Renewable Energy 35 (2010) 2739e2751

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Performance, reliability and failure analysis of wind farm in a developing Country

G.M. Joselin Herbert a, S. Iniyan b, *, Ranko Goic c
a

Department of Mechanical Engineering, Noorul Islam University, Kumaracoil, Thackalay, K.K. Dist. India Institute for Energy Studies, Department of Mechanical Engineering, Anna University, Sardar Patel Road, Chennai 600025, Tamil Nadu, India c Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, University of Split, Croatia
b

a r t i c l e i n f o
Article history: Received 13 October 2008 Accepted 20 April 2010 Available online 8 June 2010 Keywords: Wind turbine generators Performance Reliability Weibull method

a b s t r a c t
In this paper, an analysis of the performance, failure and reliability, as well as a spare parts analysis have been conducted for a wind farm, which has 15 wind turbine generators (WTGs), each of 225 kW capacity. This wind farm is located at Muppandal, Tamil Nadu, South India. The average value of performance parameters such as technical availability, real availability and capacity factor for the wind farm were 94%, 82.88% and 24.9% respectively during the years 2000e2004. This paper also deals with Pareto analysis to nd out the reduction in problems, when one problem is tackled partly and completely. The Weibull technique was also used for the reliability analysis. The reliability factor in the initial period after one year seems to be good as the wind farm has a lower failure rate of 0.000019. As a supplemental activity, spare parts optimization was also carried out for a few vital components of this wind farm and the results are presented. The failure and its nancial implications are also analyzed in this paper. 2010 Elsevier Ltd. All rights reserved.

1. Introduction Wind energy is one of the vital inputs for the social and economic development of any nation. It supplies affordable, inexhaustible energy to the economy. It is an alternative clean energy source and has been the worlds fastest growing renewable energy source with a growth rate of 28% in the last decade. Technological improvements over the last 5 years have placed wind energy in a stable position to compete with conventional power generation technologies. Due to the impending exhaustion of fossil fuels, it is crucial to develop clean wind energy as an alternative source of energy. Wind energy is eco-friendly and does not pollute the atmosphere like thermal power plants. It is to be noted that a 225 kW wind turbine installed in a moderately high wind area, generates about 600 MWh per year. The same amount of energy, when generated in a thermal power plant, consumes about 250 tonnes of coal and emits 800 tonnes of CO2 and other poisonous gases into the atmosphere. Carbon dioxide build up in the atmosphere is the single biggest contributor to global warming. The total CO2 emission from all the coal-red power plants in India is 1.1 million tonnes per day with an annual emission computed to be 395 million metric tons. Coal red plants cause diverse public health and environmental

* Corresponding author. Tel.: 91 9444007771. E-mail addresses: joselindev@yahoo.co.in (G.M.J. Herbert), iniyan777@hotmail. com (S. Iniyan), rgoic@fesb.hr (R. Goic). 0960-1481/$ e see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.renene.2010.04.023

problems. To address the problems of global warming and environmental degradation, steps need to be taken to slash the amount of CO2 that various power plants emit. The replacement of traditional fossil fuel-based energy with wind energy offers great opportunities for the reduction of CO2 emission. Enormous power can be generated by the selection of suitable wind farm sites with suitable WTGs, improved maintenance procedure, better policies from the Government and advanced technologies of wind turbines. A literature survey was conducted to review the performance and failures of WTGs in different parts of the world. Mejia et al. [1] analyzed the performance of a wind turbine generator by studying its capacity factor. Ben Amar et al. [2] calculated the monthly and annual values of capacity factor and energies of the rst wind farm section of Sidi Daoud, Tunisia. Zekai Sen [3] analyzed statistical behaviour of wind energy within the northeastern part of Turkey by employing the Chebyschevs inequality to approximate the probability distribution function of the wind energy. It provided a basis for assessing the risk and reliability of wind energy. Chi-ming Lai and Ta-hui Lin [4] analyzed wind power potential, described the practical installation, measured the actual energy output, veried the reliability of the energy output for a land aqua-farm in Taiwan and introduced the relationship between the actual energy generated and the wind speed characteristics. Dokopoulos et al. [5] proposed a Monte Carlo based method for predicting the performance of energy systems consisting of diesel generators and WTGs to eliminate problems associated with wind intermittency and to provide a source of electrical energy. Paul Denholm [6] proposed the technical,

2740

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

environmental and social performance of wind energy systems using biomass-based energy storage. Sorensen et al. [7] studied wind farms connected to weak grids in India and found that variations in grid voltage and frequency have a considerable impact on the operation of wind turbines. Jang et al. [8] described a simulation model and a case study for analyzing the probability of power supply failure in hybrid photovoltaic-wind power generation systems incorporating a storage battery bank and also analyzed the reliability of the system on the islands surrounding Hong Kong. Danmi et al. [9] suggested the stall-delay phenomenon for horizontal axis wind turbine in order to accurately predict the loading and performance of WTGs operating in stall. Thomsen et al. [10] found that the difference in energy production could be more than 50% between MW sized wind turbines installed at different sites. Peter Fuglsang and Kenneth Thomsen [11] had compared a 1.5 MW stall regulated wind turbine in normal onshore at terrain and in offshore wind farm and showed a potential increase in energy production of 28%. Kathleen ODell [12] investigated new materials and new manufacturing processes to increase the performance of WTGs. Idriss Ammara et al. [13] had found that the inefcient spacing between the turbines reduced the performance associated with the wake effect. Energy is lost due to wake interactions and in electrical losses. Ssu-yuan Hu and Jung-ho Cheng [14] evaluated performance of pairing between sites and wind generators and estimated energy output performance. Bhatt and Jothibasu [15] studied the prediction and enhancement of the performance of wind farm in India and found that there was scope for improving the grid and WTG availability. This paper is an attempt to study the performance, reliability and failure analysis of wind farm located at Muppandal (South India).

