See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/282504299

Cyclic Load Profiles for Offshore Wind Farm Cable Rating

Article  in  IEEE Transactions on Power Delivery · August 2015

DOI: 10.1109/TPWRD.2015.2469538

CITATIONS READS
7 346

5 authors, including:

Simon Catmull Richard D. Chippendale

RES Offshore University of Southampton
3 PUBLICATIONS   23 CITATIONS    24 PUBLICATIONS   67 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Simon Catmull on 01 April 2016.

The user has requested enhancement of the downloaded file.


Cyclic Load Profiles for Offshore Wind Farm
Cable Rating
S. Catmull, R. D. Chippendale, J. A. Pilgrim, Member, IEEE, G. Hutton and P. Cangy
1
Abstract— By its nature, the power generated by offshore wind
farms varies substantially with time. Despite this fact, the cable
systems which connect such wind farms back to the onshore grid
are typically sized based on a maximum, continuous current
rating. In order to develop techniques by which the sizing of
such cable connections might be optimized, this paper presents a
method for deriving equivalent cyclic load curves from realistic,
time variant generation data. These cyclic load profiles may be
used with the conventional IEC 60853-2 cable rating calculation,
allowing the cable to be sized for a more realistic duty cycle. The
example calculations presented show that this reduction in
conservatism permits the use of smaller cable sizes, helping to
drive down project costs. The implications of using such methods
are discussed, highlighting the tradeoff between cost and
conservatism.
Index Terms-- Power cable insulation, Power transmission, Statistical
analysis, Statistical distributions, Underwater cables, Wind energy,
Wind energy integration, Wind farms.
I. INTRODUCTION
H IGH voltage AC subsea cables represent a significant
proportion of the capital expenditure associated with
offshore wind farms. As wind farms go further offshore
and the distance to the landfall of the export cable increases,
the proportion of total capital expenditure becomes greater.
Raising the transmission voltage or moving to HVDC
transmission both offer the possibility of reducing the number
of cables or, at the least, minimising the conductor size of the
cables needed to carry the rated output of a wind farm. An
alternative approach, and one potentially requiring less of a
technical advance, is to take advantage of the inherent
variability of the yield from a wind farm. Numerous published
studies have proposed different methods to predict short term
and long term wind farm yields from weather forecasts [1, 2,
3]. Whilst these studies have investigated the wind farm yield,
they have not considered how the fluctuating yield impacts the
thermal rating of the export circuit. Therefore another
approach is to find the appropriate cable size according to the
load that they are likely to experience.
A. Conventional Cable Sizing Approach
The conventional approach to sizing cable systems for
offshore wind farms has been to use a steady state, continuous
1
R. D. Chippendale, J. A. Pilgrim and P. Cagny are with the Tony Davies
High Voltage Laboratory, University of Southampton, SO17 1BJ, UK (e-mail:
jp2@ecs.soton.ac.uk).
S. Catmull and G. Hutton are with RES Offshore, Kings Langley,
Hertfordshire, WD4 8LH, UK (email: simon.catmull@res-offshore.com).
1
current rating. This is defined as the current which, if applied
on a continuous basis, would ensure that the conductor
temperature reaches but does not exceed its operational limit.
For the majority of cross-linked polyethylene (XLPE) cables
in use today, this will equate to 90°C, the maximum
temperature at which such cables should be operated to give
an acceptably low rate of thermal ageing of the insulation.
If the transient thermal behavior of common export cable
systems is considered, the cable thermal response to a step
change in load is relatively slow. This is particularly true for
sections of the route which are the most thermally limiting,
such as deeply buried cable sections at landfalls. The main sea
section of the route is frequently not thermally limiting due to
lower burial depths and better thermal resistivity compared to
soils at the landfall. In the case studied here, the rating of the
cable in the sea bed is 40% higher than at the deeply buried
landfall section. It will be shown in Section V that achieving
the steady state maximum temperature can take 20 years.
Given the inherent intermittency of wind generation, such
steady state conditions will never in reality be attained. This
leads to the cable being sized for an overly worst case
scenario, which can result in larger conductor sizes than are
truly necessary.
One previous approach to reduce this conservatism is to
dynamically predict the transient thermal rating based on the
local weather forecast [4, 5, 6]. This approach has been used
for overhead lines [7, 8], where such approaches have
obtained 120% increase in rating [7]. There is however an
added complexity of applying these methods to buried cables,
due to importance of accurately predicting the thermal
properties of the burial environment [5]. Whilst these methods
could be used for wind farm export cables, they do not take
into account the fluctuating generation of wind farm itself, and
assume a maximum possible circuit load.
B. Overview of Proposed Methodology
This paper begins by defining an example wind farm,
followed by a method for the statistical analysis of a wind
speed time series to obtain an equivalent cyclic load profile for
use in IEC60853-2 calculations [9]. Although it is a
generalization to approximate wind farm output by a
cyclically varying load, it is a closer reflection of reality than
assuming a constant output. This equivalent load cycle may
then be used to size the export cable, with due consideration
for the thermal behavior of the cable system itself. Such wind
speed time series are now commonly used in the evaluation of
potential wind farm sites, hence the data is readily available at
the design stage of the project. The final part of the paper
compares the predicted cyclic load against traditional ratings,
 2
Insulation screen
analyzing the importance of the wind speed assumptions. Insulation Water Tape
Conductor screen Sheath
Conductor Inner Sheath
II. EXAMPLE EXPORT CIRCUIT
The method will be demonstrated by carrying out the initial
sizing of the export system for a hypothetical 300MW
offshore wind farm. Long term historical wind speed data will
be obtained from a meso-scale wind model of the UK and its
surrounding waters [10]. The wind speed data shall be
converted into a time series of wind farm power output using a
generic wind turbine power curve with an appropriate number
of turbines assumed. Account shall be taken of the wake losses
associated with the turbine array, but for the sake of simplicity
it will be assumed that there will be no losses associated with Filler
Armour
the reliability of the turbines or the balance of plant. Serving
A. System Architecture
The transmission architecture will be assumed to be Fig. 2. Simplified cross section of 3 phase export cable
composed of two offshore transformers to step up the array TABLE I
voltage of 33kV to a transmission voltage of 132kV and that Cable dimensions and material properties
an export cable per transformer would be required. The time Volumetric
Thermal
Heat Radius
series of the wind farm power output at 132kV will be Geometry Material Conductivity
Capacity (mm)
converted to a current on the basis that the landfall is the (Wm-1K-1)
(MJm-3K-1)
thermally limiting location and that the reactive power Conductor Copper 320 3.45 20
compensation has been designed so that the power factor at Conductor Semicon
0.5 2.4 21.5
the landfall minimises the power loss. A sketch of the wind Screen XLPE
farm transmission system is shown in Fig. 1. Insulation XLPE 0.286 2.4 38.5
Inulation Semicon
0.5 2.4 40
Screen XLPE
Unknown
Water Tape 0.286 2.4 41.5
Polymer
Sheath Lead 35.3 1.45 43.8
Inner
Semicon PE 0.5 2.4 46
Sheath
Bedding Polypropylene
0.123 1.9 101
and filler yarn
Armour Steel 18 3.8 106.5
Serving PE 0.2 1.7 111

