Failure rate and downtime survey of wind turbines located in Spain

Correspondence Estefania Artigao, Renewable Energy Research Institute, UCLM, Albacete 02071, Spain. Email: estefania.artigao@uclm.es Abstract Publicly available wind turbine failure rate and downtime studies are scarce in the scientific literature. Recent surveys in the field reveal that only 20 studies of wind farms based in Europe, Asia and the U.S. have been published, but there are no studies analysing wind farms located in Spain. With Spain holding the second place in terms of installed capacity in Europe, and the fifth worldwide, the importance of the Spanish wind power sector is evident. The present study comprehensively analyses the Spanish WT fleet in terms of yearly installed capacity and number of machines installed per technology and capacity ranges. The current fleet age ranges are also analysed. Under this scenario, the present work aims to contribute to the literature with a new failure rate and downtime study, where 75 WTs from six wind farms located in Spain are analysed over a period of 11 years. The most critical assemblies are identified per WT technology type and in general for the whole data set. Finally, the failure rate and downtime results obtained for the Spanish survey are compared against a selection of similar previously published studies.

ered by wind, with Denmark having the largest share (with 41%), followed by Ireland, Portugal and Germany (with 28%, 24% and 21%, respectively) [2]. The yearly share of electricity demand covered by wind in Spain for the past 7 years can be seen in Table 1. Both the wind energy (WE) generation and load remained fairly constant, with variations of around 4-5%. In Spain, wind power is the second largest contributor to the total electricity demand, after nuclear power.
Understanding how WTs fail and the consequences associated with different types of faults [5], that is, the downtime these cause, is a key aspect across the whole WT life cycle [6]: from new developments, appropriate operation and exploitation of existing WTs, both onshore and offshore [7], to extending the expected 20-year lifetime of WTs [8]. However, very few reliability and availability studies of wind turbines have been published in the scientific literature [9][10][11]. Under this scenario, the aim of the present paper is to provide a new failure rate and downtime survey of a set of wind turbines located in Spain. The content of the present study is compared against those previously published from Europe, Asia and America. Furthermore, a comprehensive analysis of the Spanish WT fleet is presented to highlight the relevance of the present study in this context. The remainder of the paper is structured as follows: in Section 2, the wind power sector in Spain is analysed. In Section 3, an overview of worldwide existing failure rate and downtime studies is presented. The characteristics of the wind turbine fleet used for the present study are described in Section 4, and the failure rate and downtime results are shown in Section 5. A selection of previously published studies is compared against the current work, and the outcome is discussed in Section 5.4. Finally, the conclusions extracted from this study are presented in Section 6.

