Professional Documents
Culture Documents
PII: S2590-1230(21)00002-5
DOI: https://doi.org/10.1016/j.rineng.2021.100201
Reference: RINENG 100201
Please cite this article as: K.K. Poleo, W.J. Crowther, M. Barnes, Estimating the Impact of Drone-based
Inspection on the Levelised Cost of Electricity for Offshore Wind Farms, Results in Engineering, https://
doi.org/10.1016/j.rineng.2021.100201.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
HIGHLIGHTS
• A new cost model for predicting the Levelised Cost of Electricity from offshore wind
• Model informs wind farm operators on the costs of drone inspection techniques
• Model informs Drone Service Providers on the cost of higher levels of autonomy
• Greatest cost benefit comes from changing ropes to drones for visual inspection
• High levels of autonomy offer marginal benefit and is currently not worth pursuing
ABSTRACT
of
Using drones for infrastructure inspection is becoming routine, driven by the benefit of reducing risk and costs. In this
paper, the business case for drone-based inspection is examined from the perspective of the wind farm operator and the
ro
Drone Service Provider (DSP). A physical and financial model of an offshore wind farm is built using techno-economic
analysis and activity-based costing, and data from the open literature. Drone operational models are developed based on
domain specific knowledge of operation practices and the predicted physical environment. Rope-access inspection is used as
-p
a baseline and accounts for 0.7% of the wind farm operational expenditure. Replacing rope-access inspection with drones
reduces costs by up to 70% and decreases revenue lost due to down-time by up to 90%. Increasing autonomy of drones
re
increases the speed at which inspections can be performed but increases costs and complexity. For wind farm operator there
is marginal economic benefit (2% reduction in inspection costs) in moving towards a fully autonomous drone-based
lP
inspection system from the current visual line of sight operation of single drone. However, from the point of view of the
DSP, fully autonomous operations allow greater scalability of the business and enables higher utilisation of the fleet.
KEYWORDS:
na
LCOE; Offshore Wind; Unmanned aerial vehicles; Cost benefit analysis; Visual Inspection
1. INTRODUCTION
ur
The use of drone technology for infrastructure inspection and maintenance is becoming increasingly routine. In the
Jo
construction industry drones have gone from being an innovative solution for surveying large areas, to a routine tool for
local onsite surveys [1]. In the energy industry, drone adoption has been driven by the immediate safety benefit of
eliminating the need for working at height and the promise of reduced costs through faster inspection times and elimination
of specialist access equipment and training. Example cost saving opportunities due to drone use specific to the energy sector
are highlighted in [2] and [3]. These studies, however, only show direct costs associated with drone inspection techniques
and do not consider the overall economic impact of the Operation and Maintenance (O&M) strategy. We aim to understand
the economic impact of drone-based inspection for offshore wind by evaluating the effects of changing inspection
technology on the overall operational expenditure and on revenue, thus defining the link between technology
implementation and the Levelised Cost of Electricity (LCOE). The methodology is relevant to offshore wind farm operators,
as it describes the gains and costs avoided by adopting autonomous systems and is relevant to drone service providers in that
it provides insight into the links between automation and business growth. The work is novel in that to the authors’
knowledge it documents the first quantitative, model-based study to assess economic benefits of drone use for offshore
inspection. Previous studies have highlighted applications for drones [4], [5], [6] and others changes in working practices
that they introduce [2], however these studies are unable to translate general findings into specific changes in LCOE due to
absence of appropriate techno-economic models.
LCOE is used as a standard measure of what it costs to produce a unit of electrical energy, usually quoted in £ per MWh.
It considers all the associated costs of producing power over the lifetime of the asset. LCOE is often used to develop policy
on implementation of different methods for electricity production [7],. Calculating LCOE for energy projects requires an
estimation of Capital and Operational Expenditure (CAPEX and OPEX), as well as the total revenue generated over the life
of the project. This information is often commercially sensitive and not openly available. To overcome this, we have
developed and evaluated mathematical models for estimating LCOE based on individual system expenditure as a result of
1
failure [9]. Other studies have derived probabilistic models for possible OPEX and CAPEX costs and applied Monte Carlo
techniques to investigate how varying inputs affects LCOE for various energy sectors like wind [10], natural gas [11], wave
[12], and hydropower [13]. These studies show how a range of LCOE values can be estimated and compared against
different production technologies and how external influences, like current policies and government targets, can affect costs.
However, further work into understanding the costs of specific aspects of OPEX would still need to be done to understand
how processes and technologies would affects LCOE. In the present paper, we use LCOE as a performance indicator for
specific methods and technologies. The methodology described herein makes no attempt to derive a precise LCOE
estimation for wind, instead, it allows for technologies to be compared by understanding the influence of a certain process
on the LCOE and thus allowing comparison of new technology scenarios. Specifically, it analyses the costs and revenue
deltas for performing visual inspection of wind turbines using Unmanned Aircraft Systems (UAS), also referred to as
drones. The case study considers three technological adoption scenarios: direct control of a drone by an operator within
visual line of sight, a multi-vehicle system supervised from a remote station and a fully autonomous system. The method for
estimating baseline LCOE is similar to techniques that have been successfully applied in the oil and gas industry [14].
The cost requirements for drone-based inspection systems are evaluated first via Techno Economic Analysis (TEA) and
then by analysing how the technology would be implemented in practice via a time-driven Activity Based Costing (ABC).
The application of these tools in isolation is well proven. In [15], TEA techniques are used for evaluating wind farm cost
of
saving opportunities based on geographical features and layouts for a Round 3 offshore site in the UK. In [14], ABC
techniques are used for evaluating OPEX in the offshore oil and gas industry. The method works well when the tasks
ro
required for evaluation are well specified. A drawback of ABC is that it can be expensive in terms of resources when a
comprehensive assessment of the task is needed [16].