TABLE II
Cable System Properties
Fig. 1. Export system sketch. Parameter Value
Conductor Strand Fill Factor 0.8
B. Cable Systems Conductor Resistivity 1.72 x 10-8 Ωm
Conductor kp 0.8
This study considers a wind farm export system with two Conductor ks 1
export cables which consist of 132kV 3 phase XLPE insulated Conductor Coefficient Of Resistance 3.93 x10-3 K-1
SL-type (each phase in separate lead sheath) cable with Sheath Resistivity 2.14 x 10-7 Ωm
1000mm2 conductors. The three phases are protected and held Sheath Temp Coefficient Of Resistance 4x10-3 K-1
Armour Resistivity 1.38 x 10-7 Ωm
together by layers of PE and steel armour wires. A cross Armour Temp Coeff Of Resistance 4.5 x 10-3 K-1
sectional illustration of the cable layout is presented in Fig. 2. Armour kp 1
The cable dimensions and thermal material properties are Armour ks 1
presented in Table I. The materials properties are considered Relative Permittivity Of Insulation 2.5
as per relevant IEC values, summarized in Table I and II [9]. Insulation Tan Delta 0.001
Voltage of cable 132kV
The filler region is typically comprised of a mixture of plastic Frequency 50Hz
ropes with air voids between the strands. Therefore the
material properties for the filler region are given by a TABLE III
summation of individual material properties weighted by their Seasonal soil and ambient conditions
volume fraction. The burial depth at the landfall is 4m with a Soil
Ambient Air
circuit spacing of 20m. The seasonal soil and ambient Thermal Volumetric
Season Temperature
Conductivity Heat Capacity
conditions are presented in Table III. (Wm-1K-1) (MJm-3K-1)
[ºC]
Summer 0.83 2 15
Winter 0.95 2 10
Dry (above 50ºC) 0.4 1 -

