV array rating and wind turbine rating in Fig. 16 with coecient of correlation equal to zero. It can be observed that for a wind-battery (y-axis) system, the increase in the wind turbine rating is more than three times when the required reliability changes from 50% to 80%, while for a solar-battery system (x-axis), the increase in PV rating is only about 1.5 times. The eect on the type of hybrid system with change in the cost of individual components is studied. The change
30
alpha=0.8 alpha=0.7

Fig. 14. Variation of sizing curves with coecient of correlation.

windsolar system with 6 kW wind turbine is shown in Fig. 14. The optimum conguration corresponding to each sizing curve is also shown. The type of hybrid system which is cost eective for different reliability of compliance and coecient of correlation is given in Fig. 15. As the reliability of compliance increases, a predominantly wind system shifts to a predominantly solar system. This can be attributed to the dierence in variability of the power available from the two sources. For low values of alpha (0.50.6), the optimal system is a wind-battery system (variability of power from wind turbine is more than that of solar PV) while for a high reliable system with condence level of 90%, its a solar alone system. Windsolar hybridisation is a cost eective solution for moderate values of alpha. The cost and size of power sources and battery reduces for highly negatively

25

alpha=0.6 alpha=0.5

Wind turbine rating (kW)

20

15

10

alpha=0.8 alpha=0.6
alpha=0.5

alpha=0.7

0 0 2 4 6 8 10 12 14 16 18

solar array rating (kWp)

Fig. 16. PV array capacity vs. wind capacity for various values of reliability of compliance.

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136

1.5 1.4

1135

Wind only
1.3 1.2 1.1 1 0.9

Wind-solar
0.8 0.7 0.6 0.5

PV only
0.6 0.8 1 1.2 1.4 1.6

actual cost of wind turbine/cost considered

Fig. 17. Type of hybrid system which is cost eective as the cost of PV array and wind turbine varies (alpha is 0.7 and zero coecient of correlation).

in cost of power sources aects the type of system which can produce power at the lowest possible cost and hence, the eect of variation of cost of wind turbine and PV array is studied. The type of hybrid system which is cost eective as the cost of PV array and wind turbine varies is shown in Fig. 17 with alpha of 0.7 and coecient of correlation of 0. For the base cost (considered in Table 3), the optimum conguration is a wind-battery system. The ratio of the varied cost of the wind turbine and PV array to the base cost is plotted in x- and y -axis of the gure respectively. It shows that the system varies from a wind only system (when solar cost is high and wind cost is low) to a solar alone system (when solar cost is low and wind cost is high). For small cost variations around the actual values of cost of power sources considered, its a hybrid system. 4. Conclusions Renewable energy based isolated hybrid systems are potential alternatives to grid extension for remote electrication. The random nature of the renewable resources and the non-linearities involved in the modeling of power sources makes the design and sizing of hybrid systems challenging. Deterministic or probabilistic approach is generally used to size the system and both have its own merits and demerits. The method proposed in this paper combines the advantages of both these approaches and takes uncertainty of resources into account. Design space approach for designing and optimizing stand alone windsolar hybrid power generating system has been demonstrated in this paper. A methodology is proposed to nd the minimum battery capacity required when the ratings of renewable energy sources are known. This methodology is used to derive the sizing curve and the design space. The methodology also helps to plan further load growth as it can

nd out the durations for which more load can be added without any change in system sizing. A case study from Sukhalai in Hoshangabad district of Madhya Pradesh, India to illustrate the proposed methodology for a windsolar hybrid system, is presented. Using deterministic procedure (as described in Section 2), the cost of energy from a PVbattery system is determined to be $0.38/kW h and the same for wind-battery system is $0.24/kW h. The wind power generating system is a cheap source of electric power compared to solar PV, but the variability of the power available from the wind is greater than that of the solar PV. Hence, if the reliability requirement is low, wind-battery system is cost eective. On the other hand, if the reliability requirement of the power produced is high, solar-battery system is cost eective. For moderate values of reliability requirement, the optimum choice is a windsolar hybrid system. Applying the chance constraint programming, the optimum conguration for 80% reliability consists of 8 kWp of solar PV, 7 kW of wind turbine, and battery capacity of 44.29 kW h, and the corresponding COE is $0.47/kW h. Among the various available renewable options for hybridization at a location, a proper choice of combination of power sources and their sizing depends upon various factors. From the case study performed in this paper, it can be concluded that the characteristics of the available resources at that location, the quality and reliability of the power supply required and the cost of various components decide the optimum conguration. The characteristics of the available resources can be captured by two parameters, the variability associated with the power available from each source and the coecient of correlation between the powers available from various sources. Negatively correlated power sources, as shown in Fig. 1, are good choices for hybridization. A highly variable power source with low cost can be a better choice compared to a low variable higher cost power source if the power supply reliability requirement is small and the reverse is true for high reliability requirement. For moderate values of reliability, hybridization of these sources will lead to cost eective solution. References
Abousdar, I., Ramkumar, R., 1990. Loss of power supply probability of stand alone wind electric conversion systems: a closed form approach. IEEE Trans. Energy Convers. 7 (3), 445452. Abousdar, I., Ramkumar, R., 1991. Loss of power supply probability of stand alone photovoltaic systems: a closed form approach. IEEE Trans. Energy Convers. 6 (1), 445452. Ai, B., Yang, H., Shen, H., Liao, X., 2003. Computer-aided design of PV/ wind hybrid system. Renew. Energy 28 (10), 14911512. Arun, P., Banerjee, R., Bandyopadhyay, S., 2007. Sizing curve for design of isolated power systems. Energy Sustain. Dev. 11 (4), 2128. Arun, P., Banerjee, R., Bandyopadhyay, S., 2008. Optimum sizing of battery integrated diesel generator for remote electrication through design-space approach. Energy 33, 11551168. Arun, P., Banerjee, R., Bandyopadhyay, S., 2009. Optimum sizing of photovoltaic battery systems incorporating uncertainty through design space approach. Sol. Energy 83 (7), 10131025.