-p
In this paper TEA and ABC tool is used for: (i) understanding the capabilities and requirements that drone inspection
techniques will bring, (ii) estimating the cost of inspecting the offshore wind turbines in a representative environment, and
re
(iii) estimating the cost of the current rope-access inspection process.
lP
2.1. OVERVIEW
na
Robotics and autonomous systems are anticipated to be an integral part of the operations and maintenance processes for
future wind farms [17]. This section is to provide a succinct summary of current and future trends in drone systems used for
inspection and maintenance to provide a foundation for choice of case study groupings as part of the LCOE analysis in
ur
section 4. The review is organised under the headings of Vehicles, Sensor Payloads, Command and Control (C2) and
Computing.
Jo
2.2. VEHICLES
The description “drone” has now been widely adopted as a generic term for a flying robot. More formally, Remotely
Piloted Air System (RPAS), is used by the International Civil Aviation Organisation (ICAO), and Small Unmanned
Surveillance Aircraft (SUSA) is the official term adopted in the UK’s Air Navigation Order [18], [19], [20].
In Europe, unmanned aircraft systems over 150kg are governed by the European Union Aviation Safety Agency (EASA)
and they have the same certification requirements as any other manned aircraft [21].. In the UK, the Air Navigation Order
(ANO) separates drones into 2 categories: (1) below 20 kg take-off mass, including batteries and fuel, and (2) between 20 to
150 kg take-off mass [20]. Legislation is more stringent on the heavier class of vehicles. In this paper we will investigate
technologies and methodologies most applicable to aircraft below 20 kg take-off mass, due to their greater ubiquity for
industrial inspection tasks.
Figure 1 shows the three main types of drone systems encountered. (a) Rotary wing (helicopter or multirotor) (b) fixed
wing (conventional aeroplane), and (c) hybrid systems (c).
2
Figure 1: Rotary wing (a), Fixed wing (b), and hybrid (c) aircraft examples.
Rotary wing vehicles have been adopted by industry faster than fixed wing [22], [23]. The advantages of rotary wing
vehicles over traditional fixed wing aircraft are that they can hover, the footprint required to operate is smaller, and the skills
required to operate them are significantly reduced. On the other hand, fixed wing aircraft offer greater forward flying
efficiency which translates to longer range and higher endurance. Hybrid vehicles attempt to combine the best attributes of
fixed and rotary wing vehicles, typically at the expense of increased structural weight and complexity for the same payload
capability [24].
of
2.3.1. Payload sizing
. The endurance and payload capacity that some typical battery powered sub 20 kg commercial drones is shown in Figure
ro
2 [25], [26], [27]. A 20-minute mission duration is typical. This relatively short operational time is limited by currently
available battery technology.
50
-p
re
Alta 6
Hover endurance at sea level [Minutes]
45
Wind 4
40
M600
lP
35
Wind 8
30
M200 V2
25
na
20
15
10
ur
5
0
0 1 2 3 4 5 6 7 8 9 10
Jo
Figure 2: Payload-Endurance chart showing the payload weight that can be carried by typical commercial drones within the 20 kg max
all up weight limit.
Thermal
Current systems offer full HD or 4K RGB systems and 640x512 thermal sensors collocated, allowing visual and thermal
images to be superimposed. Thermal systems usually will have fixed optics (no optical zoom) meaning the drone must be
close to the target to achieve a high sample resolution. In addition, sensors with increased resolution or framerate may be
subject to ITAR restrictions.
3
2.3.3. Active sensors
Lidar
Light Detection and Ranging (LIDAR) sensors are becoming increasingly available for sub 20 kg aircraft systems. They
are mainly used for surveying or navigation (sense and avoid). Survey grade LIDAR are normally paired with a differential
or Real-Time Kinematic Global Positioning System (RTK-GPS) to improve geolocation of the data, and newer systems are
equipped with cameras to add texture to point cloud data.
Radar
Use of radar-based inspection tools is of increasing interest as the typical weight of an operationally useful system is now
within the SWAP capability of sub-20 kg drones. These systems work by comparing the radar return against a library of
data from different characterised materials [4]. Radar-based inspection systems offer information on subsurface features that
complements optical inspection data [31]. For drone application, the technology is limited by the relatively narrow field of
view of flight capable sensors and the relatively slow capture rate (of the order of seconds for radar compared with
microseconds for vision).
of
2.4. COMMAND AND CONTROL
Drone command and control (C2) requires low latency, high integrity but relatively low bandwidth communications. 5G
ro
services could in the future be used for both C2 and data transmission for drones. [32], [33], [34]. Table 1 shows the
minimum bandwidth required for transmission of data from a range of sensors on a drone to a remote monitoring station.
-p
The reference data rate for 5G is thought to be at least 25Mbit/s in uplink and at least twice that for downloads, with a
latency of around 10ms [35]. This would allow each drone to be able to stream 2 x 4K camera streams at 30 fps, a front FPV
re
camera of lower resolution, the flight control system telemetry from the aircraft, and C2 signals. Work on demonstrating the
use of 5G infrastructure for offshore wind farm monitoring is already underway [36].
lP
Table 1: Typical bandwidth requirement for various services which are needed for visually operating a drone.
Table 2: Example embedded computing boards suitable for on board data processing in sub 20 kg drones. The computing power required
for flight control functions is typically orders of magnitude smaller than that required for image processing.
4
The main use case for drones in offshore power is typically remote observation. Drones are currently operated within
Visual Line Of Sight (VLOS) by a pilot with direct control of the aircraft [48]. Historical trends in the development of flight
control and mission capability for commercial drones are shown in Table 3. With each iteration more tasks are offloaded to
onboard computing, freeing the pilot to perform the mission critical aspects of the operation. However, we are still multiple
generations away from a fully autonomous drone able to perform missions in which the drone itself can understand its task
in the same way a human operator can [49], [50]. This table consider consumer level drones as there is available historical
information regarding their software capability.