2. Wind energy programme in India India is endowed with an abundance of natural wind energy resources, which can supply the energy needed even in remote areas. Considerable efforts have been made for the sustainable development of wind energy in India. The wind power programme in India is promoted and coordinated by the Ministry of New and Renewable Energy (MNRE), Government of India. The MNRE has issued guidelines to all the States on the general policies and the facilities to be offered for wheeling, banking and purchase of power from such projects. The Indian Renewable Energy Development agency (IREDA) plays a signicant role in promoting wind energy projects in India. Wind power generation has grown at an alarming rate in the past decade because of the advancement in power electronic technology. Wind power installations worldwide have reached 159,213 MW producing more than 340 TWh of electricity annually. The wind power programme in India was initiated towards the end of the 6th plan in 1983e1984. The Government of India established four demonstration wind farms in the year 1985 at Tuticorin (Tamil Nadu), Okha (Gujarat), Puri (Orissa) and Deogarh (Maharashtra). A detailed study of the establishment of wind farms on a scientic basis was started in the year 1986. The major demonstration wind farms in India were established by MNRE with the assistance of the Danish International Development Agency in the year 1989e1990. The installation of private wind farms was started in 1991 and more than 95% investment is made by the private sector. India has earned recognition as a new Wind Super Power as per the 1998 World Watch Institutes Report. The growth of the wind energy scenario accelerated from 1999 because of technological maturity, the introduction of machines suitable for Indian conditions and the Governments declaration of incentives to private entrepreneurs,

concessional import duty on specied wind turbine parts, a veyear income tax holiday and sales tax and excise tax relief. Indias preliminary gross estimated potential stands at 45,195 MW, but recent scientic surveys re-assessed it as 60,000 MW. The State-wise capacity addition during the rst half of the year 2007e2008 was 566 MW. In the last 10 years, wind power development in India has been promoted through Research and Development (R&D) support, policy support and development of infrastructure capability. The scal year 2006e2007 was a landmark year for the wind sector in India with a capacity addition of 1771.5 MW and this was higher than the last years installed capacity of 1746.3 MW. The installed capacity of wind power in India has reached 10,925 MW till December 2009 with more than 10,000 numbers of WTGs of different capacities. About 33.5 billion kWh of electricity have been fed to the grid so far. The percentage contribution of wind energy installed capacity is 5.82% of the all-India total installed capacity (134717 MW) of power. Since 1980, advances in aerodynamics, structural dynamics, and micrometeorology have contributed to a 5% annual increase in the energy yield of the turbines. Current research techniques have led to the production of stronger, lighter and more efcient blades for the turbines. These costs are expected to decline as protability increases with technology improvements. Wind energy has been proven as a reliable and cost effective energy source and it is expected to create maximum impact in the production of electricity. The rapid increase of wind capacity in India is likely to continue, and is expected to reach a cumulative installed capacity of 18,028 MW by 2011. A diagram of WTG is illustrated in Fig. 1 to identify the technology. A modern WTG is composed of ve basic subsystems, 1) The rotor consists of blades mounted on a hub and includes aerodynamic braking systems and pitch control. 2) The drive train including gearbox, hydraulic systems, shafts, braking systems and nacelle, encases the actual turbine. 3) The yaw system positions the rotor perpendicular to the wind stream. 4) Electrical and Electronic systems include the generator, relays, circuit breakers, wiring, controls and electronic sensors and 5) The tower. 3. Performance analysis The performance of the wind farm was analyzed to rectify the failures and improve the reliability of the system. The performance of the wind turbine is proportional to Wind Power Density (WPD). The pattern of monthly variation in WPD of the Muppandal wind farm is shown in Fig. 2. This curve indicates the maximum WPD of 723.3 W/m2 in the month of June. The annual average WPD in this wind farm is 361 W/m2. High WPD was recorded from April to September. Low WPD was recorded during the monsoon (OctobereDecember). The wind power density varies with different locations. The maximum extrapolated mean annual WPD of a site (Perampukettimedu, Kerala) at a 50 m height amongst the MNRE sites is 721 W/m2. The number of wind monitoring stations in India with an annual WPD of more than 200 W/m2 at a 50 m height is 212. The monthly generation curve for one WTG of 225 kW for the period from April 2002 to March 2003 is shown in Fig. 3. It is found that the energy generation varies due to wind speed and technical factors. From April to July, it shows an upward trend and afterwards it shows a declining trend. The upward and downward trends are caused by a monthly variation in wind power density. The power generation curve is correlated with the wind power density curve. The WTGs should be properly maintained for sustained operation and generation of optimum power. To ensure optimum power generation, an undisturbed ow of air is required by the wind turbine.

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

2741

Fig. 1. Diagram of Wind Turbine Generator.