actual cost of PV array/cost considered

1136

E.S. Sreeraj et al. / Solar Energy 84 (2010) 11241136 Kolhe, M., Kolhe, S., Joshi, J.C., 2002. Economic viability of stand-alone solar photovoltaic system in comparison with diesel-powered system for India. Energy Econ. 24 (2), 155165. Kulkarni, G.N., Kedare, S.B., Bandyopadhyay, S., 2007. Determination of design space and optimization of solar water heating systems. Sol. Energy 81 (8), 958968. Kulkarni, G.N., Kedare, S.B., Bandyopadhyay, S., 2008. Design of solar thermal systems utilizing pressurized hot water storage for industrial applications. Sol. Energy 82, 686699. Kulkarni, G.N., Kedare, S.B., Bandyopadhyay, S., 2009. Optimization of solar water heating systems through water replenishment. Energy Convers. Manage. 50 (3), 837846. Li, P., Arellano-Garcia, H., Wozny, G., 2008. Chance constrained programming approach to process optimization under uncertainty. Comput. Chem. Eng. 32 (12), 2545. Linnho, B., 1993. Pinch analysis: a state-of-the-art overview. Chem. Eng. Res. Des. 71, 503522. Linnho, B., Townsend, D.W., Boland, D., Hewitt, G.F., Thomas, B.E.A., Guy, A.R., Marsland, R.H., 1982. User Guide on Process Integration for the Ecient Use of Energy. The Institution of Chemical Engineers, Rugby, UK. Mani, A., Rangarajan, S., 1982. Solar Radiation Over India. Allied Pub., New Delhi. Natarajan, E., Rajendra Prasada, A., 2006. Optimization of integrated photovoltaicwind power generation systems with battery storage. Energy 31 (12), 19431954. Papoulis, A., Pillai, S.U., 2002. Probability, Random Variables and Stochastic Processes, fourth ed. Tata McGraw-Hill. Pillai, H.K., Bandyopadhyay, S., 2007. A rigorous targeting algorithm for resource allocation networks. Chem. Eng. Sci. 62, 62126221. Powell, W.R., 1981. An analytical expression for the average power output of a wind machine. Sol. Energy 26 (1), 7780. Protogeropoulos, C., Brinkworth, B.J., Marshall, R.H., 1997. Sizing and techno-economical optimization for hybrid solar photovoltaic/ wind power systems with battery storage. Int. J. Energy Res. 21, 465479. Rao, S.S., 1980. Structural optimisation by chance constrained programming techniques. Comput. Struct. 12 (6), 777781. Roy, A., Arun, P., Bandyopadhyay, S., 2007. Design and optimization of renewable energy based isolated power systems. SESI J. 17, 5469. Roy, A., Kedare, S.B., Bandyopadhyay, S., 2009. Application of design space methodology for optimum sizing of wind-battery systems. Appl. Energy 86 (12), 26902703. Shenoy, U.V., 1995. Heat Exchanger Network Synthesis: Processes Optimization by Energy and Resource Analysis. Gulf Publishing Company, Houston. Shenoy, U.V., Sinha, A., Bandyopadhyay, S., 1998. Multiple utilities targeting for heat exchanger networks. Chem. Eng. Res. Des. A 76, 259272. Singhvi, A., Shenoy, U.V., 2002. Aggregate planning in supply chains by pinch analysis. Chem. Eng. Res. Des. 80 (6), 597605. Singhvi, A., Madhavan, K.P., Shenoy, U.V., 2004. Pinch analysis for aggregate production planning in supply chains. Comput. Chem. Eng. 28 (67), 993999. Smith, R., 1995. Chemical Process Design. McGraw-Hill, New York. Wang, Y.P., Smith, R., 1994. Wastewater minimization. Chem. Eng. Sci. 49 (7), 9811006. Yang, H.X., Lu, L., Burnett, J., 2003. Weather data and probability analysis of hybrid photovoltaicwind power generation systems in Hong Kong. Renew. Energy 28 (11), 18131824.

