You are on page 1of 4

Analysis of the Benefits of Fault-Tolerant

Converters in Offshore Wind Turbines


John Warnock David McMillan Max Parker
Wind Energy Systems DTC University of Strathclyde, University of Strathclyde,
University of Strathclyde, Glasgow, Scotland Glasgow, Scotland
Glasgow, Scotland david.mcmillan@strath.ac.uk max.parker@strath.ac.uk
john.warnock@strath.ac.uk

Abstract – The converter in a large variable- that fault-tolerance is introduced into these
speed wind turbine has amongst the highest turbines.
failure rate of any component, but in an onshore Generally the converter set up within any fully
setting it has not received much attention as rated wind turbine will include a series of parallel
repairs can be carried out quickly and with switching modules which will house the converter
minimal equipment. In an offshore wind farm, components such as the IGBTs, gate drives and
access is by boat or helicopter, which is capacitors. This will typically be connected to a
significantly more expensive and may be general system controller and cooled via an air or
restricted by high winds or wave heights. This liquid cooling system, which will depend on the
means that converter failures offshore will cost turbine manufacturer and the location of the
more to fix, and potentially also lead to a greater turbine.
downtime, leading to loss of revenue. Converters To include fault-tolerance a local controller
can be designed to be modular and to feature will be implemented alongside the converter
fault-tolerance, where the turbine can continue switching modules; this will allow the turbine to
running at reduced power after a fault in one operate under certain faults at a lower rating. Fault-
converter module, or fully redundant, in which tolerance will come with a cost however this paper
the turbine can continue at full power. The investigates the costs and benefits associated with
purpose of this project is to investigate the implementing fault tolerance in several offshore
effects of using such converters on the turbine wind turbines of different ratings.
downtime, maintenance cost and lost revenue. Another method to reduce downtimes, which
has already been implemented in wind turbines, is
Index Terms - Converters, Cost benefit analysis, the concept of converter redundancy [2]. In a
Fault tolerance, Redundancy, Reliability redundant set up there will be one or more
analysis additional converters which are not operating and
will only begin to operate should one of the already
I. INTRODUCTION operational converters fail.

In large offshore turbines there is typically a II. RELIABILITY MODELLING


need for several converters and since the converters
in said large turbines have amongst the highest In any kind of reliability engineering the model
failure rates of all of the turbine components, can be defined as a combination of series and
failures will occur and eventually lead to turbine parallel reliability blocks. If a component is
downtime. This can prove problematic offshore, connected in series then should it fail then the
especially in the larger turbines being proposed in entire system in which it is in series with will fail.
the round 3 wind farms such as Dogger Bank Should a component be connected in parallel and
which is 3,343 square miles in size and at its experiences any kind of failure, provided the other
closest point to shore is 78 miles and at its furthest components which are also connected in parallel do
is around 180 miles from shore [1]. At such a huge not fail then the system will continue to operate.
distance from shore and with such a vast area the Using this, a model of the converters within various
downtimes will depend largely upon the sea different wind turbine set ups could be modelled
conditions and the availability of transport to the and analysed.
location of the fault to carry out maintenance. This
could potentially lead to a massive loss of energy III. WIND TURBINE
production. To combat this it has been proposed CONFIGURATION
Throughout the analysis three different wind be considered is how to model the series
turbine configurations were considered, namely a components of the model [5],
1.5MW turbine, a 3.6MW turbine, which was 𝑅𝑠 = 𝑅1 ∗ 𝑅2 ∗ 𝑅3 ∗ ⋯ ∗ 𝑅𝑛 {2}
modelled on the Siemens 3.6MW offshore wind Since this is a simple multiplication, equations 1
turbine and then altered to incorporate fault- and 2 can be combined to give
tolerance and redundancy, as was the case for the 𝑅(𝑡) = 𝑒 −(𝜆1+𝜆2+𝜆3+⋯+𝜆𝑛)𝑡 {3}
6MW turbine. This representation of the series reliability can be
The 1.5MW turbine configuration was particularly useful when there are multiple
modelled such that without introducing the capacity components as it will save coding and
for redundancy there would be a general controller, computational time. It is also important when
a cooling system and three identical branches each modelling the reliability of a system is how to
of which including a local controller and a represent the parallel reliability [5]:
switching module on the generator side. There 𝑅𝑝 = 1 − (1 − 𝑅1 )(1 − 𝑅2 ) ⋯ (1 − 𝑅𝑛 ) {4}
would also be an identical number of parallel The values of R1 etc. are not to be confused with
branches on the grid side. The switching module those of the series system. Given each of the
essentially includes all of the converter components parallel branches are identical this can be
such as IGBTs, capacitors etc. for this reason the simplified to
reliability block diagram simply includes the 𝑅𝑝 = 1 − (1 − 𝑅𝑠 )𝑛 {5}
reliability of the switching module as opposed to Where n is the number of parallel branches being
overcomplicating the model by introducing the modelled.
components within the switching modules. To add Using the failure rates taken from
redundancy to this configuration one parallel literature with the model of the system described
branch including a local controller and a switching previously the following results were observed for
module was added in parallel with the existing the overall system for a varying number of parallel
branches, both on the generator side and the grid branches:
side. Reliability of Proposed System
In the case of the 3.6MW wind turbine the 0.8