THE SPANISH WIND TURBINE FLEET
Most Spanish wind farms have a rated capacity of 50 MW or below, as historically this was a requirement to receive Feed In Tariffs (FiTs) [12]. The evolution of installed wind capacity in Spain is depicted in Figure 1, in MW (Y-axis) and in number of WTs (bubble size). This evolution was marked by regulatory changes to FiTs, where three changes were performed in 1998, 2004 and 2007, generating significant growth in installed capacity. The growth from 1997 to 2007 (only 10 years) went from 1.5 to 16.7 GW, reaching 22 GW later in 2012. Under the Spanish subsidy support regime, wind farm operators could choose between two options: a guaranteed FiT or a bonus paid on top of the electricity price achieved on the day-ahead market. The last FiT and bonus values in 2012 were 73 € /MWh and 29€/ MWh, respectively [13].
In 2012, there was a moratorium on the installation of new renewable power plants after a repeal of FiTs for new facilities [14], whose impact can be clearly appreciated in Figure 1. In 2014, the incentives were modified (for every renewable power plant), now being mainly for installed capacity with a small energy component associated with maintenance [15], as opposed to the previous incentives per MWh (production) case. These incentives are variable according to the age of the plant and the operation equivalent hours in order to obtain a cost effectiveness of 7.5% [16]. This reform left 6.3 GW wind energy capacity commissioned prior to 2004, more than 300 wind farms, without any type of incentive. Thus, they were obliged to operate with only electricity market revenues. The average investment incentive from 2015 to 2018 was significantly lower than the repealed FiTs. Prior to April 2014, it was common for the day-ahead market to clear at 0 € /MWh in hours of low demand and high wind production. The regulatory changes made it possible for wind farms to participate in balancing markets [17], also eliminating the obligation of renewable energy to bid at 0 € /MWh, and removed the priority dispatch of wind and other renewable energy sources in the day-ahead market [15]. Since April 2014, the minimum day-ahead market price has been 2.3 € /MWh, due to higher wind energy bids [18].
The first wind turbines installed in Spain were squirrel cage (SC), also known as Type I. Shortly afterward, the trend was to install doubly-fed induction generators (DFIG) or Type III, which is currently the most common technology [19], with over   70% of the Spanish fleet being DFIG. Full converter (FC), or Type IV, is the minority technology, covering less than 5% of the Spanish WTs. Figure 2 shows the evolution of WT technologies installed both in terms of capacity and number of machines for currently operating wind farms.
Wound rotor induction generators (WRIG), or type II, are generally rare. In fact, very few commercial Type II machines exist, these being pitch regulated with limited variable speed operation. According to the latest JRC Wind Energy Status Report [20], the share of Type II WTs has been negligible worldwide since 2015. Only 29 Type II WTs exist in Spain, 27 of which were installed in 1997 and the other 2 in 2000. For reasons of simplicity, Type II WTs are merged with Type I in the present study.
The distribution of WT capacity and number of WTs in Spain per age range is shown in Figures 3a and 3b, respectively. As can be seen, there is a small (but not negligible) percentage of WTs operating after the end of their expected 20-year lifetime. A significant proportion is operating with an age range between 15 and 20 years. The 10 to 15 years range is the largest category both in terms of installed capacity and number of machines. Although the 5 to 10 years age range is smaller than 10 to 15 or 15 to 20 in number of WTs, its share with regards to capacity is bigger than the 15 to 20 category. It can be concluded from Figure 3 that over 75% of the Spanish WTs are over 10 years old, representing around 60% of the installed capacity.
The current age range distribution of the Spanish WT fleet is analysed in Figure 4 from two perspectives. First, the distribution of the different technologies installed is shown in Figure 4a. As can be seen, DFIGs show a good spread across the three main age ranges (i.e. 5 to 20 years). This agrees intuitively with the previous Figure 2, where it can be observed that the yearly installations of DFIGs remained fairly constant from 1999 to 2011. The scenario is completely different for SC and FC WTs, being, in fact, the opposite. More than 50% of SC WTs are over 15 years old, whereas over 50% of FC WTs are younger than 15 years old.
Secondly, Figure 4b shows the age range distribution depending on the WT rated power. As can be seen at a glance, over 50% of the fleet are less than 1 MW WT. Of these, the smaller share belongs to the <0.5 MW WTs, which is 15 or more years old. The majority of the WTs with a power range of 0.5 to 0.75 MW are again 15 or more years old. The 0.75 to 1 MW power range has the biggest share. Of these, the majority are between 10 and 20 years old. It can be clearly appreciated that WTs are younger as they scale up, and so, the remaining categories (from 1 MW onward) are mainly formed by WTs aged between 5 and 15 years.

