Mitigation of assembly constraints for floating offshore wind turbines using discrete event simulation

There is a large and increasing pipeline of floating offshore wind projects with total global floating offshore wind capacity projected to grow year on year by, on average, between 59 and 104 % in the 2020s. This will lead to competition for infrastructure resources, in particular, port facilities for the construction and marshalling of the floating foundations and turbines. It is likely that multiple ports will need to be combined to provide the necessary fabrication capacity for a floating offshore wind farm of commercial scale. To enable an efficient and coordinated utilisation of multiple fabrication ports, it is crucial to understand the likely duration of different assembly and construction activities at different locations. However, at present this task is difficult to perform using top-down estimation models, as commercial-scale floating offshore wind farms comprising many tens of units have not been built to date. In this work we present a methodology, based on discrete event simulation (DES) and time series analysis, to produce an explicit simulation-based estimate of assembly activity durations, which are sensitive to setting specific factors. Three example case studies are outlined to demonstrate the ability to capture the variation in activity duration due to resource availability, and the season and location of activity. The methodology will be of use to project planners as it can be used at an early stage in the project life-cycle to appraise and adopt different construction strategies.


Introduction
Floating offshore wind is a technology set to undergo a rapid and large-scale uptake.At the end of 2022 there were less than 30 floating offshore wind turbines (FOWTs) deployed across the world, totalling a cumulative capacity of under 200 MW [1].Despite the limited operational knowledge and refinement of this nascent technology, governments have set increasingly ambitious targets for floating offshore wind deployment by 2030.In the UK, the government has set a target of 5 GW of floating offshore wind capacity by 2030 [2], and, at the recent ScotWind round of seabed leasing in Scottish waters, 15 GW of floating offshore wind capacity was awarded [3].Globally, the compound annual growth rate of floating offshore wind capacity is projected to be between 59 and 104% over the next 10 years [4].Monopile [5] 750 Jacket [6] 270 Floating semi-submersible foundation [7]    The increase in the number of FOWTs to be deployed over a relatively short time span will inevitably lead to supply-based challenges.An area which poses substantial risks to floating offshore wind projects is the construction and installation phase [8], as the nature of the technology introduces novel challenges.Although FOWTs and established bottom fixed offshore wind turbines are similar in many ways, from a logistical point of view one of the most significant differences is the size and footprint of the foundations (see table 1).As an example of the implications of this difference, consider the transport arrangement of the two structures, shown in figure 1.The smaller size of the jacket foundations allows approximately ten to be transported to the marshalling port at once (figure 1a).By contrast, only a single FOWT semi-submersible foundation can be carried by a transport barge at a time (figure 1b).A similar size issue affects the storage capacity of the marshalling port, as the number of FOWT foundations which can be stored at one time is a fraction of the number of jackets.
To overcome the additional space demands, floating offshore wind projects will need to consider using multiple ports concurrently in the construction phase to meet deployment targets.This approach presents a much greater logistical challenge compared to the construction of bottom fixed offshore wind projects, as synchronising multiple construction activities in parallel will require a detailed understanding of FOWT construction processes, including required resources and the likely duration of each activity.The construction of FOWTs for a project represents a significant planning and logistics problem as three main assembly activities need to be performed on every FOWT before it can be transported and installed at the wind farm site, namely (a) assembly of the foundation, (b) loadout of the foundation with possible subsequent transit between different construction sites by sea, and (c) wind turbine generator (WTG) and tower integration with the floating foundation [11,12] (figure 2).To assist with the planning of Figure 2: Assembly activities in the construction of a floating offshore wind foundation these phases, it will be crucial for project planners to understand the likely duration of different assembly activities across ports over the year.However, it is difficult to get an estimate of this kind using conventional top-down approaches, as there have not been any floating offshore wind farms constructed in this manner to date.In this work we accordingly present detailed projectspecific methodology using discrete event simulation (DES) which results in a probabilistic representation of the duration of different assembly activities, accounting for the effects of resource availability, and the likely weather conditions at different ports in different seasons.Case studies are presented for each stage of the construction process, demonstrating the value of the method in understanding the effect of the these port-specific external factors.
Discrete event simulation is a well-established methodology which has been used for several decades to model process-orientated systems in the areas of logistics, construction and transportation [13].It is therefore well suited to the simulation of offshore wind construction and many commercial and academic tools have been created which address both the construction and maintenance of offshore wind farms.A review of state of the art models was performed in [14] and although there has been continuing activity in the area since that publication, the focus of these tools to date has been on bottom fixed offshore wind projects, and little attention has so far been paid to floating offshore wind farm construction, with the authors aware of only two papers which explicitly consider the topic.[15] models the construction of a floating offshore wind farm, and does account for weather-induced delay for transit between ports and WTG and tower integration.However, there is no consideration at all of the fabrication process of the foundation, and no representation and investigation of resource scarcity on total durations for different activities.[16] does consider the fabrication of the foundation explicitly.However, there is no attention to the effect of weather on activity duration, and again no consideration of resource availability.The approaches taken by both papers are therefore not sufficient to assess the multi-port construction approach proposed above, as such assessment requires explicit consideration of all stages of FOWT construction, accounting for the different key characteristics of each port, namely the level of scarce resource available and the location-dependent weather.This paper seeks to address this gap by proposing and demonstrating a methodology which will allow for such comparison of different ports, providing a valuable planning tool in a multi-port approach.
The remainder of the paper is structured as follows.Section 2 presents the methodology developed, and introduces the case studies which are used to demonstrate the value of this method when comparing project decisions.Section 3 presents the results of these case studies, which are then discussed in section 4, along with some current limitations of the approach.Finally, a conclusion is drawn in section 5, with reflections on the benefits of the method and outline of future work.