Table 3: Historical delegation of tasks from Pilot to Flight Control System (FCS) for a commercial drone product line ( DJI Phantom)
[51], [52].
of
Flight feature Responsibility
ro
Low level control: FCS
Altitude hover hold: Pilot FCS
Position hover hold (GPS): Pilot FCS
Position hover hold (no GPS):
RTL failsafe action:
Pilot
Pilot
-p FCS
FCS
re
Battery management: Pilot FCS
Obstacle avoidance (forward): Pilot FCS
lP
3. METHODOLOGY
Eq. (1) represents the operational expenditure of a given year () as the sum of all work package element () in table 4,
where is the installed power factor eq. (2), is the nameplate capacity of the turbine, is the number of turbines, and
is the nominal capacity from the data, in this particular case 500MW, and is a weighting, between 0 and 2, to define
the expected cost between the quoted maximum and minimum values eq. (3).
5
Table 4: Operations and Maintenance costs by work package for a 500MW offshore wind farm in the UK [53].
of
Export cable surveys and repairs £50 to £200
Weather forecasting £40 to £90
Onshore electricals £10 to £100
ro
Total = £24,660 to £57,290
-p
+ 2 −
= 1
2
re
!" #
lP
= (2)
!
0 ≤ ≤ 2 3
na
In addition to OPEX, the total amount of revenue from electricity generation must be understood. Energy generated over
a time interval is a function of the nameplate capacity of the asset and the weather conditions on site (capacity factor).
Jo
Practically, other factors, like the availability of the asset must also be considered. Eq. (4) is used to estimate the average
revenue generated on a given time interval,
where
' is the energy generated over a time period (*), ) is the capacity factor, and ( is the assumed availability of
the turbine. Availability is defined as the percentage of the time the farm is operational over a given time period.
If wind data is available, eq. (4) can be rearranged for estimating capacity factor over the life , of the farm, eq. (5).
∑
.
'
) =
5
(
Revenue for given year 01 is a function of the energy generated
' and the current selling price of electricity 2'3
eq. (6).
Analysis of revenue also allows for the economic effect of shutdown to be measured.
6
3.2.1. Example: Revenue loss due to unscheduled shutdown for an offshore 3.6MW turbine.
Consider a turbine with power and wind availability curves shown in Figure 3. The turbine is part of a farm located 20km
off the coast of Kent, UK, for which wind data between 2004 and 2010 is available [54].
0.12 4.0
3.5
0.10
Historical 3.0
0.08 wind data
Pout [MW]
Probability
2.5
0.06 Wind turbine 2.0
Power Curve
1.5
0.04
1.0
0.02
0.5
0.00 0.0
Vw [m/s]
1 3 5 7 9 11 13 15 17 19 21 23 25 27
Figure 3: Assumed power output as a function of wind speed for a 120m 3.6MW offshore turbine [55], and Distribution of historical
of
wind data in 10 minute intervals between 2004 and 2010 at 82.3 mAMSL from [54].
Using the information from Figure 3, equations (4) through (6), and an assumed wholesale price of £65/MWh [56], the
ro
revenue loss due to an unscheduled random shutdown can be estimated.
Figure 4 shows the range of potential loss of revenue per turbine per hour of shutdown.
0.35
-p
re
0.3
0.25
lP
Probability
0.2
0.15
na
0.1
0.05
0
ur
Figure 4: Likely loss of revenue during random unscheduled shut down per hour per turbine for an asset as described in
Figure 3.
'5!67859:; = 3,216,672 ?
ℎ
7
CAPEX will also include the residual value of the asset after the useful life of the farm has elapsed and any investment
made during the life of the farm eq. (7). For this paper, we assume that CAPEX is composed entirely of repayment of a loan
acquired to construct the farm (L2 ). For a more complete analysis, a cash investment (LM ), and that the farm’s
residual value or remediation cost (L2N ) can be taken into account. The projected life of the farm is given by ().
of
Table 5: Loan terms for a £2B wind farm project
ro
Variable Value Notes
Total loan = £2,000,000,000 Assumes a 100% loan
Terms = 25 years Projected life of the farm
Interest rate =
Payments per year =
O
PB = 9
'
na
is the expenditure incurred each year brought to a net present value eq. (10) and
' is the electricity generated
during the life of the farm as calculated using eq. (4), and Q is the discount rate of the farm. We assume for the initial
ur
analysis that the discount rate is equal to the yearly inflation rate.
Jo
L
+
= 10
1 + Q
8
' = 200 ∗ 3.6 ∗ 0.95 ∗ 0.51 ∗ 25 ∗ 365 ∗ 24 MWh
' = 76,395 U
ℎ, over 25 years.
= £2.9 V
Therefore,
£W.XY
O
= = £37.60 / MWh
Z[.\]^
These results are comparable with the latest strike prices for offshore wind generation published as part of the Contracts
for Difference (CfD) round 3 auctions of £39.65 /MWh [58]. Under these conditions, operators could see a profit of just
over £2 / MWh.
of
available OPEX data offers some useful insight into where most of the money is spent. However, there is no data available
which gives an understanding on how much specific tasks (i.e. wind turbine inspection) cost. To estimate this, a process
ro
called time-based Activity Based Costing (ABC) is used [14]. ABC is applied by understanding the cost per unit time
incurred when performing a task. In this paper the method is applied to understand the cost of various inspection methods,
including rope-access which will serve as a baseline.