The most important performance criterion is the power curve. The power curve is the ofcial certicate of performance of a WTG. On the basis of the power curve the annual energy output can be calculated for a given site. The determination of the power curve is connected with both the technical properties of the turbine and the wind data on which the WTG design is based. The rst factor determining the performance of the WTG is the aerodynamic characteristic of the rotor. The power curve for a stall regulated 225 kW capacity WTG is shown in Fig. 4. The International Electro technical Commission standard, IEC 61400-12, species the procedure for measuring the power performance characteristics of WTG systems. These performance characteristics are dened by the measured power curve and the estimated annual energy production. The technical specications of this WTG are 3 blades, 29.8 m rotor diameter, 45 m hub height, stall regulation, tubular tower, asynchronous dual speed generator coupled with gear, rated power of 225 kW and rated wind speed of 15 m/s. The cut-in speed is 4 m/s

and the cut-out speed is 25 m/s. The turbine is designed to operate between the cut-in speed and cut-out speed. The performance data of a wind farm was collected from the Muppandal wind farm, in South India. This wind farm has a cluster of 15 identical wind turbine generators. The data pertaining to power generation, machine failure, grid failure, wind speed and maintenance were collected for 5 years from 2000 to 2004. The total installed capacity of this wind farm is 3375 kW i.e., (15 225 kW). The annual wind power generation is shown in Fig. 5. It is found that there is a uctuation of generation from year to year due to climatic and technical factors. The generation reached a maximum (7480 MWh) in the year 2002 because of higher turbine availability. The annual average generation was 7370 MWh. There was minimum generation (7215 MWh) in the year 2001 because of increased break down and maintenance time. The accumulated dirt, grime, and insect deposits on the blade slightly impaired and reduced the power generation of WTGs. The

800

700

Wind power density (W/m 2)

600

500

400

300

200

100

0 April May June July Aug. Sept. Oct. Nov. Dec. Jan. Feb. March

Fig. 2. Pattern of monthly variation in Wind Power Density. Source: Wind energy resource survey for India-I by Anna Mani.

2742
100000 90000 80000

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

Electricity generation (kWh)

70000 60000 50000 40000 30000 20000 10000 0 April May June July Aug. Sept. Oct. Nov. Dec. Jan. Feb. March

Fig. 3. Monthly Generation of 225 kW WTG for the year 2002e2003. Source: Log book maintained at Muppandal wind farm in Tamil nadu.

blades should be periodically cleaned with a silicon-based solution to decrease bug accumulation. The capacity factor, real availability and technical availability are computed and their variations in the ve years of operation are shown in Fig. 6. The technical availability was almost steady throughout the period and its average value was 94%. It is possible to increase the technical availability by reducing technical failures of the wind farm. It is observed that the average real availability was 82.88%, because the turbine does not operate total time due to grid failure time, turbine failure time and low wind speed (less than 4 m/s) time. The wind farm had very low real availability in the year 2003 due to an increase in down time because of grid failure and low wind speed. By reducing the failures due to grid and components, the real availability can be increased and thereby the reliability factor can also be improved. The capacity factor was found to vary from 24.41% in 2001e25.3% in 2002. The average capacity factor for ve years was 24.9%. There is scope for improvement in the performance of the WTG by improving the grid availability and turbine availability. The availability of the WTGs can be increased by proper maintenance procedure.

The down time is used to identify the current maintenance condition. Table 1 shows the annual comparison of total time, generation time and stoppage time for the wind farm. The summation of total generation time and stoppage time gives the total time. The total time, generation time and stoppage time for the ve years of operation were 657,720 h, 517,394 h, and 140,326 h respectively. The stoppage time consists of stoppage due to low wind, grid failure, control panel fault, electrical and mechanical failures, and maintenance works. Performance data of 3.735 MW wind farm is given in Table 2. It is observed that the down time due to low wind (100,887 h) was higher than the down time due to grid failure. Turbine fault time and maintenance time were less than the grid failure time. The low wind time (23,408 h) was more in the year 2003 due to climatic variation in the atmosphere. The average stoppage time due to low wind, grid failure and mechanical failure were 20,177, 5813 and 1011 h respectively. Overall, 21% of total time, wind farm was stopped. The average stoppage time due to preventive maintenance, electrical failure and control panel failures were 522, 470 and 72 h respectively. The percentage distribution of total stoppage

250

200

Rotor Power (kW)

150

100

50

0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Wind Speed (m/s)

Fig. 4. Power curve for 225 kW WTG. Source: Indian Wind power Directory e 2004.

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

7500 7450

2743

Electricita generation (MWh)

7400 7350 7300 7250 7200 7150 7100 7050 2000 2001 2002 2003 2004

Year
Fig. 5. Annual wind power generation.

is shown in Fig. 7. The stoppage time due to mechanical failure is high in 2001 (8%) as well as in 2004 (7%). It has been noted that the average time taken for replacement of electronic components is much less than the time taken for replacement of mechanical components. The wind farm was stopped for 4%, 2% and 21% of total stoppage time due to mechanical failure, electrical failure and grid failure respectively during the ve years of operation. Grid drop, frequency fault, asymmetric current, overvoltage and low voltage are major causes of grid failure. Wind power generation can be increased through better operation and maintenance activities.

failure rate. In this section, a Weibull distribution technique is presented to study the reliability of the wind farm. 4.1. Estimation of properties of Weibull distribution The reliability estimation methodology reects the actual eld failure rate and defect density. It can be a good indicator of eld reliability. For the reliability analysis of WTGs, the Weibull model is chosen. The Weibull analysis is a powerful technique for identifying the reliability characteristics of WTGs. The estimation of the parameters of the Weibull distribution can be done graphically via probability plotting or analytically, using least square or maximum likelihood methods. In this study an analytical method is used to determine the Weibull parameters. The Weibull methodology is a general purpose reliability-engineering model, which depending on the shape parameters, assumes exponential, normal as well as other types of distribution. The properties of the Weibull distribution are as follows. The two-parameter Weibull cumulative density function is given as below:

