*3.3. Infrastructure Pre-Selection*

3.3.1. Feasible Infrastructure

On a first stage, feasibility functions are used to assess whether vessels, ports and equipment, listed in the DTOceanPlus catalogs, meet the absolute minimum requirements for the job (e.g., sufficient vessel deck area, adequate terminal dry dock dimensions, acceptable ROV depth rating, etc.). These functions are simple mathematical Boolean formulations that filter out the maritime infrastructure noncomplying with the previously defined logistic requirements (see Section 3.2). A default safety factor of 10% is implemented in the feasibility functions to reflect uncertainties and account for potential margins of of error, although this value may be modified by the user.

The identification of feasible ports and equipment follows a simple elimination process, where instances of the port terminals and equipment databases are discarded based on the operation requirements. The port terminal database consists of 203 terminals, from 12 different EU countries, with 21 parameters, including name, type, country, location, terminal entrance width, draught, maximum load, and terminal area, to name a few (see Table 5). Similarly, six main types of equipment (e.g., cable burial tools, piling hammers, and ROVs, to name a few) are considered in DTOceanPlus and included in the equipment catalogs. However, the fleet selection process is slightly more complex.

#### 3.3.2. Fleet Selection Methodology

There are numerous approaches to conduct a given offshore operation. Devices may be transported from port to site, on deck of a large crane vessel with adequate cranes to carry out the installation procedures. Alternatively, they may be loaded at port onto a transport barge, which would in turn be towed a set of tugs or a Anchor Handling Tug and Supply (AHTS) vessel. In case devices were structurally designed for the purpose, they may be individually wet-towed directly to site. Low draft floating converters may be transported using a semi-submersible vessel, with the capability of ballasting down and submerging its deck to unload the converter in the water. Based on the existing experience in the offshore renewable energy sector, a vessel combination (VC) database was developed, namely, for the device installation operation (see Table 6), featuring combinations of different vessel types in different quantities and under different roles. For simplicity, for each vessel combination, three major vessel roles were defined, with different evaluation criteria:


The fleet selection algorithm follows a two-stage elimination process. It starts by discarding unsuitable vessel combinations that do not meet project requirements or user preferences (e.g., wet tow is not allowed). Then, for each feasible vessel combination, the fleet selection algorithm searches in a vessel cluster database for vessels that meet the technical requirements associated with the attributed roles (e.g., sufficient deck area, sufficient bollard pull to wet tow the device). A given vessel is deemed "feasible" if capable of performing the minimum work criteria (e.g., in case of on deck transportation, vessel must have sufficient deck area to accommodate at least one device per trip).

In DTOceanPlus, a vessel cluster database was compiled based on the statistical analysis of an original database with 14,847 vessels and 46 technical parameters. The very large size of the original database ensured the representativeness of the considered vessel list, although prohibited directly using it in DTOceanPlus due to data privacy issues and to keep computational requirements manageable. In the vessel cluster database, vessels sharing a large number of similar characteristics were grouped into clusters, using the K-Means unsupervised machine learning algorithm [48]. For each technical parameter of each vessel cluster, the p25, p50, and p75 statistical values were computed and stored. Deliverable D5.8 [46] provides more information about the vessel clustering process.


**Table 5.** Example port terminal entry of the DTOceanPlus port catalog.


**Table 6.** Example vessel combinations for the device installation operation, stored in the DTOceanPlus Vessel Combinations catalog.

#### 3.3.3. Infrastructure Matching

Once feasible infrastructure have been identified, infrastructure-matching functions assess the compatibility between each infrastructure type in the context of an integrated solution. In this step, independently feasible but incompatible infrastructure solutions are discarded (e.g., port entrance width must be larger then vessel beam, port draught must be compatible with vessel draft, etc.). Once the infrastructure matching algorithm has been run, suitable infrastructure combinations are produced to be further analyzed in respect to total operation duration and ultimately, costs.

#### *3.4. Definition of Activity Sequence*

In the LMO module, operations (e.g., foundation installation) are broken down into smaller, uninterruptible tasks that must be carried out, referred to as activities (e.g., mobilization, transit, and positioning), with specific durations and weather restrictions. For each operation, activity flowcharts were developed, featuring the activity blocks, precedence rules (i.e., which activity comes next), and condition nodes which define multiple potential paths that an operation may take. Condition nodes were defined as static, when based on previously defined component types and operation methods (e.g., foundation type is a pile, transportation method is dry), or dynamic, when dependent on the considered infrastructure solution and operation stage (e.g., vessel is already full or not). In the flowcharts, activities may have a constant duration, or a dynamic duration when the activity length depends on external criteria such as distance and vessel transit speeds (for transits and tows) or soil conditions (e.g., for piling activities such a "Seafloor drilling"). In Table 7, the activity flowchart of the foundation installation operation is presented as an example.

