Peterson, Steve with Brian Bush, Emily Newes, Danny Inman, David Hsu, Laura Vimmerstedt and Dana Stright "An Overview of the Biomass Scenario Model", 2013 July 21 - 2013 July 25

Online content

Fullscreen

An Overview of the Biomass Scenario Model

Steve Peterson, Lexidyne LLC

Emily Newes, NREL

Danny Inman, NREL

Laura Vimmerstedt, NREL

David Hsu, NREL

Corey Peck, Lexidyne LLC

Dana Stright, Lexidyne LLC

Brian Bush, NREL

August 2013

Submitted to
The 31st International Conference of the System Dynamics Society
Cambridge, Massachusetts USA

July 21 - July 25, 2013

An Overview of the Biomass Scenario Model | August 2013

Introduction

Biofuels are promoted in the United States through aggressive legislation, as one part of an
overall strategy to lessen dependence on imported energy as well as to reduce the emissions of
greenhouse gases (Office of the Biomass Program and Energy Efficiency and Renewable Energy,
2008). For example, the Energy Independence and Security Act of 2007 (EISA) mandates 36
billion gallons of renewable liquid transportation fuel in the U.S. marketplace by the year 2022
(U.S. Government, 2007). Meeting such large volumetric targets has prompted an
unprecedented increase in funding for biofuels research, much of it focused on producing
ethanol and other fuel types from cellulosic feedstocks' as well as additional biomass sources
(such as oil seeds and algae feedstock). In order to help propel the biofuels industry, the U.S.
government has enacted a variety of incentive programs (including subsidies, fixed capital
investment grants, loan guarantees, vehicle choice credits, and aggressive corporate average fuel
economy standards) -- the short- and long-term ramifications of which are not well understood.
Efforts to better understand the impacts of incentive strategies can help policy makers to
develop a policy suite which will foster industry development while reducing the financial risk

associated with government support of the nascent biofuels industry.

Purpose and overview

This paper describes the Biomass Scenario Model (BSM), a system dynamics model developed
under the support of the U.S. Department of Energy (DOE). The model is the result of a multi-
year project at the National Renewable Energy Laboratory (NREL). It is a tool designed to

better understand biofuels policy as it impacts the development of the supply chain for biofuels

' These feedstocks, such as agricultural and forestry residues, perennial grasses, woody crops, and municipal solid
wastes, are advantageous because they do not necessarily compete directly with food, feed, and fiber production.

An Overview of the Biomass Scenario Model | August 2013 2

in the United States. In its current form, the model represents multiple pathways leading to the
production of fuel ethanol as well as advanced biofuels such as biomass-based gasoline, diesel,

jet fuel, and butanol).

The BSM uses a system dynamics modeling approach (Bush et al., 2008), developed using the
STELLA software platform (isee systems, 2010) to model the entire biomass-to-biofuels supply
chain. In order to gain a clear view into the evolution of the supply chain for biofuels, BSM
focuses on the interplay between marketplace structures, various input scenarios, and

government policy sets, as shown in Figure |.

Government Policies
Marketplace Structure Analysis
Producer/C onsumer exchanges 7 ~~» * Implications
Investment A _ : \ Inclusion decisions /scope
Financial decisions t ‘\

for Biofuels

Input Scenarios

Feedstock demand
Oil prices
Learning curves

Figure |. BSM strategy and approach

An Overview of the Biomass Scenario Model | August 2013

In this paper we begin with a description of the BSM architecture. Then, we provide a more
detailed view of the sectors and modules which comprise the BSM. Third, we outline the “back-
end” system that we have developed in order to support analysis efforts with the BSM. Fourth,
we describe a set of scenarios used as the basis for policy exploration with the model. Finally,
we provide a summary of current and potential uses for the model. A set of appendices detail

important structures relating to pricing, investment, and vehicle vintaging.

An overview of the BSM architecture

BSM has been designed in a top-down, modular fashion which allows material (feedstocks) to
flow down the supply chain and be converted into various types of biofuels, with feedback
mechanisms among and between the various modules. In developing the model, we have taken
care to create a structure that is transparent, modular, and extensible, enabling standalone
analysis of individual model components as well as testing of different module combinations. As

shown in Figure 2, the model is framed as a set of interconnected sectors and modules.

External Data
and Scenarios

Biomass Scenario Model f

¥
Feedstock
Supply and
Logistics
* Supply
* Logistics ~<}++--

Conversion
+ Algae
* Oilcrops
* Cellulose to “Infrastructure
Compatible”

* Cellulose to Butanol
* Cellulose to Ethanol
* Starch to Ethanol

Petroleum Industry
Vehicles (car, light truck)
* Gasoline, Diesel, FFV, PEV,
H2, CNG, HEV, C3H8, other
* Vehicle miles traveled
+ Efficiency
* Fuel demand

Downstream Ethanol
and Butanol

* Distribution Logistics
* Dispensing Stations
+ Fuel Use

= Material flow (e.g., Feedstock, fuel)

Information flow (e.g., relative fuel price)

Figure 2. Overview of BSM structure

An Overview of the Biomass Scenario Model | August 2013 4

Feedstock supply and logistics

The feedstock supply and logistics sector captures the dynamics of cellulosic, oil crop, and
starch feedstock supply from agricultural lands within the context of the operation of the U.S.
agricultural system. It incorporates harvesting and transportation logistics associated with
cellulosic feedstock, as well as feedstock supply and logistics associated with forest, urban, and

agricultural residues.

Feedstock production from agricultural land occurs against the backdrop of other uses of the
agricultural land base. These uses include commodity crop production (corn, wheat, soybean,
small grains, cotton), hay, pasture, and Conservation Reserve Program (CRP) land. The
agricultural production system is disaggregated regionally into 10 production regions taken

from the U.S. Department of Agriculture (USDA), as shown in Figure 3.