4. Reliability analysis The Wind Turbine Generator (WTG) has a different impact on the reliable performance of a generating system. Concern for the reliability of the wind turbine with its complexity has grown, and this requires better maintenance skills. Reliability estimation is highly desirable for such equipment and is directly related to their

100

80

Percentage (%)

60

40

20

0 2000 2001 2002 Year Technical Availability Real Availability Capacity Factor 2003 2004

Fig. 6. Comparison of Technical availability, Real availability and Capacity factor of wind farm.

2744

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

FT 1 eT=hb

(1)

where F(T) e probability of failure; T e Time; b e shape parameter; h e scale parameter Reliability is dened as the probability in which an item or an entity performs its intended function over a period of time under stated conditions. The reliability function for the two-parameter Weibull distribution is given as

B (X) life is dened as the estimated time when the probability of failure will reach a specied percentage point (X %). To be specic, in the bearing industry, it is the normal practice to check for the life deterioration of 10% of bearings and a check is kept on the time at which this occurs. This duration of the failure of 10% of bearings is normally termed as B(10) life of the product. B(10) life is equivalent to a warranty time for 90% reliability, i.e., 10% of the products are expected to fail within a specied time. 4.2. Failure rate and reliability factor It is an essential tool in discovering the appropriate probabilistic model for a set of data. By appropriately transforming the failure time values, the exact quality of tness tests can be constructed for the model. The plot of the data resembles a straight line which can identify the quality of tness of Weibull distribution. The data collected from the wind farm are fed into the Weibull plot and presented in Fig. 8. In the case of this wind farm the software package Weibull 6th version of Reliasofts was used for the analysis. The failure time data were fed into the package and the Weibull properties such as failure rate, reliability, mean, median, mode, warranty and B(X) life information were obtained, and are given in Fig 9aee. The initial failure rate after one year (8760 h) was 0.000019 while the failure rate was 0.000086 after ve years (43,800 h). The reliability factor was observed as 0.9173 after one year (8760 h) and 0.1419 after ve years (43,800 h). The reliability factor in the initial period after one year seems to be good, as the wind farm has a lower failure rate of 0.000019. But the reliability factor becomes very low after ve years because of the frequent occurrence of blade failure, followed by the occurrence of gearbox failure. By tackling the blade and gearbox failure the reliability of the wind farm could be improved. The mean life, median life and mode of failure intensity of the wind farm were 27,503, 25,668 and 21,325 h respectively. The required warranty time for a reliability factor of 0.7, 0.8 and 0.9 would be 18,218, 14,302 and 9710 h respectively. The warranty time can be given for more than one year for a reliability factor of 0.9. Preventive maintenance is to be carried out by trained employees to prevent failures and increase the life of the wind turbine generators. By reducing the failure rate, the warranty time can be increased further; thereby, customer satisfaction and continuous improvement can be obtained. The B(X) life information gives the time in hours for different percentages of failure. In this wind farm, the percentage of failure varies from 1% to 30%. For example, 10% of failures occur at 9710 h. 5. Failure analysis Failure is an event in which an item does not perform its required functions within the specied limits under specied conditions. Failure analysis is an attempt to determine the causes of failure. The number of failures over the years is shown in Fig. 10. In the year 2000, the frequency of failure is 6 which is low for a ve year period. The failures in the year 2003 were 3.83 times more than in the year 2000. The higher frequency of failures in the year 2004 was due to blade failure. The number of failures in each year is not constant because of variation in external load conditions. These failures reduced the overall generation and caused excessive stoppage time. Failures due to the blades, gearbox, hydraulic unit, yaw unit and brake pad are considered as mechanical component failures and those of the control panel, capacitor panel and generator failures are considered as electrical and electronic component failures. Failures due to mechanical components are 79% and failures due to electrical and electronic components are only 21%. It is inferred that

RT eT=h

(2)

The mean life or mean time of failure (MTTF or MTBF) is dened as the average time of failure-free operation up to a failure event calculated from a homogeneous lot of equipments under operation. The MTTF or MTBF of the Weibull pdf is given as

  1 T h G 1

(3)

where

 1

is called gamma function The two-parameter Weibull probability density function f(t) is given as.

f t

b T h h

 b1

eT=hb

(4)

The Weibull failure rate function is dened as the number of failures per unit time that can be expected to occur for the product. It is given as.

lT

  b T b1 f T h h RT

(5)

The median is dened as the failure density function equivalent to 50% probability. Mode is dened as maximum failure intensity of the probability density function. For Weibull the median and mode are given as
b Median h ln 21=

(6)

  1 1=b Mode h 1

(7)

Warranty time is dened as the estimated time when the reliability will be equal to a specied goal. The reliable life T(R) of a unit for a specied reliability is given by
b TR hf ln R TR g1=

(8)

This is the life time for which the unit will function successfully with a reliability of R (TR).

