I spy with my little eye, or: using satellite data to investigate performance of offshore wind farm installation campaigns
Offshore wind is rapidly becoming a major source of energy. With more than 5700 offshore wind turbines currently operational, a significant amount of experience has been gathered by industries and nations alike. The wind turbines currently being operated in the world's seas are, however, far from homogenous
To ensure reproducibility, the complete chain of processing is available via a shell script (run_analysis.sh).
To run the analysis from scratch, please run the following command in the base of this repository. Beware:
this may take a considerable amount of time.
./run_analysis.sh --verboseThe corresponding paper will be build using latexmk in the subdirectory manuscript
This section describes the types of data this project is concerned with, the structure of the data and the relationships between the different data types. After a short overview, each data set is described in depth.
Central to the project are tracks of offshore installation vessels, spanning several years, covering more than a dozen offshore wind farm installations. These wind farms in turn form a second (meta) data set, as each offshore wind farm consists of a different type of turbine, has a different location, a different number of turbines and a different layout. Finally, for each offshore wind farm installation campaign, metocean data, such as wind speed and direction, wave height, wave direction and wave period, are compiled.
Automatic Identification System (AIS) tracks of nine offshore wind farm installation vessels have been procured from a data broker (marinetraffic.com). A track refers to a collection of unique AIS records. Each record contains a unique time stamp and the vessels ID as well as latitude, longitude, speed, heading and the vessel's status. The vessel AIS tracks spread over several years and thus contain the port calls, sailing legs and of course offshore wind farm installation campaigns. Each vessel is identified by it's unique ID. There are several types of IDs available: IMO, MMSI, the vessels call sign and it's name. In this project, we use the MMSI ID in conjunction with the vessel's name to uniquely identify a vessel. Note, that the MMSI can change, if the vessel is reflagged and that there several vessels can carry the same name.
The following vessels and their respective tracks are available
| MMSI | Name | data availability |
|---|---|---|
| 218389000 | Thor | 2010 - 2021 |
| 218657000 | Vole au Vent | 2013 - 2021 |
| 253355000 | Vole au Vent | 2013 - 2021 |
| 219019002 | Sea Challenger | 2013 - 2021 |
| 229044000 | Brave Tern | 2012 - 2021 |
| 229080000 | Bold Tern | 2013 - 2021 |
| 235090598 | Blue Tern | 2015 - 2021 |
| 215655000 | Blue Tern | 2015 - 2021 |
| 245179000 | Aeolus | 2010 - 2021 |
| 245924000 | MPI Adventure | 2010 - 2021 |
| 246777000 | MPI Resolution | 2010 - 2021 |
| 253609000 | Taillevent | 2010 - 2021 |
The raw vessel data as delivered by the data broker has been split into at least one csv file per year,
e.g.: 2010.csv. For some years, however, the data has been split into two files per year, e.g. 2012A.csv and 2012B.csv.
The data files are structured into eight columns and as an example, the first three lines from the file 2010.csv are shown:
MMSI;LAT;LON;SPEED;HEADING;COURSE;STATUS;TIMESTAMP
245924000;53.53467;0.284;81;511;303;0;2010-01-01 06:54:38.000
245924000;53.5505;0.1855;61;511;258;0;2010-01-01 07:25:38.000
MMSI is the ID of the vessel, LAT and LON correspond to the vessel's latitude and longitude,
SPEED is the vessel speed in knots, HEADING the orientation of the vessel with respect to due North,
COURSE the vessels current planned direction of motion. STATUS is unknown and TIMESTAMP corresponds to the
date and time of the record in UTC.
The raw data can be found in the data/marine-traffic/raw directory.
