RAM of Entire LNG Supply Chain

Background

An Operator was evaluating the possibility of building a new LNG plant in N Africa for the liquefaction of gas from onshore and offshore fields with characteristics suitable to serve both the European and American markets.

I undertook a RAM (Reliability, Availability and Maintainability) study to evaluate the entire LNG supply chain for the development to maximise operability and availability of LNG. The study also included marine transit analysis and ports delay analysis for the proposed 8 ports where the LNG products would be offloaded.

Objective - the main objective of the base case models was to determine the required storage tanks volume. 

Basis

In order to obtain a true representation of the availability, a RAM model for the complete LNG supply chain from the LNG liquefaction plant to the different offloading ports were considered. The RAM model therefore included the following:

• LNG liquefaction plant;
• LNG bulk storage tanks;
• Loading arms;
• Jetty;
• Ship and ship size;
• Ports delay analysis;
• Marine transit analysis.

Port Delay Analysis

The berthing and loading/unloading of ships at the ports can be delayed due to wind, tide or fog. The port weather data was used to create a Weibull probability distribution for the events. A Weibull curve was used so that the effect of clustering (i.e. bad weather is more likely the day after bad weather) can be modelled.

 Return Journey Ship Transit Time

Transit data to each gasification port were provided along with probabilities for transit durations. The return journey distributions were determined by the calculating the probability and journey time for every possible combination of transit time to port, transit time in port and delays in port. The journey time was then assigned to the arrival data and all the probabilities summed together. An example of return journey transit time distribution is shown below.

Reliability Block Diagrams

Upstream of the storage tanks the model consists of two parallel, identical trains of acid gas removal, dehydration and liquefaction. A reliability block diagram representation of one of the ELNG trains, as included in the RAM model, is shown below.

In addition to the liquefaction trains, the model also includes the main utility and support systems, which are shared by both trains. These included:

• Wet system flare
• Dry system flare
• Instrument air
• Main power generation
• Fuel gas
• Hot oil

The OPTAGON screenshot of the reliability block diagram downstream of the LNG tank(s) is shown below. It shows the tank, eleven transfer pumps, three loading arms (with logic for the vapour return arm), a block to model spurious ESD and a block to model power cuts.

Failure and Repair Times

Failure rates and repair times of equipment critical to production was based on industry standard data, either from OREDA or from an in-house reliability database.

 Conclusions

Base Case

• Four ships are required to fulfil the shipping schedule taking in to account the transit times to the ports.
• A ship + 5 days storage (307,205 m3) would result in less than one tank top or tank empty incident per year.

Number of Trains

• The availability of the liquefaction process is ~99.50% with one 100% train compared to ~99.88% with two 50% trains.
• The train configuration does not significantly affect the tank events and so does not influence the tank storage volume.

 Driver Type

• There was minimal benefit from using an electric driven generator instead of a gas turbine driven. This is due to the high availability already obtained for the liquefaction trains using a gas turbine driver.

 US Spot Market

• Utilising the US Spot Market was not recommended as it would require both an extra ship and extra storage tank capacity.

 Early Production System

• There is no significant change in the results when modelling a 3 MTA case with 6.67 MTA capacity assuming an early production system from the gas network to the LNG liquefaction terminal.

I think this is a good example of needing to model an entire system to obtain the result for just a small aspect of the overall system. This is relevant to many studies and not just RAM where the interfaces throughout the system can have an impact at the part you are looking at.