An Overview of the Biomass Scenario Model | August 2013

USDA Farm Production Regions

a \,
, eu Lake )"
Pacific States
Plains = >}
\ } 5
a | i s ‘
_| (ees
Southern States Southeast \”
Plains )
som ow
» .
a, 8

Figure 3. USDA farm production regions

Conversion
The conversion sector is composed of six different modules, each corresponding to a different

set of pathways for production of biofuels.

e Starch to ethanol: This module represents the conversion capacity acquisition and
utilization dynamics associated with the existing starch (corn) ethanol industry. The
industry is considered to be mature; hence, the module provides a simple
representation of the financial logic that controls acquisition and utilization of
commercial scale corn ethanol facilities. This module is disaggregated by USDA

production regions.

An Overview of the Biomass Scenario Model | August 2013

¢ Cellulose to ethanol: This module captures the development of the cellulose-to-ethanol
conversion industry. Biochemical and thermochemical conversion options are
considered on a USDA-regionalized basis. The module represents pilot, demonstration,
pioneer-commercial and full-commercial scale facilities. It includes learning curve
dynamics, investment decision logic, and utilization logic for both pioneer- and full-
commercial scale facilities.

¢ Cellulose to butanol: This module captures the development of the cellulose-to-butanol
conversion industry. In BSM, butanol serves as an industrial solvent and as a substitute
for ethanol in the oxygenate market. A single, regionally-disaggregated cellulose-to-
butanol conversion option is captured in the model. The module represents pilot,
demonstration, pioneer-commercial and full-commercial scale facilities. It includes

learning curve dynamics, investment decision logic, and utilization logic for both pioneer-

and full-commercial scale facilities.
¢ Cellulose to “refinery ready:” This module captures the industry development of
cellulose-to-refinery-ready “infrastructure compatible” conversion processes. The
model structure can accommodate the following conversion options:
co Fast pyrolysis
o Fischer-Tropsch
o Methanol to gasoline
o Catalytic pyrolysis
o Fermentation

o Aqueous phase reforming.

An Overview of the Biomass Scenario Model | August 2013 7

As with other cellulosic conversion modules, this module is disaggregated by USDA
regions. It provides a representation of pilot, demonstration, pioneer-commercial and
full-commercial scale facilities. It includes learning curve dynamics, investment decision
logic, and utilization logic for both pioneer- and full-commercial scale facilities. Multiple
products or product substrates can be produced, including gasoline, diesel, and jet fuel.

The “drop-in point” for various products is determined as a scenario variable.

¢ Oil crops: The oil crop module captures development of conversion capacity for soy-to-
refinery and “other” oilseed-to-refinery processes. Oil crop conversion facilities are
represented as U.S. aggregates (rather than disaggregated by USDA production region).
The module represents pilot, demonstration, pioneer-commercial and full-commercial
scale facilities. It includes learning curve dynamics, investment decision logic, and
utilization logic for both pioneer- and full-commercial scale facilities.

e Algae: The algae model represents open pond, photobioreactor, and heterotrophic
conversion options. It is not geographically disaggregated. Algae feedstock production is
presumed to be vertically integrated in the algae to refinery-ready system. The module
represents pilot, demonstration, pioneer-commercial and full-commercial scale facilities.
It includes learning curve dynamics, investment decision logic, and utilization logic for

both pioneer- and full-commercial scale facilities.

In addition to the six conversion modules, the conversion sector includes a simple module that
knits together the “attractiveness” of the various investments in conversion options, allocating
limited facility construction capacity among these options based on their perceived relative

economic value.

An Overview of the Biomass Scenario Model | August 2013 8

Petroleum industry

The petroleum industry sector comprises scenario inputs around crude oil prices, providing

logic that translates these prices into price inputs for the various refinery ready conversion

modules as well as the pricing/inventory module of the downstream ethanol/butanol sector.

Additionally, the petroleum industry model provides accounting logic that captures

displacement of crude by biofuel-derived infrastructure compatible fuels.

Downstream ethanol and butanol

The downstream ethanol and butanol sector is composed of a set of four modules. These

modules capture activities “downstream” of conversion, for ethanol and butanol.

An Overview of the Biomass Scenario Model | August 2013

Pricing/Inventory: This module captures pricing and inventory dynamics for both ethanol
and bio-based butanol. Ethanol flows into two distinct but coupled markets: the “low-
blend” oxygenate market and the “high-blend” market associated with flexible-fuel
vehicles (FFV). Bio-butanol is assumed to serve as a substitute for ethanol in the
oxygenate market, and also can supplant butanol produced by other processes in the
industrial market.

Distribution logistics: This module provides a very simple representation of the regional
build-out of the distribution network for fuel ethanol.

Dispensing stations: The dispensing station module addresses the regional acquisition of
tankage and equipment capable of dispensing high ethanol blends into flexible-fuel
capable vehicles. Build-out of E85-capable stations is driven by economic considerations,

and is constrained by regional availability of ethanol from the distribution network.

e Fuel use: The mix of low-ethanol-blend vs. high-ethanol-blend consumption is
determined by the relative economics of the two products as constrained by the

regional availability of ethanol for high-blend consumption through dispensing stations.

Vehicles

The vehicle scenario module functions primarily as an accounting structure, which is used in
BSM to keep track of the cumulative effect of multiple scenarios around volume, vehicle mix,
vehicle efficiency, and vehicle miles traveled for the car and light duty truck sectors. Its
structure captures acquisition, aging, and retirement of vehicles, as well as the translation of

vehicles into potential demand for fuel.

Ethanol import

The ethanol import module provides a simple representation of the evolution of non-domestic
ethanol production capacity. It generates imports of ethanol into the United States based on a
price differential as perceived from abroad. This structure enables the model to capture
historical patterns of growth and decline in imports of fuel ethanol. It is structured to facilitate

exploration of multiple scenarios around production cost.

A more detailed view of BSM”

Feedstock supply and logistics
The feedstock supply and logistics sector is responsible for generating cellulosic, starch, and oil
crop feedstocks for the conversion sector in BSM. The U.S. agricultural system forms the

context for the production of a significant portion of these feedstocks. Accordingly, in

? The following sections provide a “deep dive” into specifics related to each sector/module in the BSM. The casual
reader may wish to move ahead to the “Interconnections between sectors” section.