-p
It is assumed that the overall cost of a process can be split into two: (i) the Directly Incurred (DI) cost of doing the task,
re
i.e. workforce, transportation of people and parts, and consumables; and (ii) the change in revenue resulting from changing
processes.
lP
DI cost and revenue change is then broken down into components and assign a cost per unit time eq. (12). For generic
offshore operations the main direct costs drivers are: workforce salaries, transportation, and consumables [14]. Revenue lost
per unit time is presented in section 3.2.1.
ur
Where e is the workforce cost derivative per unit time and is mainly made up from the salaries of the people involved. *f
is the transportation cost derivative per unit time, and comprises the boats, helicopters, or any other vehicle involved in
transporting people or equipment. g_ is the consumables cost derivative per unit time, in this case, tools, safety equipment,
or consumables used to complete the task. * is the time allocated to complete the activity.
3.5.1. Example: Rope-access inspection costs for a 500MW offshore wind farm.
In this example the DI cost associated with performing a rope-access based inspection of a wind farm is estimated.
Assume a wind farm with 139 turbines and a nameplate capacity of 500 MW, and that the turbines have the same
characteristics as those used in previous examples. Each turbine is visually inspected once a year by three technicians
employed by the operator, and the turbine is shut down during inspection. Furthermore, assume that 7 hours are allocated to
perform the inspection [3] and that the team carries out one inspection per day. Estimated directly incurred costs are shown
in Table 7. The same data is shown graphically in Figure 5 to illustrate the relative contribution of each element to the
overall inspection cost.
9
Table 7: DI cost of inspecting 139 individual wind turbines once a year: OPEX and revenue lost.
19%
Workforce
59% 21%
Transportation
of
Consumables
Downtime
ro
2%
Figure 5: Distribution of direct costs and revenue lost for inspecting a 500MW farm once a year using rope-access methods.
process. TEA is used to derive the costs under the same terms of workforce, transportation, and consumables used for ABC
analysis. Furthermore, 3 operating conditions will be investigated: (i) Operator controlling a single drone within Visual Line
of Sight (VLOS), (ii) Single operator controlling multiple drones within Extended VLOS (EVLOS), and (iii) Single operator
na
managing a fleet Beyond VLOS (BVLOS). These operating conditions, represented by the pictograms in Figure 6,
represent the current and future trends for the technology.
ur
Jo
Figure 6: Pictogram representations for Single drone within VLOS (i), Multiple drones within EVLOS (ii), and multiple drones BVLOS
(iii) operating conditions.
1
Based on 3 technicians, inspecting 139 turbines and assuming each turbine takes 7 hours to inspect.
2
Based on utilising 7 hours a day for 139 days.
3
1 day per turbine.
4
1 per person involved (3 wind turbine technicians and 1 boat operator).
5
Based on shutting down the turbine.
10
compliance with the ANO, (c) appropriate level of insurance, (d) a permission from the UK Civil Aviation Authority to
operate the drone in a commercial environment, (e) the expectation that any person involved in operating the drones will
keep themselves current in their abilities to fly and up to date with legislation and best practices [20], [21], [60].
(a) is normally achieved by receiving training and passing an assessment by a National Qualifying Entity (NQE), (b) is
achieved by producing an Operating Safety Case which specifies the companies policies and their operational procedures,
(c) procuring the right insurance level compliant with regulation (EC) 785/2004 [61]. (d) submitting an Operating Safety
Case (OSC) to the CAA along with the necessary forms and (e) by attending further training sessions and ensuring all
elements of the operation are regularly checked and updated where needed.
of
Table 8: Aircraft and sensor parameters for simulating the time it takes to survey a 120m wind turbine
ro
Parameter Value
Aircraft = DJI Matrice 210 V2
Max horizontal speed = 17 m/s
Max Ascent speed =
Max Descent speed =
Estimated flight time =
-p
5 m/s
3 m/s
26 min
re
Desired GSD = 0.11 cm/px
Camera FoV = 72 degrees
Camera resolution = 20.8 MP
lP
The asset used is a 120m 3.6MW turbine model. Human operators were able to complete an inspection with complete
na
coverage of the turbine in around 15 to 30 minutes. A drone flying automatically, with no human intervention was able to
carry the inspection is just over 6 minutes, including travel time to the boat located 50 meters away from the turbine. The
automatic inspection was performed by setting waypoints around the turbine and having the drone fly whilst taking the
ur
photos. In addition, ray casting was used to measure the GSD of images being taken and a threshold was set and overlaid in
the image to get feedback on coverage. Green was set for anything better than a GSD of 0.11cm/px, up to 0.2cm per pixel
and white for unsuccessful or missing images. These values were arbitrarily selected based on the requirement of being able
Jo
Figure 7: Simulation environment in Unity showing wind turbine model with a path defined for the drone to follow and image quality
feedback overlay.
11
Transportation: Currently, VLOS requirements are met by having the pilot near the wind turbine being inspected.
Therefore, there will still be a requirement for a workboat to provide transport to and from shore. The unit cost of
transportation will be comparable to that for rope-access inspection.
Workforce: Current best practices dictates that a second person is required to help the pilot during operations, acting as a
spotter/ground station operator. It is assumed that base salaries for drone operators are close to that of wind turbine
technicians, based on 2019 market rates. Hence, the overall personnel cost per unit time will be similar to that for rope-
access but will require fewer people.
Consumables: The cost of a drone able to operate within the physical environment and able to take images at the
required resolution will change depending on the specific requirements. As an example, drones mentioned in Figure 2
average around £15,000 in 2019 once a suitable optical payload is installed. A challenge when assigning a time unit cost to
the drone is that there are no definitive studies or information available on what the lifespan of these systems are under
offshore operating conditions. For this study we will assume drones have 4000 flight hours (fhrs) lifespan, based on flying
5h per workday for 3 years and that batteries and propellers are be replaced every 500 fhrs at a cost of £800 per set.