Data/marine-traffic/sanitized contains the sanitized AIS tracks per vessel as
csv files. These files span all available years for a given vessel and are organized into 7 columns. As an
example the first three lines from data/marine-traffic/sanitized/215644000_blue-tern.csv are shown:
timestamp,epoch,latitude,longitude,speed,heading,course,status
2020-06-15 06:52:10+00:00,1592203930,53.45238,6.81926,0.0,287.0,132.0,5
2020-06-15 07:13:11+00:00,1592205191,53.45235,6.819262,0.0,287.0,132.0,5
Central to extracting offshore wind farm installations from vessel tracks is knowledge about the wind farms and their
corresponding installations. This required meta-data must include the wind farm name, location
(latitude and longitude of wind farm's centroid), number and type of turbine, turbine manufacturer, installation
beginning and end, type of foundation and installing companies.
Wikipedia provides a non-exhaustive list of
offshore wind farms per world region. The python script get_windfarms_wikipedia.py downloads the latest version
of the wikipedia tables, converts them into a machine readable format and saves the table as a csv in the data directory.
Wind and waves have a significant impact on offshore wind farm installations. Wind and waves cause forces, commonly referred to as environmental loads, acting on the installation vessel, the crane and, of course, the wind turbine components undergoing installation. These loads in turn lead to adverse structural responses and may even cause collisions between components when assembling turbines. Offshore wind farm installations are thus subject to wind and wave limits during installation operations. Different steps in the installation procedure can have different limits. It is vital to understand though, that wind and wave limits are only a proxy to estimate the real limits, imposed by the structural response of the components being installed.
Only a limited number of measurement stations are available world wide, hence, the usage of hind-cast models has seen a increase in usage over the last years. These models incorporate real-world, sparsely-scattered measurements into a numerical model of the sea and the atmosphere. Running these models then yields metocean data for any place on the earths surface with varying accuracy.
A commonly used model is the model developed and provided by the European Union's European Centre of Medium-Range Weather Forecasts (ECMWF). The data produced by ECMWF's numerical model is commonly referred to as ERA5 data. The ECMWF provides a volume-limited API to download ERA5 data. A more thorough description of ERA5 data can be found here. ECMWF also provides a python package that can be used to request data. More information on how to access ERA5 data can be found here
data
installations
installations.sqlite # sqlite data base containing all successfully identified installations
windfarm-name_vessel_start_end.csv # installation data as csv
...
marine-traffic # contains AIS vessel tracks
raw # raw AIS data
2010.csv
2011.csv
...
clean # clean AIS data decomposed into per vessel time series
215644000_blue-tern.csv
218389000_thor.csv
...
clustered # clustered AIS vessel tracks
215644000_blue-tern # folder per vessel
clusters
215644000_blue-tern_cluster_0.csv
215644000_blue-tern_cluster_1.csv
215644000_blue-tern_cluster_2.csv
...
windfarms
windfarm_0
215644000_blue-tern_windfarm_0.csv
215644000_blue-tern_windfarm_0_turbine_0.csv # AIS vessel tracks of single turbine 1
215644000_blue-tern_windfarm_0_turbine_1.csv # AIS vessel tracks of single turbine 2
...
windfarm_1 # wind farm candidate 2
215644000_blue-tern_windfarm_1.csv
215644000_blue-tern_windfarm_1_turbine_0.csv
215644000_blue-tern_windfarm_1_turbine_1.csv
...
...
218389000_thor
clusters
cluster_0.csv
cluster_1.csv
cluster_2.csv
...
windfarms
windfarm_0
218389000_thor_windfarm_0.csv
218389000_thor_windfarm_0_turbine_0.csv
218389000_thor_windfarm_0_turbine_1.csv
...
windfarm_1
218389000_thor_windfarm_1.csv
218389000_thor_windfarm_1_turbine_0.csv
218389000_thor_windfarm_1_turbine_1.csv
...
metocean # metocean data
metocean-data.sqlite # sqlite database listing available metocean data
windfarm-name_coord-lat_coor-lon_start_end # metocean record per wind farm
wind-farms # wind farm data
raw # raw wind farm data from e.g. wikipedia
clean # clean wind farm data as csvs
wind-farms.sqlite # sqlite database containing all wind farm data