02635571011055090

Industrial Management & Data SystemsNew procedure for wind farm maintenanceJose Antonio Orosa Armando C. Oliveira Angel Martn Costa

Article information:To cite this document:Jose Antonio Orosa Armando C. Oliveira Angel Martn Costa, (2010),"New procedure for wind farmmaintenance", Industrial Management & Data Systems, Vol. 110 Iss 6 pp. 861 - 882Permanent link to this document:http://dx.doi.org/10.1108/02635571011055090

Downloaded on: 14 April 2015, At: 19:05 (PT)References: this document contains references to 47 other documents.To copy this document: [email protected] fulltext of this document has been downloaded 741 times since 2010*

Users who downloaded this article also downloaded:Idriss El-Thalji, Jayantha P. Liyanage, (2012),"On the operation and maintenance practices of wind powerasset: A status review and observations", Journal of Quality in Maintenance Engineering, Vol. 18 Iss 3 pp.232-266 http://dx.doi.org/10.1108/13552511211265785Ingrid Bouwer Utne, (2010),"Maintenance strategies for deep-sea offshore wind turbines", Journal of Qualityin Maintenance Engineering, Vol. 16 Iss 4 pp. 367-381 http://dx.doi.org/10.1108/13552511011084526

Access to this document was granted through an Emerald subscription provided by 478311 []

For AuthorsIf you would like to write for this, or any other Emerald publication, then please use our Emerald forAuthors service information about how to choose which publication to write for and submission guidelinesare available for all. Please visit www.emeraldinsight.com/authors for more information.

About Emerald www.emeraldinsight.comEmerald is a global publisher linking research and practice to the benefit of society. The companymanages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well asproviding an extensive range of online products and additional customer resources and services.

Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committeeon Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archivepreservation.

*Related content and download information correct at time of download.

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

New procedure for wind farmmaintenance

Jose Antonio OrosaDepartment of Energy and Marine Propulsion, University of A Coruna,

A Coruna, Spain

Armando C. OliveiraFaculty of Engineering, University of Porto, Porto, Portugal., and

Angel Martn CostaDepartment of Energy and Marine Propulsion, University of A Coruna,

A Coruna, Spain

Abstract

Purpose Conditions monitoring system (CMS) is a tool for describing the present condition of thecomponents of a system. To achieve this objective, there is a need to develop an efficient faultprediction algorithm. This paper seeks to address this issue.

Design/methodology/approach The paper analyses four real wind farms with control charts ofindices derived from UNE EN15341:2008 standard indicators, as the main CMS algorithm to definewhich index must be considered for improving wind farm maintenance and related costs.

Findings The findings show that climatic conditions are related to maintenance cost indices.Employing the statistical control process of various wind energy converter (WEC) indices proposed bywind farm operators is an adequate procedure to monitor and control wind farm performance. Inparticular, only the maintenance cost index and the hourly maintenance cost index presented a clearrelationship with respect to weather conditions.

Practical implications Climatic conditions must form the basis for organising maintenanceactivities. Despite this, future maintenance models must be centred on indices obtained fromexperience, like the maintenance cost index and hourly maintenance cost index, and not solely ingeneral indicators defined by standards.

Originality/value A practical case study of wind farm maintenance based in the new UNEEN15341:2008 standard and wind farm operators experience is shown, defining real indices to beemployed in future maintenance procedures.

Keywords Maintenance, Wind power, Climatology

Paper type Case study

1. IntroductionA recent review work (Alsyouf and EI-Thalji, 2009) on wind farm maintenance,classified it according to relevant life cycle processes of wind power systems: designand development, production and construction, diagnostic, autonomous, proactive,predictive and preventive maintenance.

Design and development maintenance shows the importance of consideringmaintenance when designing wind turbines (Kuhn et al., 1999; van Bussel and Zaaijer,

The current issue and full text archive of this journal is available at

www.emeraldinsight.com/0263-5577.htm

The authors wish to thank University of A Coruna for the financial support of project number5230252906.541A.64902.

Wind farmmaintenance

861

Received 16 November 2009Revised 4 January 2010Accepted 22 March 2010

Industrial Management & DataSystems

Vol. 110 No. 6, 2010pp. 861-882

q Emerald Group Publishing Limited0263-5577

DOI 10.1108/02635571011055090

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

2001; Teresa, 2007), and production and construction maintenance is related withtechnical failures like manufacturing problems, weather conditions and storingproblems (Wood, 2004).

On the other hand, the diagnostic group (Aroatia.lbizu et al., 2004) is based onon-line condition monitoring of components like, for example, induction generators.The autonomus (TPM) group (Krokoszinski, 2003; Tavner et al., 2007) is based onmathematical models that quantify wind farm production losses in terms of plannedand unplanned downtimes. In particular, this last group analyses how weather canaffect the observed results.

The proactive group (Rademakers et al., 2010; Braam et al., 2010) is based on astructure approach of operation and maintenance (O&M) issues and towardsoptimisation of maintenance strategies in wind farms. This group showed the need of atool that assists operators in taking cost effective decisions in their day-to-day work.

Finally, the preventive group (Iniyan et al., 1996) showed that downtimefluctuations of a wind farm depend on wind velocity, and the predictive group showsthe integration of condition monitoring systems (CMS) in wind farm technology. Itshowed that current maintenance planning is not optimised (Caselitz et al., 1994;Jefferies et al., 1998; Wilkinson and Tavner, 2004; Khan et al., 2006; Nilsson andBertling, 2007).