OVERVIEW OF EXISTING FAILURE RATE AND DOWNTIME STUDIES
Three WT reliability review articles that include all publicly available studies in the scientific literature have recently been published [9][10][11]. Combining these article reviews, 20 studies were identified. Their main characteristics are summarised in Table 2, including country, WT fleet, location (onshore/offshore) and available information for each study (failure rate and/or downtime).
Twelve of the studies refer to European countries, six to Asia (four China, one Japan and one India) and two to the United States. As can be seen, there are important differences across the studies in terms of WT capacity ranges, WT age, length of period analysed, population size and location. In most cases, such differences are found also within each single study, in which several power capacities, technologies (geared and gearless, pitch and stall), WT age etc., are mixed and analysed jointly [10]. The majority of the studies include data for 5 years or less and WTs younger than 6 years old (on average). The number of WTs used in previous studies varies largely, between as few as 15 units (MUPPANDAL, India [33]) up to as many as 4285 (WSD, Germany [35]). Similarly, regarding WT capacity, the range goes from 0.1 MW up to 3 MW, for onshore developments (like the present one). This top end reaches 4 MW for studies developed using offshore WTs.
In the present work, a failure rate and downtime study was carried out, where 75 WTs were analysed over a period of 11 years. In order to frame the present work within the existing studies, it has been included in the first row of Table 2. Although the WT capacity and number of WTs included in the study falls within the low range compared to the rest of the studies, the period analysed is long, and includes between 0 and 16-year-old WTs. Thus, for the first time in the scientific literature, a failure rate and downtime study is presented for wind turbines located in Spain. Moreover, unlike the majority of the existing studies, the results are presented separately per WT technology, as well as jointly. Further details of the data used for the analysis are described in the following section (Section 4).

DATA USED FOR THE PRESENT ANALYSIS
The population analysed in the present work is formed by 75 WTs from six wind farms located in Spain over a period of 11 years. These turbines were commissioned from 1995 to 2001. Figure 5 shows the percentage of WTs used per commissioning year, capacity and type. As can be seen, 20% of the turbines are DFIG and the remaining 80% are SC. The WT capacities range between 150 and 900 kW, with 60% of them over 600 kW. Further details such as WT model or the exact wind farm location cannot be provided for reasons of confidentiality.
The period analysed in the present study goes from 2001 to 2011. However, data protection and confidentially agreements existed that have prevented the publication of these results until  very recently, when the embargo period of the mentioned contracts came to its end. Despite the fact that the mentioned embargo period finished, the calculated wind turbine failure rate and downtime results had to be normalised into percentages for publication, explained as follows.
Each of the six wind farm owners provided information on the number of alarms and the downtime associated with each alarm for the individual components. As expected, the data had to be unified and clustered into subassemblies and assemblies so that they could be jointly analysed. Eleven assemblies were defined, some of them including several components, based on the available data. The WT taxonomy defined for the present study for each of the technologies (DFIG and SC) is detailed in Table 3. As can be seen, there is more information available for the DFIG sub-set than for the SC one, as it is a more complex machine.
In the present work, the failure rate was calculated as the total number of failures per wind turbine assembly divided by the total number of wind turbine years, for the whole set of 75 wind turbines. For that, the failure rate per assembly f i was calculated as: where n i,t is the number of failures per assembly in a period of time T , N t is the total number of wind turbines, and T t the total number of hours in the period T . Similarly, the downtime per assembly was calculated as the total downtime of an assembly i in a period of time T , divided by the number of failures of that assembly in that period of time, as: Once f i and d i were calculated for each assembly of the whole data-set, both values were normalised into percentages following the confidentiality agreements signed with the wind farm owners providing the data.

RESULTS
This section presents the results after analysing the failure rate and downtime data collected from the 75 WTs over a period of 11 years. The data were classified into 11 assemblies common for the two technologies, as described in the previous section, some of which have a further breakdown depending on the WT type and the available data. The results are presented separately for each technology, with the breakdown described in Table 3. Then, a combined plot of failure rates and downtime for the whole data set is included, to extract general conclusions. Finally, a comparative with previous (similar) studies is carried out.