Table 1 Annual comparison of total time, generation time and stoppage time. Year 2000 2001 2002 2003 2004 Total (h) Total time (hrs.) 131,760* 131,400 131,400 131,400 131,760* 657,720 Generation time (hrs.) 103,943 101,625 106,558 100,694 104,574 517,394 (79%) (77%) (81%) (77%) (79%) (79%) Total stoppage time (h) 27,817 29,775 24,842 30,706 27,186 140,326 (21%) (23%) (19%) (23%) (21%) (21%)

8784 15 131,760* (leap year); 8760 15 131,400.

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751 Table 2 Performance data of 3.735 MW wind farm. Year Generation (kWh) Generation Time (hrs.) Grid Failure Time (hrs.) Low Wind Time (hrs.) Control panel failure Time (hrs.) Electrical Failure Time (hrs.) Mechanical Failure Time (hrs.) Preventive Maint. Time (hrs.) 2000 7,246,119 (19.7%) 103,943 (20.1%) 5924 (20.4%) 20,188 (20.0%) 39 (10.9%) 736 (31.3%) 381 (7.5%) 547 (21.0%) 2001 7,215,905 (19.6%) 101,625 (19.6%) 5963 (20.5%) 20,126 (19.9%) 91 (25.4%) 634 (27.0%) 2280 (45.1%) 679 (26.0%) 2002 7,480,740 (20.3%) 106,558 (20.6%) 4796 (16.5%) 19,179 (19.0%) 63 (17.6%) 233 (9.9%) 216 (4.3%) 354 (13.6%) 2003 7,430,162 (20.2%) 100,694 (19.5%) 6165 (21.2%) 23,409 (23.2%) 107 (29.9%) 268 (11.4%) 352 (7.0%) 404 (15.5%) 2004 7,480,229 (20.3%) 104,574 (20.2%) 6216 (21.4%) 17,986 (17.8%) 59 (16.5%) 477 (20.3%) 1826 (36.1%) 621 (23.8%) Total

2745

36,853,155 517,394 29,065 100,887 358 2349 5056 2607

Mechanical 1% Electrical 3%

Preventive Maint. 2%

Mechanical 8% Preventive Maint. 2%

Electrical 2%

Control Panel 0%

Grid 21%
Control Panel 0% Low Wind 68%

Grid 20%

Low Wind 73%

2001

2000

Electrical 1% Control Panel 0%

Mechanical 1%

Preventive Maint. 1%

Electrical 1%

Mechanical 1%

Preventive Maint. 1%

Grid 19%

Control Panel 0%

Grid 20%

Low Wind 78%

Low Wind 77%

2003

2002
Mechanical 7% Electrical 2% Preventive Maint. 2%

Control Panel 0%

Grid 23%

Low Wind 66%

2004

Fig. 7. Percentage distribution of stoppage time in wind farm.

2746
99.90 90.00 50.00
Unreliability, F(t)

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

10.00 5.00 1.00 0.50 0.10 1000

10000 Time, (t)

100000

Fig. 8. Weibull plot of wind farm data.

a wind farm experiences more failures due to mechanical components than electrical and electronic components. Reliable components and effective preventive maintenance procedure can reduce failures.

The percentage distribution of failures of mechanical and electrical and electronic components from the year 2000 to 2004 is shown in Fig. 11. The inexibility of the blade rope and improper crimping, rubbing of the blade rope against its support and lack of lubrication of the rope resulting in its erosion, are the main causes of blade rope failure. Proper heat treatment of the blade rope will strengthen it. The force at a right angle to the plane of rotation, attempts to bend the blade back against the tower. The clearance between the rotor and tower depends on the distance between the blade tip and the tower at no load condition. Due to wind force, creep, shrinkage and temperature deformation cause the blade to bend backward over time. Tip deection is caused by the interaction of dynamic loading and structural response. Care should be taken during designing, to avoid gearbox failure. The dynamic loading on gear teeth due to frequent interruptions of grid supply should be reduced. The high temperature developed in the gearbox, can be avoided by using good lubricants, and by adequately cooling the gearbox oil. Adequate stock of high quality gear oil should always be maintained for periodical replacement of the old oil. The nature of blade failures was blade tip failure, blade rope failure and blade cylinder failure. Inexibility of blade rope and bending of

a
Failure rate (t)

0,0001 0,00009 0,00008 0,00007 0,00006 0,00005 0,00004 0,00003 0,00002 0,00001 0 8760 17520 26280 35040 43800 0,000019 0,000054 0,000037 0,00007 0,000086

1 0,9 0,8 0,7 0,6 0,9173 0,7184

R(t)

0,5 0,4 0,3 0,2 0,1 0 8760 17520

0,4841 0,2817 0,1419

26280

35040

43800

Time (hrs.)
Failure Rate

Time (hrs.)
Reliability

c
Time (hrs.)

30000 25000 20000

27503

25668 21325

d 20000
15000

18218 14302 9710

15000 10000 5000 0 Mean Median Mode

Time (hrs.)

10000

5000

0,7

0,8 Reliability, R(t)

0,9

Mean Median and Mode

Required warranty time

e
Time (hrs.)

20000 18218 15000 10000 6697 5000 2888 0 1 5 10 Percentage 20 30 9710 14302

BX information
Fig. 9. Weibull properties of wind farm.