Conditions monitoring system (CMS) is a tool for describing the present condition ofthe components of a system. CMS is being used today in many other applications, butin the wind power industry it is relatively new, so it is very interesting to analyse apractical case study that allows to define the main parameters to be considered(Verbruggen, 2003). It plays an important role in establishing a condition-basedmaintenance and repair (M&R), which can be more beneficial than corrective andpreventive maintenance. To achieve this objective, there is a need to develop anefficient fault prediction algorithm and this algorithm shall be the basis of CMS.

The development and application of algorithms is one way to establish an efficientand reliable operational and repair (O&R) system (Hameed et al., 2009). Thesealgorithms are developed and then implemented by keeping in view the maincharacteristics of the WEC. In this sense, one of the main algorithm sources is thequality control process of different indices.

Once an algorithm is proposed, it must be employed over some indices. Theseindices were usually obtained from operators experience and are usually defined as afunction of standard indicators. An indicator is a numerical parameter that providesinformation on critical facilities identified in the processes or individuals with regard totheir expectations or perceptions of cost, quality, and lead times. Care must beexercised in choosing various indicators, as there is a risk involved in using a lot ofnumbers that do not provide any useful information.

When a wind farm has a computerised maintenance management system (CMMS),the calculation of these indices is often much quicker. An additional advantage is thatonce they are automated, we can periodically generate reports with minimal effortaccording to our needs. In recent years, several studies have been carried out on theapplication of these indices to maintenance (Pastor Calvo and Sacristan, 2005; MazaSabalote, 2007), particularly, wind power generation (Moratilla, 2008; Ro Chao, 2004and Bilbao et al., 2005). For example, (Krokoszinski, 2003) concluded that the definitionof a Layout Factor (LF), the adoption of the Planning Factor (PF) and the Overall

IMDS110,6

862

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Equipment Effectiveness (OEE), represent the external losses and the technical lossesof wind farms. This enables a systematic description and quantification of the lossesthat reduce the overall output capacity of wind farms and, therefore, it is of interest tounderstand these indices.

For wind farms, one defines the layout factor (LF) as the maximum possible outputof electrical energy that could be fed into the grid per year, if the complete referenceelectrical energy (Eref) was transformed (transported to the grid connection) withexactly the specified reference productivity. It combines the specified wake losses dueto the arrangement of WECs in the wind farm and the calculated electrical losses(described by the electrical efficiency) of the cables and devices, all the way down theline from the WEC-terminals to the grid connection point (Krokoszinski, 2003).

The planning factor PF is the ratio between the available electrical energy and thetheoretical electrical energy of a wind farm, so that:

Eavailpark E theopark PF 1

The overall equipment effectiveness (OEE) is the ratio between valuable productiontime and available production time. Hence, compared with the theoretical productiontime (the theoretical maximum of deliverable electrical energy) the actual valuableproduction time (corresponding to finally sold electrical energy) is described by thetotal overall equipment effectiveness, defined through:

Tvalu T theo TotalOEE 2

The total efficiency of the wind farm describes the losses that are already determinedin the engineering and operation planning phases of the wind farm, i.e. the parkefficiency and the electrical efficiency due to positioning and cabling of the WECs(combined in the layout factor) and the planned downtimes, as shown by:

Efficiency LF PF 3

Finally, despite the fact that these indices are commonly employed in maintenancestudies, they are very difficult to be employed with the different algorithms of a controlsystem to assist operators. As a consequence, new and easier indices are needed. In thissense, wind farm operators experience enables us to define new indices. A practicalcase study to analyse new indices derived from standard indicators is needed.

In this sense, recent studies presented different models for monitoring wind farmpower (Orosa et al., 2009a; Orosa et al., 2009b, c). In particular, these models employedwind speed as the input to predict the total power output of a wind farm, based onreal-time measured data (Kusiak et al., 2009; SIAM, n.d.; Caselitz et al. 1994). Thesestudies are based on statistical studies of wind energy (Sen, 1997) and particularlyemployed control charts of measured wind characteristics to detect abnormal or wrongdata (Kusiak et al., 2009).

In this study, we intend to analyse the main wind farm indices with control chartsmethods as control algorithm, to define the consumption of resources in all processes,the performance of the organisation, the cost of quality achieved and, in general, todefine the procedure that must be followed in real wind farms to improve maintenance.


863

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

2. Equipment and methodologyIn this section, the maintenance of four real wind farms will be analysed, based onindicators proposed by standards and literature, and based on five indices derivedfrom these indicators proposed by wind farm operators experience.

2.1 Wind farmsThe useful lifetime of wind farms is supposed to be about 20 years. Each plant isinstalled in a similar manner and all of them are exposed to the same weatherconditions. Currently, maintenance is carried out in technical plants, when the plantfails or according to a mandatory preventive schedule imposed by law. In our casestudy the electricity companies, however, schedule maintenance and control cycles,specifying a daily inspection during the operation period and, afterwards, oneinspection every two to three weeks.

Normal maintenance cycles are scheduled two or three times/year. They mostlyinvolve periodic inspections of equipment, oil and filter changes, calibration andadjustment of sensors and actuators, and replacement of consumable such as brakepads and seals. Finally, housekeeping and blade cleaning generally fall into thisscheduled (preventive) maintenance.