Table 9: List of DI costs associated with drone-based inspection within VLOS per unit time.
of
Item Cost Frequency
ro
Workforce (WT technicians) = £18 Per person per hour
Workforce (boat) = £18 Per person per hour
Transport (boat) = £600 Per boat per day
Consumables (Drone) =
Consumables (Spares) =
Revenue lost (worst case) =
Training =
-p
£15000
£800
£237
£1500
Per drone per 4000fhrs
Per drone per 500fhrs
Per turbine per hour
Per pilot, one-off
re
Insurance = £1500 Per year
CAA Fees = £750 Per year
lP
By using the key derived costs listed in Table 9 and applying time-driven ABC, it can be concluded that inspection cost
could be cut by around 90%. This is mainly driven by the reduction in revenue lost achieved by reducing the inspection time
from 7 hours, to half an hour. Directly incurred costs and revenue lost results are shown in Table 10. However, for
na
infrequent drone inspections (one per turbine per year), it is more likely that wind farm operators would utilise Drone-as-a-
Service Providers (DSP) to undertake inspections.
ur
Table 10: Directly incurred cost of inspecting 139 individual wind turbines once a year: OPEX and revenue lost for drone based VLOS
case.
Jo
6
Based on 2 drone operators being able to survey a 139 turbine farm in 18 days (8 turbines per day).
7
18 days for 7 hours each.
8
Assumes turbine needs to be shut down for 30 mins for inspection.
12
Table 11: Daily rate estimation for a DSP operating VLOS based on directly incurred costs and assumed revenue, service uptake, and
drone reliability.
Cost per
Category Line item Qty Sub-Total Notes
year
Workforce Drone operators / technicians £30,000 12 £360,000 Based on 2019 market rates from
Workforce Management £45,000 3 £135,000 Accounts for management, HR, and PR roles
Workforce Senior leadership team £75,000 4 £300,000 Accounts for CEO, CTO, COO, and CFO
Workforce Support staff £25,000 3 £75,000 Accounts for supporting roles at 2019 market rates
Workforce Pension contribution £8,700 Pension contribution from the employer at 1%
Infrastructure Rent £12,000 1 £12,000 Based on 2019 market rates
Infrastructure Services £8,400 1 £8,400 Accounts for utilities, phone, and internet connections
Infrastructure Van £3,600 4 £14,400 Based on 2019 lease deals for work van style vehicles
Consumables Drones £36,000 4 £144,000 Assumes £9000 per drone per 250fhrs and 1000fhrs/year
Consumables Drone spares and tools £3,600 4 £14,400 Assumes 10% of the cost of the drone per year.
Consumables Fuel £2,838 4 £11,352 At 25k miles per vehicle at 50mpg at £1.25/l of fuel
Other Commercial van insurance £900 4 £3,600 Based on 2019 insurance rates for commercial vehicles
of
Other Licensing fees and PfCO £2,500 1 £2,500 Based on 2019 CAA rates and CPD training for operators
Other Comprehensive drone insurance £1,200 4 £4,800 Based on 2019 quotes for VLOS operations
ro
Total operating cost = £1,094,152 Sum of all operating costs
Desired revenue = 30% 1 £328,245 Assumes a 30% profit over operating cost
Tax = 30% 1 £98,473 Assumes a 30% tax on the revenue
Total required income =
Service days available = 988
-p
£1,520,871
Based on 4 teams working 247 days in a year
re
Estimated daily rate = £1,540 Per day for VLOS operating conditions
3.7. OPERATING COSTS FOR A DRONE-AS-A-SERVICE PROVIDER (DSP)
lP
For a wind farm operator, to procuring services from a DSP reduces indirect costs. Furthermore, investment on
technology and responsibility for keeping current with legislation lies with the DSP. On the other hand, availability,
responsiveness, and flexibility of the services offered by DSPs might not suit wind farm operators during responsive
na
maintenance operations. In addition, if the service is being used often, there might be economic advantages for wind farm
operators to develop their own in-house capability. This will depend on how often the service is being used, and/or if other
activities within the wind farm operations would see a benefit in employing the skills from the workforce, or the drones.
ur
As the responsibility to deliver a service lies with the DSP, the economic argument for higher levels of autonomy must
Jo
now be made from their point of view. Therefore, the effect of technology on the daily rate must be investigated.
3.7.1. Example: DSP daily rate estimation and the effect of reliability.
This example introduces the main operating costs of a DSP to derive a daily rate, table 11. Costs are grouped into
categories, namely (i) workforce, (ii) infrastructure, (iii) consumables and (iv) others. Incurred costs are then analysed to
understand the key drivers of a daily rate.
A typical DSP will comprise of pilots, technical staff, and full-time administration support staff as their workforce;
infrastructure comprises of a location where they operate and vehicles used for transportation; consumables are drones
which have a finite useful life, spare parts, fuel (for transportation), etc.
The highest uncertainty around DSP costs lies with the reliability of the drones. The rate at which new products are
release to market means that drones are replaced long before they might fail due to fatigue. However, the effect of reliability
on the cost of the service, Figure 8, shows that the overall cost would change up to 10% if time between replacement, due to
failure or technology upgrade, lies within 250 to 1000 flight hours.
13
6%
Workforce
Infrastructure
22%
Consumables
22% 6% Other
1% 58% Revenue
1% Tax
11% 2%
4%
64%
3%
Figure 8: Cost breakdown of a DSP daily rate taking into account the influence of drone reliability assuming (i) 1000 flight hours, outer
chart, and (ii) 250 flight hours, inner chart.
Table 12: Daily rate estimation under (i) VLOS single drone operations, (ii) Multiple drones within EVLOS, and (iii) Multiple drones
of
operating BVLOS
ro
Category
Table 12 shows an estimate for daily rates of operation for a DSP where the workforce, infrastructure and other non-
drone related expenses are kept and the change in technology allows for single operator to fly multiple drones with the
objective of increasing the capability and productivity of the team. For a wind farm operator this translates to a higher
number of inspections per day of operation. For a DSP this translates to higher operational capability.