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

2747

25
23 21

20

No. of failures

15

14 12

10
6

0
2000 2001 2002 2003 2004

Year
Fig. 10. Number of failures over the years.

blade at high speed thus hit the tower are the causes of blade failure. It is suggested that manufacturers should take necessary steps to improve the design. Sufcient cooling and constant air gap between the stator and rotor and proper alignment of the driver and driven shaft, and grid supply without tripping will reduce generator failure. By providing sufcient cooling and protection to electronic components, the life of the components can be increased. The yaw unit of wind turbines is subjected to dynamic load and fatigue load. The yaw unit is subjected to loads which can be eliminated by means of an extra set of bearings, so that all the radial loads are led directly to the nacelle frame. The life of the WTGs can be increased by proper maintenance and good controls.

condition, after 50% of the problem is solved, are shown in Fig. 12bee. If 50% of blade, gearbox, hydraulic unit and yaw unit failures are removed, the overall reduction in failure will be 13.46%, 12.5%, 12.5% and 11.54% respectively, which is quite signicant.

7. Spare parts analysis Scientic spare parts management is of vital importance in wind turbine maintenance, where huge investments are made on spare parts. Any delay in replacement of failed components will result in a huge loss of power generation. By maintaining an optimum number of spares, the availability of the wind farm can be improved. So, a scientic spare parts optimization study was undertaken, and is presented in this section. The spare parts optimization technique helps us to arrive at the optimum number of spare parts, by striking a compromise between over-stocking of spare parts involving heavy expenditure and under-stocking of spare parts, running the risk of non-availability of spare parts. The spares optimization of wind turbine generators is very essential for ensuring high operational readiness. With an increase in the basic quality level of the components, the repair rate can be reduced considerably. Based on the collected data it was understood that certain policy decisions have to be taken to arrive at the various factors, viz. risk, replacement period and inventory procedure, that were required for the analysis. The failure rate is the main driver for the spare parts analysis. The failure rate was calculated using Weibull software for four vital mechanical units of WTGs and nally converted into failure per million hours for easy compatibility with the software calculation. For this study the software package on spare parts optimization, Spare was utilized. Inputs like failure rate, risk, unsupported period, and average utilization were fed into the spare software and the number of required spare parts was obtained. The spare parts analysis has been done for four vital units of the wind farm such as the blade, gearbox, yaw and hydraulic units.

6. Pareto analysis The Pareto diagram is one of the seven quality circle tools used to solve problems, to classify data, and to rank categories in the descending order of occurrence to separate signicant categories from trivial ones. It helps in prioritising the different categories taken into account for analysis. This also helps in identifying the minimum number of factors required to full a function from a lot. An added advantage is, getting the percentage reduction in the overall scenario when reduction is carried out in one or two of the factors. Even though one may want to solve all the problems, it may not be possible in reality. Hence, prioritisation is essential. All the problems identied have to be categorized in terms of the extent of damage they cause to the system. The percentage values can be used to nd out the percentage reduction in the overall problems, when one or two of the prioritised problems are solved. Pareto analysis has been carried out in this study to prioritise the vital factors which caused failures in the system. This analysis was extended to nd out the reduction in problems, when one problem is tackled partially or completely. These failures reduce considerably, the availability of the wind farm and its generation. By tackling a few vital failures, failure time can be minimized to a great extent. Pareto analysis has been done for four vital components of WTGs such as the blade, gearbox, hydraulic unit and yaw unit, and the frequency of failures is shown in Fig. 12. It reveals that in ve years of operation, the frequency of failure of these components was 26.92%, 25%, 25% and 23.08%, respectively. Among these mechanical components, the frequency of blade failure is high during the ve years of operation. The Pareto diagrams of the

7.1. Risk % analysis A major risk in stocking spare parts is an insufcient spare parts during the unsupported period to meet the demands due to component failure. This risk is xed by engineering judgment based on previous experience and non-availability of components. Taking

2748
Gear box; 0% Others; 17%

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

Others; 14% Brake pad; 0% Gear box; 29%

Brake pad; 17%

Blade; 49%

Hydraulic unit; 21%

Blade; 14%

Hydraulic unit; 17%

Yaw unit; 0% 2000

Yaw unit; 21% 2001

Gear box; 17%

Others; 13%

Gear box; 13%

Blade; 0% Others; 42%

Blade; 9% Brake pad; 22%

Yaw unit; 25%

Yaw unit; 17%

Brake pad; 8%

Hydraulic unit; 8% 2002

Hydraulic unit; 26% 2003

Gear box; 19% Others; 24%

Brake pad; 5%

Hydraulic unit; 10% Yaw unit; 10% 2004

Blade; 33%

Fig. 11. Percentage distribution of failures in wind farm.

the following factors into account, the probability of risk can be assessed. 1) The cost of items going down while more spare parts are procured. 2) The possible long delay if more spare parts have to be procured (specialized spare parts may be unobtainable after an initial production run). 3) The importance of maintaining good customer relationship. The spare holding is determined for three different risks individually, for comparison and better understanding. A single risk cannot be decisive as it involves many policy decisions to be taken at various management levels of the wind energy industry. Hence the risks of 90%, 80% and 70% are chosen to arrive at the results with marked differences in the total spare parts requirements, so that comparison can be easily made between the various risks.

7.2. Unsupported period (days) This is the expected operational period of wind farms during which no replenishment of spare parts is needed. In other words, it is the period starting right from initiating the indent of requirement of a spare, to the actual time of receipt of the spare parts from a source. This has a direct bearing on the total number of spare parts required. As the unsupported period increases, the spare parts requirement also increases. The unsupported period can be between 1 and 999 days. In this study it is taken as 45 days, based on previous experience.

