Modeling the North American Market for Natural Gas Liquids with NGL-NA™

Robert E Brooks
Founder & President, RBAC, Inc.
(Sherman Oaks, CA)

This paper concerns NGL-NA™, a new model designed to analyze and forecast market fundamentals in the North America natural gas liquids (NGL) market.  A principal impetus for this model is the expanding production of natural gas and natural gas liquids due to the “shale gas revolution”.  This phenomenon has stressed existing NGL infrastructure and caused unanticipated declines in NGL prices.  The industry requires a realistic scenario analysis tool to help it identify and evaluate investment opportunities in this fast growing but risky market.

Description of Market

The NGL market originates with production of crude oil, condensate, and natural gas.  This is a somewhat crude classification into three categories of a whole spectrum of hydrocarbons produced by the industry, from the heaviest and most complex (crude oil) to the lightest and simplest (natural gas).  Condensate resides in the middle zone, consisting of those lighter hydrocarbons that are yet liquid at temperatures and pressures found on the ground above the formations from which they are produced.  Any given well could produce one or more of these products which are then generally separated on the leasehold for transport to different markets.  A well is usually classified by which of these three is its most predominant product.

Crude oil produced in the US or imported from other countries is shipped to refineries via pipeline or rail car.  Heavy crude such as that produced in the oil sands of Alberta requires blending with lighter hydrocarbons to enable or improve transportability.  The blending material is called diluent.  This material may consist of condensate produced in the same or other fields or may be natural gas liquids produced by natural gas processing plants or even liquefied refinery gases produced by refineries.

Condensate is shipped by pipe, rail, or possibly truck to refineries or to heavy oil fields or to “splitters” which separate the lighter natural gas liquids from the heavier hydrocarbons such as naphtha.  From the splitters, the NGLs are then sent to fractionators and the naphtha to refineries.  Condensate is sometimes sold as diluent.

Natural gas is mostly methane but also contains ethane, propane, isobutane, normal butane, and “natural gasoline” (pentane, hexane, etc.)  These are termed “natural gas liquids” even though all but natural gasoline are gaseous at room temperature and pressure.  If the gas is very “dry”, it can be inserted directly into natural gas pipelines for transport to market.  If it is “wet”, it is usually sent to natural gas processing plants where the liquids are separated from the dry gas.  The exceptions occur when there is insufficient gas processing capacity or when NGL prices are too low to justify extraction.  In these cases, producers sell the wet gas at dry gas prices, or extract only the heavier NGLs, and deliver the resulting gas to pipelines.  Pipelines, however, limit the wetness of the gas so it will not adversely affect their operations.

From the processing plant, the dry or “residue” gas is delivered to a gas pipeline and liquids can be loaded into mixed NGL (Y-Grade) pipelines, rail cars or trucks.  Wetness of gas is measured in gallons of mixed NGLs per thousand cubic feet of wet gas (“GPM”).  The following figure shows substantial regional variation in wetness of gas.


  Figure 1:  Natural Gas Liquids Content by Basin (National Petroleum Council, 2011)


Mixed NGLs are generally delivered to fractionation plants where the Y-Grade is separated into the individual “purity” products often labelled as C2, C3, N-C4, I-C4, and C5+.  Both NGL Mix and purity products are stored in separate caverns in underground salt dome storage fields (or above ground tanks when volumes are small enough).  Injection and withdrawal from these fields is accomplished using immiscible brine which is stored in above ground ponds when not injected underground.  The amount of usable storage is limited by both the underground capacity and the above ground brine pond capacity.