C. Power Generation Time Series
A time series of the current at the landfall was obtained by
modeling the power output from a large array of wind turbines
with respect to an 11year wind speed and direction time series.
Estimates of wind speed variation across an offshore wind
farm site on an hourly time resolution were obtained through
the use of the WRF Mesoscale Model [10]. This industry-
standard model uses a combination of measured data and
meteorological theory to estimate long-term mean wind speeds
in each direction sector for points on a grid of 2km resolution.
By interpolating between these grid points the wind speed at
each turbine location can be estimated. The mechanical power,
Pturbine(t), captured by a turbine is a function of the wind speed
at hub height, given by
1
𝑃𝑃𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 (𝑡𝑡) = 𝐶𝐶𝑝𝑝 𝜌𝜌𝜌𝜌𝑣𝑣 3 (𝑡𝑡) (1)
2 cable system
Where Cp is the turbine performance coefficient, ρ is the air
density, A is rotor area and 𝑣𝑣(t) is the wind speed. The
velocity deficit in wind speed downstream of a turbine, also
known as its wake, is accounted for by the Ainslie eddy model
[11] with overlapping wakes from multiple turbines calculated
using the maximum wake deficit and Root-Sum-Square
methods in combination, as outlined in [12]. The wind speed
time series was entered into the wind farm model to obtain the
real power, P(t), as a function of time (t). The power was
converted to a current time series, I(t), using
𝑃𝑃(𝑡𝑡)
𝐼𝐼(𝑡𝑡) = (1 + ℎ𝑓𝑓 ) (2)
√3𝑉𝑉𝑟𝑟𝑟𝑟𝑟𝑟 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
Where Vref is 132kV, cos𝜙𝜙 is the power factor and hf is the
harmonic current allowance. The typical assumed values for
these parameters are stated in Table IV.
TABLE IV
Assumptions used in derivation of current time series
Parameter Value
Power Factor at Landfall (cos φ ) 0.98
Harmonic Current Allowance (ℎ𝑓𝑓 ) 2.5%
Transformer losses 0%
Wind farm reliability 100%
Wind farm rating (P(t)) 310 MW
Load division between cables 50%
The variation of current with respect to the wind speed for the
duration of the time series is shown in Fig. 3. The maximum
output from the wind farm sets a target rating for the export
system of 710A. This would be the typical value of current
used to size the cable system under the conventional steady
state assumptions.
III. CURRENT RATING CALCULATIONS
This section presents a summary of the thermal current
rating methods used for the analysis of Section VI.
A. IEC 60287
The international standard method for calculating the
continuous seasonal ratings for cables installed in a variety of
installations is IEC 60287-1 [13]. This method has proven to
be suitable for predicting the continuous seasonal rating of
buried cables, however its analytical nature does force some
assumptions [14].
3
Fig. 3. Single cable output load as function of wind speed for a dual export
B. IEC 60853-2 Cyclic Ratings
The IEC standard 60853-2 presents an analytical method to
calculate the cyclic rating for a buried cable. The method
requires a continuous rating to be calculated using IEC 60287,
which is then multiplied by a cyclic rating factor M, defined as
𝑡𝑡𝑚𝑚𝑚𝑚𝑚𝑚 −1 −1�2
𝜃𝜃𝑅𝑅 (𝑖𝑖 + 1) 𝜃𝜃𝑅𝑅 (𝑖𝑖) 𝜃𝜃𝑅𝑅 (𝑡𝑡𝑚𝑚𝑚𝑚𝑚𝑚 )
𝑀𝑀 = � � �𝑌𝑌𝑖𝑖 � − �� + 𝜇𝜇 �1 − �� (3)
𝜃𝜃𝑅𝑅 (∞) 𝜃𝜃𝑅𝑅 (∞) 𝜃𝜃𝑅𝑅 (∞)
𝑡𝑡=0
Where 𝜃𝜃𝑅𝑅 (𝑡𝑡) is the conductor temperature rise above
ambient in ºC at the next time step (𝑖𝑖 + 1) or at the end of the
cyclic profile. The given standard equation only takes into
account the 6 hours before the peak value of the cyclic curve,
thus the equation shown in equation (3) is modified to take
into account ‘i’ hours for the full length of the load cycle for
the time considered. The final terms which need to be defined
are the loss-load factor (μ), which is expressed as the average
of all the ordinates of the loss factor load cycle (Yi), by
𝑡𝑡𝑚𝑚𝑚𝑚𝑚𝑚−1
1
𝜇𝜇 = � 𝑌𝑌𝑖𝑖 (4)
𝑡𝑡𝑚𝑚𝑚𝑚𝑚𝑚
𝑡𝑡=0
The value of Y is calculated by taking each element in the
cyclic load and dividing it by the maximum of the cyclic load
and squaring the value. The method proposed in (3) and (4)
are for a non-drying out case, and [9] proposes an equivalent
equation to include the effect of soil dry out.
C. Finite Element Analysis
Finite Element Analysis (FEA) has previously been shown
to be a useful tool for the rating of cables, allowing for better
representation of physical processes than the analytical IEC
standards [15, 16]. This paper considers a de-coupled
transient thermal response of the cable due to the thermal
losses from the applied load. Heat sources within the cable are
defined using conventional IEC equations following the
principles outlined in [13]. Fig. 4 shows an illustration of the
two dimensional (2D) landfall geometry implemented, along
with the external boundary conditions. Due to the symmetry
plane between the two export cables, the FEA model only
needs to consider one half of the circuit i.e. only one cable.
The right hand boundary is set as thermal insulation, as this
boundary is far enough away from the cable that there is no
heat flux passing through the boundary.