7.3. Average utilization (%) This is the average percentage of time for which the wind farm is operational. The average utilization percentage is taken as 100% as

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

2749

120 100
100%

b
Cumulative Frequency (%)

100 90 80 Cumulative Frequency (%) 86,54%

Frequency (%)

Frequency (%)

80 60 40 20 0

70 60 50 40 30 20 10 0

Blade

Gearbox Hydraulic unit

Yaw unit

Gearbox

Before problem solving

Hydraulic Yaw unit Blade unit Reduction of 50% failures in Blade

100 90 80
Cumulative Frequency (%)

d
87,5%

100 90 80 Cumulative Frequency (%) 87,5%

Frequency (%)

Frequency (%)

70 60 50 40 30 20 10 0 Blade Hydraulic Yaw unit Gearbox unit Reduction of 50% failures in Gearbox

70 60 50 40 30 20 10 0 Blade Hydraulic unit Reduction of 50% failures in Hydraulic unit Gearbox Yaw unit

100 90 80 88,46% Cumulative Frequency (%) Blade Hydraulic Yaw unit unit Reduction of 50% failures in Yaw unit Gearbox
Fig. 12. Pareto analysis of wind farm.

Frequency (%)

70 60 50 40 30 20 10 0

there is always a demand for continuous power generation from wind farms. 7.4. Spare parts requirement An optimized spare parts requirement is essential for quick replacement of failed components to avoid loss of power generation. The optimization is done for 45 days of unsupported period and 100% average utilization. The risks on the operation of wind

farms have been considered at three levels of 90%, 80% and 70% and spare parts requirements have been determined. The required number of spare parts for 15 WTGs is shown in Table 3. In the case of 90%, 80% and 70% risk, the spare parts requirement was 10, 8 and 5 respectively. It was found that the requirement of gearbox bearing was the highest in comparison with the requirement of other sub-components. It was observed that when risk decreases, the requirement of spare parts also decreases. In order to increase wind farm availability, necessary steps are to be taken for

2750 Table 3 Number of spare parts required for 15 WTGs. Risk Sub-units % 90% Blade cylinder Gearbox bearing Yaw pinion Hydraulic spring 80% Blade cylinder Gearbox bearing Yaw pinion Hydraulic spring 70% Blade cylinder Gearbox bearing Yaw pinion Hydraulic spring

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751 Table 5 Loss of power generation and consequential cost due to repair. Critical mechanical component repaired Blade cylinder Yaw pinion Gear box bearing Hydraulic spring Total Repair time/ repair (h) 72 60 25 50 No. of repairs Loss of power in ve years generation due to of operation repair in million (kWh) 10 5 6 8 0.051 0.021 0.011 0.028 0.111 Consequential cost due to repair in million Rupees 0.24 0.09 0.05 0.13 0.51

No. of spare parts required for Total no. of Spare parts 45 days Required 2 3 3 2 1 3 2 2 1 2 1 1 5 8 10

optimum periodical maintenance. In this spare parts analysis the optimum number of spare parts required was determined to maintain minimum shutdown period and to increase power generation. This spare parts analysis will be very useful for the effective maintenance of a wind farm. 8. Financial implications on failures The failure and its nancial implication are analyzed to determine the loss of power generation, consequential cost of failure, repair cost and return on investment. This wind farm is operated for 79% of total time and produced 36.8 million kWh during ve years. The total stoppage time due to various failures including low wind speed period is 140,326 h during ve years of operation. The low wind speed is considered as one of the critical stoppage parameters of the wind farm. Due to low wind, the loss of power generation was 7.2 million kWh and its consequential nancial loss was Rs. 33 million. It is an indication that the wind speed plays a vital role in the selection of right location. It is recommended that the periodical maintenance can be arranged during the periods of low wind. The critical stoppage parameters, stoppage time, loss of power generation and consequential cost of failure are given in Table 4. The loss of power generation and consequential cost e of
Table 4 Loss of power generation and consequential cost of failure due to stoppage. Critical Stoppage parameters Stoppage time in ve Loss of power years of operation generation in (h) million (kWh) 7.2 2.1 0.025 0.17 0.36 2607.64 10.045 Consequential cost of failure in million Rupees 33.0 9.5 0.11 0.77 1.66 0.19 45.9

failure during the period of stoppage are 10.045 million kWh and Rs. 45.9 million respectively. The loss of power generation due to mechanical components failure is higher than the electrical components failure to an extent of 116%. If 50% of mechanical failure stoppage time is reduced then Rs 0.83 million revenue can be gained. The mechanical components which are critical such as blade cylinder, yaw pinion, gearbox bearing, and hydraulic spring are repaired frequently in comparing with other components. The critical mechanical components repaired and their repaired time, loss of power generation and consequential cost due to repairs are depicted in Table 5. The repair time caused generation loss of 0.111 million kWh which also led to revenue loss to an extent of Rs.0.51 million. In the case of blade failure, if the blade failure is eliminated completely, the power generation can be improved to an extent of 0.21% and the return on investment will also be

Table 6 Components failure and replacement cost for 15 wind turbine generators. Components Blade Sub-components Blade up Blade rope Blade cylinder Total Shrink disc Gearbox shaft Gearbox support Pinion Bearing FAn cooling Gear oil seal Total Spring Bearings Pinion Look screw Bed Bolt Total No. of failures Total replacement cost Rs 2 2 10 14 2 1 1 1 6 1 1 13 1 1 5 2 3 12 1 1 1 8 2 13 3,800,000 2000 300,000 4,102,000 36,358 31,949 4900 13,000 15,000 475 500 102,182 365 500 60,000 500 300 62,665 90 500 5500 1200 30,000 37,290 32,000 106,800 22,500 62,500 10,000 233,800 Rs. 4,537,937

Gearbox

Yaw unit

Low wind 100887.90 Grid failure 29065.01 Control panel 358.00 failure Electrical 2349.78 failure Mechanical 5056.73 failure Preventive maintenance 0.86 Total 140326

Hydraulic unit Damper bush Filter Burst disc Spring Accumulator Total Others

Brake pad 8 Control panel (MCCB) 6 Capacitor panel 3 Generator 5 Anemometer 2 Total 24 Over all failures 76

[Note: 1 USD Rs. 48.00].