model was built using the configuration of the 0.7 Series System
Siemens equivalent and altered slightly to introduce System with 2 parallel branches
System with 3 parallel branches
fault tolerance. As with the 3.6MW turbine, the 0.6
System with 4 parallel branches
Reliability (Failures/Turbine)

6MW turbine was based on the Siemens equivalent 0.5


System with 5 parallel branches

and modified to include fault tolerance as well as


redundancy. 0.4

Firstly the impact of introducing 0.3


redundant fault-tolerant converters into a system
was analysed such that the reliability of the system 0.2

for each different configuration could be


0.1
established. The failure rates of the various
components had to be established from literature as 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
this is a theoretical system. The failure rates taken Time (Years)

from “Reliability of Megawatt Drive Concepts” by


Klug and Mertens, 2003 [3] had to be converted Figure 1: Reliability of proposed system given a
from the value given as failures in time (FIT) to changing number of parallel branches
failures per turbine per year. This was relatively
straight forward and simply required basic It can be seen that there is a large difference in
arithmetic. Other figures, such as the failure rate of the reliability of a purely series system and one
the switching modules were taken from which has parallel components as expected.
“Investigation of converter failure in wind However it can be seen that the number of branches
turbines” by Fischer et al, 2012 [4]. is increased the increase in reliability is
Using these failure rates and assuming that diminishing and as such there will be a point at
each of the components reliability deteriorates which the increased costs outweigh the increased
exponentially over time the overall reliability of the reliability. For example in this instance this would
system could be calculated. Firstly the individual be at four parallel braches as the difference
component reliabilities had to be calculated using: between the reliability of four branches and five
𝑅(𝑡) = 𝑒 −𝜆𝑡 {1} branches is almost unnoticeable.
Where R(t) is the reliability of a component as a
function of time, λ is the failure rate and t is quite IV. COST BENEFIT ANALYSIS
simply time [3],[5]. Using this value of reliability
the model can start to take shape, what must then Once the optimal number of parallel branches had
been determined a cost benefit could be carried out.
Firstly assumptions had to be made regarding the remainder of the turbines experienced two
costing of various different components; these converter related downtimes. In most cases it was
figures were deduced from figures given in assumed that the nature of the fault would result in
Comparison of Direct-Drive and Geared Generator either fault-tolerant operation or would require a
Concepts for Wind Turbines by Polinder et al. [6]. level of redundancy. Any other faults were
Using these figures the overall converter cost for assumed to result in turbine downtime. It was from
the turbine could be calculated through simple this assumption that the additional uptime and as
arithmetic. The average annual energy capture was such energy captured during this time was
then calculated for a turbine with a normal calculated for each of the converter configurations.
converter set-up, a set-up which included fault- The final and perhaps most important assumption
tolerance and finally including both fault-tolerance made was regarding the downtimes of the onshore
and redundancy. To do this some assumptions had data provided and how to relate these to an offshore
to be made: environment. Whilst it has been discussed
The initial assumption was that the previously that the very naute of the offshore
capacity factor remained constant throughout the environment could lead to several weeks
entire year. Whilst this is clearly not the case the downtime, it was assumed that any failure which
annual capacity factor was taken to be the Danish occurs onshore would lead to three times the
average offshore capacity factor between July 2013 downtime offshore.
and June 2014 of 46.9% [7]. Another of the The average energy capture was then
assumptions made was that as the turbine size and multiplied by the wholesale price of energy which
rating increased, the cost of the general controller was taken to be 181€/MWh [8] and the renewable
and cooling system would scale linearly. Whilst obligation certificates (ROCs) were taken into
the system would be fault-tolerant it is unclear at account at the value of 51.33€/ROC [8] where
which rating it would operate given it was in a offshore ROCs were taken as 2 ROCs/MWh of
fault-tolerant state and as such the converter was electricity produced as per the Department of
taken to be 50% operational under any given fault. Energy and Climate Change bandings [9]. The
Another of the key assumptions was due monetary output of the turbine with each different
to the nature of the data provided. Whether or not a converter configuration was then compared. The
converter – or turbine in general – is capable of difference in converter costs was then subtracted
operating under a fault would of course depend from the difference in energy output given the
upon the nature of the fault and how many faults additional energy capture with the alternative
have occurred. From the onshore data provided configurations and the results shown in figure 2
over 90% of the turbines only experienced one were observed.
downtime related to converter faults whilst the