Fig. 4. Numerical geometry with boundary conditions
Both the top and bottom boundary are held at a constant
temperature, with the bottom boundary being 10°C and the top
boundary temperature is defined by the ambient seasonal
temperature [14].
IV. CYCLIC LOAD DERIVATION
This section presents the method by which cyclic load
profiles are derived for subsequent use in rating calculations.
A. Load Profile Shape
A cyclic load is characterised by the ratio of time spent at
high power to low power and the magnitude of high and low
loads as is illustrated in Fig. 5. There are two main steps for
deriving an equivalent cyclic load for a wind farm. The first is
to determine a threshold by which a time series can be
separated into periods of high and low load, this is based on
the assumption that the cable temperature will rise more
rapidly than the burial environment when the cable is subject
to load. The second is to determine the duration of periods of
high load relative to low, based on the ratio of the average
period of high to low load for a long term load time series.
A. Calculating Current Thresholds
A threshold for separating the high loads (Irate) and low load
(Ilow) was determined through calculation of the steady state
temperature rise in a cable over a range of currents. The low
current level (Ilow) was initially determined by calculating the
load on the 1000mm2 cable which produces a steady state of
30ºC, which corresponds to a current which is mainly heating
the cable and not having a significant impact on the
temperature of the burial environment, which is demonstrated
in Fig. 6. The impact of assuming this 30ºC threshold is
investigated in detail in Section V part C. The steady state
conductor temperature due to a range of wind speed and
current load are presented in Fig. 6. This figure shows a
30 °C conductor temperature is obtained from a load of 302 A,
which from Fig. 3 equates to an approximate wind speed of
9 ms-1. This sets the wind speed bounds for a high load
between 9 ms-1 and 25 ms-1, the wind speed at which the
turbine cuts out as a form of mechanical protection. Wind
speeds in this range are assumed to generate an output load of
Irate. Any wind speed below this bound is considered to result
in a low load (Ilow) of 302 A, which is equivalent to a steady
state current that would cause a 30 °C temperature. With the
4
lower current level fixed, the high load current (Irate) is then
varied to find the thermal rating of the cable due to the
application of the cyclic load.
Load (A)
Δtrate
Irate
Ilow
Δtlow
Time
Fig. 5. Generic cyclic load profile
Fig. 6. Circuit load for various Conductor temperatures
B. Determining Durations of Spells
The duration for which the high (Irate) and low (Ilow) loads
are likely to persist was determined through a simple analysis
of the long term wind time series and the amount of time spent
within and without a 9 to 25ms-1 range. In addition to basing
this analysis on the hourly wind speed data from the WRF
model, the use of a rolling average filter to smooth the wind
speeds by removing short term fluctuations before assessing
persistence was investigated. The length of the filter affects
the persistence of the series; the longer the averaging period,
the smoother the series and so the longer the load cycle. An
appropriate filter length for smoothing was determined
through an assessment of the underlying structure of the wind
data in terms of its serial correlation. Fig. 7 shows the
autocorrelation function of the wind speed data and the clear,
cyclic, pattern which corresponds to seasonal variations of
wind speed.
This seasonality was removed to expose the underlying
structure via a multiplicative decomposition by moving
averages which takes the form of
𝑣𝑣(𝑡𝑡) = 𝑆𝑆(𝑡𝑡) 𝑇𝑇(𝑡𝑡) 𝑅𝑅(𝑡𝑡) (5)
Where, v is the wind speed and S, T and R respectively
denote the seasonality, trend and irregular components of the
time series. Removing the seasonality also gets to a structure
that remains constant throughout the year. Further details of
the method can be found in [17].
The effect on the autocorrelation function of removing
seasonality in the time series is shown in Fig. 8. The short-
term persistence of the deseasonalised wind data is then
 5