9
Per team per year.
14
The importance to understand the reliability of the drone will obviously increase as technology progresses and allows
multiple drones to be operated BVLOS. The data presented in Table 12 is also shown in Figure 9 to graphically demonstrate
the influence of the drone cost on the daily rate for the operation. The error bar in the figure only takes into account the
uncertainty around the reliability of the drones, the error around uncertainty from legislation, insurance, and operations
needs further investigating.
£4,500
£1,500
£0
(i) (ii) (iii)
Figure 9: DSP daily rate for technology cases (i), (ii), and (iii). assuming between 200 and 1000 flight hours between replacement with a
of
nominal of 250 flight hours.
To bring Figure 9 into the context of the wind farm we can present the data based on inspection cost per turbine. Figure
ro
10 shows the cost per inspected turbine. The next section analyses the impact on the LCOE from these technologies.
£200
-p
Inspection cost [£/turbine]
re
£150
£100
lP
£50
na
£-
(i) (ii) (iii)
ur
Figure 10: Inspection cost per turbine under various technology cases (i), (ii), and (iii) for flight times of between 200 and 1000 flight
hours between replacement.
Jo
Revenue gained can also be plotted as a function of the installed capacity to better understand the potential of increasing
the inspection speed as a tool to decrease downtime. Figure 11 shows how introducing drones enables an extra £375 per
year per installed MWh to be generated. For the London Array this would mean an extra revenue of over £230,000 per year.
£400
Additional revenue per year [£/MW]
£300
£200
£100
£-
(i) (ii) (iii)
Figure 11: Additional revenue generated per technology case due to increase in availability.
15
OPEX
Rev. Lost
47%
53%
of
Case (i) = £ 81k Case (ii) = £ 28k Case (iii) = £ 25k
ro
Figure 12: Comparison of OPEX and Rev. Lost contribution to the overall inspection cost of each technology.
The economic value of a technology can be assessed by understanding the effect that the technology will have on the
-p
LCOE. Figure 13 shows the change in LCOE due to introduction of different drone inspection techniques compared to the
baseline rope-access methods. This result is valid for the case of outsourcing the service to a DSP and assuming a single
re
inspection is required per year. If the inspection rate is increased, the benefit will be enhanced. Note that this LCOE
reduction estimation is a direct result generated by the savings in inspection cost: it does not take into account any other of
lP
the potential benefits of the technology, such as the decrease in the risk associated with manual inspection and the cost
avoided by being able to spot problems earlier. In addition, it does not consider the added benefit from utilising the
technology for other tasks, for example as part of a blade repair system and transportation of small parts. Those benefits will
na
need to be studied further to understand the wider effect of drone-enabled services for offshore wind.
0.20%
ur
Reduction on LCOE over the baseline
0.15%
Jo
0.10%
0.05%
0.00%
(i) (ii) (iii)
Figure 13: Reduction in Levelised Cost of Electricity for offshore wind energy resulting from the introduction of drone visual inspection
when using drone service providers.
5. CONCLUSION
A transparent, comprehensive semi-empirical model for predicting the Levelised Cost of Electricity (LCOE) from
offshore wind has been developed for the purposes of evaluating the impact of novel drone-based inspection strategies. The
model is transparent in that it uses publicly available data sources and clearly defined analytical methods based on
engineering physics and economics. It is comprehensive in that it allows meaningful analysis of the relative impact of
changes in inspection strategy on overall costs. The model has been shown to be broadly valid in that it predicts LCOE for
offshore wind in line with present industry Contracts-for-Difference bids based on currently available data.
16
The cost of traditional manual inspection of turbines using rope access is estimated to account for 0.7% of the wind farm
operational expenditure and around 0.2% of the LCOE. Replacing manual inspection with remotely piloted drones reduces
inspection costs by 70% and could decrease revenue lost during inspection by 90% due to reduced downtime. Increasing
autonomy of drone operations increases the speed at which inspections can be performed but introduces additional costs in
terms of complexity and compliance with regulations. As such, there is currently marginal economic benefit (2% reduction
in inspection costs) in moving towards the envisaged end goal of fully autonomous operations with a drone fleet
commanded by single remote supervision.
Drone inspection has been considered as both an in-house capability of an existing wind farm maintenance team and as a
service provided by a dedicated drone operator. It is shown that fully autonomous operations allow much greater business
scalability for drone service provides based on much higher fleet utilisation. Hence, it is likely that the drones as a service
inspection model will predominate as the technologies and legislation to support fully autonomous operations mature.
Acknowledgements
The work presented is supported by the UK Engineering and Physical Sciences Research Council (EPSRC) HOME-
Offshore project (EPSRC Reference: EP/P009743/1)
of
References
ro
[1] C. Anderson, “Drones Go to Work,” Harvard Business Review, May 2017.
[2] -p
Measure, “The Case Drones in Energy,” 2018. [Online]. Available: https://www.measure.com/the-case-for-drones-
energy-download.
re
[3] C. Stout and D. T. March, “UAV Approaches to Wind Turbine Inspection Reducing Reliance on Rope-Access,”
2019.
lP
[4] J. F. Skov, M. D. Ulriksen, K. A. Dickow, P. H. Kirkegaard, and L. Damkilde, “On structural health monitoring of
wind turbine blades,” Key Eng. Mater., vol. 569–570, pp. 628–635, 2013, doi:
10.4028/www.scientific.net/KEM.569-570.628.
na
[5] J. Igba, K. Alemzadeh, C. Durugbo, and K. Henningsen, “Through-life engineering services: A wind turbine
perspective,” Procedia CIRP, vol. 22, no. 1, pp. 213–218, 2014, doi: 10.1016/j.procir.2014.07.021.
ur
[6] A. Reddy, V. Indragandhi, L. Ravi, and V. Subramaniyaswamy, “Detection of Cracks and damage in wind turbine
Jo
blades using artificial intelligence-based image analytics,” Meas. J. Int. Meas. Confed., vol. 147, p. 106823, 2019,
doi: 10.1016/j.measurement.2019.07.051.