An Overview of the Biomass Scenario Model | August 2013 10

developing the feedstock supply and logistics sector we have taken care to respect both the
physical (land use) and economic aspects of U.S. agriculture. The sector is divided into two

modules: feedstock supply and feedstock logistics.

Feedstock supply
Feedstock supply refers to the production of different feedstocks required as substrate for

conversion, as summarized in Table |.

An Overview of the Biomass Scenario Model | August 2013

Table |. Summary of feedstocks produced by feedstock supply module

Feedstock Source Use Notes
Corn Cropland Ethanol
Soy Cropland “Refinery-ready” fuels

Other oil seed

Cropland (small
grains)

“Refinery-ready” fuels

Model does not explicitly represent land
allocation to other oil seed (e.g.,
rapeseed)

Crop residue

Cropland

Ethanol | butanol |
“refinery-ready” fuels

Model allows residue collection from
corn, wheat, other grains, cotton

Herbaceous
cellulosic energy
crop

Cropland
pastureland

Ethanol | butanol
“refinery-ready” fuels

Woody cellulosic Cropland Ethanol | butanol
energy crop pastureland “refinery-ready” fuels
Pasture Pastureland Ethanol.| butanol

“refinery-ready” fuels

Urban residue

Urban areas

Ethanol | butanol
“refinery-ready” fuels

Represented as simple price-response
supply curve

Forest residue

Forest lands

Ethanol | butanol
“refinery-ready” fuels

Represented as simple price-response
supply curve

As indicated in Table |, urban and forest residue feedstocks are generated using simple price-

supply relationships. All other feedstocks are produced by the agricultural land base. Figure 4

identifies the different land categories represented within the feedstock supply module.

An Overview of the Biomass Scenario Model | August 2013

All Cropland

Excluded from BSM BSM Cropland Used & CRP Land Cropland Used
Feedstock Analysis for Crops for Pasture

BSM Cropland Allocation
Annuals Perennial Energy Crop BSM Pasture Allocation
a he = Corn = Herbaceous Perennial = Pasture (used as pasture or
u Rscy Energy Crop harvested as energy crop)
A Wheat = Woody Perennial Energy = Herbaceous Perennial
Crop Energy Crop
= — Small grains (Sorghum, ial
Oats, Barley, “other” oil crop) Biwcody Get Oniclene ley)
Crop

= Cotton

Figure 4. Land categories represented within feedstock supply module

Within each of the ten USDA regions represented in the model, land is divided among three
high-level categories: cropland used for crops, Conservation Reserve Program (CRP) land, and
cropland used for pasture. These land bases are typically treated as static quantities over the
course of a simulation run. However, as indicated in Figure 4, BSM structure supports scenarios
that will cause land to move from CRP or pasture into cropland used for crops. Within each
land base, land is allocated among different uses based on expected relative per-acre grower
Payment accruing to producers from the various products. Land allocation is region-specific,
reflecting the production economics of different crops in different regions. Allocation of land to
cellulosic crops is more restrictive: only those producers who have adopted the practice of
producing cellulosic products (either residue or perennials) consider cellulosic grower

payments in their decision making. “New practice” producers can grow over time based on the

An Overview of the Biomass Scenario Model | August 2013 13

potential profitability of cellulosics, but constrained by the requirements of the existing and

prospective conversion facilities, as shown in Figure 5.

Region-specific calculations for each potential land use

Expected Per Acre Expected Per Acre
Production Cost Net Grower Payment

Expected Yield ---5» Expected Per Acre”
Grower Payment
Land allocation logic

eT
ee
Expected Per-Unit Compare Net Grower Determine desired allocation of
Grower Payment Paymentacrossall -----— > land among potential land uses
potential land uses (nested logit function)
1
Notes: t
+ Cellulosic Energy crops use annualized NPV to calculate expected net per acre revenue
+ Production cost and yield inputs inputs sourced from USDA, Oak Ridge National Allocate land among different
Laboratory. fee

Yield inputs sourced from USDA, Oak Ridge National Laboratory
Grower payment reflects price and feedstock logistics costs for cellulosic feedstocks

Figure 5. The agricultural land allocation algorithm

Not shown in Figure 5, but essential to the dynamics of BSM, is the logic surrounding pricing for

the various commodity crops, cellulosic products, and hay. This logic is central to a feedback
mechanism that uses land allocation to equilibrate production and consumption across all

product categories in the model. A more detailed treatment of the pricing structure used in

BSM is provided in Appendix A. Figure 6 shows in simple terms the feedbacks around price in

the feedstock supply module.

An Overview of the Biomass Scenario Model | August 2013

Expected Yield — eS
Sy, Supply

Consumption

Production Relative to
Demand
Land Allocation ( 3 ) (= )
Expected Per-Unit Price

Grower Payment

Figure 6. Price feedbacks in feedstock supply module

Feedstock logistics

Demand (via
scenario or other
sectors)

The feedstock logistics module provides a simple accounting structure that captures the

following costs:

e Harvesting and collection

e Transport from “farmgate” to “plantgate”

e Storage, queuing, handling, and pre-processing between farmgate and plantgate.

These costs are used to translate the per-ton price of cellulosic feedstock at the plantgate into

a per-ton grower payment at the farmgate. In developing the feedstock logistics module, we

have drawn from analyses of the Biomass Logistics Model (BLM) developed at Idaho National

Laboratory (INL). In its current form, the feedstock logistics module supports cost accounting

for both pioneer and advanced storage, pre-processing, and queuing/handling processes.

The feedstock logistics module underscores the high degree of interplay among different cost

components. For example, truck transport is viewed as a primary mechanism for moving

feedstock from farmgate to plantgate. Depending upon the feedstock involved, the mass

transported on the truck varies, with residue resulting in significantly lighter loads than woody

An Overview of the Biomass Scenario Model | August 2013

cellulosic crops. Other things equal, this implies a higher logistics cost per ton for residues than

for woody cellulosic crops.