Methodology
To estimate the durations of different assembly activities, two techniques are used.First, each assembly activity is modelled using discrete event simulation (DES) [17].This model is then combined with time series analysis of historical weather data.Using Monte Carlo analysis, this provides a statistical understanding of activity sensitivity to the prevailing weather conditions at different locations and for different seasons.

Discrete event simulation (DES)
A DES models the system of interest as a collection of entities which can interact with one another.These entities each follow a process: a sequential set of actions which are triggered when the system reaches a certain state, known as an event.This event can be predetermined, such as when the simulation clock reaches a specific time, or can be dependent upon other entities performing actions, such as finishing an action and causing a resource to become available.A DES proceeds by way of a "simulation executive", a computational process that assesses the "future" time of all triggering events in each entity's process, advancing the simulation time to the earliest event, and then triggering the relevant action(s).This is repeated until all processes are complete or a predetermined simulation time has been reached [18].To model the construction of a FOWT using this framework, each component is modelled as an entity, which each follow a process to assemble themselves into a containing FOWT entity.The events which trigger the assembling actions are the completion of any predecessor activities (such as scaffolding erected, other components in place, etc.), availability of one or more required finite resources (such as welders, cranes, riggers, etc.) and a suitable future weather window long enough to perform the action.Each action's duration can be estimated in a number of ways, including bottom-up, parametric, or Delphi methods.Once an action is complete, the system is updated, triggering other actions.Each port is modelled as an encapsulating entity which possess all the required resources.The level of these resources at each port are specified by the user.

Time series analysis with Monte Carlo
As noted above, some tasks are dependent upon certain weather conditions to proceed.Typically, these limits are safety-related, such as wind limits for cranes, and wave limits for sea-based operations.To investigate the impact of weather-related delay on the assembly activities, timeseries analysis is performed on one-hour resolution historical weather data retrieved from the ERA5 data service [19], and incorporated into the DES.An example of an acceptable weatherdependent window identification strategy is summarised in algorithm 1.For any task with a weather limit, the algorithm checks if there is a suitable weather window forecast from the present time to cover the entire duration of the task including contingency, taking into account relevant alpha factors (a factor applied to the weather limit to account for uncertainty in the forecast), and waits until the weather is suitable before proceeding.In this way, the time taken to wait for a suitable weather window can be estimated.
Using a Monte Carlo technique, the above process can be implemented for many thousands of starting points in the historical time series.Statistical analysis of the results provides probability distributions of delay for the activities for different seasons and locations.

Case studies
Three case studies are considered to demonstrate the value added from analysis using the methodology.Each case study provides an example of the impact of different project choices on Algorithm 1 Establishing a suitable waiting period T waiting for a task INPUTS : Initial simulation time t 0 , simulation duration t sim , simulation timestep ∆t, wind velocity magnitude limit U max , duration of weather window required d, task starting time t start , wind data time series u(t, T ), where t the start time of a time series spanning a period T .
if max (u(t, d)) > U max then: end if 8: end while 9: T waiting = t start − t 0 the duration of different assembly activities.