[7] M. Lerch, M. De-Prada-Gil, C. Molins, and G. Benveniste, “Sensitivity analysis on the levelized cost of energy for
floating offshore wind farms,” Sustain. Energy Technol. Assessments, vol. 30, no. March, pp. 77–90, 2018, doi:
10.1016/j.seta.2018.09.005.
[8] M. Ram et al., “A comparative analysis of electricity generation costs from renewable, fossil fuel and nuclear
sources in G20 countries for the period 2015-2030,” J. Clean. Prod., vol. 208, pp. 687–704, Oct. 2018, doi:
10.1016/j.jclepro.2018.07.159.
[9] C. D. Dao, B. Kazemtabrizi, and C. J. Crabtree, “Modelling the Effects of Reliability and Maintenance on Levelised
Cost of Wind Energy,” in Proceedings of ASME Turbo Expo 2019: Turbomachinery Technical Conference and
Exposition GT2019, 2019, vol. 9, pp. 1–8, doi: 10.1115/GT2019-90015.
[10] A. Ioannou, A. Angus, and F. Brennan, “Stochastic Prediction of Offshore Wind Farm LCOE through an Integrated
Cost Model,” Energy Procedia, vol. 107, no. September 2016, pp. 383–389, 2017, doi:
10.1016/j.egypro.2016.12.180.
[11] N. Heck, C. Smith, and E. Hittinger, “A Monte Carlo approach to integrating uncertainty into the levelized cost of
electricity,” Electr. J., vol. 29, no. 3, pp. 21–30, 2016, doi: 10.1016/j.tej.2016.04.001.
[12] D. A. Sunter et al., “Two-Stage Monte Carlo Simulation to Forecast Levelized Cost of Electricity for Wave
17
Energy,” in 6th International Conference on Renewable Energy Reseach and Applications, 2017, pp. 638–641.
[13] D. R. Chalise, P. O’connor, S. Deneale, R. Uria-Martinez, and S.-C. Kao, “LCOE Uncertainty Analysis for
Hydropower using Monte Carlo Simulations,” pp. 1–14, 2018, [Online]. Available:
https://dolraj.wordpress.ncsu.edu/files/2018/09/3-LCOE-Uncertainty-Analysis-for-Hydropower-using-Monte-Carlo-
Simulations-20150801.pdf.
[14] M. J. Kaiser, “The role of factor and activity-based models in offshore operating cost estimation,” J. Pet. Sci. Eng.,
vol. 174, no. October 2018, pp. 1062–1092, 2019, doi: 10.1016/j.petrol.2018.10.093.
[15] V. Mytilinou and A. J. Kolios, “Techno-economic optimisation of offshore wind farms based on life cycle cost
analysis on the UK,” Renew. Energy, vol. 132, pp. 439–454, 2019, doi: 10.1016/j.renene.2018.07.146.
[17] D. J. Willis et al., “Wind energy research: State-of-the-art and future research directions,” Renew. Energy, vol. 125,
pp. 133–154, 2018, doi: 10.1016/j.renene.2018.02.049.
of
[18] Google LLC, “RPAS vs SUSA vs Drone Search trends,” Google Trends.
https://trends.google.com/trends/explore?date=today 5-y&geo=GB&q=RPAS,SUSA,Drone (accessed Sep. 17,
ro
2019).
[19] International Civil Aviation Organization (ICAO), “Remotely Piloted Aircraft System (RPAS) Concept of
Operations (CONOPS) for International
-p
IFR,”
https://www.icao.int/safety/UA/Documents/RPAS CONOPS.pdf.
p. 22, 2014, [Online]. Available:
re
[20] UK CAA, The Air Navigation Order 2016. United Kingdom, 2016.
lP
[21] EASA, “A proposal to create common rules for operating drones in Europe,” no. September, p. 8, 2015, [Online].
Available: http://easa.
na
[23] M. Klaassen, “2018 Commercial Drone Industry Trends,” 2018. [Online]. Available:
https://cdn2.hubspot.net/hubfs/530284/10M_Acre_Report_2017_.pdf?blog.
Jo
[24] Martek Drones Lt, “Best Commercial Drones For Beginners,” 2018. https://www.coptrz.com/best-commercial-
drones-for-beginners/ (accessed Sep. 02, 2019).
[25] Heliguy, “DJI Wind Series Drones,” 2019. https://www.heliguy.com/dji-wind-series-i86 (accessed Oct. 25, 2019).
[27] Freefly, “What is the max flight time on the ALTA 6?,” 2019. https://freeflysystems.com/knowledge-base/what-is-
the-max-flight-time-on-the-alta-6 (accessed Oct. 25, 2019).
[28] F. Giones and A. Brem, “From toys to tools: The co-evolution of technological and entrepreneurial developments in
the drone industry,” Bus. Horiz., vol. 60, no. 6, pp. 875–884, 2017, doi: 10.1016/j.bushor.2017.08.001.
[29] S. Watkins et al., “Ten questions concerning the use of drones in urban environments,” Build. Environ., vol. 167,
no. August 2019, p. 106458, 2020, doi: 10.1016/j.buildenv.2019.106458.
[30] DJI Innovations, “Mavic 2 Pro & Mavic 2 Zoom.” https://www.dji.com/uk/mavic-2/info#specs (accessed Dec. 03,
2019).
[31] J. Moll et al., “Radar imaging system for in-service wind turbine blades inspections: Initial results from a field
installation at a 2MW wind turbine,” Prog. Electromagn. Res., vol. 162, pp. 51–60, 2018, doi:
18
10.2528/pier18021905.