On the other hand, unscheduled (failure related) maintenance is anticipated withany project based in previous failure data but not in a particular indicator. Failure ormalfunction of a minor component will frequently shut down the turbine and requirethe attention of maintenance personnel. Additional maintenance is planned accordingto common international practices and suggestions of the companys personnel(Ardente et al., 2008).

In our case study each of the four wind farms shows 24 wind turbines of fivedifferent technologies that exist in Galicia with a total nominal power of 17.56 MW andan annual production of 38.500 MWh. All wind turbines present a horizontal axis rotorand, consequently, their wind power conversion will begin at 3m/s and they will bedisconnected from the electrical network when winds reach values higher than 25m/swith Southwest and Northwest predominant winds. All data of power production isstored in a control center with a ten-minute time frequency. Maintenance is mainly acorrective maintenance and a preventive maintenance based in a wind farm operatorsexperience. Finally, a predictive maintenance based in the integration of conditionmonitoring systems (CMS) is being considered for future deployment.

2.2 Maintenance cost indicatorsThe development and application of algorithms is one way to establish an efficientand reliable operational and repair (O&R) system. These algorithms are developedand then implemented by keeping in view the main characteristics of the WEC,previous failure data, identifying components which will cause more downtime,components which are more prone to the initiation of crack, wear, misalignment, etc.Focus should be given to rotating components, and the structures that directlysupport those rotations, like the bearing which supports the generator shaft(Hameed et al. 2009).

In accordance with Krokoszinski (2003), the external losses, which are analysed onthe basis of a general theory, can be classified into three types: external downtimelosses, external speed losses, and external quality losses. Among these types, only a

IMDS110,6

864

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

few of the causes of external losses are related to the operation and maintenance ofwind farms. For example, external downtime losses are caused by planned downtimefor preventive maintenance and is called scheduled maintenance of systems, such aswith auto control stops for regular maintenance. External speed losses are onlyexternal losses related to operation and maintenance by the excessive input due towind velocity and are caused by blade- and cable-overload protection. Finally, externalquality losses do not include external losses related to operation and maintenance.

Consequently, operation and maintenance scheduling calls for planning downtimefor preventive maintenance and for the losses related to excessive input due to windvelocity, call for predicting climatic conditions, particularly when many plantsdistributed in the territory are considered (Concetti et al. 2009).

In March 2007, a newmaintenance-related British Standard was published BS EN15341, maintenance maintenance key performance indicators. It describes a systemfor measuring maintenance performance. The standard aims to help organisations inall sectors to appraise and improve their asset maintenance efficiency and effectivenessin pursuit of better global performance and competitive advantage.

The new standard presents a superb set of indices for measuring the outcome ofcomplex maintenance activities. It brings a welcome degree of clarity, order andauthority to this crucial, yet insufficiently understood area of maintenancemanagement.

The standard defines a structure of key performance indicators (KPIs) 24economical, 21 technical and 26 organisational. Each operator is urged to selectthe indicators that align directly with each business objectives, and then apply them tothe management of maintenance activities within the organisation using a CMMS.

To implement this schedule, the analysis of the main maintenance indicators listedbelow is required. As a first step of the analysis of the measured data, maintenance costindicators listed in Table I must be described in accordance with the UNE EN15341:2008 (AENOR, 2008) and UNE EN 13306 (AENOR, 2003):

(1) Mid time to repair (MTTR): it is a basic measure of the maintainability ofrepairable items. It represents the average (mean) time required to repair afailed component or device.

Maintenance cost indicator

1 MTTR2 Total cost of corrective maintenance due to faults3 Time spent on preventive maintenance4 Expenditure on time spent on preventive work (in euros)5 Expenditure on the material used for preventive maintenance (in euros)6 Total cost of preventive maintenance7 Total maintenance cost (TMC)8 Time devoted to other work9 Available maintenance time

10 Time actually spent on maintenance-related activities11 Time of stoppage of production machinery and equipment12 Time of work by production personnel13 Cost of downtime losses due to corrective maintenance caused by failures14 Total production cost

Table I.Maintenance cost

indicators


865

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

This value considers the hours spent to address failures in differentproduction processes. Faults arising from production and earlier maintenanceactivities for repairing the damage account for some of these hours. Theyinclude the total hours spent on repair, corrective maintenance, and breakage ordamage. This indicator is related with the mean time between failure (MTBF)that expresses the frequency of failures.

(2) Total cost of corrective maintenance due to faults: this value is obtained byconsidering the labour and materials employed for corrective action.

(3) Time spent on preventive maintenance: this value is obtained by totalling thehours spent in each preventive maintenance task. Each task must include thetotal number of hours devoted to preventive maintenance or inspection.

(4) Expenditure (in euros) on time spent on preventive work: this is the product ofthe number of hours dedicated to interventions for preventive maintenance (5)and the cost per hour allocated to maintenance services.

(5) Expenditure (in euros) on the material used for preventive maintenance: thisvalue summarises the cost of the materials used in preventive inspectionscarried out at production centres.

(6) Total cost of preventive maintenance: this is the total sum of the value of labourdeployed and materials consumed, expressed in euros (4 5).

(7) Total maintenance cost (TMC): this value represents the total expenditure oncorrective and preventive interventions by maintenance personnel and thematerials used (2 6).

(8) Time devoted to other work: this summarises the hours devoted to special workrequests (changes, implementing improvements, etc) and the maintenance tasksundertaken by maintenance services.

(9) Available maintenance time: these are the hours spent by skilled service andmaintenance personnel divided by the hours spent on other work and thenumber of hours actually spent on corrective and preventive maintenance.

