G2M2 Wiki

One of the most difficult calibration challenges with any network model is to match historical flows between various sources and destinations. It is believed this is due to the fact that there are generally so many globally equivalent alternate optimal solutions to min-cost or partial equilibrium class models. Additional constraints appear to be required in order to try to match historical flows, even when historical supply and demand numbers match well.

In G2M2 we use several features of the market to try to accomplish this task. The first involves modeling LNG supply contracts. For each publicly available contract, we establish the Seller and the Buyer, the annual contract volume, the contract start year and duration, and various other factors, and then use constraints in G2M2 to try to enforce satisfaction of the contract. The maximum contract volume is modeled as a hard constraint. The Min Pct parameter multiplied by the annual contract volume is used to represent a Take-or-Pay quantity for the contract. This is modeled as a soft constraint (i.e. each unit of violation is given a penalty cost in the objective function.)

Some contracts specify the source LNG facility. The sum of the annual volume times the min pct for all such contracts for each facility provides a minimum requirement for that facility’s production. Even though this is modeled as a soft constraint, it still provides reasonably good support for each such facility’s utilization. The same cannot be said for LNG destinations (i.e. import facilities). The reason is that a large percentage of LNG contracts have destination flexibility. For some of these, the source is defined, but not the destination. For others, only the buyer and seller are named; both source and destination are not named. For such contracts, the buyers and sellers have flexibility for where the LNG is produced and where it ends up being delivered. But this is not true for all contracts. Many contracts are made with buyers which do want to receive the LNG at specific import facilities or countries. Using GIIGNL reports from earlier years in combination with assistance from colleagues in Japan, we have identified those contracts with “destination restrictions”. As long as those contract records also identify a destination country (or subregion or region), G2M2 will attempt to route LNG to that destination on that contract.

Many countries have policies discouraging reliance on only one or a few suppliers for strategic commodities such as LNG. G2M2 provides a means for modeling such policies. It has a table called LNGDestSharesData with records specifying both a minimum and maximum fraction (share) of total LNG imports from origin countries to each destination country. These min and max shares can be different for different time periods. Here’s an example of data for Japan for year 2025:

Note that Australia is allowed to have up to 49% of the market (and a minimum of 28%) but no other country is allowed to have more than 25%. This constrains the solution to require at least four suppliers in 2025.

The next table shows the yearly shares that US exporters would be allowed to have of the Japan market: Here note the historical share of between 1 and 2% in 2017 growing to between 6 and 7% in 2020. For our 21Q3base case, we projected the evolution of the min-max range for these shares by decreasing the min share and increasing the max share by 2% each year until the min share went to zero and the max share went to 20%. In other words, we designed a scenario where the market share in Japan for US LNG was allowed to deviate from the historical share by 2% per year until it reached a min of 0% and max of 20%. We allowed a gradual evolution from the past while also reducing risk of dependence on the US by capping its maximum share at 20%.

We use this same concept to compute min-max share ranges for all source-destination country pairs. For each historical year, we use historical LNG flows reported in the BP Statistical Review to compute source shares for each destination in each year. We then set the Min Share to the integer percent value just less than the computed share and the Max Share to the integer percent value just greater than the computed share. For example, a share of 5.67% would be bounded by 5% and 6%.

Using this methodology, we compute LNG Destination Share Data for each pair for each year from 2011 through 2020. For 2021 and future years, we decrease the min and increase the max share values by a certain amount each year. In the example above, we used 2% per year. If the computed min value is less than 0%, we set it to 0%. If the computed max value is greater than 100%, we set it to 100%. We can then choose to reduce specific max shares which are very large. In the example above, based on information from our Japanese colleagues, we capped the US to Japan share at 20%.

When creating scenarios, you are free to choose other criteria for evolving these destination shares. For example, you could change the min and max by only 1% per year rather than 2%. You can also choose to establish more severe restrictions on future max share values for certain countries. You can run various such scenarios and compare results to see how sensitive results are to these choices.

While constraining the destination shares in a G2M2 scenario to values near the historical shares helps, calibration could be improved by also restricting total imports to be closer to historical levels. This could be implemented with a min-max approach as above. The BP Statistical Review could be used as guidance for computing these parameters. In the future, these constraints could be loosened to represent future changes in contracting and sourcing of supply.

Please contact customer support if you have any further questions - RBAC, Inc. support line (281) 506-0588 ext. 125, from 9:00 am to 5:00 pm CT