In Tables 8 and 9, the cable burial and piling speeds are presented, respectively, for different soil conditions. Activities, such as piling and cable burial, have specific speeds that are highly dependent on the seabed geology. In the Logistics and Marine Operations module, the piling and cable burial speeds were compiled and adapted from the literature review carried in the original DTOcean project [45,49], for the considered soil types, as defined by Kervella, Y. [50].

Based on the operation flowcharts, specified project characteristics, and infrastructure solution, a sequencing algorithm computes the full sequence of activities, from start to finish, that must be carried out to perform a given operation. Flowcharts are stored as tables in the operation catalogs, allowing for modifications to the durations, weather limits, and sequencing, by advanced users. This activity sequence is then fed into the weather window model described in Section 3.5.


**Table 7.** Flowchart of the foundation installation operation, featuring activity blocks, precedence rules, and condition nodes (shown in italic).

**Table 8.** Cable burial speeds (m/h), for different soil types and cable burial methods, stored in the DTOceanPlus operations catalog.



**Table 9.** Pile installation speeds (m/h), for different soil types and piling methods, stored in the DTOceanPlus operations catalog.

#### *3.5. Weather Window Model*

Weather window analysis is a crucial step in the strategic planning of marine operations in order to estimate potential weather delays and operation costs. The most common approach is to simulate a project subject to several years of historical environmental conditions, being commonly referred to as hindcast analysis [51]. Given the random nature of the met-ocean conditions at a given site, sample size must be sufficiently large to appropriately capture the potentially large annual variability. Even though more is better, 20-years of continuous weather data is a commonly accepted reference. As maritime operations are typically planned on hourly basis, DNV standards recommend linearly interpolating the raw met-ocean conditions when necessary to generate an hourly time series [52]. Subsequently, the time series of met-ocean conditions can be analyzed as a single continuous record.

The environmental conditions observed at a given offshore location can be understood as a multivariate stochastic process [53–55], whereas each environmental parameter (wind speed, significant wave height, peak wave period, and current speed) is interdependent and can be described by statistical distributions with specific joint probabilities but clear ensemble seasonal trends [56,57]. Even though cyclic patterns may be observed throughout the year (e.g., the summer season is typically calmer than winter, even though summer storms should not be overlooked), it may be reasonable to assume data stationarity for smaller time periods [58]. This method is known as piecewise stationarity and consists of grouping the entire met-ocean time series by seasons or months and carrying out separate calculations. The assumption weather data stationarity implies that the statistical properties (e.g., mean, variance, and autocorrelation) of the historical dataset are constant, and is typically assumed reasonable for fixed monthly blocks.

Following a hindcast simulation approach, the underlying principle of the LMO's weather window algorithm consists of attempting to schedule the specified operations in the past. Once the sequence of activities, durations, and weather restrictions have been specified for each operation (see Section 3.2), the algorithm attempts to iteratively initiate the operation in different time-steps of the historical time-series of met-ocean conditions, each iteration corresponding to a different simulation. Initial time-steps are randomly selected using a Monte Carlo approach, taking as user input the percentage of time-steps to analyze in each month (where 100% corresponds to analyzing the entire time-series). For each simulation, in case both the first and subsequent time-steps are deemed workable (i.e., OLCs are met) for a period that is equal or longer than the entire operation duration, then the operation can be carried out without any delays. Otherwise, waiting on weather is required, which may include waiting at port (WAP), and/or waiting on site (WOS), i.e., between consecutive activity blocks. The waiting on site is defined with a maximum duration and weather limit, which may not be exceeded.

For an operation with *n* activities, starting in time-step *t*, the total operation duration *dop*.*total* would be defined as shown in Equation (1) below, where *dnet*,*<sup>i</sup>* refers to the net duration of activity *i* of the operation.

$$d\_{op.total}(t) = \text{WON}(t) + \sum\_{i=1}^{n} d\_{net,i} = \sum \text{WAP} + \sum \text{WOS} + \sum\_{i=1}^{n} d\_{net,i} \tag{1}$$

Assuming monthly stationarity for the weather conditions, the waiting times calculated for each initial time-step can be grouped and statistically analyzed in monthly blocks. Given that the monthly waiting on weather values do not follow a normal distribution, the statistical parameters such as median (p50) and the interquartile ranges (p25–p75) can be used to estimate the expected value and quantifying statistical dispersion. As an example, Figure 4 shows a hypothetical non-exceedance distribution plotted for a given operation "op.A", considering all WOW values that occurred in every month of February of the entire 20-years long time series. As shown in Figure 4, there is a 50% probability that the waiting time for the specified operation will be equal or lower than about 28 h, whereas the p25 and p75 values are equal to 22 and 38 h, respectively. According to the estimated interquartile range, there is a 50% probability that the waiting time for this operation will be in the range of 22–38 h, for the month of February.

**Figure 4.** Illustrative representation of the non-exceedance probability of waiting on weather for an example operation in a given month.

For each operation, the weather window model thus computes monthly weather window statistics featuring the expected weather delay (p50) and resulting total operation duration for the different months of the year. The advantage of calculating the weather delays for each month of the year is that potential cost-reduction approaches, such as changing the starting month or optimizing the sequence of operations, may be unveiled to the user. The monthly weather window statistics are illustrated in Table 10 for a given operation with a net duration of 30 h.

