Utilizing Liquefied Natural Gas as Marine Fuel in
Liner shipping
Dennis van Duren
July 19, 2016
Student number: s1916076 Supervisor: Dr. E. Ursavas Second supervisor: Dr. X. Zhu AbstractContents
1 Introduction 1
2 Literature review 4
2.1 Economical aspects of LNG in container shipping . . . 4
2.2 Technical aspects of LNG in Container Shipping . . . 7
2.3 LNG infrastructure . . . 10
2.4 MARPOL and Emission Controlled Areas . . . 12
2.5 Speed Adjustment and Port Congestion Based Time Windows . 14 2.6 The liner shipping network design problem . . . 16
3 Problem Formulation 19 3.1 Notation . . . 19
3.2 Rotation generation model . . . 22
4 Experimental Design 26 4.1 Dataset . . . 26 4.1.1 Demand Data . . . 26 4.1.2 Port Data . . . 27 4.1.3 Fleet Data . . . 27 4.1.4 Distance Data . . . 28 4.2 Experiment Sets . . . 29
4.2.1 Experiment set: low SOx fuel . . . 29
4.2.2 Experiment set: LNG - half round trip capacity . . . 29
4.2.3 Experiment set: LNG - Reduced Capacity . . . 30
4.2.4 Experiment set: LNG - Containerized LNG . . . 30
4.3 Experiment Setting . . . 31
5 Computational Results 32 5.1 Comparison of the LNG configurations . . . 32
5.1.1 Present scenario . . . 33
5.1.2 Future scenario . . . 33
5.1.3 Present scenario - Including CAPEX . . . 34
5.1.4 Future scenario - Including CAPEX . . . 34
5.2 Route and speed adjustment . . . 35
5.2.1 Route: Pacific Tradelane - Current scenario . . . 35
6 Discussion 36 7 Appendix A: The Neighbourhood Search Heuristic 37 8 Appendix B: Tables 39 8.1 Asia Europe Trade lane . . . 39
8.2 Europe Mediterranean . . . 41
8.3 Singapore Feeder . . . 43
1
Introduction
Environmental sustainability is becoming increasingly important in the business world, and liner shipping is no exception. Various regulations by the Interna-tional Maritime Organization (IMO) force shipping companies to reduce their emissions so they can continue to transport in global waters. The most strin-gent IMO regulations regarding sulphur (SOx) and nitrogen oxides (NOx) are for Emission Control Areas (ECA), which are areas where shipping is concen-trated and pollution is more likely to be a problem (Burel et al., 2013). Future regulations are expected to become even stricter, because of the climate change problem (Gilbert et al., 2014). To comply with these regulations, it is generally accepted that liner shipping companies have three feasible options available to them: Switch to liquefied natural gas (LNG), use low SOxfuel or use scrubbers
(Wang and Notteboom, 2014; Brynolf, Magnusson, Fridell and Andersson, 2014; Gilbert, 2014; Brynolf et al., 2016).
low SOxfuel is the easiest solution, as it requires no technical change.
How-ever, low SOx fuel is expensive and does not reduce NOx or CO2(Burel et al.,
2013; Wang and Notteboom, 2014). SOx scrubbers require a capital
invest-ment and use additional energy, while also providing no reduction in NOx or Co2(Burel et al., 2013; Wang and Notteboom, 2014). Both approaches are short term solutions that will require additional technology to comply to future regu-lation (Burel et al., 2013; Wang and Notteboom, 2014). This leaves LNG as the other main option. Using LNG will not only reduce SOx by 90%, it will also reduce NOx by about 80-90%, PM by 100% and CO2 by -15 to -20% (Burel et al., 2013; Bengtsson et al., 2011; Brett et al., 2008; Xu et al., 2015; Anderson et al., 2015). Using LNG ensures compliance to all the upcoming SOx and NOx emission regulations. Various life cycle assessments confirm the superior overall environmental performance of LNG (Brynolf, Magnusson, Fridell and Anders-son, 2014; Bengtsson et al., 2012; Lowell and Bradley, 2013; Thomson et al., 2015). Burel et al. (2013) describe an expected reduction in operating
expen-ditures (OPEX) when using LNG, not only compared to low SOxfuel but also
compared to regular heavy fuel oil (HFO) (Levander and Sipila, 2008; Wang and Notteboom, 2014; Thomson et al., 2015). This is due to its lower price and lower maintenance costs as LNG is easier on the engine. Another benefit is the discount to port call costs at certain ports when LNG is used(Green Award, n.d.). By 2030 LNG is expected to be a main bunker source for liner shipping (Lloyd’s Register, 2012).
short as 2 years for new build and 2-4 years for retrofitting (adapting existing vessels to use LNG) are mentioned (The Danish Maritime Authority, 2012). However, these studies assume LNG can be used non stop and significant (or all) time is spend in ECA. These assumptions do not hold in actual practice. The limited infrastructure makes the high capital costs hard to justify as it is more difficult to benefit from the low operational costs. To deal with this issue several approaches are considered in this research. Liner shipping companies can use LNG dual fuel engines or containerized LNG, larger tanks and speed adjustment to reduce fuel consumption rates and thus dependency on LNG bunker facilities. Dual fuel engines allow vessels to switch between LNG, HFO or marine gas oil (MGO)(Brynolf, Fridell and Andersson, 2014). Using basically the same technology as regular HFO vessels use to switch to low SOxfuel to comply with
ECA regulations. They provide great flexibility as a vessel can use LNG and switch to HFO if necessary. Introducing containerized LNG tanks is another option. This concept entails storing the fuel in a container and bunkering by loading tanks on board with a container crane (Wang and Notteboom, 2014). These can be shipped to any port, thus allowing the use of LNG as the only fuel. A downside is the higher price and the loss of capacity as they claim space used by regular containers. This concept is also very new and has not been implemented in practice yet. Adjusting speed also reduces dependency on LNG infrastructure as slow steaming will decrease fuel use at the cost of longer travel times. The final option contains the use of bigger LNG tanks that come at the cost of loss of capacity and higher capital costs due to the increase in the size of the cryogenic tank.
The scale of containerized shipping is huge, 60% of the total value of goods transported by sea is done through containerized shipping (Stopford, 2009). The focus of this paper is specifically on liner shipping, which is known for having a published itinerary and schedule, where companies can book their cargo onto. Golias et al. (2009), Stopford (2009), Lofstedt et al. (2010) describe how fuel usage is responsible for up to 60% of the vessel operating costs of liner shipping
companies. Maritime routing and scheduling methods are thus essential to
LNG fuel usage, while having minimum effect on liner service schedules. The aim of this research is to fill this gap by considering the utilization of LNG as marine fuel when designing the liner shipping network, while also considering the limited LNG infrastructure, compliance to present and future ECA and IMO regulations, high capital investments, capacity loss and environ-mental discounts. Both LNG dual fuelling and the use of containerized LNG are considered and compared to the scenario of using low SOx fuel. As speed
adjustment is an important tool when dealing with the limited availability of LNG, varying speeds between ports will also be included. However, this intro-duces an unrealistic scenario as vessels are inclined to optimize LNG usage and disregard travel time. To prevent this, time windows based on port congestion are introduced, forcing vessels to be in a specific port in a specific time frame, as is common in practice (Notteboom, 2006). These time windows depend on how busy a certain port is on average. This is in line with what Notteboom (2006) has found as he describes how berth availability cannot be guaranteed when time slots in ports are missed. In addition, he found that 65.5% of schedule disruption in liner shipping is caused by port congestion. Port congestion based time windows thus bring the model closer to practice, while still allowing the management of LNG bunker through speed adjustment.