Additionally, the logistics module emphasizes the importance of the travel distance from farm
to conversion facility. The model estimates farm-to-plant distances regionally, by considering

the following components:

e The total number of cellulosic plants requiring agriculturally-produced feedstock

¢ The total volume of agricultural land allocated to producing cellulosic feedstock

e The aggregate average yield of those producing acres

e An estimate of the fraction of land within the “plant-shed” that is available for cellulosic
harvesting

¢ Geometric factors that relate the resultant plant-shed area to average travel distance

from farm to plant.

Conversion sector

The conversion sector is responsible for transforming feedstock into liquid fuels, including
ethanol, butanol, and refinery-ready fuels (gasoline, diesel, and jet fuel) suitable for insertion
into the existing fuel infrastructure as refinery feedstocks, blendstocks, or finished products. In
BSM, the conversion module comprises a significant fraction of the overall model structure. It
consists of seven modules. Six of these modules look at the dynamics of industry development
for sets of conversion pathways. These dynamics include operations at different scales, learning
along multiple dimensions, logic surrounding the attractiveness of investment in new facilities,

and utilization of existing facilities. A seventh module compares investment attractiveness

An Overview of the Biomass Scenario Model | August 2013 16

across all conversion options, allocating scarce investment capacity among these options based

on their net present value.

As indicated in Table 2, there is significant overlap among the different industry development

modules. In particular, most modules share the following characteristics:

e Multiple conversion options, represented using an arrayed variable structure

e Regional disaggregation, following the feedstock supply module’s use of ten USDA
production regions

e Incorporation of pre-commercial pilot- and demonstration-scale operations

e Representation of pioneer-commercial-scale operations

e Representation of full-scale operations

e Learning curve dynamics.

An Overview of the Biomass Scenario Model | August 2013 17

Table 2. "Dimensionality" of conversion sector

learning curve)

Within the sector, there are a few departures from the generic structure. For example, in the

oil crop and algae modules, regional production of feedstock is of secondary importance. As a

result we have chosen not to disaggregate these modules by region. The starch-to-ethanol

industry, to take another example, is assumed to have reached maturity. Hence, there is no

need to represent the dynamics of pilot-, demo-, or pioneer-scale operations, nor is there a

requirement to represent learning curve dynamics. In the algae module, feedstock production is

considered as endogenous to the algae system rather than produced by the feedstock supply

sector. Algal feedstock production costs are subject to learning curves in the algae module.

An Overview of the Biomass Scenario Model | August 2013

Within the typical conversion module, there are multiple processes that govern the

development of the conversion options under consideration and their production of fuel. These

processes, as shown in Figure 7, are centered on:

e Pilot- and demonstration-scale operations

e Pioneer-commercial scale operations

e Full-commercial scale operations

e Expected economic value of the “next” investment

e Allocation of scarce capital in the investment decision
e Learning along multiple dimensions

e Industry aggregate average utilization rates for existing facilities.

Industry Development Pathway

Pilot = Demo > Pioneer > Full Commercial

Learning
* Process Yield Investment
¢ Technical Failure Attractiveness
* Input Capacity
* Capital Cost Growth
Utilization

Investor Risk Premium

Access to Debt Financing eign \
< Feedstock Cost > < Product Price>

Figure 7. Key interactions within the typical conversion module

An Overview of the Biomass Scenario Model | August 2013

Pilot- and demonstration-scale operations
Pilot and demonstration-scale operations are represented with a simple stock/flow structure in

the model, as shown in Figure 8.

Pilot
Pilot Projects on Line Completed 7
in Development Pilot Scale Operations Pitot Operations Illustrative Output
€ 1
sit comping competi Initiation Scenario | °On Line" Operations .
initiation i ot A ——$<__——
scenario development operations :
1, In Development *.
. Completed Operations
Demo
Demo Projects ontine Completed
in Development Demo Scale Operations ‘Demo Operations
© S 5
ema completing completing are wee Teas ced mare

ination lemo
use Pit tpt

demo
development operations

Figure 8. Pre-commercial structure and illustrative output

There are several important features of the pilot- and demonstration-scale structures in the
model. First, the structure for each pre-commercial-scale operation is arrayed, based on the
number of conversion options at play within the module. Technology on/off switches, set by the
user, enable the activation or deactivation of each conversion option. Second, both pilot- and
demonstration-scale operations are specified as exogenous scenario inputs. These scenario
inputs enable an arbitrary pattern of initiation to be specified by the end user as a scenario.
Third, the model explicitly represents the dwell time between initiation and completion of
development using conveyors. Fourth, the time for which operations are active—used in the
model to generate learning—is limited in duration. Finally, note that cumulative completed
operations are tracked by the structure. Figure 8 provides illustrative output that translates an

arbitrary initiation scenario into development, “on line,” and completed operations.

An Overview of the Biomass Scenario Model | August 2013 20

Pioneer-commercial-scale operations

The model accommodates two commercial scale operations: Pioneer- and full-commercial-scale
operations. In the model, pioneer-commercial-scale facilities are often the first commercial-
scale plants to come on line. These facilities have a smaller (about 1/3) capacity than full-
commercial-scale facilities. They do not take full advantage of economies of scale. Hence in
typical simulations of BSM, subsidies are required to stimulate investment in pioneer plants.

(Note that the starch-to-ethanol module excludes pioneer facilities from analysis.)

As shown in Figure 9, two stock/flow chains are used to account for pioneer-scale plants.
Depending on the module in question, these chains are arrayed by conversion option and/or by
region (see Table 2 for details). The top chain represents the number of plants in design and
construction, in startup, and in use. The bottom chain is a co-flow structure that is used to
account for the process yield (gallons of output per ton of feedstock input). These two

concepts—facilities and process yield—jointly determine the output capacity for pioneer

facilities in the aggregate. Output capacity is a reflection of the total ability of pioneer-scale

facilities to produce fuel via a particular conversion.

The co-flow structure here is essential for the accurate accounting of facilities and their
associated process yields. Whenever a new facility enters the system through the initiating flow,
the model samples the current state of the industry process yield for the associated conversion
option. This process yield then moves along with the facility through the development process,
eventually being used as an input for the average process yield of on-line facilities. This
structure enables the model to dynamically track the cumulative impact of growth in process

yields for “new” plants, as the industry moves from “blue sky” to “nth plant maturity.”