Azaiez, M.N., Hariga, M., Al-Harkan, I., 2005. A chance-constrained multi-period model for a special multi-reservoir system. Comput. Oper. Res. 32, 13371351. Bagul, A.D., Salameh, Z.M., Borowy, B., 1996. Sizing of stand-alone hybrid windphotovoltaic system using a three-event probability density approximation. Sol. Energy 56 (4), 323335. Bandyopadhyay, S., 2002. Eect of feed on optimal thermodynamic performance of a distillation column. Chem. Eng. J. 88 (13), 175186. Bandyopadhyay, S., Malik, R.K., Shenoy, U.V., 2003. Feed preconditioning targets for distillation through invariant rectifyingstripping curves. Ind. Eng. Chem. Res. 42 (26), 68516861. Bandyopadhyay, S., Mishra, M., Shenoy, U.V., 2004. Energy-based targets for multiple-feed distillation columns. AIChE J. 50 (8), 18371853. Bandyopadhyay, S., Ghanekar, M.D., Pillai, H.K., 2006. Process water management. Ind. Eng. Chem. Res. 45, 52875297. Beyer, H.G., Langer, C., 1996. A method for the identication of congurations of PV/wind hybrid systems for the reliable supply of small loads. Sol. Energy 57 (5), 381391. Bucciarelli Jr., L.L., 1984. Estimating loss-of-power probabilities of standalone photovoltaic solar energy systems. Sol. Energy 32 (2), 205 209. Celik, A.N., 2002a. Optimization and techno-economic analysis of autonomous photovoltaic wind hybrid energy systems in comparison to single photovoltaic and wind systems. Energy Convers. Manage. 43 (18), 24532468. Celik, A.N., 2002b. The system performance of autonomous photovoltaicwind hybrid energy systems using synthetically generated weather data. Renew. Energy 27 (1), 107121. Changchit, C., Terrell, M.P., 1993. A multiobjective reservoir operation model with stochastic inows. Comput. Ind. Eng. 24 (2), 303313. Charnes, A., Cooper, W.W., 1959. Chance-constrained programming. Manage. Sci. 6, 7379. Chedid, R., Rahman, S., 1997. Unit sizing and control of hybrid wind solar power systems. IEEE Trans. Energy Convers. 12 (1), 7985. Deshmukh, M., Deshmukh, S., 2008. Modeling of hybrid renewable energy systems. Renew. Sustain. Energy Rev. 12 (1), 235249. El-Halwagi, M.M., Manousiouthakis, V., 1989. Synthesis of mass exchange networks. AIChE J. 8, 12331244. Gavanidou, E.S., Bakirtzis, A.G., 1992. Design of stand-alone system with renewable energy sources using trade o methods. IEEE Trans. Energy Convers. 7 (1), 4248. Government of India, 2001. Tables on Houses, Household Amenities and Assets (H Series Tables). Census of India, New Delhi. Government of India, 2003. The Electricity Act, 2003, The Gazette of India, Extraordinary, 2003, Part II Section 3 Sub-section (ii). Ministry of Power, New Delhi. Government of India, 2005. National Electricity Policy, The Gazette of India, Extraordinary, Part I Section 1. Ministry of Power, New Delhi. <http://www.powermin.nic.in/whats_new/ national_electricity_policy.htm>. Hennessey Jr., J., 1977. Some aspects of wind power statistics. J. Appl. Meteorol. 16 (2), 119128. International Energy Agency, 2006. World Energy Outlook, second ed. IEA Publications. Justus, C.G., 1978. Wind energy statistics for large arrays of wind turbines (New England and central US regions). Sol. Energy 20, 379386. Karaki, S.H., Chedid, R.B., Ramadan, R., 1999. Probabilistic performance assessment of autonomous solarwind energy conversion systems. IEEE Trans. Energy Convers. 14 (3), 766772. Khallat, M.A., Rahman, S., 1986. A probabilistic approach to photovoltaic generator performance prediction. IEEE Trans. Energy Convers. 1 (3), 3440.