[32] N. Hosseini, H. Jamal, J. Haque, T. Magesacher, and D. W. Matolak, “UAV Command and Control, Navigation and
Surveillance: A Review of Potential 5G and Satellite Systems,” IEEE Aerosp. Conf. Proc., vol. 2019-March, no.
March, pp. 1–10, 2019, doi: 10.1109/AERO.2019.8741719.
[33] G. Yang et al., “A Telecom Perspective on the Internet of Drones: From LTE-Advanced to 5G,” pp. 1–8, 2018,
[Online]. Available: http://arxiv.org/abs/1803.11048.
[34] G. Geraci, A. Garcia-Rodriguez, L. Galati Giordano, D. Lopez-Perez, and E. Bjornson, “Understanding UAV
cellular communications: From existing networks to massive MIMO,” IEEE Access, vol. 6, pp. 67853–67865, 2018,
doi: 10.1109/ACCESS.2018.2876700.
[35] J. Rendon Schneir et al., “A business case for 5G mobile broadband in a dense urban area,” Telecomm. Policy, vol.
43, no. 7, p. 101813, 2019, doi: 10.1016/j.telpol.2019.03.002.
of
[37] Intel Corportation, “Intel RealSense Depth Camera D400-Series,” 2017.
ro
[38] Flir, “Flir a35/a65 Tm Visualize Heat.” p. 2, 2018, [Online]. Available: www.flir.com/automation.
[39]
-p
TERADEK, “Bitrates,” 2019. https://teradek.com/blogs/articles/what-is-the-optimal-bitrate-for-your-resolution
(accessed Aug. 30, 2019).
re
[40] Dronecode Project, “PX4 Development Guide (v1.9.0),” 2019. https://dev.px4.io/v1.9.0/en/ (accessed Oct. 25,
2019).
lP
[42] SCADA, “SCADA Training Session for HOME Offshore,” Manchester, 2017.
ur
[43] H. Zhao, H. Liu, W. Hu, and X. Yan, “Anomaly detection and fault analysis of wind turbine components based on
deep learning network,” Renew. Energy, vol. 127, pp. 825–834, 2018, doi: 10.1016/j.renene.2018.05.024.
Jo
[44] M. N. Scheu, L. Tremps, U. Smolka, A. Kolios, and F. Brennan, “A systematic Failure Mode Effects and Criticality
Analysis for offshore wind turbine systems towards integrated condition based maintenance strategies,” Ocean
Eng., vol. 176, no. January, pp. 118–133, 2019, doi: 10.1016/j.oceaneng.2019.02.048.
[45] Google LLC, “Coral Dev Board datasheet,” 2019. https://coral.withgoogle.com/docs/dev-board/datasheet/ (accessed
Aug. 29, 2019).
[47] NVIDIA Corporation, “Jetson AGX Xaview Developer Kit,” 2019. https://developer.nvidia.com/embedded/jetson-
agx-xavier-developer-kit (accessed Aug. 29, 2019).
[49] M. Elbanhawi, A. Mohamed, R. Clothier, J. L. Palmer, M. Simic, and S. Watkins, “Enabling technologies for
autonomous MAV operations,” Prog. Aerosp. Sci., vol. 91, no. March, pp. 27–52, 2017, doi:
10.1016/j.paerosci.2017.03.002.
19
[50] A. Atyabi, S. MahmoudZadeh, and S. Nefti-Meziani, “Current advancements on autonomous mission planning and
management systems: An AUV and UAV perspective,” Annu. Rev. Control, vol. 46, no. July, pp. 196–215, 2018,
doi: 10.1016/j.arcontrol.2018.07.002.
[51] S. French, “DJI Market Share: Here’s Exactly How Rapidly It Has Grown In Just A Few Years,” The Drone Girl,
2019. http://thedronegirl.com/2018/09/18/dji-market-share/.
[53] GL Garrad Hassan, “A Guide to UK Offshore Wind Operations and Maintenance,” 2013.
[54] The Crown Estate, “Wind Data,” 2019. http://www.marinedataexchange.co.uk/wind-data.aspx (accessed Jul. 30,
2019).
of
https://www.businesselectricityprices.org.uk/retail-versus-wholesale-prices/ (accessed Nov. 07, 2019).
ro
[57] J. Carroll, A. McDonald, I. Dinwoodie, D. McMillan, M. Revie, and I. Lazakis, “Availability, operation and
maintenance costs of offshore wind turbines with different drive train configurations,” Wind Energy, vol. 20, no. 2,
pp. 361–378, 2017, doi: 10.1002/we.2011.
[58]
-p
UK Gov., “Contracts for Difference Allocation Round 3 Results,” 2019. [Online]. Available:
re
https://www.gov.uk/government/publications/contracts-for-difference-cfd-allocation-round-3-results.
[59] PayScale, “Average Wind Turbine Technician Salary in United Kingdom,” 2019.
lP
[60] CAA, “Unmanned Aircraft System Operations in UK Airspace – Guidance,” Cap 722, no. 5, p. 165, 2015.
na
[61] The European Parliament and the Countil of the European Union, “Regulation (EC) No 785/2004 of The European
Parliament and of The Council of 21 April 2004,” Off. J. Eur. Union, vol. L–138, 2004.
ur
Jo
20
HIGHLIGHTS
• A new cost model for predicting the Levelised Cost of Electricity from offshore wind
• Model informs wind farm operators on the costs of drone inspection techniques
• Model informs Drone Service Providers on the cost of higher levels of autonomy
• Greatest cost benefit comes from changing ropes to drones for visual inspection
• High levels of autonomy offer marginal benefit and is currently not worth pursuing
of
ro
-p
re
lP
na
ur
Jo