The following research question is answered in this research:
How should LNG be utilized as a marine fuel in liner shipping considering limitations on infrastructure, ECA compliance, capital investments and capacity loss?
To answer this question, relevant literature will be reviewed, and the inclu-sion of LNG as marine fuel and incorporation of adjustable speeds and port congestion based time windows will be modelled. The liner shipping network design problem (LSNDP) will then be solved using a column generation based neighbourhood search (NS) heuristic. Experiments comparing the use of LNG dual fuelling and containerized LNG to low SOxfuel dual fuelling will be
2
Literature review
This chapter reviews relevant literature on LNG use in container shipping, LNG infrastructure, MARPOL and Emission Control Areas, speed adjustment and port congestion based time windows, and finally the Liner Shipping Network Design Problem (LSNDP).
2.1
Economical aspects of LNG in container shipping
An important driver for LNG to be usable in containerized shipping is the economic viability. As was already mentioned before, there is an increase in CAPEX with the use of LNG, since engines need to be modified to be able to use it (Gilbert, 2014; Kumar et al., 2011; Seo et al., 2014). Sophisticated LNG engines and cryogenic fuel tanks are required, thus explaining these high capital costs. The specific costs that are reported in literature are varying, as it depends on the type of engine (dual fuel or single LNG), the ship itself and the tank size (Wang and Notteboom, 2014). These in turn depend on the function of the ship and other elements like available infrastructure, capacity, etc. Wang and Notteboom (2014) estimate based on reviewing a large amount of LNG literature, that using LNG instead of HFO will increase CAPEX by about 20-25%, this is in line with for example Gilbert (2014) who mentions an additional 25-30% CAPEX. According to Stopford (2009) capital cost is 30% – 45% of a vessel’s cost, bunker cost is 35% – 50% , OPEX is 6% – 17%, port cost is 9% – 14%. As CAPEX makes up a big part of the total vessel cost, the increase in CAPEX can have a significant affect on total costs.
There is a trade-off between the higher CAPEX and lower OPEX when using LNG. This is caused by the fact that the price of LNG is low compared to HFO and especially low compared to low SOxfuel. While it is true that price
differ-ences can change, it is important to note that the price of LNG has proven to be quite strong compared to HFO (Thomson et al., 2015). Wang and Notteboom (2014) describe, in their review of 33 LNG related studies, how most of these studies have a positive outlook on future development of the price difference. An example of this is the The Danish Maritime Authority (2012) who describes a competitive price development compared to regular fuels. However, Wang and Notteboom (2014) note that with increasing demands, lack of infrastructure and LNG supply chains, it is hard to make an accurate prediction. The actual difference between LNG and HFO varies as fuel prices fluctuate daily. The joint study of Germanischer Lloyd and MAN Diesel examining the costs and benefits of LNG as ship fuel for container vessels used a price difference between LNG and HFO of 60-80% on an energy basis (Sames et al., 2013).
that could make LNG economically interesting. An important initiative is the Green Award who certify ships that are extra clean and safe (Green Award, n.d.). In turn these ships are provided with financial benefits in the form of a discount on port call costs. A variety of countries has joined this initiative and provide discounts varying from 10 and 20 % at all their ports. Countries that are part of this initiative are Belgium, the Netherlands, Portugal, South Africa, Canada, Latvia, Lithuania,Oman and New Zealand. An example of this is the port of Rotterdam, where they use discounts on seaport dues paid for seagoing vessels that score 31.0 points or more in the Environmental Ship Index (ESI)(Port of Rotterdam, n.d.). Additionally, most ports that have LNG available as bunker offer a similar discount on port call costs. An example is the port of Bremerhaven which offers a 15% discount if LNG is used (Bremenports, n.d.).
Brynolf, Magnusson, Fridell and Andersson (2014) describe how there are different payback times for rebuild and retrofitting ships with LNG technology. Retrofitting is the adaptation of ships to be able to use LNG, and rebuild means using LNG equipment when building a new vessel. These payback times are based on a report by The Danish Maritime Authority (2012) who analyzed pay-back times. They concluded that pay-back times are short namely 2 years for new build vessels and 2-4 years for retrofitting a vessel that is not older than 10 years. However, they assume ships to operate entirely within ECAs. Since low SOx fuel is expensive, not having to use it all the time decreases
payback periods significantly. While this assumption will only hold in certain areas, it does illustrate the economic potential that LNG has. Adachi et al. (2014) confirm the economic viability of using LNG in their economic analysis of a trans-ocean LNG-fueled 9,300 TEU container ship using a dual fuel engine. It is based on an existing container ship that runs on HFO and has similar capacity. Their analysis predicts that it is not only an environmentally friendly investment to comply with IMO regulations, but also economically attractive compared to a NOx Tier III complied oil-fuelled container ship. The joint study by Germanischer Lloyd and MAN Diesel is also concerned with the economic benefits of LNG for container vessels (Sames et al., 2013). They concluded that for a new build container ship as small as 2400 TEU, LNG would be feasible. Payback time for larger vessels is even shorter. All these studies assume that LNG can be used full time without restriction of lack of infrastructure, which is certainly not in line with the real life situation. Still it illustrates the power that this trade-off of higher CAPEX and lower OPEX can have.
Model implications
solutions. The price difference used in this research is the average price difference used in the joint study by Germanischer Lloyd and MAN Diesel. Thus LNG is 70% cheaper than HFO on an energy basis. Also, an environmental discount on port call costs for the use of LNG is implemented. All ports located in countries that are part of the Green Award initiative provide this discount. In addition, every port that has LNG bunker available offers this discount. As the exact increase in CAPEX depends greatly on a variety of factors, separate experiment sets are included that include an average increase of 25% in CAPEX. The increase in CAPEX for these scenarios is varied with the tank size used in the specific experiment set using the investment data provided by The Danish Maritime Authority (2012). The expected decrease in maintenance costs is also included in these scenarios as maintenance cost is decreased by 10% for vessels using LNG. The profitability of switching to LNG also depend on the amount of time spend in ECA zones, since using LNG in these areas means costs related
to low SOx fuel are not incurred. To be able to compare the use of LNG to
the main alternative low SOx an experiment is included that forces vessels to
2.2
Technical aspects of LNG in Container Shipping
There are several technical aspects that are important when using LNG in liner shipping. This section discusses the different types of engines that use LNG, the relationship between speed and fuel consumption, capacity loss, LNG tank sizing, and the Boil Off Gas (BOG).The LNG engine concepts that are in use as of 2014 are gas-only engines, dual fuel 4 stroke engines, and dual fuel 2 stroke engines (Deniz and Zincir, 2016). Brynolf, Magnusson, Fridell and Andersson (2014) add to that the possible use of Otto engines and fuel cells. Dual fuel engines can run on LNG, marine gas oil (MGO) or HFO, thus making them very flexible. These engines use small amounts of HFO for ignition. The dual fuel 4 stroke engines comply with the newest NOx tier 3 Limits (Brynolf, Fridell and Andersson, 2014; Brynolf et al., 2016). Thus in the model we can assume usage of dual fuel 4 stroke engines, since it is important that switching to LNG when entering an ECA where NOx tier 3 applies (e.g. United States) should be sufficient to comply. As varying speeds are included in the model it is important to know the relationship between speed and consumption of LNG and how it relates to HFO. Cheenkachorn et al. (2013) studied this effect of LNG on a 4 stroke dual fuel engine. Figure 2 below illustrates this relationship. Both LNG operation and HFO operation (Diesel Operation in the graph) show a very similar relationship between speed and fuel consumption. A good approximation that is often used is a cubic funtion(Stopford, 2009). Since both LNG and HFO operation share the same structure the cubic function can be used to determine the bunker consumption for both LNG and HFO at varying speeds. Aldous and Smith (2012) work on speed optimisation for LNG carriers confirms that the cubic relationship holds for LNG.