assessed via the order of a simple autoregressive time series by considering the probability density function of the wind
model, taking the form speed durations, which is shown in Fig. 10.
𝑣𝑣(𝑡𝑡) = 𝛼𝛼1 𝑣𝑣(𝑡𝑡1 ) + 𝛼𝛼2 𝑣𝑣(𝑡𝑡2 ) + … 𝛼𝛼𝑛𝑛 𝑣𝑣(𝑡𝑡𝑛𝑛 ) + 𝜖𝜖𝑡𝑡 (6)

Fig. 9. Partial autocorrelation of deseasonalised wind speed data.

TABLE V
Fig. 7. Autocorrelation of wind speed data. Distribution of wind speed with different filter lengths
Persistence of periods Persistence of periods
Filter
Length
inside 9-25 m/s range below 9 m/s
(hrs) Mean Max Mean Max
length (hr) length (hr) length (hr) length (hr)
1 13.62 191.00 13.58 293.00
3 21.44 259.00 21.49 387.00
6 28.83 333.00 29.03 448.00

From Fig. 10, it is clear that the wind speed threshold of

9ms-1 is approximately the peak of the wind speed distribution
with occurrence of wind speeds both above and below this
threshold approximately equal. That means durations are also
approximately equal, which indicates that the persistence is
stable across the range of observed wind speeds

Fig. 8. Decomposition of wind speed data

The order of the autocorrelation, n in (6), was identified

through visual inspection of the partial autocorrelation
function of the deseasonalised time series, as shown in Fig.9.
This suggests that the data are at least an autoregressive
process of order 1, with potentially significant lags at orders 2,
3 and 4, but no clear structure beyond lag 1. On this basis
filter lengths, of 1hr (corresponding to no smoothing or
persistence) and 3hr (corresponding to persistence over 3
hours) have been chosen. An additional filter of 6hr was used
to further demonstrate the effect of an increased filter length
and to capture the maximum persistence in the time series that
could be interpreted from the data. Carrying out a persistence
analysis on each of the filtered time series returned high and
low load durations with the characteristics detailed in Table V.
As the moving average filter length increases the persistence
of the series also increases, as expected. The ratio of mean
periods inside the wind speed limits to those outside them is
consistently 1:1 for each of the filter lengths. This is an artifact
of the chosen wind speed threshold which is ultimately a result
of the steady state current that would cause a 30°C
temperature rise in the conductor. This can be demonstrated
Fig. 10. Probability density of wind speed
Finally it is important to consider the frequency of
occurrence of periods above the 9ms-1 threshold, as shown in
Fig.11. From this figure, it is apparent that the majority of
occurrences when the wind speed is above the 9ms-1 threshold
are for durations less than, or equal to the mean (Table V), i.e.
as a measure of the persistence in time of high load periods the
mean gives a conservative estimate. This is important since
the frequency of occurrences where the mean duration is
exceeded and therefore the likelihood of a steady state thermal
rating of the cable being exceeded is small but probable over
the operational lifetime of the modeled offshore wind farm.
V. CALCULATION OF CYCLIC RATINGS
In the previous section three cyclic load profiles were
derived on the basis of the following:
1. Steady state load that caused a 30°C temperature rise.
2. The ratio of the mean length of periods where the wind
speed would cause a greater than 30°C temperature rise to