Foundation assembly: activity procedure and resource availability
The foundation under consideration is a three column semi-submersible, with identical sides made up of two bottom braces joined with a K-node, two diagonal V-braces, and an upper brace, as in figure 3.Each component is assumed to be positioned in place using either a self propelled modular transporter (SPMT) or a crane, depending on the altitude of its final alignment, and then tack welded into position.The assumed process sequence for each side of the foundation is as follows: (i) Position corner columns (ii) Position and tack weld K-node (iii) Weld and test both ends of K-node (iv) Position and tack weld V-braces (v) Weld and test upper end of V-braces (vi) Weld and test lower end of V-braces (vii) Position and tack weld upper brace (viii) Weld and test both ends of upper brace The crane or SPMT is assumed to only be required until the tack welding has been completed.The cranes are subject to a 9 m/s wind limit, and the SPMTs are subject to a 20 m/s wind limit.The welding duration is calculated according to the following equation [20]: where T W is the time to weld, a w is the weld area (equal to plate thickness for butt welds), and L w is the weld length.Two resources are required for this process: welding teams and equipment to position the components.Therefore, to investigate the effect of resource availability, the simulation was performed iteratively with varying numbers of cranes and welding teams.After welding, nondestructive testing (NDT) such as ultra-sonic or magnetic particle inspection is required to verify weld quality, ensuring it is free from defects.According to applicable technical standards, a minimum period of time must pass to allow hydrogen to dissipate from the weld before NDT testing can be performed.This lasts 48 hours as standard, but can be reduced to 24 hours if "consistent low failure rate of delayed cracking has been documented for the materials and welding consumables in question" [21].Simulations with both 48 and 24 hour waiting times were therefore performed to investigate the effect of this reduction.

Transit between ports: seasonal variation in duration
The expected delay for transit is calculated for a route between a construction port in the north east of Scotland and a construction port in the Orkney Islands.The foundation is assumed to be carried on a semisubmersible barge, as in figure 1b.The assumed speed of the barge is 5 knots, with a wind and significant wave height limit of 15 m/s and 2.5 m respectively.The appropriate alpha factor (used to account for forecast unreliability) was applied to these limits when running the weather forecasting subroutine [22].The route has a calculated distance of 129 nm.To account for the large geographical area of interest, all weather nodes along the route are considered with the appropriate node checked to account for the vessel's progress along the route.

WTG and tower integration: spatial variation in duration
The total duration for the integration of the WTG and supporting tower with the foundation is calculated by the method described.In this case study the tower supporting the WTG is assumed to be made of three sections.The activity is modelled using the following assumed sequence of tasks: (i) Position and attach tower section 1 (ii) Position and attach tower section 2 (iii) Position and attach tower section 3 (iv) Position and attach nacelle (v) Position and attach blade 1 (vi) Position and attach blade 2 (vii) Position and attach blade 3 All tasks are assumed to be performed using a crawler crane and as such are subject to a 9 m/s wind limit.

Results
Results are presented for each case study considered.4 highlights the effect of resource constraints on total foundation construction time.The construction time is shown to be very sensitive to the number of welding teams working on the foundation, but there is almost no effect on the total time when the total number of cranes available is increased.This holds even when there are large numbers of weld teams.Although there is no occurrence here, figure 4 therefore demonstrates the ability of the proposed method to investigate resource level interaction.Reducing the time to wait before NDT is performed from 48 hours to 24 hours leads to a reduction in total duration of 13% when 6 welding teams are available to work on the foundation.

Transit between ports: intra-annual delay
The distribution of delay calculated from simulations of the transit case study starting in either in the months of January or July are shown in figure 5a.The large majority (83%) of delays calculated for the month of July are equal to zero, compared to 23% of simulations resulting    5b, where the boxes represent the interquartile (p25-p75) range, and the horizontal lines are the median (p50) delay.The p50 delay ranges from 61 hours in January, to zero in the summer months (June, July, August).

WTG and tower integration: spatial variation
Figure 6 shows the variation in the summer contingency factor, α Summer , for performing the WTG and tower integration activity at different locations around the British Isles.α Summer is calculated follows: where D M w is the average duration calculated for the month M , and D nw is the duration of activity with no weather delay.There is a clear variation in this factor for the locations considered, ranging from the lowest value 1.42 in the north east of Scotland, to 2.4 on the northern island archipelagos of Orkney and Shetland.

Discussion
The approach adopted in this work is discussed by exploring each of the above case studies in turn.

Exploring the effects of assembly procedure and resource availability
Consideration of figure 4 shows the types of useful information that can be produced by a DES model of a complex construction process.Although weather is not considered in this scenario, there are several factors which affect the duration of the activity which are directly under a project planner's control.In the case study under consideration, the total time for construction is affected very little by the number of cranes which can be used in the construction, despite the fact that nine of the components require a crane to be lifted into position.This is due to the assumption that the crane is only required whilst the component is moved into position and tack welded, a process assumed to last 10 hours.In contrast, the time to perform the weld is far greater, at over 100 hours for most connections.As each welding procedure requires a weld team, and many welding operations can be performed simultaneously, the reduction in total time with increasing weld teams is the governing factor here.
It should be noted that a similar trend will not necessarily exist for other FOWT foundation designs.Designs which require cranes for longer periods, or with shorter welding periods, may exhibit a different relationship between the total foundation assembly time and the number of both crane and welder resources available.To capture these effects for different designs, a unique DES of that assembly procedure should be carried out.
The benefit of such an analysis to a project planning process is that at an early stage in the project plan it is easy to estimate the optimal number of resources required.