Each of these products has its own set of markets.  Ethane is almost exclusively used in the manufacture of ethylene.  Propane can be used for cooking and heating in areas where natural gas (methane) distribution is not available, in transportation, in agriculture and in ethylene production.  Normal butane is used by oil refineries as blending material for gasoline and for making iso-butane which is also a blending component.  The demand for isobutane and normal butane in gasoline blending is quite seasonal.  This is due to EPA regulations that impose RVP (Reid Vapor Pressure) restrictions during the summer months (typically Jun-Sep) in order to reduce transportation related air emissions.  These summer restrictions increase the demand for iso-butane and correspondingly reduce the normal butane demand for motor gasoline blending purposes.  As ambient temperature changes occur during the fall and winter months, these restrictions are eased which allows for greater normal butane blending (i.e. increasing normal butane demand). Note that specialized “isomerization” units are also employed to convert normal butane to iso-butane to help satisfy this seasonal demand. Butane is also used to create butadiene, a chemical used in production of synthetic rubber.  “Pentane +” or “natural gasoline” is typically used in gasoline blending but can also be used as diluent for heavy oil transportation.  The market for diluent is growing due to increasing bitumen production in the oil sands region of Alberta.

Production of ethylene, and to a lesser extent, propylene, butadiene, and other chemicals is accomplished in petrochemical plants known as “steam crackers”.  More ethylene is produced worldwide than any other organic chemical.  The composition of the inputs is controlled to yield the desired outputs at the lowest possible input cost. 

Demand for ethylene is the primary driver for the petchem industry as it is the principal ingredient for polyethylene, the most important starting product for production of plastics.  About two-thirds of all propylene is used to make polypropylene, the second most important starting product in plastics.  Though propylene is typically a by-product of ethylene production, propane dehydration plants (PDH) exist to produce “on-purpose” propylene.  Similar BDH (butane dehydration) plants are used to make “on-purpose” butadiene.


NGL-NA Product Flow Model

Figure 2 below shows a node-arc representation of the NGL-NA product flow model.  Nodes (circles) represent operations and arcs (directed arrows) represent product flow.  The basic operations considered are Production (including separation and conditioning), Condensate Splitting, Crude Oil Refining, NGL Fractionation, and NGL Steam Cracking.  Subsequent to any operation, the resulting product can be stored or moved to one or more of the other operations by pipeline, rail, truck, or barge.  Implicit in the diagram below is that storage of both inputs and outputs can occur wherever some type of conversion operation occurs.  Potential bottlenecks in the system are production and storage capacity in each operation in the supply chain as well as transportation between operations.


 Figure 2:  NGL-NA Product Flow Model

Product can enter the system as imports or leave as exports using tankers.  Imports and exports among the US, Canada, and Mexico are considered to be endogenous to the system whereas those from or to other countries are computed as excess demand or supply in various regions.  Import and export prices are exogenously set by the user as scenario assumptions.


Modeling Considerations

Oil and Gas Production

Levels of oil, gas, and condensate production are user inputs.  These can be outputs from a natural gas market model scenario or any other source.  North American oil production is an important input, because the changing quality of oil is already making a difference in terms of refinery outputs, including that of LPG (liquefied petroleum gas), a mixture of propane and butane.

Wet gas in each producing region is characterized by its chemical composition.  This can be modeled as composition by weight or by wetness content (GPM, or gallons of liquids per thousand cubic feet of gas) and per cent composition of the liquids by weight.  Assumptions about composition may change over time to reflect a trend toward greater dryness or wetness of the gas as the play is developed. However, model size grows proportionally with number of distinct compositions.  Though composition data is not generally available from public or industry sources, it can be inferred from data on natural gas and liquids production.

Crude Oil Refining

Crude oil comes in many different compositions.  For the purposes of NGL-NA, we need to be able to characterize a variety of crudes produced and imported into North American refineries by their yields.  The model must then be able to compute how much LPG is produced per barrel of each type of crude produced or imported and how much NGL purity product is needed for gasoline blending.  This is clearly a very complex operation and is unique to each plant and crude used by each plant.  The challenge for NGL-NA is generalizing this operation using a small number of parameters to adequately represent the process of converting a barrel of crude into those refined products which affect the NGL market, namely LPG and gasoline, as well as the required volume of butane and “natural gasoline” required for gasoline blending.

Condensate Splitting

Condensate can be utilized as is or can be “split” into light and heavy streams.  The light steam is an NGL Mix and the heavy stream is composed of naphtha and other compounds.  Each condensate is unique.  In NGL-NA, each must be characterized in a reasonable way.  This means the model will need parameters representing the split between light and heavy and some idea of the composition of the NGL mix.

