GreenSteam, a marine data intelligence company with offices in Denmark, the UK, and Poland, offers a service to build digital models of how a specific vessel’s fuel consumption changes based on different parameters. This model can be used to give advice to seafarers about how to adjust speed or trim and when to clean the hull for optimum fuel consumption.
The company is majority owned by BP Castrol.
The software “ingests” all of the relevant available data about the ship to build the model, distilling large amounts of data into something manageable.
Typically it needs 3 months of data to build the model, so it has data about the vessel’s performance in a range of sea conditions.
Currently there are about 500 vessels on GreenSteam’s platform. Half of those vessels just send noon day report data, showing the fuel consumption every over the past 24 hours, says Simon Whitford, COO of GreenSteam.
It also integrates the 24h data with high frequency AIS data and meteorological data describing the sea state, wave height, sea surface temperature, for each point of the ocean at different times during the vessel’s operation.
For the most sophisticated vessels deploying GreenSteam’s dynamic trim optimiser, wave height can be measured with a radar device on the bridge wing, which is more accurate than hull-mounted pressure sensors.
In this way, GreenSteam builds a model of the vessel’s performance at 10 minute intervals. GreenSteam also has many vessels which automatically collect high frequency data direct from sensors – the key is being able to encompass every ship regardless of the number of sensors or the frequency of data collection – that is GreenSteam’s design principle.
That said, “the quality of the model ultimately depends on the quality and frequency of the data you have,” Mr Whitford says.
Many vessels have torque meters measuring the force to rotate the shaft, taken from a sensor in the shaft. This can be used to improve the model. “We gather the data whenever we can get it,” he says.
Around 70 per cent of a vessel’s fuel consumption is unavoidable – what a perfectly optimised vessel on the calmest of seas would consume. A further 15 per cent of consumption is the vessel overcoming the impact of the waves and wind, and the last 15 per cent is down to vessel optimisation - the impact of fouling and trim (how the vessel sits in the water, whether the bow or stern are higher) and the speed of the vessel.
“It is a very tricky computational task to work out how fuel consumption is affected by speed or trim when these are just 2 of around 13 factors affecting the vessel which are in constant flux but splitting up or as we say, decomposing these various factors with accuracy is unavoidable to competently target vessel performance optimisation.” Mr Whitford says.
The model continually evolves as data is added.
“It ingests data, and keeps trying to find correlations, which get better over time,” he says. “After a reasonable period of study – looking at the vessel’s performance in many different sea states and operating parameters you get a very accurate vessel specific picture.”
“For example, when it has enough data, the model has calibrated how each of the 13 factors drive vessel fuel consumption, even when (of course) there are all changing at the same time – that is why we need machine learning.”
Return on investment
Mr Whitford believes that its customers usually see a 5 x return on investment, based on the cost of subscribing to GreenSteam, and the fuel savings they achieve.
“If you are already getting data from a vessel, you don’t need to make any further capital investment to use the service, you just need to start to share your data with GreenSteam.”
In the past, one of the hardest tasks has been to persuade people to act on the insights that the model provides, particularly when it challenges some strongly held beliefs.
Fuel level data
The most critical piece of information is the vessel’s fuel consumption.
GreenSteam is currently developing a mobile phone app which seafarers can use to easily capture fuel consumption data from the vessel and send ashore.
The app turns manual gauge data into a few kilobytes of data – which can either be sent immediately to shore, or incorporated into the noon day report.
The app avoids manual reading / recording errors and timestamps consumption levels automatically.
The project is still in the testing phase. “We’ve tried it on 10 different variants of fuel gauge so far, at 3 major shipping companies,” Mr Whitford says. “Some of the gauges are dirty, we want to prove that it’s a robust solution.”
It can also be used to reduce intentional and unintentional misreading of the fuel gauge.
Trim advice
GreenSteam’s trim optimiser can advise the seafarer how the trim could be adjusted to get the best fuel consumption.
Trim can be described as the “slant” of the vessel – whether the stern is sitting higher than the bow of the ship in the water, or the opposite.
The trim can be measured on the ship either using a “trim sensor”, which compares the water pressure beneath the bow and beneath the stern or with sensor equipment mounted in the vessel.
In some vessels, the trim of the vessel can be automatically adjusted from the bridge, by shifting ballast water fore or aft.
GreenSteam is able to train its vessel specific model to learn how the fuel consumption varies with different trim, also taking into consideration the loading of the vessel (and so the draft).
There is a “pre-departure” trim planner, which can calculate the best trim for the specific vessel before you depart, based on the draft (cargo loading) and vessel speed.
GreenSteam has also developed a system somewhat akin to autonomous vehicle technology which, when deployed onboard captures high frequency performance data and integrates this with the vessel’s digital model to continually assess the most optimal trim for the vessel – this is called “dynamic trim optimiser”. The vessel’s trim is adjusted during the voyage, by transferring ballast fore and aft from the bridge.
The master has a visual display showing whether the vessel is in the “green” or “red” zone, empowering the vessel to optimise each voyage in near real time.
Better alerting
GreenSteam’s software generates a range of “alerts” direct to the vessel in near real time. A basic alert might inform the vessel crew that the last 6 hours average fuel consumption has strayed outside an expected range. It can be sent to the captain as an e-mail, with a link to see further insights. “We call those regular alerts,” Mr Whitfield says.
Whilst this is useful, the problem with so-called regular alerts is that there may be a simple reason for the higher fuel consumption, such as strong winds. So we risk distracting the “time-starved” crew with too many alerts, sometimes just telling the Captain something he or she already knows.
By the end of 2020, GreenSteam plans to produce “smart alerts”.
A “smart alert” would only send an alert to a vessel when there is something useful to say, a specific actionable and useful change they can make.
“For example, it could indicate, “if you were to adjust your speed by this much, or adjust your trim by this much, you can still make your arrival laycan, but you will cut your fuel consumption this much,” Mr Whitford explains.
The alerts rely on the latest developments in GreenSteam’s machine learning platform and can be configured for the various needs of people in different roles, such as the vessel operator and technical manager.
Longer term decisions
The data in Green steam’s models can also be helpful in longer term evaluations such as measuring fuel consumption by exhaust gas scrubbers.
Scrubbers will consume fuel, usually 1-3 per cent of the fuel required for vessel propulsion.
Mr Whitford says that based on his discussions with shipping companies in Greece on a trip held in January 2020, companies can see a price differential between heavy fuel and low sulphur fuel emerging of $350 per ton.
“If you believe this differential will continue for the next decade, and know how much a scrubber will cost to build, you can see whether it would be a good investment.”