those that caused less.
3. Calculate the rated current (Irate) on the basis of Ilow and
cycle duration.
A. Steady State Ratings to IEC60287
The continuous seasonal ratings at the landfall, calculated
using IEC 60287, are 604 A in summer and 660 A in winter.
Using (2), it is possible to calculate that the total output load
from the considered wind farm is 1420 A. Assuming this load
is equally divided between the two export cables, each cable
needs to carry 710 A. Based on IEC 60287 continuous ratings
above, it is evident that under this conservative scenario, three
rather than two export cables at 1000mm2 would be required
to avoid exceeding the thermal limit. The incremental cost is
of the order of £1 million per route mile.
Fig. 11. Probability density of wind speed durations above the 9ms-1
wind speed threshold when the filter length is 1hr
B. Finite Element Results
Using the 2D FEA model the transient response of the
conductor temperature due to a single step load can be
investigated. Using the summer continuous rating from IEC
60287 as the step load, the conductor temperature response is
presented in Fig.12. From Fig.12 it is evident that there is an
initial rapid increase in temperature, up to approximately 89%
of the thermal limit after 1.75 years. After this, there is a much
slower increase in temperature as the soil heat capacity begins
to dominate. It is therefore evident that with a step load at the
continuous rating, it would take the circuit approximately 20
years to reach a steady state temperature.
Fig. 12. Conductor temperature response due to step load of 604A
With the application of a cyclic load, the same global
conductor temperature trend as presented in Fig.12 is visible.
However due to the comparably high frequency cyclic load
profile, there is also a short duration temperature oscillation
within the cable. The quasi-steady state conductor profile due
to the application of the cyclic load profile with a filter length
6
of 6 hrs is presented in Fig.13, which shows that there is a
continual temperature oscillation due to the cyclic load. The
model is assumed to be in a quasi-steady state when the peak
conductor temperature for each oscillation is approximately
constant with neighbouring peaks. The quasi-steady state
solution occurs after approximately 15 to 20 years, with the
oscillation height being approximately 26ºC.
Fig. 13. Conductor temperature response due to a cyclic profile with a 6hour
filter length.
C. Comparison of cyclic rating methods
Table VI compares cyclic ratings obtained from both the FE
model, which provides a full transient temperature
distribution, and the simpler IEC method which uses (3) to
obtain a multiplier of the continuous rating. It is also possible
to calculate a full time domain solution using IEC 60853-2 to
avoid the simplifications inherent in the use of (3).
TABLE VI
Comparison of seasonal cyclic rating for various filter lengths predicted by
FEA and IEC 60583
Filter Summer Rating, Irate (A) Winter Rating, Irate (A)
length
FEA IEC 60853 FEA IEC 60853
(hr)
1 762 738 828 815
3 745 728 819 805
6 735 725 814 801
Table VI shows that as the filter length increases there is a
decrease in cyclic rating from both IEC and FEA. This is a
caused by the increased high and low levels as the filter length
is increased. The result indicates that as the duration of the
high (Irate) and low (Ilow) load level decreases, the current
rating will increase. Further it is apparent that the ratings
obtained from FEA are 1% to 3% greater than by IEC60853,
primarily due to the more detailed model of the temperature
distribution within the soil. Such differences are within the
bounds of uncertainties associated with measuring the soil
thermal properties. By comparing the cyclic ratings in Table
VI with the continuous ratings in Section V A, it is apparent
that by using a cyclic rating there is a 22% increase in
maximum rating. Whilst the cyclic ratings predict an
increased peak rating it is important to ascertain if the
temperature from the representative wind farm load exceeds
the conductor temperature from the cyclic rated profile. The
temperature exceedance is investigated through the difference
in conductor temperature caused by the realistic wind farm
load minus the temperature caused by the rated cyclic load.
For this comparison a five year period is chosen and the
conductor temperature difference is presented in Fig.14.
 7

threshold reaches 10 ms-1 there is a decrease in cyclic rating,

which is caused by the increase in low level load, which as the
low duration exceeds the high duration, begins to dominate.