(10) Time actually spent on maintenance-related activities: the total number of hoursspent for corrective maintenance due to troubleshooting and preventivemaintenance.

(11) Time of stoppage of production machinery and equipment: it is the totalnumber of hours during which the machines are stopped in a production linedue to breakdowns, resulting in production losses. Preventive maintenancemust bring a reduction in the time spent on these activities.

(12) Time of work by production personnel: these include operators belonging todifferent production lines.

(13) Cost of downtime losses due to corrective maintenance caused by failures: it isthe product of the number of hours of stoppage due to failures and the cost perhour allocated to each product line.

(14) Total production cost: it is the product of the number of hours of productionstaff turnout and the cost per hour allocated to each product line.

IMDS110,6

866

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

2.3 Maintenance indicatorsTable I shows performance data for maintenance cost indicators for the four windfarms, expressed in different units of measure (UM).

Once the indicators noted in the previous section are listed, we need to developindices that facilitate the monitoring of maintenance management. Due to theexcessive information obtained in a reduced period of time, it is very difficult forwind farm operators to take decisions. These indices must be obtained fromprevious experiences in real wind farms, as those shown in the following equations(4 to 8):

(1) Index of staff actually utilised in maintenance activities (indicator 1): this isexpressed as a percentage of the total working time spent on maintenanceactivities, as expressed by:

ID1 Time actually spent on maintenance2 related activitiesTime worked by production personnel

100 4

(2) Index of extension of preventive maintenance beyond the hours available formaintenance activities (indicator 2): this index expresses the percentagebetween the number of hours spent on preventive maintenance and thedifference between maintenance time (hours) and time spent on other works:

ID2 Time spent on preventive maintenanceTime actually spent on maintenance2 related activities

100 5

(3) Ratio of interventions by faults to the total available hours (indicator 3): it isexpressed as the relationship between the hours spent on correctivemaintenance for breakdowns and the number of available maintenance hours:

ID3 MTTRAvailable maintenance time

100 6

(4) Maintenance cost index: this index expresses the relationship between the totalcost of troubleshooting and preventive maintenance and the number of hoursactually spent on corrective and preventive maintenance (indicator 4):

ID4 TMCTime actually spent on maintenance2 related activities

7

(5) Maintenance cost index related to the cost of production (indicator 5): itexpresses the relationship of the sum of the total cost of production forcorrective maintenance, preventive maintenance, and production stoppages forcorrective action with the total cost of production:

ID5 TMC Cost of downtime losses due to corrective maintenanceTotal production cost

100 8

Table II lists the maintenance cost indicators employed.


867

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

2.4 Limitations of process control: variables control chartsIn any production process, there is always some unavoidable variation. This is anormal variation and the cumulative effect of many small uncontrollable causes. Whenthe variation is relatively small and associated with unforeseen causes in a stablesystem, it is considered acceptable in the course of normal operation of the process andtreated as if it is within the statistical control limits. In contrast, there are other causesof variation arising from the process that are derived from three different sources:

(1) Rectifying malfunctioning machines.

(2) Human errors of the people who operate the machines.

(3) Defective raw materials.

The variations produced by these assignable causes are usually large compared tonormal process variations. Consequently, the process attains an unacceptable level ofperformance and is treated as a process out of control. Statistical process control isbasically intended to detect the presence of assignable causes calling for correctiveaction. In particular, mid-range control charts are used when the controlled qualitycharacteristic is a continuous variable.

As has been explained, the control chart is one of the most important and commonlyused the statistical quality control (SQC) methods for monitoring process stability andvariability (Montgomery, 1991). It is a graphical display of a process parameter plottedagainst time, with a centre line and two control limits ( Jennings and Drake, 1997). Inour case study, we can measure some continuously varying quality characteristics ofinterest (EN15341 indicators) and eventually the variable control charts were selected.

Once we measured the time evolution of different indicators, a statistical processhad to be developed to establish the control limits. These limits are usually set aboveand below the mean value equivalent to three times the standard deviation of theprocess.

The calculation of the average of all measures, and the upper and lower controllimits are given by:

UCL m 3 s 9

LCL m2 3 s 10where:

m is the mean of each indicator.

s is the standard deviation of each indicator.

Indices of costs

1 Index of programmed maintenance2 Index of extension of preventive maintenance beyond the hours available for maintenance

activities themselves3 Ratio of interventions by faults on the total available hours4 Maintenance cost index related to the cost of production5 Index of cost of maintenance hours related to production

Table II.Cost indices

IMDS110,6

868

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

After setting the control limits, it is necessary to define when a process is said to be outof control. To do this, we define A, B, and C as the regions between one, two, and threetimes the standard deviation over and below the mean and apply the following rules:

. two of three points in a row in Area C.

. four of five points in a row in zone B or beyond.

. six straight points up or down.

. eight consecutive points outside the area A, on both sides of the centre line.

In any case, we must consider the presence of patterns or trends in control charts, 2009(Molinero, 2003). The procedure is illustrated schematically in Figure 1.

Figure 1 shows the typical procedure that must be employed with variable controlcharts. Once the information is analysed in a control chart, we may determine if each ofthe indicators is under or out of control, in accordance with the previous rules. If theprocess is out of control, we only have to eliminate the assignable causes. If the processis under control, we must assess the capability to control the process. If the process isnot capable to control each indicator within control limits, we must take a generaldecision over the process. On the other hand, if the process is capable to control theindicators within the control limits, we must try to optimise the process. If the processis centred within the control limits, the optimisation was obtained.