Accounting for seasonal weather in project planning
As demonstrated by the analysis on transit delay, the time taken to perform an assembly related activity at one point of the year is not the same as at another.This type of information can be very useful to anyone planning marshalling activities as it gives an insight into the relative rates of progress of turbines through the whole construction process in different seasons.The variations can then be incorporated into the marshalling plan, to make optimum use of resources.Additionally, knowledge of the variation in rate is useful to ensure that sufficient storage space is available to deal with the queue of partially assembled turbines forming in the system.Knowledge of the underlying distribution of delay is also valuable, as planners can determine what level of risk of delay (p50, p75, p90, etc.) is acceptable, and design the system accordingly.

Accounting for weather when selecting assembly locations
Comparing the effect of weather on activity duration for different locations shows a significant impact on the resulting times, even when considering a reasonably small geographical area such as the UK, and in months with less adverse weather conditions.The knowledge gained from an analysis of what ports are more adversely affected by weather could be extremely useful when deciding what ports to use for a project, and deciding which tasks to perform at which location.As an example, if a project was planning to use a port on the south west of Wales (α summer = 2.15), and a port on the south east of Ireland (α summer = 1.69), it might make sense to perform the highly wind sensitive WTG integration in the Irish port, and a less wind-limited activity like the foundation assembly at the Welsh port.In this way, it is possible to mitigate some project risks during the planning process.

Factors affecting accuracy of results
Whilst the results of the methodology can yield valuable results, care should be taken to understand the limitations of the methods and data used.

Limitation of DES
As shown in the foundation assembly case study, the sequence of events modelled, and their inter-relatedness, has a very significant impact on the resulting durations.Therefore, in order to present a realistic estimate of the different processes, the sequence should be established to a low level of detail, i.e. with high specificity of the required tasks, including scheduling interaction with other components, likely duration and resource requirement.This is not always possible in the early stages of a project, when the approach in this work would be most useful.Therefore, additional and out of sequence work may need to be performed at an early stage in the project life cycle to refine the assembly activity procedures to the level of detail required for the DES analysis, so that the added value of this approach can be realised.

Limitation with weather data
The weather data used for the prediction of weatherrelated delay is historical weather data from the ERA5 weather model.There are problems related both to the fact that the data is model derived and that it is historical.First, the wind data used in the analysis is derived from the ERA5 model, a mesoscale model "produced using 4D-Var data assimilation and model forecasts in CY41R2 of the ECMWF Integrated Forecast System (IFS)" [23].As such, the data produced is not an exactly accurate representation of historical weather data for a single point, particularly if, as in the case of wind data, there are significant changes due to the local topography [24].Second, even if this problem is overcome by using local data, a problem may remain in using historical weather data as a basis for making predictions about future weather patterns, in the context of climate change.If climate change increases average wind speed, or increases the occurrence of storm events, then the assumption that weather dependent durations will be statistically similar in the past and the future may not be valid.However, based on the timescales which are considered to be relevant for this work (10-20 years from present), and the lack of consensus around the effect of climate on future wind resource [25], no correction is made for the effects of climate change at this stage.

Conclusion
We have demonstrated a methodology to estimate, using an explicit representation of the problem under consideration, the durations of different activities in the assembly of FOWTs.Using DES combined with a Monte Carlo approach to a time series analysis of reanalysed weather data, the expected duration of foundation assembly, transit between assembly ports, and the integration of a WTG and tower with a FOWT foundation has been calculated.Additionally, as demonstrated by the case studies, not only the mean duration, but also the underlying distribution of durations for activities in different settings can be revealed by the method.These results will be of great value to project planners, and will allow for some project risk to be designed out.
Further work will address some of the limitations highlighted by a validation of the method's assumptions, potentially using data from industries adjacent to floating offshore wind, such as oil and gas, and bottom fixed offshore wind.Further research effort will also be spent on linking the models of construction activities into a comprehensive model of the entire assembly system for multiple units.This will enable better understanding of the total flow of floating offshore wind units through the construction phase, allowing planners to better understand the risks associated with certain plans. 3500 (a) Transit and storage of jacket foundations.Photo: OHT[9] (b) Transit of FOWT foundation.Photo: Boluda[10]

Figure 1 :
Figure 1: Logistical arrangements for marshalling different offshore wind turbine foundations

Figure 3 :
Figure 3: Foundation design considered in simulation

Figure 4 :
Figure 4: Case study results demonstrating the effect of resource constraint on foundation assembly duration, showing 24 and 48 hour hydrogen dissipation times (a) Distribution of delay for transit starting in January and July (b) Comparison of the distribution of delay for a transit starting in a given month across the year

Figure 5 :
Figure 5: Results of transit simulation

Figure 6 :
Figure 6: Mean contingency factor for WTG integration for locations across the British Isles

Table 1 :
Footprint of different offshore wind foundation technologies