DFIG analysis
The failure rate and downtime results when analysing DFIG type WTs are presented in Figures 6a and 6b, respectively. The electric system is the biggest contributor to both failure rates and downtime, with a notably large share (40.08% and 43.11%, respectively). Of this assembly, the highest sub-category is other electric, which includes alarms such as synchronization faults, pre-load fault, instantaneous line tripping, melted fused, high temperature, fan fault, group does not disconnect, ground emergencies, input power fault or other tripping. The next critical assembly in terms of failure rate is the control system (19.85%), closely followed by the generator (15.61%) and the yaw system (9.39%). The remaining categories show lower percentages. The control system is also second in the downtime ranking, contributing with 17.40% of the total, followed, in this case, by the other and yaw system assemblies (9.21% and 8.10%, respectively). The generator is in fifth position in the downtime ranking, with a 5.86% share, closely followed by the pitch system and sensors assemblies (5.48% and 5.38%, respectively).
The gearbox and hydraulic assemblies show low percentages (around 1% and 2%) both for failure rates and downtime. Previous studies show low failure rate gearbox percentages but higher downtime percentages, in contrast to the present study. This and other findings will be further analysed and discussed in Section 5.4. The braking system and structure assemblies are the lowest contributors, both in terms of failure rates and downtime, and are thus the most reliable ones.

SC analysis
The failure rate and downtime analysis for SC type WTs are presented in Figures 7a and 7b, respectively. The results differ from those for the DFIG type. As can be seen in Figure 7a, over 50% of the total failure rates are caused by the control system, mainly under the controlling board reset and negative power alarms. Although the failure rate is very high, the contribution to the total downtime caused by the control system is not, being third in the downtime ranking with a 14.14% share (Figure 7a). In this case, the electric system is second in the failure rate ranking, closely followed by the yaw system and the structure, with 9.49%, 8.17% and 8.11%, respectively. The generator, hydraulic system and gearbox assemblies follow, with mediumlow percentages. Very low percentages are obtained for the remaining assemblies. It is unusual to find the structure category among the top-five contributors to failure rates. This will be further analysed and discussed in Section 5.4.
Analysing Figure 7b, it can be observed that the electric and hydraulic systems together account for nearly 50% of the total downtime. Third, as previously mentioned, is the control system, followed by the generator (with 8.11%). It must be noted that, for both WT types and both the failure rates and downtime, the generator is among the top-five contributors. It can thus be considered a critical component. The gearbox and yaw system show similar percentages (6.75% and 6.07%, respectively). Similarly to the DFIG type WT, the braking system and structure show the lowest percentages with regards to downtime. As previously discussed, however, the SC type WT showed higher structure failure rates. Therefore, in this case, the most reliable assembly appears to be the braking system.
The main differences between DFIG and SC WT types are found in the hydraulic system, structure and gearbox assemblies. The hydraulic system shows much higher downtime in SC (22.72%) than DFIG (1.54%) WT type. The structure failure rate share is higher in SC (8.11%) than in DFIG (0.38%) WT type; however, the contribution of the structure assembly to the total downtime is in the low range for both WT types. With regards to the gearbox, higher percentages are seen, both in terms of failure rates and downtime, for the SC (4.00% and 6.75%, respectively) than for DFIG (1.10% and 2.35%, respectively) WT type. The electric and control systems present significant differences, in terms of percentages, between DFIG and SC WT types. However, these two assemblies are in the topthree contributors of failure rates and downtime for both WT types, being, therefore, the most unreliable categories overall.