3. ResultsHaving defined the benchmarks and indicators normally used for maintenanceanalysis, we applied them to our case study of the four wind farms, for a period of twoyears. Consequently, the data of Table I and each index of Table II were calculated foreach month, and are represented in Tables III and IV and Figures 2-8.

4. DiscussionAs a first step, we analysed the average personnel index and its corresponding controllimits, as shown in Figure 2. This index represents the need for manpower dedicated tocurtailing the operation cost of production systems. Comparing this index with themonthly value revealed that in December it was clearly lower and, consequently, thenumber of hours of work spent for total maintenance had been reduced. Once theyearly weather conditions are observed as shown in Figure 7, we can confirm that thisbehaviour was related to weather conditions. For example, in Figure 7 we observe thatlow wind speeds prevailed in the period from October to December and, therefore,maintenance needs were low. On the other hand, when we examine the months ofhighest wind speed, such as March, a distinct increase in working hours is visible.

In Figure 4, a similar correlation is observed in the increase of number of hoursspent in troubleshooting and available hours. This figure shows a rise in the number ofhours spent in rectifying damage from December to March and then a sharp decline. Inparticular, plant reports showed that these hours were spent on replacing systemcomponents causing errors and system failures. This same effect can also be seen inFigure 3, which shows a decrease in the number of hours spent on preventivemaintenance with respect to the total time spent from November to March.

If we convert these indices to economic terms, we can say that, in most cases, thecost of maintaining production related to the cost of maintenance hours reveals an


869

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

annual figure varying about the mean in the range of the standard deviation. Only inFebruary does the maintenance cost exceed the control limits, and can therefore beassociated with an assignable cause. On the other hand, the number of hours of failurein Figure 4 indicates that preventive maintenance requirements in those months werelower. In particular, this problem is associated with poor weather conditions as thewind speed causes transient phenomena associated with equipment breakdown.

Figure 1.Method of analysingcontrol chart variables

IMDS110,6

870

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Prior to this research work, trends have been observed in all indices due to weatherconditions. To demonstrate the existence of assignable causes, a correlation analysisbetween the monthly mean value of each index and the mean monthly wind velocitywas developed. As new added knowledge, this correlation factor is plotted in Figure 8and shows that despite most indices being employed to evaluate the performance ofwind farms, only the maintenance cost index and the index of maintenance hourly costpresent a clear relationship with weather conditions. In particular, these parametersreached values of 0.78 and 0.82, which are considered as acceptable for defining astatistical relationship between the variables.

Some solutions for this problematic situation are proposed. As pointed out before,we can conclude that the maintenance cost of a wind farm depends on the weatherconditions, and these conditions can be predicted once the weather patterns are known.In this sense, the first solution is to raise the maintenance on these components basedon weekly weather conditions. This consideration enables weekly modifications ofpreventive maintenance work, which can result in only ten more minutes per machine.The expected result is the dramatic improvement in maintenance levels that willsignificantly reduce the number of errors and failures.

Indices of costs Total UM

1 Index of staff actually utilised in maintenance activities 36.04 %2 Index of extension of preventive maintenance over the hours available for

maintenance activities themselves 64.98 %3 Summary of interventions by faults on the total available hours 7.57 %4 Maintenance cost index related to the cost of production 7.72 %5 Index of maintenance hours cost related to production 52.89 e

Table IV.Indices of costs for a

month

Maintenance cost indicators Total UM

1 Time (hours) devoted to correct failures 9,672.92 Hours2 Total cost of corrective maintenance due to faults 347,357.20 e3 Time (hours) spent on preventive maintenance 18,516.53 Hours4 Expenditure (euros) on time spent on preventive work 2,873,287.89 e5 Expenditure (euros) on the material used for preventive

maintenance 960,868.95 e6 Total cost of preventive maintenance 3,834,156.84 e7 Total cost of maintenance 4,181,514.04 e8 Time (hours) devoted to other work 52,704.00 Hours9 Available maintenance time (hours) 131,760.00 Hours

10 Time (number of hours) actually spent on maintenance-relatedactivities 28,495.45 Hours

11 Time (hours) of stoppage of production machinery andequipment 22,048.57 Hours

12 Time (hours) worked by production personnel 79,056.00 Hours13 Cost of downtime losses due to corrective maintenance caused

by failures 220,485.65 e14 Total production cost 570,074.35 e

Table III.Maintenance cost

indicators for a month


871

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Figure 2.Index of personnel (I1)

IMDS110,6

872

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Figure 3.Index of extension

preventive maintenance(I2)


873

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Figure 4.Index of repairs due tofailures (I3)

IMDS110,6

874

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Figure 5.Maintenance cost index

related to the cost ofproduction (I4)


875

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Figure 6.Index of hourlymaintenance cost relatedto production (I5)

IMDS110,6

876

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Figure 7.Wind velocity during the

year


877

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

A second option would be to stop the operation of the equipment during periods ofmaximum wind instability, but this would waste precious wind energy when the windcan transfer energy at a greater intensity. Therefore, a complementary option is to adda new control system to avoid these faults before the abrupt changes of wind speed.This control system must be tuned to climatic conditions.