LNG is natural gas which is liquefied, thus reducing its volume and mak-ing it possible to use natural gas without losmak-ing ridiculously large amounts of space (Anderson et al., 2015). However, LNG tanks still take up more space than regular diesel fuel tanks, thus they may slightly reduce room for contain-ers on container carricontain-ers. The exact space reduction caused by the cryogenic tanks depends on the vessel and its function, which makes it hard to estimate. Added to that are the conflicting reports in literature on capacity loss. Adachi et al. (2014) showed in their analysis that there is no difference in capacity be-tween a LNG-fuelled dual fuel engine container ship and a similar oil fuelled container ship. Wang and Notteboom (2014); Panagakos et al. (2014) contrast this and describe that space required for LNG is three to four times more than conventional fuel usage. This is due to the volume of LNG compared to diesel (1.8 times larger), the engine, and the cylindrical-shaped tank. Possible miti-gation for this is smaller tanks, which means regular refuelling. Another option they propose is a hull-integrated tank, which is still in development. The issue of fitting LNG technology when retrofitting is an even bigger problem as the integration of such a big LNG system has not been taken into account when originally designing the ship (Panagakos et al., 2014; Wang and Notteboom, 2014; Brynolf et al., 2016). Again conflicting reports are described in literature. The Danish Maritime Authority (2012) describe how retrofitting a container vessel to use LNG reduced capacity by 4%, but another retrofit of an ocean go-ing tanker vessel did not reduce cargo capacity. This variation is acknowledged by Wang and Notteboom (2014) who describe it depends on configuration of the ship and ship age. The joint study performed by MAN Diesel and Turbo and Germanischer LLoyd states that in general some loss of capacity is to be expected (the largest being 3% for medium sized container vessels), which in turn results in lost earnings (Sames et al., 2013).
For LNG to remain liquefied, special tanks called cryogenic tanks are re-quired (Brynolf, Magnusson, Fridell and Andersson, 2014). These tanks have to deal with increasing pressure and the occurrence of the Boil Off Gas (BOG) effect. Seo et al. (2014) describe this process for LNG carriers, and describe how this process is very similar to what happens in the tank of a vessel using LNG. This is confirmed by Wang and Notteboom (2014) who describe the 40 years of experience of these carriers powering their ships using the LNG they transport, as the main driver of the well developed technology. As LNG is stored over pressurisation of LNG storage tanks must be handled. This is caused by heat ingress and while these tanks are heavily isolated, it is simply not possible to prevent some heat from transferring into the tank. This causes LNG to expand, which in turn puts high pressure on the tank. This heated LNG needs to leave the tank, to deal with this the gas is released to the atmosphere. If LNG is not properly released and pressure builds up issues with safety arise. This effect is called the Boil Of Gas effect (BOG). This means some of the LNG is lost ev-eryday a container vessel is in operation. This is not only a waste of LNG, but can also have a large impact on the global warming potential of the operating vessel. However, as of 2013 there are already non-LNG carrier gas-fuelled ves-sels that have the technology to use BOG as part of the ship propulsion (Wang and Notteboom, 2014). This means BOG no longer leads to a loss of LNG, nor does it significantly impact global warming.
Model implications
2.3
LNG infrastructure
One of the biggest disadvantages of using LNG is the limited bunkering infras-tructure and distribution networks. This section discusses the infrasinfras-tructure problem, the current situation, and the available LNG bunkering modes.
The LNG infrastructure problem is also described as a ‘chicken-and-egg’problem, and it is common for every new fuel introduction (Sperling and Kitamura, 1986). This means that suppliers are waiting for demand to increase before investing in infrastructure, while shipping companies do not invest in LNG fueled vessels since LNG infrastructure is not readily available. Since neither side is motivated to move forward this problem is not easy to solve. Fortunately several exter-nal forces are driving the development of LNG infrastructure. The European Commission announced on January 2013 its Clean Power for Transport Pack-age. According to this all maritime ports of the trans-European core network should have LNG refuelling stations installed by 2020 (Wang and Notteboom, 2014; Panagakos et al., 2014). Another important driving force is the Interna-tion Maritime OrganizaInterna-tion who are enforcing increasingly stringent regulaInterna-tions. Their plans to enforce strict SOx regulations globally in 2020 will force every
vessel to comply. This is seen as a main driver for LNG infrastructure (Wang and Notteboom, 2014). While promising, these drivers will need time to have a significant effect on LNG infrastructure development.
Figure 2 LNG refuelling possibilities now and planned (DNV-GL, 2015).
shipping with dense regular liner traffic. This is a logical place to start, as there are more possibilities to refuel reducing the dependency on infrastructure. Fortunately LNG infrastructure is not solely found in this area, around the world several ports already have LNG bunker available, and many more have already planned the development of their LNG bunker infrastructure. Important ports like Buenos Aires, Los Angeles, New Orleans and Piraeus have or will have LNG infrastructure in the near future. The light blue nodes in figure 2 are the ports that are currently discussing plans for LNG infrastructure. Major ports like Algeciras, Jebel Ali, Shanghai, Busan, and Barcelona are seriously considering the development of LNG infrastructure.
There are four LNG bunkering modes available to vessels (Wang and Not-teboom, 2014; Skramstad, 2016). Namely ship-to-ship bunkering by using a LNG bunker ship, truck-to-ship bunkering, transfer via a pipeline and loading arm from a terminal, and finally using portable LNG tanks that are loaded on board and used as fuel. Truck-to-ship is viable for small amounts, while ship-to-ship bunkering requires a large amount of LNG. The main benefit of ship-to-ship-to-ship-to-ship bunkering is that it can be done during cargo handling, which saves time. Using transfer via a pipeline is a suitable option for limited demand, high frequency, and less strict timetables. The use of portable LNG tanks has not been im-plemented yet, so there is no information regarding its feasibility and benefits. Although in theory it could provides great flexibility, especially in the early years when LNG infrastructure is still limited. These containers could simply be shipped to a port and LNG can be utilized the entire voyage even when no other bunker possibilities are available.
Model implications
The implications for the model are as follows. A distinction is made between LNG bunkering being directly available from the port and the use of portable
LNG tanks. No differentiation is made between the various forms of LNG
2.4
MARPOL and Emission Controlled Areas
The International Maritime Organization (IMO) is a United Nations (UN) spe-cialized agency that is responsible for the prevention of marine pollution and the security and safety of shipping (International Maritime Organization n.d.). They are the main regulatory body for the shipping industry and provide global standards regarding pollution. In order to promote sustainable shipping they work on a sustainable global maritime transport system. MARPOL 73/78 be-ing one of the most important environmental laws created by IMO and is part of these efforts, and its goal is to preserve marine environment by eliminating pollution by oil usage and harmful substances. Over the years new and stricter MARPOL annexes have been adopted, most notably the revised MARPOL Annex VI (2008) which includes not only significantly lower emission limits, but also includes sulphur oxides (SOx) and nitrous oxides (NOx) limits (Inter-national Maritime Organization, 2014). This annex also introduced emission control areas (ECAs) which are specific areas that demand an even larger re-duction of emissions (International Maritime Organization, 2016). The global change of SOx to 0.50 will go into effect either in 2020 or 2025 depending on a review to be performed by IMO in 2018. The table below shows more details on these specific regulations and how they are set to be lowered even further. SOx limits apply to every vessel, since it directly relates to the fuel used (Burel et al. 2013). The global SOx limit is 3.5% since 2012 and the ECA limit is currently 0.1% (International Maritime Organization, 2014)).