Whole data-set
After analysing the results obtained separately for each WT technology and establishing the main differences, a common plot is produced, combining all failure rates and downtime per assembly ( Figure 8). It should be noted that, in the present work, the population of SC type WTs is larger (80%) than the DFIG one (20%), as stated in Section 4. As can be seen, the electric system shows the highest downtime, whereas the control system shows the highest failure rate. Only these two systems, together, account for over 60% of the total failure rates and over 50% of the total downtime. These results might be due to the novelty (at that time) in the power converter, controller and communication systems, with many prototypes (not fully validated) being installed back then. The hydraulic system shows a medium-low failure rate percentage (4.13%), but causes 16.74% of the total downtime, being the second largest contributor. The next sub-assemblies contributing most to both failure rates and downtime are the generator and yaw/flap system (yaw in case of DFIG WT type, flap for the SC WT type). The gearbox holds a central position, being sixth is the downtime ranking (5.51%) and seventh in the failure rate ranking (3.55%). The other, pitch system and sensors assemblies show similar failure rates (around 1%-2%) and downtime (around 5%), being in the bottom half. The failure rate of the structure is nearly 7%, positioned in the fifth place in the failure rate ranking. However, it is the assembly with the lowest contribution to the total downtime, closely followed by the braking system, which shows the lowest failure rate share. The WTs analysed in this study are not located in complex terrain where high turbulence is experienced, thus the drive train is under less stress compared to other WT sites. It should also be considered that around 60% of the WT fleet analysed is less than 10 years old for almost the whole period analysed, and thus less prone to develop critical faults within the drive train.

Comparative with previous studies
The results obtained for the present study show certain similarities, as well as significant differences, for a number of assemblies compared to previous studies carried out in different countries. In order to further analyse them, the most similar studies in terms of WT capacity and age, period analysed and location (onshore/offshore) were compared against the present study ( Figure 9). The selected studies are the last nine in Table 2; of these, only six included information on the downtime. The results obtained in the current study for the pitch, gearbox, mechanical brake, sensors and other assemblies are in the low range compared to the selected studies, both in terms of failure rates and downtime. The control system is well above the top limit compared to the other studies in terms of failure rates; for the downtime, however, it falls within the limits (although slightly above the average). The structure is also above the top limit for failure rates (with a smaller difference), but in the low range for downtime. The remaining components, hydraulic, yaw and electric systems and generator, are around the average failure rate in the selected studies. The yaw system and generator fall around the downtime average too. With regards to the hydraulic and electric systems. However, higher downtime per-centages are obtained compared to the rest of the studies, especially for the electric system.

CONCLUSIONS
With Spain being second in Europe and fifth in the world in terms of cumulative wind power installed, its importance within the wind power sector is clear. A comprehensive study of the Spanish WT fleet was performed, showing that over 75% of the WTs currently installed are over 10 years old, accounting for 60% of the installed capacity. The most common WT technology is DFIG, representing over 70% of the fleet, with the majority of these WTs being over 10 years old. Second in terms of type of WT technology installed is SC (around 25%), which shows an older age range distribution, the majority being over 15 years old, with an appreciable share currently over 20 years old. FC is the minority technology installed in Spain, less than 5%, which is also a much younger fleet. Only 20 reliability studies have been published worldwide, with important differences between them (as discussed in the overview presented); none of these was carried out solely in Spain. Under this scenario, for the first time in the scientific literature, a new failure rate and downtime survey carried out in Spain is also now provided. The study was performed using 75 WTs located in six wind farms, of two different technologies, over a period of 11 years, with ages up to 16 years. The results show that the most critical assemblies are the electric and control systems for both WT technologies. The generator is among the top-five contributors to failure rates and downtime for both technologies, and is thus also critical. The main differences are obtained for the hydraulic system, structure and gearbox assemblies. With regards to the total amount of downtime caused by an assembly, the hydraulic system is much more problematic in SC than in DFIG technology (with 22.72% against 1.54%), which is similar for the gearbox, with 6.75% for SC against 2.35% for DFIG WT types. The results after combining both technologies show that the electric, hydraulic, control and yaw/flap systems and the generator are the most critical assemblies.
Finally, in order to put the present survey in context with the existing ones, our results were compared against a selection of the most similar existing studies. This analysis showed that the failure rate and downtime values obtained for the Spanish survey fall within the limits of previous studies for most of the assemblies. In terms of failure rates, the control system is an exception, with a percentage well above the top limit. Regarding downtime, exceptions are found for the hydraulic system (with a percentage slightly over the top limit) and the electric system (far above the top limit).