An Overview of the Biomass Scenario Model | August 2013 21

pioneer plantinitaion oie

Facilities

InDesignsCons ° Staring UpP onineP

competing statu P

vensinoscr C)

‘hug PlantOuputgatiday® OuputC apactyP

joni

Ang PY DEC P
a ‘Avg Plant Output ally P

Process Yield

Oo
PY ite?

Figure 9. Accounting for pioneer facilities, process yield, and output capacity

Note that structure has been provided to account for retirement of plants. In the current

version of the model, we assume a retirement fraction of zero, which implies that no plants are

taken permanently off line over the course of a simulation. The logic that controls utilization
factors, discussed below, accounts for the dynamics of short-term plant idling in response to

market forces.

Full-commercial-scale operations

Figure 10 shows the structure that accounts for full-commercial-scale operations in the model.
As with pioneer plants, commercial plants use two stock/flow chains to represent the design
and construction, the start-up, and the online phases of the facility life cycle. As with pioneer
plants, the structure for commercial operations is arrayed by conversion option and/or by

region within each module.

A comparison of Figure 9 and Figure 10 will reveal two notable differences between the
pioneer and commercial accounting structures. First, note that in contrast to pioneer facilities,

commercial facilities in the start-up phase are assumed to contribute to the overall output

An Overview of the Biomass Scenario Model | August 2013 22

capacity. A utilization rate (less than |) is assumed for plants during the period that they are in

startup.
Second, note that a new flow has made its way into the process yield chain for commercial-
scale facilities. This flow enables the model to capture the effect of process yield improvements
to be incorporated into the existing capital stock. The user of the model can specify the specific
rate at which a yield gap—measured as the discrepancy between the state of the industry

process yield for a particular conversion option and the existing industry average process yield

for that conversion option—is eliminated.

Facilities
Line Equivalent ¢

Starting Up

Ratnnnen rection F
| ave Plant Outpt gstcay ©

commercial plant |
initiation logic

Process Yield Days per Year Online

PY in Startup C

rate of industy process
yield adjustment

b)
State of Indust Process Yield
Figure 10. Accounting for commercial operations, process yield, and output

capacity

An Overview of the Biomass Scenario Model | August 2013

Expected economic value of the “next” investment
In order to represent economic valuation of potential investments, it was essential to develop a
simple, defensible mechanism for determining the viability of investment in the “next” plant, at

either pioneer- or commercial-scale, for the various conversion options within the regions

under consideration. We needed a dynamic economic mechanism to facilitate industry growth,
ultimately achieved through a structure culminating in a net present value (NPV) calculation. At
any point in simulated time, this structure captures important streams of costs and revenues
associated with a prospective project investment. By discounting these streams to the present,
it captures the dynamics of an evolving industry using a simple metric that enables comparison

of prospective investments across multiple conversion options, regions, and scales. In turn, this

metric enables the model to allocate scarce capital toward its highest valued uses.

A parallel algorithm is used for NPV calculations within each conversion module. As
appropriate to each module, the algorithm reflects conversion options, regional considerations,
and scale. Wherever possible, the algorithm operates at the highest possible degree of
aggregation by rolling up sub-categories into high-level summaries. For example, for purposes of
the NPV calculation, factor inputs and expected per-gallon revenues are held constant over the
plant lifetime. Figure 11 provides a simple influence diagram showing the logic flow leading to

the NPV calculation.

An Overview of the Biomass Scenario Model | August 2013 24

Expected Revenue NPV Revenue
Power Sales Expected Revenues —
m (net of Op Cost (net of Op Cost)

Other Co-product Sales ee {

Expected Sale
Price (incl subsidy) —————> pre Taxable Income

——_ nv Taxes

Expected

Expected production a
— Operating Cost

Feedstock Expense

Interest

Annual Loan

PV Loan Payment ——— NPV Investment

Feedstock cost
(incl subsidy)

Expected Other
Vbl Op Cost Fawment
Loan Principal
Depreciation f
Expected Fixed
eo Equity Fraction, ——________ Initial Equity Investment

Op Cost

Expected FCI
{incl Subsidy)

Figure |1. Logic of NPV calculation

In developing the NPV logic, we have adopted some important simplifications. In addition to
simplifications around revenue streams, we assume straight-line depreciation of the plant in
question. This significantly reduces detail complexity in the model. Additionally, we divide the
overall project life cycle into distinct phases, as shown in Figure 12. In the model, NPV

calculations are made for each phase of the project life cycle, and then rolled up to create an

overall NPV for the plant.

Plant Lifetime |

Loan |

i i}

Total Project Length |

Figure 12. Phases of project life cycle

Allocation of scarce capital

Within the BSM conversion sector, then, multiple opportunities present themselves to potential
investors at any point in time. At the extreme, thirteen conversion options can be active,

An Overview of the Biomass Scenario Model | August 2013

competing across ten regions and often at both pioneer- and commercial-scale. Within the
model, each conversion module uses NPV as a basis for determining the attractiveness of the
various investment options under consideration using a logit function. (See Appendix B for
more details.) The resultant attractiveness metrics are then compared within the relative
attractiveness module, which also includes a default “other” investment category. The relative
attractiveness for each alternative is then applied to a scenario-driven maximum construction
capacity, which generates a platform and scale-specific yearly start rate. This “desired” start flux
is communicated back into the conversion modules, where it is allocated regionally, if required,

and “batchified” so as to send a discrete signal to begin plant development, as shown in Figure

13.