The main implication of these new models, based on real time measured data bynetworks (Walsh et al., 2000; Kehoe and Boughton, 2001), is that it can be sent to newsoftware tools (Sahay and Gupta, 2003; Robert and William, 1999; Schmidth, 1999)designed and selected to implement the maintenance of a specific wind farm. Thesesoftware tools can propose the optimum working periods of each wind farm inaccordance with weather conditions and historical information of the wind farm.

From this, we may conclude that the analysis of real measured data reveals theappropriate characteristics of wind farms (Apt, 2007 ). Furthermore, this new conceptof weather maintenance could be the basis of a new wind farm control system. Forexample, the creation of a tele-maintenance intelligent system, based also on neuralnetworks and marked by result management logic, called GrAMS (GrantedAvailability Management System), was recently evaluated as a real possibility(Concetti et al. 2009). GrAMS is a system for the technical administrative managementof a technical plant system, with differing technical natures, distributed in the territory,which gives the management and maintenance service contractor (ManagementOrganization) the possibility of being able to guarantee, even through monitoring andcontrol from a remote centre, a service characterized by total availability of plantsand zero failures.

As a consequence, this future tele-maintenance system can be improved by weathermaintenance indices. Moreover, these maintenance index models lead to reduction inmaintenance costs. However, care must be exercised in choosing them, because of therisk of using a lot of numbers that do not provide any useful information. It is in thissense that the indices of percentage of repair and preventive maintenance and thepercentage rate of corrective repair costs for damage are not suitable for the study ofcontrol charts and, consequently, only instantaneous values clearly identify thebehaviour of these variables.

Figure 8.Correlation coefficient ofeach index with respect toclimatic conditions

IMDS110,6

878

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

On the other hand, another way to define wind farm models is to correlate the windturbine power curve with wind characteristics. These models must be developed on thebasis of real scenarios, and the statistical techniques employed must be adapted tofilter the data obtained from real wind farm operation (Sainz et al., 2009). Furthermore,it is well known that actual measurements are liable to include wrong data; so otherautomatic filtering techniques are essential to deal with this problem(Llombart-Estopinan, 2008).

Finally, it is interesting to consider as future research work that cultural dimensionsinfluence management behaviour in different countries (Ellemose Gulev, 2008) and thatthere are a few real case studies about real measured data on wind farms, itsmaintenance and the sustainability of the wildlife (Solari and Minervini, 2004;Anderson et al. 1998), to consider the bio-naturalistic components when preventingenvironmental impacts.

5. ConclusionsPrevious research works showed that apparently the maintenance cost of a wind farmdepends on weather conditions, which could be predicted once the future weatherpatterns are known.

To prove the real relationship between wind farm maintenance and weatherconditions, this work analysed four real wind farms from a total quality perspective,helping to improve maintenance based on operators experience.

The findings showed that operators experience is a good method to select wind farmmaintenance indices derived from standard indicators. In this sense, employing thestatistical control process of these new indices resulted as an adequate algorithm tomonitor and control wind farm performance. Furthermore, once trends in all the newindices have been observed, assignable causes can be established to explain variations. Inthis sense, a correlation analysis between the monthly mean value of each index and themean monthly wind velocity was carried out. Only the maintenance cost index and thehourly maintenance cost index presented a clear relationship with weather conditions.

Despite these indices being adequate to monitor and control wind farmmaintenance, care must be exercised in choosing them, because of the risk of usingmany numbers that do not provide any useful information. In this sense, it wasrevealed that the indices of percentages of repair and preventive maintenance and thepercentage rate of corrective repair costs are not suitable for the statistical controlprocess of wind farms.

Finally, this could be the basis of a new control system and of the indices to beconsidered. Therefore, it is suggested that future research works and innovations (Orosaet al., 2009a; Orosa et al., 2009b, c) define wind farm models by correlating the windturbine power curve to the wind characteristics based on real scenarios and statisticaltechniques. In these models, actual measurements are liable to include wrong data, andconsequently, the application of other automatic filtering techniques is essential to dealwith this problem.

References

AENOR (2003), UNEEN 13306, Maintenance Terminology, UNI EN Standard, AENOR, Madrid.

AENOR (2008), UNE EN15341:2008, Maintenance Maintenance Key Performance Indicators,AENOR, Madrid.


879

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Alsyouf, I. and EI-Thalji, I. (2009), Maintenance Practices in Wind Power Systems: A Review andAnalysis, School of Technology and Design Vaxjo University, Vaxjo.

Anderson, R., Erickson, W., Strickland, D., Tom, J. and Neumann, N. (1998), Avian monitoringand risk assessment at Tehachapi Pass and S. Gorgonio Pass wind resources areas,California: phase 1, preliminary results, Proceedings of National Avian-wind PowerPlanning Meeting III, San Diego, CA, May.

Apt, J. (2007), The spectrum of power from wind turbines, Journal of Power Sources, Vol. 169No. 2, pp. 369-74.

Ardente, F., Beccali, M., Cellura, M.V. and Brano, L. (2008), Energy performances and life cycleassessment of an Italian wind farm, Renew. Sust. Energ. Rev., Vol. 12 No. 1, pp. 200-17.

Aroatia.lbizu, I., Tapia, A., Saenz, J., Mazon, A.J. and Zamora, I. (2004), Online stator windingfault diagnosis in induction generators for renewable generation, paper presented atIEEE MELECON 2004, Dubrovnik.

Bilbao, M., Terradillos, J., Ciria, J.I. and Malaga, A. (2005), Analisis del aceite como herramientade mejora del comportamiento de las multiplicadoras de aerogeneradores: principalesproblemas detectados a traves del lubricante, Mantenimiento: ingeniera industrial y deedificios, Vol. 189, pp. 6-16.