Gas Plant Processing

The figure below gives a very good generalized picture of “gas processing”.  It includes the separation operation discussed earlier under Oil and Gas Production, elimination of non-hydrocarbon compounds and elements such as water, CO2, sulfur, helium, and nitrogen (conditioning), removing methane from the resulting hydrocarbon stream and then “fractionating” the resulting mix into the five component NGL purity products.  In the NGL-NA model, the primary function of gas plant processing is separating wet gas into the dry gas (mostly methane) and mixed NGLs.  It assumes non-hydrocarbon compounds have already been netted out of the initial specification of “wet gas” production volume.  Separation of mixed NGLs occurs in a separate process called “fractionation”.

 Figure 3:  Generalized Natural Gas Processing Schematic (EIA)


NGL Fractionation

Fractionation separates NGL “purity” products out of NGL mix.  The resulting products are ethane, propane, normal butane, iso-butane, and a mix of pentane and other higher order hydrocarbons called “pentanes plus” (or “natural gasoline”).

Some fractionators do not produce ethane because the NGL mix they receive as input material lacks it.  An example of this is Dominion’s Hastings Extraction / Fractionation plant in West Virginia.  Such situations occur when there is neither a petrochemical plant nearby to consume the ethane or an ethane pipeline to move it to such a plant.

Some fractionators produce a combined product known as “E/P Mix” consisting of ethane and propane, rather than fully separating them.  This mix can be used as feedstock in steam crackers. 

Steam Cracking

Petrochemical plants are often known as “steam crackers”.  They use the energy in high temperature steam to convert NGLs such as ethane (C2H6) into ethylene (C2H4) or propane (C3H8) into propylene (C3H6) or other products.  In general, a steam cracker is primarily in the business of creating ethylene from whatever inputs it can utilize most economically.  While doing so it also produces propylene, butadiene, and other by-products. 

A steam cracker located near a fractionator will use the feed material which will produce the highest margin, given the market price of ethylene and any by-products as well as the various feedstock prices.  Some crackers are optimized to use “light” feeds such as ethane, propane, and butane, whereas others also have the flexibility to use “heavy” feeds such as naphtha produced in oil refineries.  In each case, the market prices of the feeds and products and the flexibility of the plant will determine feed and product slates actually scheduled.

A cracker located some distance from the fractionator often has more restricted options.  For example, Westlake Chemical’s ethylene plant in Kentucky uses propane off of the TEPPCO pipeline as its only feed material. However, due to the huge increase in natural gas production in the Appalachian Basin, it is now planning to convert the plant to use ethane delivered via a new pipeline from Pennsylvania to Texas. 

NGL Transportation

NGL pipelines can move mixed NGLs or purity products or both.  Other specialized pipelines move petrochemical products such as ethylene and propylene.  When a pipeline can move more than one kind of product, it must be operated in a batch mode:  it sends a certain amount of product of one type and then another type. 

NGL-NA’s database contains tables of data on  each NGL pipeline, its various receipt and delivery points and the capacity and cost of transporting each product.  NGLs can also be moved by rail or truck.  NGL-NA also contains data forrail connections and costs and truck transportation rates for movements of NGLs.

NGL Storage

NGL mix and NGL purity products are largely stored underground in salt domes.  The largest of these is located under Mt. Belvieu, Texas.  Thus this is the most important receipt point for NGL mix from gas processing plants and delivery point for purity products to steam crackers and other downstream markets and export facilities.  Other large centers are located in Kansas, Oklahoma, Louisiana, Mississippi, and Sarnia, Ontario.

NGL-NA contains capacity and cost information on all important salt dome storage, as well as aboveground tank storage, used in the storage of NGL mix and purity products. 

Imports and Exports

The large increase in natural gas production over the past several years has created the   potential for a substantial increase in exports of NGLs and petrochemical products.  NGL-NA models export terminals and demands for these products. 