Source: Log book maintained at Muppandal wind farm in Tamil Nadu.

G.M.J. Herbert et al. / Renewable Energy 35 (2010) 2739e2751

2751

increased to an extent of 0.28%. The components failure and replacement cost for 15 wind turbine generators are represented in Table 6. The total replacement cost for 76 failures of 15 WTGs is Rs. 4,537,937. About 80% wind turbine components are used by indigenous components such as anemometer, gear box, yaw unit and control card. The turbine requires periodic attention and proper maintenance for trouble free operation. By improving the electricity network and adopting planned maintenance during periods of low wind could reduce tremendous amount of loss of energy and cost. 9. Conclusion The performance, reliability, failure and spare parts analysis of a wind farm were carried out in this paper. An effort was made in the present study to estimate the reliability of a wind farm using Weibull distribution as a measure of performance. The different types of failure and the causes of failure were determined. In order to increase the availability of wind turbine generators, spare parts analysis was also done to determine the optimum number of spare parts required at the wind farms. Important conclusions drawn from the present analysis are as follows.  The performance analysis reveals that the technical availability was found to be 94%, real availability was 82.88%, and the capacity factor was 24.9%.  The reliability analysis reveals that the reliability of this wind farm was high during the rst year of operation and with time it decreases drastically beyond economical viability and is reduced to 13% at the end of the fth year of operation.  Pareto analysis reveals that if blade unit problems are solved completely, 26.92% of the total problems will be reduced and if 50% of blade failure, gearbox failure, hydraulic unit problems and yaw unit failures are removed, the overall reduction in failure will be 13.46%, 12.5%, 12.5% and 11.54% respectively.  In this spare parts optimization study it was found that at 90%, 80% and 70% risk, the number of spare parts required for a wind farm was 10, 8 and 5 for an unsupported period of 45 days. It is hoped that these studies will be useful for WTG manufacturers

and policy makers for the sustainable development of wind energy in India.  The failure and its nancial implication revealed that if blade failure is eliminated the power generation will be improved to an extent of 0.21% and the increase in return on the investment will be 0.28%.

References
[1] Mejia Juan M, Chejne Farid, Smith Ricardo, Rodriguz Luis F. Simulation of wind energy output at Guajira, Colombia. Renewable Energy 2006;31:383e99. [2] Ben Amar F, Elamouri M, Dhifaoui R. Energy assessment of the rst wind farm section of Sidi Daoud, Tunisia. Renewable Energy 2008;33:2311e21. [3] Sen Zekai. Statistical investigation of wind energy reliability and its application. Renewable Energy 1997;10:71e9. [4] Lai Chi-ming, Lin Ta-hui. Technical assessment of the use of a small-scale wind power to meet the demand for electricity in a land aquafarm in Taiwan. Renewable Energy 2006;31:877e92. [5] Dokopoulos PS, Saramourtsis AC, Bakirtziz AG. Prediction and evaluation of the performance of wind-diesel energy systems. IEEE Transactions on Energy Conversion 1996;11(2):385e93. [6] Paul Denholm. Improving the technical, environmental and social performance of wind energy systems using biomass-based energy storage. Renewable Energy 2006;31:1355e70. [7] Sorensen P, Unnikrishnan AK, Mathew SA. Wind farms connected to weak grids in India. Wind Energy 2001;4(3):137e49. [8] Jang HX, Lu L, Burnett J. Weather data and probability analysis of hybrid photovoltaic-wind power generation systems in Hong Kong. Renewable Energy 2003;28:1813e24. [9] Hu Danmi, Hua Ouyang, Du Zhaohui. A study on stall e delay for horizontal axis wind turbine. Renewable Energy 2006;31:821e36. [10] Thomsen K, Fuglsang P, Schepers G. Potentials for site-specic design of MW sized wind turbines. Journal of Solar Energy Engineering 2001;123:304e9. [11] Fuglsang Peter, Thomsen Kenneth. Site-specic design optimization of 1.5e2.0 MW Wind turbines. Journal of Solar Energy Engineering 2001;123:296e303. [12] ODell Kathleen. Improved blade designs and manufacturing process reduce the cost of wind energy. Journal of Solar Energy Engineering 2001;123:264e8. [13] Ammara Idriss, Leclerc Christopher, Masson Christian. A Viscous threedimensional differential/actuator-disk method for the aerodynamic analysis of wind farms. Journal of Solar Energy Engineering 2002;124:345e56. [14] Hu Ssu-yuan, Cheng Jung-ho. Performance evaluation of pairing between sites and wind turbines. Renewable Energy 2007;32:1934e47. [15] Bhatt MS, Jothibasu. Prediction and enhancement of performance of wind farm in India. Journal of Scientic and Industrial Research 2002;61 (12):1056e62.