Braam, H., Rademakers, L.W. and Verbruggen, M.M. (2010), T.W. CONMOW: conditionmonitoring for offshore wind farms, available at: www.ecn.nl

Caselitz, P., Giebhardt, J. and Mevenkamp, M. (1994), Online fault detection and prediction inwind energy converters, EWEC94 Thessaloniki, pp. 623-7.

Concetti, M., Cuccioletta, R., Fedele, I. and Mercuri, G. (2009), Tele-maintenance intelligentsystem for technical plants result management, Reliability Engineering & System Safety,Vol. 94, pp. 63-77.

Ellemose Gulev, R. (2008), Cultural repercussions. An analysis of management behaviourthrough the lens of European cultural variation, Industrial Management & Data Systems,Vol. 109 No. 6, pp. 793-808.

Hameed, Z., Hong, Y.S., Cho, Y.M., Ahn, S.H. and Song, C.K. (2009), Condition monitoring andfault detection of wind turbines and related algorithms: a review, Renew SustainableEnergy Review, Vol. 13 No. 1, pp. 1-39.

Iniyan, S., Suganthi, L. and Jagadeesan, T.R. (1996), Fault analysis of wind turbine generators inIndia, Renewable Energy, Vol. 9 No. 1, pp. 772-5.

Jefferies, W.Q., Chambers, J.A. and Infield, D.G. (1998), Experience with bicoherence of electricalpower for condition monitoring of wind turbine blades, IEEE Proc. Image Signal Process,Vol. 145 No. 3, pp. 141-8.

Jennings, A.D. and Drake, P.R. (1997), Machine tool condition monitoring using statisticalquality control charts, International Journal of Mach. Tools Manufact., Vol. 37 No. 9,pp. 1243-9.

Kehoe, D. and Boughton, N. (2001), Internet-based supply chain management: a classification ofapproaches to manufacturing planning and control, International Journal of Operations &Production Management, Vol. 21 No. 4, pp. 516-25.

Khan, M.M., Iqbal, M.T. and Khan, F. (2006), Reliability and condition monitoring of a windturbine, paper presented at Electrical and Computer Engineering 2005, CanadianConference, 1978-1981.

Krokoszinski, H.J. (2003), Efficiency and effectiveness of wind farms keys to cost-optimizedoperation and maintenance, Renewable Energy, Vol. 28 No. 14, pp. 2165-78.

IMDS110,6

880

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

Kuhn, M., Bierbooms, W., van Bussel, G.J.W., Cockerill, T.T., Harrison, R., Ferguson, M.C.,Goransson, B., Harland, L.A., Vugts, J.H. and Wiecherink, R. (1999), Towards a matureoffshore wind energy technology guidelines from the Opi-OWECS project, WindEnergy, Vol. 2, pp. 25-58.

Kusiak, A., Zheng, H. and Song, Z. (2009), Models for monitoring wind farm power, RenewableEnergy, Vol. 34 No. 3, pp. 583-90.

Llombart-Estopinan, A. (2008), Improving the operation and maintenance of wind farms:determination of wind turbine performance, Icrepq08, available at: www.icrepq.com/papers-icrepq08.htm

Maza Sabalote, J.P. (2007), El mantenimiento y la productividad, Ing. Qumica, Vol. 450,pp. 64-5.

Molinero, L.M. (2003), Control de calidad (Quality control), Control charts, available at: www.seh-lelha.org/calidad.htm (accessed October 2009).

Montgomery, D.C. (1991), Introduction to Statistical Quality Control, 2nd ed., Wiley, New York,NY.

Moratilla, B.Y. (2008), Energa eolica: operacion y mantenimiento de parques eolicos,Generacion electrica con energa eolica: presente y futuro/coord, pp. 101-4.

Nilsson, J. and Bertling, L. (2007), Maintenance management of wind power systems usingcondition monitoring systems life cycle cost analysis for two case studies, IEEETransactions on Energy Conversion, Vol. 22 No. 1, pp. 229-32.

Orosa, J.A., Garca-Bustelo, E.J. and Oliveira, A. (2009a), Low speed wind concentrator toimprove wind farm power generation, 35th Annual Conference on the IEEE IndustrialElectronics Society, IECON 2009, IECON Proceedings.

Orosa, J.A., Garca-Bustelo, E.J. and Perez, J.A. (2009b), Galician climatic change effect on windpower production, POWERENG2009, 978-1-4244-2291-3/09/$25.00 q2009 IEEE,pp. 180-4.

Orosa, J.A., Garca-Bustelo, E.J. and Perez, J.A. (2009c), Wind turbine concentrator design basedon moist air phase change, POWERENG2009, 978-1-4244-2291-3/09/$25.00q2009 IEEE,pp. 180-4.

Pastor Calvo, E. and Sacristan, J.S. (2005), Avance estudios Grupos de Trabajo del Comite deMantenimiento en el Sector de la Energa (AEM): (1 parte), valor del mantenimiento ycuadro de mando; formacion y cualificacion del personal, Mantenimiento: ingenieraindustrial y de edificios, Vol. No. 184, pp. 30-42.

Rademakers, L.W.M.M., Braam, H. and Verbruggen, T.W. (2010), R&D needs for O&M of windturbines, available at: www.ecn.nl

Ro Chao, C. (2004), Mantenimiento predictivo en la energa eolica, opcion rentable ycompetitiva, Energa: Ingeniera energetica y medioambiental, Vol. 30 No. 177, pp. 118-20.