Within each Conversion Module In Relative Attractiveness Module

> Calcul Calculate “Platf ” fe i

“Other”
id + Conversion option By conversion option + Calculate sum across all platforms tiveness
Logit ——s- ‘Sale (pioneer/commercial) + For each scale (pioneer/commercial) + Calculate relative attractiveness as

Parameters | Reeen + Across all regions platform attractiveness/S(platform

Regional Distribution of Project Starts

Allocate Project Starts by Region

+ By conversion option

+ For each scale (pioneer/commercial)
+ By region (ifrequired)

Calculate Desired Project Starts
+ By conversion option
+ For each scale (pioneer/commercial)

<— Maximum
Starts/yr

Batichfy to Create Discrete Project Start Signal
+ By conversion option

+ For each scale (pioneer/commercial)

+ Byregion if required)

Figure 13. Translating NPV into project start signal

Learning along multiple dimensions
For most conversion options under consideration in BSM, the initial performance along multiple

dimensions would fall far short of expected mature industry (or “nth plant”) performance, as

An Overview of the Biomass Scenario Model | August 2013

reflected by NREL and in other design studies. Industry evolution is in no small measure the
story of performance improvement that results from learning-by-doing. Merrow’s research on
cost growth in capital-intensive industries (Merrow, 1983), for example, underscores the
important role of experience at prior scale in reducing the risk of capital cost growth at
commercial scale. Henderson’s work with the Boston Consulting Group in the 1970s (Hax &
Majluf, 1982) demonstrates the role of accelerating industrial learning as a cost-reduction

strategy at commercial scales.

Given the important connections between learning and industry evolution, we needed to
develop a simple, consistent, and defensible mechanism to translate the accumulation of
experience into a set of performance parameters to represent the current “state of the
industry” for each conversion option. Our approach, which we call “cascading learning curves”
draws upon simple learning curve principles in order to address learning for multiple
conversion options at multiple development stages, addressing multiple performance attributes.
There are three fundamental tasks, illustrated in Figure |4, involved in the cascading learning

curve approach:

e Develop separate cascading curves for each conversion option. By providing separate
structure for each conversion option, we have created the possibility to separately
characterize different initial conditions, mature industry conditions, and learning rates on
a conversion option-specific basis.

e Capture learning for each conversion option at three distinct development stages. In
BSM, we look at learning for pilot-scale operations, for demonstration-scale operations,

and for commercial-scale operations (including pilot- and full-commercial scale). A

An Overview of the Biomass Scenario Model | August 2013 27

staged approach to learning enables us to capture prior scale effects (important for

capital cost growth). It also enables us to explore the implications of stage-specific

progress rates as well as the analysis of timing and placement of policy initiatives.

e Use learning to create indices of maturity. These indices of maturity, in turn, drive

essential technology attributes that are used within BSM. Key attributes of performance

for each conversion option are:

Process yield

Likelihood of “technical failure”

Feedstock throughput capacity—the degree to which facilities are able to
perform at nameplate capacity

Capital cost growth—the premium in capital cost, beyond the nthplant estimate,
which would be observed if development of a facility was begun today.

Investor risk premium—the additional premium, beyond normal hurdle rate, that
investors would require for investment in the facility.

Access to debt financing—the portion of the expected facility capital cost that

would be financed via borrowing (vs. equity investment).

An Overview of the Biomass Scenario Model | August 2013 28

Growth in experience is driven by stock of
operations that are “on line”

Level of maturity
logging pilotemperience drives multipliers for
key technology
attributes

Op Experience Per Year

rac rate of growth Pilot Multipliers
Piloteyperience

Min Pilot Experience
for Learring

Early Pilot ulipiers 0-1 Scale. Possible to
include technology
“spillover” effects

Mature Pilot M ulipliers

Progress ratio controls
“productivity” of learning
process

Growth in maturity at each stage
driven by rate of doubling of
experience

pilot earning logic
Doubling of
experience removes
a fraction of gap
Number of Doubings

‘of Pilot Experience D
pllotmaturity gap

doubling factor

Figure 14. Industry learning curve structure

There are some differences between the representation of the learning curve structure in BM
and other, perhaps more common, formulations of learning curves. In the classic formulation of
a learning curve, for example, a power law is used to relate cumulative experience to a single
attribute such as cost. The asymptote of cost is often implicitly set to zero. By contrast, in BSM,
cumulative learning at each stage of development is reflected along a 0-1 scale in pilot, demo, or
commercial maturity. As experience accrues, the model calculates explicitly the rate at which
experience is doubling. This rate of doubling is applied to a maturity gap (simply the difference
between current maturity and full maturity) to generate learning. Maturity, in turn, drives

movement along a vector of attributes.

A second set of differences involves the development stages over which learning is applied.
While a typical learning curve analysis might consider cost reductions for relatively stable

An Overview of the Biomass Scenario Model | August 2013 29

developed industries, in BSM we consider multiple attributes over multiple development stages.

Figure 15 shows how the learning curves cascade over these development stages.

Pilot Demo Commercial Current
Industry
Stow ofthe Cure ntindustry
dusty Muli Technology Atéibutes

Pilot Manip

leary Pot Mutiptions

ature Pilot Multipliers ature Demo Maples Mature Commercial Multipliers Mature Inds ty

Technology Atibutes

Commarea Matsity

Figure 15. Cascading learning curves

At any point in simulated time, the current industry technology attributes reflect the
performance and cost characteristics associated with an investment in a pioneer- or full-
commercial-scale facility for a given conversion option. At each stage, multipliers that are
passed on to the next stage are calculated as a weighted average, with the maturity level used

as the weighting factor.

Dynamically, this structure enables BSM to jump from one performance trajectory to another
based on the behavior of pilot, demo and commercial operations. Figure 16 illustrates that
simple exogenously-defined scenarios for pilot-, demo-, and commercial-scale operations drive
learning at each stage. Cost and yield parameters follow three distinct pathways as the industry

evolves.