Benefits of Introducing Fault-Tolerance and Redundancy Offshore


60,000 € 18

50,000 € 15
Number of Parallel Branches
Overall Cost Benefit

40,000 € 12

30,000 € 9

20,000 € 6

10,000 € 3

0€ 0
Fault-Tolerance Fault-Tolerance with Redundancy

Cost Benefit - 1.5MW Cost Benefit - 3.6MW Cost Benefit - 6MW


# pBranches - 1.5MW # pBranches - 3.6MW # pBranches - 6MW
Figure 2: Benefits of Introducing Fault-Tolerance and Redundancy Offshore
V. DISCUSSION VII. REFERENCES

Considering the reliability of any given converter [1] Forewind Annual Review. June 2014 Accessed
set up within a wind turbine it is evident that as the on: 24/07/2014 Accessed from:
number of parallel branches is increased the http://www.forewind.co.uk/uploads/FOR_final%20
observed increase in system reliability is
Annual%20Review_2014.pdf
diminishing, for example, after 4 or 5 branches are
reached in the case of the 1.5MW model. This is
[2] F. Blaabjerg, M. Liserre, K. Ma: “Power
interesting as a 1.5MW turbine would typically
require at least 3 branches to operate at rated Electronics Converters for Wind Turbine Systems”,
power, and could not operate with any less than 2 IEEE Transactions on industry applications. Vol.
branches. Hence it can be deduced that in any wind 48, No. 2 March/April 2012
turbine configuration there will be a point at which
there will be a greater cost than benefit. From the [3] R.-D. Klug, A. Mertens: “Reliability of
results of the reliability calculations and cost Megawatt Drive Concepts”, IEEE 2003
benefit analysis it is the maximum number of
redundant branches appears to be 4/3 of the number [4] K. Fischer, T. Stalin, H. Ramsberg, T. Tjiringer,
of branches required for rated operation. As with J. Wenske, R. Karlsson: “Investigation of converter
any more branches there would be a cost incurred failure in wind turbines”, Nov 2012, pg. 29
whislt any fewer branches would be somehwat
illogical, including one branch of redundancy in a
[5] R. Billinton, A. N. Allan: “Reliability
6MW wind turbine which has 12 branches either
side would incur a much smaller cost but would Evaluation of Engineering Systems”, 2nd Ed, 1993
lead to larger downtimes in the event of a failure of
more than one converter branch. However it is [6] H. Polinder, F. van der Pijl, G.-J. de Vilder, &
important to note that these results rely heavily Tavner, P. J.: “Comparison of Direct-Drive and
upon the assumptions made during the calculations. Geared Generator Concepts for Wind Turbines.”,
The assumptions made, which have been listed IEEE Transactions on Energy Conversion, 21(3),
previously, could have been vastly improved. The pp. 725-733. 2006
capacity factor and ability to access the turbine
could have been improved by altering the data to [7] Capacity factors at Danish offshore wind farms.
include seasonal variatons in weather conditions. Accessed on: 04/07/14. Accessed from:
Given that the wind speed will be higher and more
http://energynumbers.info/capacity-factors-at-
constant during the winter months, the capacity
factor during this time would have been danish-offshore-wind-farms
significantly higher however the failures may have
been more frequent due to increased loading on the [8] European Commission. Energy prices and costs
converters and due to the weather conditions this in Europe. March 2014
would have lead to much greater downtimes for
each of the turbines. [9] Department of Energy and Climate Change,
New Renewables Obligation Certificates (ROCs)
VI. CONCLUSIONS table. September 2012.

The findings of this investiagtion would suggest


that there is a marginal benefit to the
implementation of fault-tolerance and redundancy.
It is thought that a combination of both would be
largely benficial offshore especially during the
winter months where access to the wind turbine for
repairs is significantly longer than that during the
summer months. The next step with this
investigation would be to use a similar method with
a slightly less crude method of estimating the
offshore downtime and obtaining more results for
more wind turbines. This could be particularly
useful when looking at the upcoming round 3
offshore wind farms.

You might also like