As of 2016 the tier III limit for NOx is into force and limits the NOx emis-sions in NOx ECAs and for new build ships. NOx emisemis-sions are related to the type of engine that is used. NOx emission limits are set as a function of the maximum engine operating speed. IMO uses a three-tier reduction programme, and currently tier II is in effect in global waters, while tier 3 will only apply to ECA zones and go into effect in 2016 (Burel et al., 2013). NOx emissions depend on the year in which the vessel was build (Anderson et al., 2015). Tier I applies globally and applies to vessels built after 2000. Tier II also applies globally and applies to vessels built after 2011. The strictest limit is Tier III which entered into force this year (2016), and applies to new build ships and only in NOx ECAs. The emission control areas (ECAs) are already enforcing these stricter rules and vessels need to find ways to comply if they want to sail in these areas. These are mostly areas where shipping is highly concentrated, and thus pollution is more likely to become a problem (Burel et al., 2013). As of 2014, there are 4 ECAs. Table 1 below mentions the exact areas and which specific limits are in effect.
Table 1: Emission Control Areas
ECA Limit on limit
Baltic Sea area SOxonly 0.10% SOx(1000 ppm) effective as of 1 January 2015
North Sea area SOxonly 0.10% SOx(1000 ppm) effective as of 1 January 2015
North American area SOx,N Ox 0.10% SOx(1000 ppm) effective as of 1 January 2015 N Oxtier III
Several areas are being considered as possible ECAs like the Mediterranean Sea, the Turkish Marmara sea Singapore Coasts, Japanese Coasts and Aus-tralian Coasts (Burel et al., 2013; Di Natale and Carotenuto, 2015). Figure 3 shows the current ECAs and possible future ECAs. It is important to mention that particulate matter (PM) limits are enforced by the SOx limits (Interna-tional Maritime Organization, 2014). Another important development in this regulatory process of increasingly stringent legislation is the involvement of the EU. The EU ratified amendments including the IMO global SOx regulations and the ECA limits into EU legislation (European Commission, 2009). This is important to liner shipping companies as the EU decided they will set a SOx
limit of 0,50% in 2020 regardless of the outcome of the IMO review in 2018. This means that all vessels that want to enter EU waters will have to comply with SOx legislation in 2020.
Figure 3 Map with current and possible future Emission Control Areas (Sys et al., 2016)
Model implications
2.5
Speed Adjustment and Port Congestion Based Time
Windows
Speed adjustment is an important strategy when dealing with limited availabil-ity of LNG infrastructure. This section discusses speed adjustment and its value in LNG bunker management, the implementation of speed adjustment in the model, and the port congestion based time windows.
Wang and Meng (2015), Yao et al. (2012) identify optimation of the sailing speed as one of the main aspects of bunker management. They describe deter-mining the proper sailing speed as making the trade-off between high bunker consumption, low transit times, and a smaller number of ships needed to main-tain frequency on one side, and lower bunker costs, longer transit times, and a larger number of ships on the other. This trade-off is mostly influenced by the bunker price, as is illustrated by the slow-steaming strategy used by liner ship-ping companies since 2007 driven by high fuel prices (Wang and Meng, 2015). When LNG dual fuelling is used speed optimization will be even more important as bunkering at any port is simply not viable because of limited infrastructure (except for the Baltic and North Sea areas). Aditionally, proper speed adjust-ment can make the difference between being able to use LNG instead of HFO thus benefiting from lower OPEX, or being able to comply to ECA regulations using LNG without having to switch to expensive low SOx fuel.
Speed optimization and adjustment has been widely addressed on various planning levels in container ship routing and scheduling literature. Reinhardt et al. (2016) address the liner shipping berth scheduling with transit times prob-lem where they focussed on optimizing an existing network by among other things adjusting speed on each leg to reduce fuel costs. Guericke and Tierney (2015a) address the optimization of speed on each leg in the liner shipping cargo allocation with service levels. The first to actually introduce variable speeds in the LSNDP are Karsten, Brouer and Pisinger (2015). They allow individual sailing speeds at each service leg. The same approach regarding speed adjust-ment will be used in this paper. Varying speeds are allowed between legs, and switching is only allowed when changing legs. However, a problem that needs to be dealt with is the non-linearity that is introduced when allowing varying speeds. In order to deal with this the sailing speed range is discretized provid-ing a set of possible speeds. This is described as a realistic approach to avoid non linearity while still allowing a wide range of speeds and associated bunker consumption rates (Alvarez et al., 2010; Gelareh and Meng, 2010). Bunker con-sumption rates are determined using the design speed and concon-sumption of a vessel and the cubic relationship between sailing speed and fuel consumption.
The introduction of variable speeds in the model will provide it with a valu-able tool to utilize LNG despite of the limited LNG infrastructure. However, vessels will be inclined to optimize LNG usage to prevent higher costs of HFO and low SOx fuel and thus disregard travel time. This is not in line with
missed. The inclusion of deadlines or maximal transit times forcing vessels to be at a specific place at a specific time, has been the focus of recent container ship routing and scheduling literature. These maximal transit times represent the commitment of the liner shipping company to deliver containers within a given time frame (Karsten, Pisinger, Ropke and Brouer, 2015). They are influenced by the type of product (e.g. perishable) or the requirements set by the client and the port (Notteboom, 2006). Therefore, from a customer point of view including these when designing the liner shipping network is essential. Recently this has been implemented in the LSNDP and in sub problems of the LSNDP by Karsten, Pisinger, Ropke and Brouer (2015); Wang and Meng (2014); Brouer et al. (2015). In these implementations the choice is made for using a maximal allowed travel time from the port of origin to the port of destination. Most implementations utilize a penalty if a time window is broken, since violating a delivery window can be preferable if profits are higher than the penalty costs and inventory holding costs incurred. Otherwise hard constraints are used forcing the model to find a solution where all commodities transferred satisfy maximal transit times constraints.
2.6
The liner shipping network design problem
To determine how to best utilize LNG in liner shipping a network design
ap-proach is used. This is done by considering all LNG related decisions and
limitations while solving the liner shipping network design problem (LSNDP). This section discusses the LSNDP itself, recent literature in this area, literature concerning LNG and network design, and finally the heuristic used to solve the model.
The objective of the LSNDP is to maximize the revenue obtained through the transport of cargo and minimize operating costs (Brouer, Alvarez, Plum, Pisinger and Sigurd, 2014), while satisfying all the constraints. The purpose is to find a set of non-simple cyclic sailing routes (rotations) that will jointly transport multiple commodities via the use of a set of container vessels (Alvarez et al., 2010). A non-simple cyclic sailing route means the start and end of a route are the same and a port may be visited twice. The LSNDP consists of solving a rotation generation problem of which the objective is to maximize profit, while satisfying the constraints. Profit is increased by satisfying more demand, which in turn leads to higher operating costs. The rotations that are the output of the rotation generation problem are used as input for the second step, i.e. the multi-commodity flow problem (MCFP). This is concerned with allocating the containers to the vessels in the network of rotations. Rana and Vickson (1991) were one of the first to present a mixed-integer programming (MIP) model for the liner shipping network design problem that uses non-simple cyclic sailing routes. Thus providing the basis for what is known today as the LSNDP. Two decades later the models have been substantially improved (e.g. eliminating non-linearity) and brought closer to practice (e.g. transhipments) (Meng et al., 2013; Christiansen et al., 2013).
The model that is used in this paper is based on the model of Brouer,
Alvarez, Plum, Pisinger and Sigurd (2014) and ´Alvarez (2009). It includes
Desaulniers and Pisinger (2014); ´Alvarez (2009) that are also included in this model are penalties for rejecting demand, tracking of transshipment and its related costs, biweekly frequency constraints and canal- and port calling costs.