An Overview of the Biomass Scenario Model | August 2013 30

Simple Exogenous Scenarios For Pilot, Demo, Commercial Resultant Indices of Commercial Maturity

Pilot (7596 PR)

" Demo (75% PR)

Cumulative Pilot Experience (years)

Cumulative Demo Experience (years)

Commercial (85% PR)

2007.00 2008.50 2010.00 201850 2017.00

2007.00 2008.50 2012.00 201850 2017.00

Cumulative Commercial Experience (bb gal) Key Technical Attributes

Process Yield Multiplier

Capital Cost Growth
Multiplier

= Pilot > i
< Pilot + Demo -> }
< Pilot + Demo + Commercial ->

2007.00 2008.50 2012.00 200850 2017.00

Figure 16. Illustrative learning curve dynamics

Learning curve dynamics, of course, do not occur in isolation from the overall dynamics of the
industry. For a given conversion option, learning curves are at the heart of feedbacks that
surround the investment process, and which can underwrite industry “take-off” (as shown in

Figure 17). Each of these mechanisms is a positive feedback loop, reinforcing development of

the conversion option in question.

An Overview of the Biomass Scenario Model | August 2013

investment
attractiveness \
a ( conversion

access to 5 - capital facilities
risk process through p!
debt financing tf premium rs) yield (+) put (+) cost (+) }

‘i production
maturity wif

Figure 17. Key feedbacks emerging from learning curve structure

Utilization of existing facilities
Multiple processes are at work in the conversion sector to generate the production of biofuels.
A final set of processes concerns the utilization of existing facilities. A fundamental premise of
basic economics is that “sunk costs” don’t matter. In BSM, conversion facilities are assumed to
follow this premise; the capacity utilization rate for each conversion option (within each region,
as appropriate) at either pioneer- or full-commercial scale is developed as a response to the
“cost-price ratio” for its products. As the price received for its product (including any
subsidies) grows relative to the per-gallon cost of producing that product (after factoring net
per-gallon co-product revenues into the mix), utilization increases to its maximum. On the

other hand, as the price-cost ratio declines below unity, utilization rates decline, as shown in

Figure 18.

An Overview of the Biomass Scenario Model | August 2013 32

Utilization

—|

Feedstock Cost —____» Feedstock Cost —H+» Cost/Price «————___ Expected Price
(including subsidy) (including subsidy) Ratio (including subsidy)
$/ton $/gallon $/gallon

Process yield

gallon/ton
Coproduct sales ————» Other revenue net
$/yr of variable cost
$/gallon

Other vbI

operating cost

$/yr

Power sales Average Plant Output

$/yr gallon/yr

Figure 18. Determining utilization from cost-price ratios

Utilization is a central element of the feedback structure within BSM, as it controls both the

production of products and consumption of feedstock.

Conversion sector summary

Each conversion module within the conversion sector is built up from multiple simpler
structures that represent pre-commercial-demonstration- and pilot-scale operations, pioneer-
and full-commercial-scale operations, the expected economic value of investment, learning, and
utilization. These structures are connected within each module in order to generate products
(diesel, jet fuel, gasoline, butanol, and ethanol). They are connected across modules via the logic

within the relative attractiveness module that allocates scarce investment capital. The

An Overview of the Biomass Scenario Model | August 2013 33

conversion sector is connected upstream in the supply chain to the agricultural system through
feedstock supply dynamics, and to both the oil industry (algae, oil crop and cellulose-to-
refinery-ready modules) and the downstream ethanol sector. These downstream sectors
determine price (and in the case of ethanol, demand) signals which are sent to the conversion

sector modules.

Downstream ethanol sector

The downstream ethanol sector comprises a set of interconnected modules that take fuel
ethanol from conversion facilities to end users, both in low-blend (E10 or EI5) and high-blend
(nominally, E85) form. Additionally, the downstream sector contains logic that controls the use
of butanol as a substitute for ethanol in the low-blend market. The model assumes that physical
characteristics of ethanol require separate infrastructure for distribution and dispensing than
for petroleum-based fuels. A significant portion of the downstream sector, therefore, is focused

on distribution and dispensing station dynamics.

Figure 19 provides a picture of the content of the downstream sector. As suggested by the
diagram, downstream dynamics focus on the build-out of distribution infrastructure, the

development of dispensing infrastructure, and decision making around fuel usage.

An Overview of the Biomass Scenario Model | August 2013 34

- Outside Production 3
= Dispensing ¢ Heese =)
Stations ‘a
=

Dispensing
AV —distevution Sabon

el Terminal
= Ee a 5
Terminal CI
Dispensing
Distribution Stations
- = Terminal

Bio
Refineries Dispensing
Stations

Regional
Hub/Terminal Nicds Legend
Primary Transport

Within Production Region

Secondary Transport

Storage (and associated processing)

==> Collection and Aggregation
=*

Figure 19. An overview of downstream dynamics

To support these dynamics, multiple modules comprise the downstream sector of BSM?,

including:

e Distribution logistics
e Dispensing station
e Fuel use

e Pricing and inventory.

Distribution logistics module
A fundamental challenge associated with ethanol as a transportation fuel is its apparent
incompatibility with existing infrastructure. The distribution logistics module provides a very

simple representation of the build-out of ethanol-friendly distribution infrastructure. Rather

* Detailed analysis of downstream ethanol dynamics can be found in (Vimmerstedt, Bush, & Peterson, 2012).

An Overview of the Biomass Scenario Model | August 2013

than speculating on the build-out of specific distribution modalities for ethanol (such as rail,
barge, or dedicated pipeline), the logistics module focuses on capturing the implications of
build-out on the rest of the downstream system. The structure focuses on the acquisition of
ethanol infrastructure for terminals within each region. The module is silent on the specific
details of infrastructure, instead focusing on the drivers, time delays, and feedback loops

associated with regional build-out, as shown in Figure 20.

Remaining Terminals
Without Ethanol Infrastructure
Within Region

Za oS

Total Terminals With Ethanol Infrastructure Pressure'to Acquire

Ethanol Infrastructure Acquisition Outside Region Ethanol Infrastructure
Ney In Other Regions

Ethanol Infrastructure
Acquisition Within Region

CC Pressure to Acquire < Regional Ethanol

Ethanol Infrastructure Production Capacity >
Within Region ee

Figure 20. Distribution logistics

()

A two-stage supply-push approach (first within a region, and then across regions) is embedded
within the module. Within a region, the model first seeks to balance ethanol production
capacity against terminal capacity to distribute that ethanol. As production capacity within a
region grows, there is pressure within the region for terminals to acquire ethanol-compatible
distribution infrastructure. Second, as build-out occurs within each region, any excess regional
production capacity creates pressure for acquisition of infrastructure in other regions, in

proportion to the terminal density within each region.