**Table 10.** Monthly weather window statistics, in hours, for a given operation with a net duration of 30 h when scheduled in different months.


**Figure 5.** Example regression of the daily charter rates for crew transfer vessels (CTVs) as a function of the vessel's length overall (LOA), based on existing database.

### *3.6. Calculation of Operation Costs*

### 3.6.1. Vessel Costs

Vessel costs play a large role in the total costs of an offshore renewable energy project. Total vessel costs can be broken down into vessel chartering and fuel expenditures. The daily operating costs per day can be calculated as shown in Equation (2).

$$
\mathfrak{c}\_{\text{vescl}} = \mathfrak{c}\_{\text{charter}} + \mathfrak{c}\_{\text{fuel}} \tag{2}
$$

#### 3.6.2. Daily Vessel Charter Rates

The cost of chartering a given vessel depends on several factors, such as vessel characteristics and capabilities, as well as surrounding market conditions. Contract duration and contractual frameworks typically also play a role. Smaller tonnage vessels such as CTVs, tugs, and survey vessels are commonly chartered out on a time charter basis (e.g., BIMCO Supply time [59]) with a clearly defined vessel day rate. However, larger vessels such as jack-up vessels, crane vessels, and cable laying vessels are mostly hired as part of comprehensive service agreements such as EPCI2 or T&I3 contracts (e.g., FIDIC or Logic [60]). In order to be able to compare different types of contracts, average daily charter costs that exclude consumables such as fuel and harbour costs, were used.

Vessel characteristics such as age, size, crane capabilities, deck area, dynamic positioning (DP) equipment, and engine power are known to have an impact on the total vessel costs. Based on guidance from Global Renewable Shipbrokers (GRS) [61], a offshore vessel broker, major cost drivers for the vessel charter rates were identified for each vessel type. Even though the vessel charter rates are dependent on a large number of variables, for simplicity and to provide a first cost estimate, vessel charter costs were modeled as a function of a single parameter for each vessel type. As shown in Figure 5, cost functions that model charter day rates for the different vessel types were then derived, based on a curve fitting applied to database points gathered from: (i) DTOcean 1.0 vessel database, (ii) WavEC's in-house vessel database, (iii) cost figures provided by *ECN* [62] and *GRS*, (iv) from industry expert experience. Different regression models, including linear, polynomial, exponential, logarithmic and piece-wise regressions, were adjusted to find a best fit based on the R-squared coefficient, while eliminating fits that result in cost inflections within the analyzed domain. The resulting cost functions were compiled in Table 11. It can be seen that even though charter price variability is not fully explained by a single parameter, important relationships were obtained, with the potential to guide vessel selection decisions.


**Table 11.** Daily charter rate regression curves, for different vessel types, in Euros, as a function of their input parameters (x).

#### 3.6.3. Daily Vessel Fuel Costs

Given that the considered vessel charter rates excluded fuel costs, vessel fuel consumption had to be estimated. Fuel consumption contributes significantly to the total operation costs, but also to the emissions and carbon footprint of the project. Total vessel fuel consumption depends on several different factors, namely number of engines (main and auxiliary), engine power, engine efficiency, operation duration, mobilized ancillary equipment, transit speed and distance, as well as weather conditions. In order to provide a first fuel consumption estimate, the LMO module calculates the average vessel fuel consumption per day as

$$f\_{\mathbb{S}^5} = \text{TIP} \cdot \text{ALF} \cdot \text{SFOC} \cdot 24 \cdot \frac{1}{1000^2} \tag{3}$$

In Equation (3), TIP is the vessel's total installed power (in kW), ALF is the average load factor, and SFOC is specific fuel oil consumption (in g/kWh) [63]. According to the ship broker's experience in vessel chartering for offshore wind projects, an average load factor of 80%, and a specific fuel oil consumption of 210 g/kWh were indicated as reference values. However, these values may be modified by the user. The daily fuel costs can thus be estimated by multiplying the daily fuel consumption *fcs* by a reference price of fuel, as shown in Equation (4). In respect to the fuel price *pf uel*, the marine diesel oil (MDO) price in the port of Rotterdam, 515 €/ton, was taken as a reference [64]. However, when

running the LMO module, this value may be modified by the user to reflect different fuel prices or even other fuel types such as heavy fuel oil (HFO).

$$
\omega\_{fuel} = f\_{\rm cs} \cdot p\_{fuel} \tag{4}
$$

#### 3.6.4. Equipment Costs

For a given operations, the equipment costs can be simply calculated as the product of the operation duration and the sum of the daily (and/or half-day) renting cost of the selected equipment for that operation. Daily and half-day renting costs figures are available in the equipment databases.

#### 3.6.5. Spare Part Costs

For O&M operations, in case of component failure, the cost of the spare components are calculated using the costs of a new component, as designed by other modules (or introduced by the user) and compiled in the Bill of Materials (BOM).