VI. IMPLICATIONS FOR WIND FARM DESIGN

The method that has been outlined in this paper is intended for
use during the Front End Engineering and Design (FEED) of
an offshore wind farm. It is dependent upon a limited amount
of data about the burial conditions of the cable and wind speed
time series data for the potential wind farm. Data that it is
anticipated would be available at the FEED stage of most
Fig. 14. Conductor temperature difference between rate cyclic load profile and
the wind farm results (6hr filter length, summer season)
wind farms.
TABLE VII
The temperature difference in Fig. 14 demonstrates that the Cyclic load profile duty durations for various wind speed thresholds (filter
conductor temperature predicted by the rated cyclic load is length is 3hrs)
generally greater than that from the real wind farm load Lower Wind Low
Mean duration Mean duration
(negative values). Despite this there are instances where the Speed Threshold current
Above (days) Below (days)
temperature from the wind farm load is greater. Fig. 15 plots (ms-1) level
an extract of the thermal time series from two FEA models. 7 1.27 0.60 120
7.5 1.15 0.66 150
The red curve shows the thermal response to the simplified 8 1.04 0.72 182
load cycle, while the blue curve shows the thermal response to 8.5 0.95 0.80 221
the full time series data from Section II C. The periods where 9 0.89 0.90 262
the blue curve shows a higher temperature than the red occur 9.5 0.84 1.01 311
during the low load levels within the cycle and hence the 10 0.79 1.13 322
10.5 0.71 1.22 360
conductor temperature remains at least 9 ºC below the thermal
11 0.67 1.39 477
limit. Although this demonstrates that the equivalent load
This method would also be dependent upon a commitment to
cycles method is still conservative, the rating calculations may
using an alternative rating methodology during the operational
be performed much more quickly, without the need to rely on
stage of the wind farm. This could take the form of short-term
FEA tools.
transient ratings, dynamic ratings or dynamic curtailment of
the wind farm output. At some point there would be periods
of high wind speed in excess of the high load duration used in
the cyclic load, potentially causing the temperature threshold
of the cable insulation to be exceeded. For a reduced cable
size and dynamic operation to be a viable choice the likely
loss of production would need to be weighed up against the
reduction in capital expenditure. This could be done on the
basis of probability of exceedance, as is commonly done for
elements of the structural design of an offshore wind farm that
are driven by wind and wave loading. Return periods for wind
Fig. 15. Conductor temperature comparison for a cyclic and a realistic load.
speed durations longer than the high load (Irate) duration can be
D. Impact of wind speed threshold on Cyclic Ratings obtained through statistical analysis of long term time series in
In previous studies the wind speed threshold has been set as the same way that the persistence analysis was done to obtain
9ms-1. The impact on the cyclic rating due to variations of the cycle durations.
wind speed threshold has been investigated. Since the wind
speed threshold is changed, the cyclic profiles need to be re-
calculated based on the method proposed in Section IV. The
variation of cyclic profile with respect to wind speed threshold
is presented in Table VII. As the wind speed threshold
increases, the ratio of mean duration above to mean duration
below decreases i.e. duration above decreases whilst the mean
duration below increases. Such a variation in the balance of
the load profile will cause a variation in cyclic rating, which is
presented in Fig.16. This shows that as the wind speed
threshold increases up to 9 ms-1 there is an increase in
continuous rating. This is explained by the mean duration of
the high load decreasing, which results in shorter high load
applications and so less time for the temperature of the Fig. 16. Cyclic rating from IEC 60853-2 as the wind speed threshold is varied
conductor to increase due to the high load. However as the
Controlling the output of an offshore wind farm on the