Other recent literature on containership routing and scheduling describes various extensions on different planning levels mostly concerned with empty containers and different solving methods. Song and Dong included repositioning of empty containers and developed a three-stage approach that solves the single liner long haul service rout design problem (Song and Dong, 2013; Dong et al., 2015). Wang, Meng and Du (2015) used a mixed integer nonlinear problem to study the seasonal revenue management and solved it with a tailored branch-and-bound method. Wang (2013) examined the impact of slot purchasing, multi-type containers, empty container and ship repositioning and integer number of
containers. Wang and Meng (2013) included the reversing of port rotation
directions of ship routes. Wang, Meng and Liu (2013) included the dependence of container shipment on transit time. Recently, Wang, Meng and Du (2015) proposed a model for altering the shipping network for changing demand.
There is minimal literature that holistically addresses LNG use and net-work design in liner shipping. One exception is the recent conference paper of Aymelek et al. (2015) who describe part of a LNG bunkering network opti-mization model of deep sea liner shipping. Their model allows the use of LNG as the only fuel and deals with the lack of LNG infrastructure through speed adjustment and selection of bunkering ports on route and off route. The work of Aymelek et al. (2015) is at a lower level of planning than the LSNDP as it focuses on a single route that is given and aims to minimize the time and cost of bunkering, considering bunkering ports off and on route in the process. The purpose of the LSNDP on the other hand is far more comprehensive as its goal is to find a set of sailing routes that will jointly transport multiple commodi-ties via the use a set of container vessels (Alvarez et al., 2010). Also, they do not consider the use of LNG dual fuelling that allows vessels to switch back to HFO if no infrastructure is available. However, they do provide some valu-able insights regarding the use of LNG in liner shipping. They underline the importance of including changeable fuel consumption rates dependent on ship speed, as lower speeds significantly decrease LNG usage, which helps address the problem of lacking LNG infrastructure and thus allows better utilization of LNG. Therefore varying bunker rates are introduced in the rotation generation model. Varying speeds and associated bunker consumption rates of all types of fuel will be addressed later in the paper.
3
Problem Formulation
The LSNDP consists of generating rotations by solving the rotation generation problem (also known as the auxiliary model) and allocating demand to the network of rotations by solving a multi commodity flow problem(MCFP). The model that is used in this paper is based on the model of Brouer, Alvarez, Plum, Pisinger and Sigurd (2014) and ´Alvarez (2009). First the notation used in the model is described. Thereafter, the rotation generation model is presented including all the proposed LNG related elements. For the MCFP, we adhere to the problem description given in Alvarez et al. (2010) and Brouer, Alvarez, Plum, Pisinger and Sigurd (2014). (2014a).
3.1
Notation
The following sets are used:
R All rotations in the model, indexed using r
P All ports in the model
E Set of all possible edges in the model. Edges are directed
and uncapacitated.
V Vessel classes in the model, indexed using ν
G Set of port pairs with demand for transport indexed using
(o, d)
S The set of sailing speeds available, indexed using s
F The set of fuel types, indexed using f
The following parameters are utilized in the MIP formulation of the rotation generation model:
aνij Canal costs for vessels of type ν when traversing arc (i, j) cν
f Capacity (in FFE) of vessels of type ν using fuel type f
dνj Port call costs for vessels of type ν when entering port j using fuel
type f
ef Fuel price per ton for fuel type f
πν
f Daily running cost for vessels of type ν over entire planning horizon
˜
πν Cost or revenue of excluding a vessel of type ν from operation. This
gνsf Fuel consumption (tons per mile) of fuel type f for vessels of type ν at speed s.
hvf Fuel consumption (tons per day) of fuel type f for vessels of type ν
when idle at the port.
kod Total demand (in FFE) to the liner company for transport from o
to d over the planning period. lv
ij Length in nautical miles of direct sailing from port i to port j using
vessel type ν. pv
j Time at port j for vessels of type ν.
qod Revenue from transport of one FFE from o to d.
˜
qod Penalty for failing to transport one FFE from o to d.
uj Cost of lifting one FFE at port j.
tj Cost of transhipping one FFE at port j.
δ Empirical parameter, estimates the amount of additional flow
flow-ing through a butterfly node, as compared to a regular node. ϕin
n , ϕoutn Empirical parameters that capture the importance of a port as an
exporter or importer.
%ij Indicating the maximal transit time of the port time window sailing
from port i to port j. The more congested port j the lower the maximal transit time.
ξj Penalty cost when the port time window is missed at port j, the
height of the penalty depends on the importance of the port as an importer, exporter, and transshipment hub.
ςj Parameter indicating if port j is located in an ECA, 1 if this is the
case, 0 otherwise.
σν Parameter indicating the LNG capacity of the tank of vessel type
ν, thus the amount of LNG that can be used on a rotation without bunkering LNG more than once.
ιj Parameter indicating if LNG bunker is available at port j, 1 if this
is the case, 0 otherwise. γν
The following variables are exclusively used in the auxiliary model to generate new rotations.
ˆ
kod Residual demand (in FFE) to the liner company for transport from
o to d. At any iteration in the heuristic, the residual demand is com-puted by subtracting demand that is carried by existing rotations from the original demand.
T Length of the planning horizon, in days.
κ Number of sister vessels on the new service.
ν Vessel type to be deployed.
In the auxiliary model the following decision variables are used:
Nj Binary variable, indicates whether port j is visited in the rotation.
Bj Binary variable, indicates whether port j is a butterfly port in the
rotation.
Ij Continuous variable, indicates the sequence of port j in a rotation
(for subtour elimination).
Cj Binary variable, indicates if port j is the master port of the route.
Aijsf Binary variable, indicates whether edge (i, j) forms part of the new
rotation, which speed s is used at that edge and which fuel type is used on edge (i, j)
Qod Continuous variable, indicates the number of FFEs with origin at
port o and final destination at port d that will be carried per sailing of each vessel in the rotation.
W1, W2 Binary variables, respectively, indicate whether the new rotation will
have weekly or biweekly call frequency.
µ Inverse of the number of trips to be completed over the entire
plan-ning horizon by each vessel on the new service.
ω Estimated cost per sailing per vessel of the new service
βij Binary variable indicating if the port time window is missed when
3.2
Rotation generation model
The rotation generation model is given by AU X(ν, κ).
M aximize ZAU X(ν,κ)= ω subject to ω = (˜πν− πν)µ (1) − X (i,j)∈E X s∈S (efhνf pν j 24+ efg ν sflijν + dνjf + aνij)Aijsf (2) + X (o,d)∈G (qod− uo− ud+ ˜qod)Qod (3) − X (i,j)∈E βjξij (4) (5) ω represents the objective function and maximizes the net revenue contribution of each sailing of the rotation. It consists of all the associated revenues and costs. Equation (1) consists of the daily operating costs of the deployed vessels. Equation (2) represents all fuel, canal and port call costs. The third equation contains all cargo related revenues and costs. Finally, equation (5) defines the penalty cost for missing a port time window.