AMPL and Gurobi

AMPL, a modeling language for mathematical programming (http://www.ampl.com), was used to interactively design the model and to provide the interface between the database containing the model information and the solver (Figure 4). 


Figure 4:  NGL-NA System Components and Data Flow

An aggregated version of NGL-NA was constructed using data available from EIA and other sources.  Natural gas supply data was gathered by PADD region (see Figure 5) for the period 2006-2012.  This was used to generate price-insensitive supply curves.  Price-sensitive PADD-level demand curves were created for each product market over the same period. 


Figure 5:  PADD Regions Used in Phase 1 NGL-NA Model

The resulting model consists of about 4,000 variables and 2,500 constraints for each period in the scenario.  Initial calibration runs to match marketed natural gas and processing plant liquids production cover the seven year period Jan-2006 through Dec-2012. 

The calibrated objection function value of $782 billion, representing the economic benefit of the US NGL market, amounts to about $4.80 per 1000 cubic feet of marketed production.  It is also equivalent to about $13.30 per bbl of NGLs produced by gas processors.


Preliminary Results

NGL-NA has a number of reports which show the resulting product flow and market price results from a scenario. 

Figure 6 shows a graph of marketed production for each PADD region and for Western Canadian Sedimentary Basin (WCSB) gas imported into the US via the Alliance Pipeline.




Figure 6: Marketed production by PADD (mmcf/day)

Figure 7 shows results from the NGL-NA processing capacity utilization report. Note the increasing utilization of capacity in PADD1 from less than 20% to close to 100%.  This is likely a misspecification of total capacity in the earlier periods. 



 Figure 7:  Cryogenic Processing Plant Utilization by PADD

Figure 8 shows a graph of the dual (shadow price) associated with capacity on the Teppco pipeline.  This graph measures the additional value to the market of an additional unit of capacity on the pipe.  Note that in early years incremental capacity is worth about $13/bbl but grows to as much as five times higher in later years.  In this scenario, Teppco’s utilization is 100% throughout 2006-2012.  In some months, alternative transportation is available to serve the same markets without driving up cost, so the dual is zero.  In other months, more expensive alternatives are required, resulting in positive duals.

Figure 8:  Dual (shadow price) of transportation on Teppco NGL pipeline.


Table 1 shows annual deliveries of ethylene to markets in each of the PADDs.  The demand functions for all PADDs were parameterized with 2% annual growth but only PADDs 2 and 3 deliveries could achieve that growth rate. 
























































Table 1:  Ethylene deliveries by PADD (lb/year)

Figure 9 shows the evolving price of ethylene over the period 2006-2012.  It is interesting that the prices diverge as relative demand for these products change.  The price for ethylene grows substantially in those PADDs where deliveries cannot keep up with demand.


Figure 9:  Ethylene market price by PADD (cents/lb)

Note that these prices have not been calibrated against historical data but are displayed as representitive of the types of results available from the NGL-NA system.  Sources for such historical data have been identified for future calibration.



In order to be practical, a model must be solvable in a reasonable time.  Linear models can be solved quickly while most non-linear models cannot.  Thus it pays to try to identify ways to keep a model linear if at all possible.  In the case of NGL-NA, by novel identification and handling of the various mixed NGL commodities in the model, an apparently mathematically difficult quadratic constraint model has been converted into one with only linear constraints.  The non-linear objective function has also been simplified by use of a piecewise-linear approximation.  As a result, a possibly intractable model has been converted into one which can be solved efficiently using modern primal and dual simplex algorithms such as CBC and Gurobi.




The author thanks Scott McKenna and David Brooks of RBAC, Inc., for identifying data sources and collecting, processing, and evaluating the extensive data required by NGL-NA.  

Data Sources

EIA (http://www.eia.gov)

CERI (http://www.ceri.ca)

Drillinginfo - DI Desktop (HPDI) (http://www.didesktop.com/)

Sulpetro - LPG Almanac (http://www.sulpetro.com/)

Oil Price Information Service (OPIS) (http://www.opisnet.com/)