basis of the cable temperature is within the reach of present
technology. It is widely understood within the offshore wind
industry that the cost of generation needs to reduce for it to
continue to be a viable part of the generation mix. This paper
has outlined a method by which the cost of a significant
contributor, the export cables, to the total capital expenditure
of an offshore wind farm may be reduced.
VII. CONCLUSIONS
A method has been presented for the sizing of a wind farm
export cable based on likely generation profiles. This is
achieved through the use of cyclic ratings, where the cyclic
load profile is derived through the statistical analysis of long
term wind time series data. Comparisons between the steady
state and cyclic rating were shown for a thermally challenging
location section of the subsea cable route for an offshore wind
farm. The cyclic rating calculated using IEC60853 was shown
to be in good agreement with results obtained through 2D
FEA. The ampacity increase of 22% as a result of considering
cyclic ratings over the traditional continuous rating method is
sufficiently great that the target rating could be met with two
export cables of conductor cross-sectional area of 1000mm2.
This would remove the need to add an extra cable, or change
the conductor size for all, or part, of the cable. Although the
use of such methods have a clear benefit in terms of reduced
costs, it must be noted that a suitable curtailment strategy must
be in place to mitigate the risk that a period of high power
generation could be longer than predicted. To fully assess the
cost reduction achieved, it would be necessary to compare the
capital saving with lost production due to output curtailment,
for which further development is presently underway.
VIII. REFERENCES
[1] J.W. Taylor, P.E. McSharry, R. Buizza, "Wind Power Density
Forecasting Using Ensemble Predictions and Time Series Models,"
Energy Conversion, IEEE Transactions on , vol.24, no.3, pp.775,782,
Sept. 2009
[2] M.C. Mabel, E. Fernandez, "Estimation of Energy Yield From Wind
Farms Using Artificial Neural Networks," Energy Conversion, IEEE
Transactions on , vol.24, no.2, pp.459,464, June 2009
[3] R. Amiri, Wang, Bingsen "A generic framework for wind power
forecasting", IECON 2011 - 37th Annual Conf on IEEE Industrial
Electronics Society, pg 796 - 801
[4] J. Heckenbergerova, J. Hosek, "Dynamic thermal rating of power
transmission lines related to wind energy integration," Environment and
Electrical Engineering (EEEIC), 2012 11th International Conference
on, pp.798,801, May 2012
[5] R. Olsen, G.J. Anders, J. Holboell, U.S Gudmundsdottir, "Modelling of
Dynamic Transmission Cable Temperature Considering Soil-Specific
Heat, Thermal Resistivity, and Precipitation," Power Delivery, IEEE
Transactions on , vol.28, no.3, pp.1909,1917, July 2013
[6] G.J. Anders, A. Napieralski, M. Zubert, M. Orlikowski, "Advanced
modeling techniques for dynamic feeder rating systems," 37th Industry
Applications Conference, vol.2, pp.1012-1019 Oct. 2002
[7] Huu-Minh Nguyen; Lilien, J.-L.; Schell, P., "Dynamic line rating and
ampacity forecasting as the keys to optimise power line assets with the
integration of res. The European project Twenties Demonstration inside
Central Western Europe," Electricity Distribution (CIRED 2013), 22nd
International Conference and Exhibition on, pp.1- 4, June 213
[8] J.K Raniga, R.K. Rayudu, "Dynamic rating of transmission lines-a New
Zealand experience," Power Engineering Society Winter Meeting, 2000.
IEEE , vol.4, pp.2403,2409
[9] Calculation of the cyclic and emergency current rating of cables. Part 2:
Cyclic rating of cables greater than 18/30 (36) kV and emergency
View publication stats
8
ratings for cables of all voltages, IEC60853-2, 1989
[10] Weather Research and forecasting (WRF): ARW User Guide V3,
September 2014.
[11] J.F. Ainslie, “Calculating the flowfield in the wake of wind turbines”,
Journal of Wind Engineering and Industrial Aerodynamics, vol 27,
Issues 1–3, pp 213-224 Jan 1988.
[12] G. Habenicht, “Offshore Wake Modelling” presented at the 10th annual
Renewable UK Offshore Wind conference, Liverpool, England, Jun.
29-30, 2011.
[13] Electric cables - Calculation of the current rating - Part 1-1: Current
rating equations (100 % load factor) and calculation of losses -
General, IEC60287-1-1, 2006.
[14] Current Ratings for Cables, TGN(T)67, 1996.
[15] D. Swaffield, P. L, Lewin, S. Sutton, "Methods for rating directly buried
high voltage cable circuits", IET Generation, Transmission &
Distribution, 2008, page 393-401
[16] Electric cables - Calculation of the current rating - Part 2: Thermal
resistance, Section 2.1 Calculation of thermal resistance, IEC 60287-2-
1, 2006
[17] E. D. Meyer. (2014, Nov 08). R-Manual. (3.2.0) [Online]. Available:
https://stat.ethz.ch/R-manual/R-devel/library/stats/html/decompose.html
IX. BIOGRAPHIES
Simon Catmull received the Bachelor’s degree in
Mechatronics from the University of Sussex in 2003 and
then went on to obtain a Master’s degree in Renewable
Energy Systems and Technologies from Loughborough
University in 2010. He has been employed by RES
Offshore since 2010 and currently works as a Project
Engineer supporting the development, construction and
operation of wind farms on- and offshore.
Richard D Chippendale received the Master’s degree in
Physics from University of Southampton in 2009. He
pursued his doctoral degree at the same university,
investigating the thermal chemical degradation to
composite materials due to a lightning strike, for which
he was awarded a PhD in 2013. He is currently employed
as a Research Fellow, working on a variety of projects
related to high voltage cable systems. He is a member of
IEEE DEIS and an associate member of the IOP.
James A. Pilgrim (M’09) received the Bachelor’s degree
in electrical engineering from the University of
Southampton in 2007, subsequently gaining his PhD in
2011 from the same institution. He joined the academic
staff of the University in 2012. His research interests
include all aspects of high voltage cables and associated
insulation systems. He is actively involved in the
development of current rating methodologies, acting as
the UK member of IEC TC 20 WG19 and Cigre Working
Group B1.35 “Guide to rating calculations”.
Gail Hutton received the Bachelor of Science degree
from Strathclyde University in 2003, and gained a PhD
from the same University in 2008 for research into
stochastic models of share prices. She worked for
National Air Traffic Services from 2007 to 2010 as a
statistician, focusing on the measurement and modelling
of aviation risk. She joined RES in 2010, providing
analysis of technical aspects of wind farm development
both onshore and offshore and with a particular interest
in characterizing uncertainty. She is a member of the Royal Statistical Society
and has attained both Chartered Statistician and Chartered Scientist status.
Priank Cangy was born in Maputo, Mozambique on
the 5th of April, 1993. He is currently studying at the
University of Southampton for an MEng in Electrical
Engineering, having recently completed an internship
at the Tony Davies High Voltage Laboratory.