The model is constrained by the following equations:
X j∈P Bj≤ 1 (10) Bi+ Ni= X j∈P X s∈S X f ∈F Aijsf ∀i ∈ P, (11) Bj+ Nj = X i∈P X s∈S X f ∈F Aijsf ∀j ∈ P, (12)
Constraints (10) ensure each rotation can have only one butterfly port. The number of edges that enter and leave any port are limited to two if that port is a butterfly port and one otherwise, this is enforced by constraints (11) and (12). Qod≤ µˆkod κ ∀o, d ∈ G, (13) Qod≤ ˆkodNo ∀o, d ∈ G, (14) Qod≤ ˆkodNd ∀o, d ∈ G, (15) X d∈P Qod≤ ϕouto c ν(N o+ δBo) ∀o ∈ P, (16) X d∈P Qod≤ ϕoutd c ν(N o+ δBd) ∀d ∈ P, (17) X (o,d)∈G lodν Qod≤ X (i,j)∈E X s∈S X f ∈F Aijsflνijc ν (18)
X j∈P Cj= 1 (19) Bj≤ Cj≤ Nj j ∈ P, (20) Nj ≤ Ij ≤ | P | Nj j ∈ P, (21) 1 + Ii− | P | Cj− | P | (1 − X s∈S X f ∈F Aijsf) ≤ Ij ∀i, j ∈ E, (22)
Constraints sets 19 to 22 are the subtour elimination constraints. Constraint 19 ensures 1 port is the master port. if there is a butterfly port, expressions (20) ensures it is the master port. Constraints (20) force the variable Ij away from
zero if and only if port j is included in the rotation. I − j will increase along the path of the rotation, indicating the sequence of ports in the rotation. This is enforced by constraints (22). W1+ W2= 1 (23) cν+ W1M ≥ 1200F F E (24) (W1− 1) + 0.91 ∗ 7κ T ≤ µ ≤ 7κ T + (1 − W1) (25) (W2− 1) + 0.91 ∗ 14κ T ≤ µ ≤ 14κ T + (1 − W2) (26)
Constraints (23) to (26) ensure all rotations havea weekly- or a biweekly sched-ule. Constraint (23) forces a rotation to have either a weekly- or a biweekly schedule. Constraint (24) states that vessels with a capacity above 1200 FFE need to have a weekly calling frequency. Constraints sets (25) and (26) build slack into the scheduled routes, as using discretized speeds will discard prof-itable routes that do not exactly match the frequency.
Constraints related to port congestion based time windows: X s∈S X f ∈F Aijsf ≤ 1 ∀i, j ∈ E, (27) %ij ≥ X s∈S X f ∈F ((l ν ij γν s )Aijsf) − M (1 − X s∈S X f ∈F Aijsf) − (βijM ) ∀i, j ∈ E, (28)
Constraints related to the use of LNG and ECA compliance: Ni+ Bi≤ X j∈P X s∈S X f ∈F Aijsf+ M (1 − ςi) ∀i ∈ P, LN G ∈ F, (29) Nj+ Bj≤ X i∈P X s∈S X f ∈F Aijsf + M (1 − ςj) ∀j ∈ P, LN G ∈ F, (30) X i∈P σν(Ni+ Bi)ιi≥ X (i,j)∈E X s∈S X f ∈F (hνLN Gp ν j 24+ g ν s(LN G)l ν ij)Aijsf (31)
Constraint sets (29) and (30) enforce the use of LNG on all edges that enter or exit a port that is located in an ECA and is part of the rotation (it is indexed as an ECA port). To be able to use LNG in a rotation, at least one port that has a LNG bunkering facility needs to be in the rotation, and for every additional port with LNG bunker facilities an additional tank of LNG for each vessel type ν is available to use on the rotation. When a butterfly port has LNG bunker available an additional tank LNG is available. Finally, the total amount of LNG consumed needs to be lower than the total amount available. This is expressed by constraint (31). Aijsf ∈ {0, 1} ∀i, j ∈ E, s ∈ S, f ∈ F, (32) βij ∈ {0, 1} ∀i, j ∈ E, (33) Nj, Bj, Cj∈ {0, 1} ∀j ∈ P, (34) W1, W2, ∈ {0, 1} (35) Ij ≥ 0 ∀j ∈ P, (36) µ, ω ≥ 0 (37)
4
Experimental Design
This section concerns the experimental setting of this research. First the exper-iment setting is described. Next the dataset that is used and the newly created data instances are discussed. Finally, the experiment sets that were conducted are described.
4.1
Dataset
To perform the experiments using the model, a dataset is needed with instances that represent the itenaries of the large liner shipping companies. In addition, ECA zones and most important ports that have LNG should also be part of the dataset. Because of these reasons it is not possible to simply use a stan-dard benchmark set like the LINERLIB-2012 dataset by Brouer, Alvarez, Plum, Pisinger and Sigurd (2014). However, the data needed to build such a dataset is not readily available, as demand information is commercially sensitive informa-tion for liner shipping companies. The LINERLIB-2012 dataset is an excepinforma-tion as it was created in collaboration with Maersk, the biggest liner shipping com-pany in the world. They provided data that represents actual real life demand structures. Therefore, the new instances that are created in this research use the LINERLIB-2012 benchmark set as its main source of data. The creation of the different types of data used in the dataset are described in this section. The creation of demand data, port data, fleet data, and distance data are discussed below.
4.1.1 Demand Data
used this on the original World Large instance, and further reducing the number of ports was not necessary. To determine the appropriate transshipment hubs when mapping demand, several ranked lists were used that rank ports based on their container volume and transshipment volumes (list, 2013).
The instances that were created can be seen in table 2 below. They are based on existing shipping lines of major liner shipping companies like Maersk, CMA CGM Group, and Evergreen Lines. In addition current and possible future LNG infrastructure and ECAs were taken into account when creating the instances. To create instances that represent the variety found in actual real life networks different types of instances are included, namely multi-hub feeder, single-hub feeder, and trade lane instances. Detailed information including the amount of unique demand pairs and total amount of ports in each instance can also be found in table 2.
Table 2: Detailed information on the created instances
Instance Description Demand Pairs Ports
Intra Asia Multi-Hub 354 31
Europe Mediterranean Multi-hub 366 39
Singapore Feeder Single-hub 55 11
Pacific Tradelane Trade lane 507 43
Asia Europe Tradelane Trade lane 46 802
4.1.2 Port Data
A 15% discount was included for fixed port call costs at certain ports when LNG was used when calling that specific port. For all ports located in Green Award countries and ports that have LNG bunker available, this discount was included. This is in line with findings in the literature review regarding environmental discounts when using LNG. The LNG port index was based on data regarding current and future LNG infrastructure by DNV-GL (2015) as discussed in the literature review. Current and possible future ECAs identified by Sys et al. (2016) were used to create the ECA port index. Seperate port data files have been created for a current scenario and a possible future scenario regarding LNG infrastructure and ECAs. For all other costs the port data provided in LINERLIB-2012 is used.
4.1.3 Fleet Data
used (Sames et al., 2013). The number and type of vessels available for the in-stances were determined based on the maximum amount of FFE, the total FFE, and the distance between ports. Several test runs for each instance have been performed to find a fleet size and composition that provides sufficient capacity and economies of scale. As the goal of this research was to determine how to best utilize LNG and not to determine the optimal fleet size and mixture, all instances have some over capacity.
4.1.4 Distance Data
4.2
Experiment Sets
Four basic types of experiments were conducted and are discussed below. For each of these basic experiment sets a current ECA and LNG infrastructure scenario, and a possible future ECA and LNG infrastructure scenario was run. In addition, seperate runs are included of the LNG experiment sets that include an estimate of the increased CAPEX associated with LNG. An exception is the experiment set using containerized LNG, for which only a future scenario was run. The reason for this is that it is not used in practice yet and no information was available on capital expenditures. All experiments performed included the option to vary in speeds and the port congestion based time windows.