Robert, C.K. and William, D.S. (1999), Software selection guide, Quality, pp. 31-5.

Sahay, B.S. and Gupta, A.K. (2003), Development of software selection criteria for supply chainsolutions, Industrial Management & Data Systems, Vol. 103 No. 2, pp. 97-110.

Sainz, E., Llombart, A. and Guerrero, J.J. (2009), Robust filtering for the characterization of windturbines: improving its operation and maintenance, Energy Conversion and Management,Vol. 50 No. 21, pp. 36-2147.

Schmidth, K. (1999), Software selection guide, Quality, Vol. 39 No. 12, pp. 5-18.

Sen, Z. (1997), Statistical investigation of wind energy reliability and its application, RenewableEnergy, Vol. 10 No. 1, pp. 71-9.


881

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

SIAM (n.d.), Environmental information system of Galicia, available at: www.meteogalicia.es(accessed October 2009).

Solari, P. and Minervini, G. (2004), Exploitation of renewable energy sources and sustainablemanagement of the territory wind farms in Regione Liguria, Management ofEnvironmental Quality: An international Journal, Vol. 15 No. 1, pp. 41-7.

Tavner, P.J., Xiang, J. and Spinato, F. (2007), Reliability analysis for wind turbines, WindEnergy, Vol. 10 No. 81, pp. 1-18.

Teresa, H. (2007), Wind turbines: designing with maintenance in mind, Power Engineering.

van Bussel, G.J.W. and Zaaijer, M.B. (2001), DOWEC concepts study, reliability, availability,and maintenance aspects, paper presented at European Wind Energy Conference andExhibition (EWEC).

Verbruggen, T.W. (2003), Wind turbine operation and maintenance based on conditionmonitoring, Final report, ECN-C-03-047.

Walsh, P., Koumpis, A. and Barziv, O. (2000), Managing real time interactions in industrialenvironments based on information supply chains: the ESPRIT ATLAS project, LogisticInformation Management, Vol. 13 No. 2, pp. 45-57.

Wilkinson, M.R. and Tavner, P.J. (2004), Extracting condition monitoring information from awind turbine drive train, 39th International Universities Power Engineering Conference(UPEC), No. 2, pp. 591-4.

Wood, J. (2004), Up and running: as the offshore wind industry develops; its learning from somepainful experiences, paper presented at IEEE Power Engineering 2004.

About the authorsJose Antonio Orosa has a PhD in Marine Engineering and graduated in Marine Engineering andNaval Architecture from the University of A Coruna. His research is related to moist air andenergy saving. Recently, he has participated in the International Energy Agency Annex 41 andcollaborated with the University of Porto in research on energy saving. Presently, he is Professorof quality control and Head of the Department of Energy and Marine Propulsion of theUniversity of A Coruna (Spain). He is a member of the Society of Naval Architects and MarineEngineers (SNAME) and ASHRAE. Jose Antonio Orosa is the corresponding author and can becontacted at: [email protected]

Armando C. Oliveira is Head of the New Energy Technologies Research Unit, which existswithin the Institute of Mechanical Engineering FEUP (Faculty of Engineering of theUniversity of Porto). He has coordinated and participated in 13 European research anddevelopment projects related to the development of new and sustainable energy systems,especially solar thermal systems (heating, cooling and CHP systems). Nowadays, he isSecretary-General of the World Society of Sustainable Energy Technologies and co-responsiblefor the conference series on Sustainable Energy Technologies, with several sessions held inEurope, Asia and America.

Angel Martn Costa holds a Masters in Marine Engineering. In previous years he has workedin ship maintenance and currently he is principally responsible for wind farms maintenance inthe northwest of Spain. He is developing his PhD in wind farms maintenances.

IMDS110,6

882

To purchase reprints of this article please e-mail: [email protected] visit our web site for further details: www.emeraldinsight.com/reprints

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

This article has been cited by:

1. Kym Fraser, Hans-Henrik Hvolby, Tzu-Liang (Bill) Tseng. 2015. Maintenance management models:a study of the published literature to identify empirical evidence. A greater practical focus is needed.International Journal of Quality & Reliability Management 32:6. . [Abstract] [PDF]

2. Mahmood Shafiee. 2015. Maintenance logistics organization for offshore wind energy: Current progressand future perspectives. Renewable Energy 77, 182-193. [CrossRef]

3. Tuomo Heikkil. 2015. A decision support system to evaluate the business impacts of machine-to-machine system. Benchmarking: An International Journal 22:2, 201-221. [Abstract] [Full Text] [PDF]

4. ngel M. Costa, Gholamreza Roshan, Jos A. Orosa, ngel Rodrguez-Fernndez. 2014. Case Study ofWeather Maintenance in Wind Power Generation. Arabian Journal for Science and Engineering . [CrossRef]

5. Phillip McKay, Rupp Carriveau, David S.-K. Ting. 2011. Farm Wide Dynamics: The Next Critical WindEnergy Frontier. Wind Engineering 35, 397-418. [CrossRef]

Dow

nloa

ded

by U

FRN

At 1

9:06

14

Apr

il 20

15 (P

T)

02635571011055090

Documents

emerald publication

maintenance strategies

real wind farms

emerald subscription

maintenance cost indices

university of porto

ptnew procedure

marine propulsion