For the scenarios that include the increase in CAPEX associated with LNG, several assumptions had to be made. To include the increase in CAPEX, the Time Charter (TC) rate for each vessel class in the dataset is used and increased. The TC rate covers operational and capital cost and depreciation of value of the vessel Brouer, Alvarez, Plum, Pisinger and Sigurd (2014). However,Brouer, Alvarez, Plum, Pisinger and Sigurd (2014) did not describe the exact build up of the TC rate. To estimate the build up of the TC rate the findings of Stopford (2009) are used. According to Stopford (2009), OPEX makes up 6%-17% of a vessels costs , while CAPEX is 30-45% of a vessels costs. Based on these findings the assumption is made that CAPEX includes depreciation and makes up the TC rate together with OPEX. CAPEX have been increased with 25%, while OPEX is decreased with 10%.
4.2.1 Experiment set: low SOx fuel
As the main alternative for ECA compliance is low SOx fuel, this is the first
experiment set. In this set a switch to low SOx fuel is demanded when sailing
to or from a port located in an ECA. The low SOx fuel that was used is Low
Sulphur Marine Gas Gil (LSMGO), as simple low SOx diesel is not sufficient
for compliance to the sulphur limit in ECAs as of 2015 (International Maritime Organization, 2014). The price difference between HFO and LSMGO is signifi-cant, and at the moment of writing LSMGO was 85% more expensive as HFO (Ship & Bunker, 2016). This price difference is used in this experiment set. LSMGO is assumed to be available at any port, so it can be used as much as required.
4.2.2 Experiment set: LNG - half round trip capacity
review. To determine the capacity loss for the other vessel classes the amount of containers lost per m3 of tank capacity was calculated and multiplied with each tank size. Tank sizes are converted from m3 to metric tonnes. This is done because of the way the price difference between HFO and LNG is implemented. The LNG price used represents the average difference identified in the literature review, namely that LNG is 70% cheaper than HFO on an energy basis. To incorporate this LNG and HFO need to use the same unit of measurement. The price of LNG is set to 70% of the price of HFO per metric tonne.
4.2.3 Experiment set: LNG - Reduced Capacity
The main difference in this experiment set with the half round trip capacity scenario is a reduced capacity of the LNG fuel tank. The four smaller vessel classes use tank sizes half of their half round trip equivalent. The two panamax vessel classes use very small LNG tanks, similarly sized as those of the 2500 TEU half round trip vessels. This is in line with work of Caballero et al. (2006). In addition, this experiment set provides insight in the trade-off between capacity and lower investment costs on one side and higher LNG utilization on the other. Capacity loss was adjusted using the same method as described in the half round trip experiment set. The price difference is the same as in the half round trip capacity scenario.
4.2.4 Experiment set: LNG - Containerized LNG
4.3
Experiment Setting
To solve the LSNDP for the newly created instances a neighbourhood search (NS) heuristic based on column generation is used. The iterative process of this heuristic was described in the literature review, but for more detailed informa-tion pseudo code can be found in appendix A. The heuristic is coded in C++ and CPLEX 12.6.1 is used to solve the MIP models. As the instances are of different sizes different approaches were used to determine run length. For the smaller instances like Intra Asia and Singapore Feeder, a fixed number of iter-ations is used, that ensures the objective value has stabilized sufficiently. The other instances require high levels of computational power and solving time of especially the rotation generation problem was unpredictable. Therefore, a run length instead of a fixed number of iterations was determined. When this run length is reached, the last iteration is finished and the heuristic terminates.
As complexity is increased significantly by the introduction of variable speeds and fuel types, solving speed slowed down significantly. To deal with this, the experiments were run on the Peregrine cluster of the University of Groningen. To solve the heuristic a single node on the cluster was used containing two Intel Xeon E5 2680v3, with 24 cores at 25 GHz. The larger instances all use 8 cores, as more did not provide any substantial improvement in solving speed.
The smaller instances use 4 cores. In addition, a limit on the
5
Computational Results
The computational results of this research are presented here. This is done by first comparing the experiment sets and data instances and each scenario based on their profit margins. Next, an actual route is discussed in more detail to illustrate how LNG is actually utilized in these networks. Tables containing detailed information on the costs and revenues, and performance indicators for all the median and best runs for every experiment set and dataset can be found in appendix C. These tables are used in describing and analysing the results but due to the size of all these tables they are not included in the main text to maintain readability.
5.1
Comparison of the LNG configurations
To compare all the different LNG and low SOx fuel scenarios the profits of the
best runs of each experiment were converted to profit margins. Profit margins provide insight on how much of the actual revenue earned is kept as profit and are useful when making comparisons based on profitability. These profit margins can be found below in table 3. Table 4 provides an overview of the ECA and LNG ports for each instance and scenario.
Table 3: Profit margins of the best runs of each experiment for each instance
Asia-Europe Europe- Singapore Intra-Asia Pacific
Trade lane Mediterranean Feeder Trade lane
Present
1. Low SOx
Best -0.6% 2.2% 49.0% 51.0% -6.0%
2. LNG half round trip capacity
Best 15.5 % 9.2% 54.0% 52.0% 17.1% 3. LNG reduced capacity Best 12.7% 7.6% 51.0% 59.0% 13.7% Future 4. Low SOx Best -2.5% 1.0% 44.0% 44.0% -1.0 %
5. LNG half round trip capacity
Best 16.9% 10.7% 51.0% 52.0% 15.2%
6. LNG low capacity
Best 17.0% 10.2% 51.0% 53.0% 16.5%
7. Containerized LNG
Best 11.2% 7.4% 49.0% 51.0% 14.1%
Present - Incl. Capital 8. LNG half round trip capacity
Best 9.5% 8.4% 50.0% 50.0% 7.4%
9. LNG low capacity
Best 10.5% 7.9% 50.0% 51.0% 11.7%
Future - Incl. Capital 10. LNG half round trip capacity
Best 9.6% 9.0% 50.0% 51.0% 13.2 %
11. LNG low capacity
Best 13.1 % 9.0% 50.0% 52.0% 16.4%
5.1.1 Present scenario
The higher profit margins for the LNG scenarios in table 3 clearly show that the use of Low Sulphur fuel as an ECA compliance strategy substantailly reduced the profitability of most of these networks. The low SOxscenario even produced
losses in the Trade Lane instances, with -6% profit margin as the highest loss in the pacific trade lane. This is explained by the larger distances covered in these instances, which cause fuel to be a significant part of total costs. Fuel costs make up 18% of total costs when using Low SOxfuel in the Pacific Trade lane
Network, while this is only 12.5 % in the half round trip capacity LNG scenario and 12.3 % in the reduced capacity LNG scenario. The switch to LNG caused profit margins to increase up to +23%. Exceptions are the instances in Asia,
which showed only slight improvements over the use low SOx. An important
reason for this is that there is currently no ECA in that region, which means low SOxfuel is not required. In addition, these instances have high demand volumes
and relatively short distances. This means fuel costs make up only a fraction of the total costs, namely roughly 10% in Intra Asia and 15% in Singapore. This makes these regions not as interesting when considering the use of LNG considering current ECA and LNG infrastructure. When considering the two LNG scenarios the half round trip scenario clearly outperforms the reduced tank capacity scenario, even though more container slots are lost due to the larger tanks. This can be explained by the limited current LNG infrastructure (see table 4 below), as larger tanks decrease dependency on LNG bunker ports.
5.1.2 Future scenario
The future scenario is characterized by an increased amount of ECA and LNG ports, as possible future ECAs and ports discussing the development of LNG infrastructure are included. Table 4 below describes this future scenario for each data instance. The Asia networks again are very limited in the amount of LNG bunker ports. The amount of ECA ports however, significantly increases in the Intra-Asia instance. In addition the main hub of the Singapore Feeder instance also becomes an ECA port explaining the decrease in profit margins from 49
to 44% for the low SOx scenario. As the amount of ECA ports substantially
increases for every instance the profit margins for the low SOx scenario went
down substantially for all instances compared to the present scenario. Similar differences that were found in the present scenario regarding LNG usage are also found in the future scenario. An important difference is the fact that in some scenarios (Intra-Asia and Pacific Tradelane), the low capacity LNG cases outperform the half round trip. Which is explained by the increase in LNG bunker ports, especially at major hubs like Shanghai. As more infrastructure is available smaller tanks can be used, which means less capacity loss. This is reflected in the increased revenue and reduction of penalty costs at these instances. Instances in Europe do not benefit as much from this developement as LNG infrastructure is already more developed and most major hubs already have
fuel, but performed worse than any of the other LNG scenarios. This implies that LNG infrastructure in the future scenario is sufficient to effectively use LNG without containerized LNG. This is also reflected in fuel costs of all LNG scenarios as HFO costs are very low and even close to zero in most instances that use half round trip capacity LNG tanks. The reduced capacity LNG scenario uses slightly more HFO, but still it makes up only a fraction of the total fuel costs.
Table 4: current and possible future ECA and LNG Ports of all the data instances
Number of ports LNG Ports - Present ECA Ports - Present LNG Ports - Future ECA Ports - Future
Asia-Europe Trade lane 46 8 10 16 25
Europe-Mediterranean 55 13 24 17 48
Singapore Feeder 11 1 0 1 1
Intra-Asia 31 1 0 3 13
Pacific Trade lane 43 2 7 6 23
5.1.3 Present scenario - Including CAPEX
The present scenario including CAPEX uses increased TC rates which makes the deployment of vessels more expensive. In the previous two scenarios CAPEX was assumed to be equal to the use of HFO. The effect of the increase in CAPEX is significant as was to be expected. Profit margins go down especially for the half round trip LNG scenario. Which is caused by the larger tanks that are used in this scenario which make up a large amount of the investment costs. The difference for the low capacity tank LNG scenario is not as substantial as profit margins only go down by a few percentage. An exception is Intra-Asia, which seems to have trouble finding an efficient route. As it uses less vessels (which makes sense when CAPEX is included), but these vessels used on average higher speeds as fuel costs are similar and the average port calls per week are similar. Also, transshipment costs are higher. This is explained by the fact that the rejection costs for both are close to zero. The choices made vary greatly, but this is believed to be a random event. As this difference is likely to disappear as more randomized runs would be performed to allow the heuristic to find better local optima.
5.1.4 Future scenario - Including CAPEX
This scenario is similar to the previous one except possible future ECAs and LNG bunker ports have been included. Again a slight reduction in profit margin is found for most instances, which is mostly explained by the increase in vessel costs caused by the increased TC rates. Intra-Asia and Singapore Feeder still show strong improvements regardless of the increased TC rates, as the increase in ECA ports and very expensive low SOx fuel drives up costs significantly.
in the future scenario. Not only do they provide more container slot and lower CAPEX, they also allow full LNG utilization with only a slight increase in LNG infrastructure. This shows that LNG does not require a fully developed LNG infrastructure.
5.2
Route and speed adjustment
In the previous section LNG utilization is considered mostly from a network per-spective. This provided insights in how various LNG configurations performed compared to the use of low SOx fuel in scenarios with varying ECA coverage,
LNG infrastructure, demand structures, and distances. As LNG utilization is also influenced by the decisions made in the actual routes, this section aims to examine the choices made. A single route of the Pacific Tradelane instance is selected from the best run of the reduced capacity LNG scenario with current ECA and LNG infrastructure.
5.2.1 Route: Pacific Tradelane - Current scenario
This specific route is selected based on the fact that it utilizes LNG the majority of the time while only bunkering once on the entire round trip journey. The route is presented below in table 5 and starts from the base port Los angeles.
Table 5: Route from the Pacific Tradelane network - Current Infrastructure - Reduced Tank Ca-pacity LNG scenario
Port(in order of port call) LNG available Located in ECA Port window broken Speed used to next port LNG used 1. Los Angeles yes yes no 16 yes
2. Oakland no yes yes 19 yes
3. Busan no no yes 15 no
4. Hong Kong no no no 16 yes
5. Shenzhen no no no 22 no
6. Shanghai no no no 16 yes
7. Los Angeles yes yes yes 12 yes 8. Long Beach no yes no 12 yes 9. Los Angeles yes yes no
6
Discussion
The main findings show that the best LNG configuration depends on the LNG infrastructure, demand, distance, and ECA presence. When LNG infrastruc-ture is limited half round trip tank sizes suffice to utilize LNG as the main fuel and allow ECA compliance. Small improvement in LNG infrastructure signifi-cantly decrease the dependence on a large tank, which allows for lower capital investments and more available container slots. All LNG scenarios that were considered provided higher profit margins than the main alternative, which is low SOx fuel.
Findings are in line with the findings in the literature review. The main difference is that the effect of the limited infrastructure is not as significant as described in current literature. Small tanks proof to be viable options, which is in line with the work of Caballero et al. (2006). However, it conflicts work of Sames et al. (2013). Implications for practice are that vessels with smaller tanks can be utilized which leads to lower capital investments and higher container slot capacity.
This research has several limitations. Due to the limited access to real
data an existing dataset was used as the main source to create new instances.
The data used originates from 2012. However, that dataset was created in
collaboration with Maersk, which means the demand structures represent real life cases. Another limitation are the assumptions made when designing the CAPEX experiment sets, as well as the containerized experiment set. These assumptions were made due to a lack of data. Results from the literature review were used to mitigate this, and provide a dataset which is as realistic as possible.
7
Appendix A: The Neighbourhood Search
Heuris-tic
Blocks of pseudo code of the NS heuristic that was used to solve the LSNDP in this reseach are provided here. The rotation generation block (AUX block), the MCFP block, and the NS heuristic block itself can be found below.
Figure 1: AUX block (Wal, 2015)
Loop until one of these conditions are met:
Condition1: The number of vessels of type v available for new rotations in iteration u is 0 for all v.
Condition2: The residual demand between all ports in cluster ˜ku od
is 0.
Condition3: There is no solution found. Compute for each cluster ηκ and select a cluster;
for All vessel types selected do
for All number of vessels within the vessel type do Solve AUX(υ,κ);
Save the solution rotation; end
end
Compare the solution rotations and save the one with the highest profit;
if The highest profit solution > 0 then
Add the rotation to the set of rotations for the MCFP; Add all costs and revenues of the rotation to the total profit function;
Subtract all used vessels (type and number of vessels) from the available vessels in this loop;
Substract all demand transported from the remaining demand to be transported;
else
There is no solution found; end
Figure 2: MCFP block (Wal, 2015)
Perform the following:
-Define a multi commodity flow graph based on the rotations created in this loop;
-Define a multi commodity flow problem from the graph; -Solve the MCFP and save its solution;
-Compute all related costs and profits;
-Update the total objective function for this iteration;
-Update the remaining demand for this iteration with the found MCFP solution;
Figure 3: NS block (Wal, 2015)
Initialize: Iteration= 0; Empty all rotations;
Set all remaining demand to total demand; AUX block;
MCFP block;
//Now there is a initial solution
//Stopping condition can either be a time limit or a number of iterations
if Stopping condition is not reached then Iteration= Iteration+1;
Set of new rotations = set of last found rotations;
Copy all costs and used vessels to new rotation’s variables; Select a rotation selection strategy according to the weights; According to the strategy picked, remove rotation(s) from current set of rotations;
Remove the related costs and add the newly available vessel to the rotation’s variables;
Compute the amount of cargo that can no longer be transported due to the removal of the rotation(s);
Add the newly untransported cargo to the remaining demand; AUX block; MCFP block; else STOP heuristic; end //End of heuristic