An Overview of the Biomass Scenario Model | August 2013 36

The result of this two-stage supply-push algorithm is an initial build-out of distribution
infrastructure in ethanol producing regions, followed by a slower build-out in non-producing
regions. Infrastructure coverage within any region constrains regional investment in ethanol
dispensing tankage and equipment, thus setting a limit on the uptake of ethanol in high-blend

form.

Dispensing station module

The dispensing station module focuses the decision making associated with the acquisition and
use of high-blend tankage and equipment by retail dispensing stations. The module considers
roughly 120,000 stations, distributed both regionally and by ownership among oil-owned
branded independents, unbranded independents, and hypermarts. The fundamental decision for
each station is the acquisition of tankage and dispensing equipment required to dispense high-
ethanol blends into FFVs. The module assumes that ten percent of stations have repurposable
mid-grade tanks. The capital cost of repurposing is assumed to be significantly lower than

investment in new tankage and equipment for high-blends ($20,000 vs. $60,000).

The basic logic within the dispensing station module combines the physics of high-blend
availability with the economics of the investment decision. Stations will not consider investment
unless distribution infrastructure is sufficient within the region. They will not invest unless the
investment makes economic sense, as reflected in a NPV calculation that captures the

discounted stream of expected costs and benefits from the investment.

Thus, two fundamental structures are at play within the dispensing station module. The first is
an accounting structure that considers the movement of stations as they adopt high-blend

tankage and equipment (Figure 21). The second provides a detailed view into the NPV

An Overview of the Biomass Scenario Model | August 2013 37

calculation that undergirds the decision to invest in high-blend tankage and equipment (Figure

22).

Not Considering Investment

_ putting dh the table

susceptitf}e for putting on tabl

Considering Investment With Hi Blend Investment

>
pleting consideration

taking off the table

Figure 21. Dispensing station accounting structure; NPV calculation captures
estimated costs and revenues of prospective investment

As shown in Figure 21, stations exist in one of three states with respect to investment in high-
blend tankage and equipment. Depending on the dynamics of regional distribution infrastructure
availability, a portion of those stations not considering investment put the investment decision
on the table each year. Based on the economic viability of the investment (as reflected in the
NPV of the decision), the consideration of investment culminates in a decision to invest or to
stop considering the decision. This investment process is disaggregated by region (so as to
account for differential degrees of distribution infrastructure within each region), by ownership
(to enable different potential affinities for high-blend ethanol sales among different ownership
types, and to account for different business details for different ownership types), and by
repurpose versus new investment (to account for different capital costs associated with

repurposing versus new investment in tankage and equipment).

An Overview of the Biomass Scenario Model | August 2013 38

expected net other expected net incremental _, Expected Net
ee

NPV Net
revenue per visit other revenue Incremental Revenue

Incremental Revenue
expected “Hi-blend”

ted margin
expecte
as x expected net incremental

incremental area
station traffic SQ Hi-blend” revenue
expected incremental
“Hi-blend” volume

Expected Taxable Income » NPV Taxes Sy

NPV of
Investment

current station traffic a
expected gasoline Interest Expense
margin
depreciation expense

Loan Principa———— Loan Payment —® PV Loan Payment

4
t loan term

equity fraction a ae

id NPV calculation for stations

expected incremental expected incremental
—

gasoline volume gasoline revenue Tax rate

Expected Fixed
Capital Investment

Initial Equity Investment

Figure 22. Logic be

As shown in Figure 22, the NPV calculation considers major categories of revenue and expense
associated with station investment. In addition to the capital cost of the investment, the NPV

calculation considers marginal cost and revenue streams associated with changes in the mix of

high-blend versus “straight” gasoline sales, changes to station traffic (to account for first-mover

advantage) and other revenues from operations.

Just as the distribution logistics module provides a context that constrains the acquisition of
tankage and equipment for stations, the dispensing station module provides a context for fuel
use. Accessibility of high blend stations within a region will constrain the potential for FFVs to

access high-blend fuels. Regional dispensing station coverage thus sets a physical limit on

ethanol uptake in the system.

An Overview of the Biomass Scenario Model | August 2013 39

Fuel use module

The fuel use module captures both the effects of regional high-blend fuel availability and the
effects of relative gasoline/high-blend pricing on the decision making for FFV owners, with
respect to the use of high-ethanol fuel blends. The module contains two major interconnected
components, as shown in Figure 23. The first component accounts for the affinity of FFV
owners toward high-blend fuels. The second uses a logit function (see Appendix B) to allocate

fuel use between for FFV owners who are “occasional” and “regular” users of high-blend fuels.

Long Tei Relative Price

preference shafe Regular Users

Non HiBlend Users

becoming occasional user Occasional HiBlend Users

‘becoming regularuser Regula HiRtend sen,

rate of becoming
reg HiBlend user

susceptible non HiBlepd users

ocdhs ional userconstrainton
becoming becffning regularHiBlend User
Potent

max rate of becomtin FAiBiend UserGap

regula rHiB lend

dropping back

‘to occasional used

HiBlend| Capable %

WiT HOUT sfation coverage
tie to drop out

Figure 23. FFV accounting structure

As shown in Figure 23, FFV users (expressed as a % of regional FFV vehicles) are divided into
three distinct categories: non high-blend users, occasional high-blend users, and regular high-
blend users. Non high-blend users do not use high blend because a) they do not have access to
stations that dispense high blend; b) they do not know they have an FFV; or c) they do not
desire to use high blend, for non-economic reasons. Based on regional dispensing station
coverage and a fraction of non-users who are assumed to be amenable to using high blends, FFV
owners leak over time from the non-user to occasional user category. Under conditions of
price parity between high blend and regular gasoline, occasional users are assumed to fill 20% of

An Overview of the Biomass Scenario Model | August 2013 40

their fuel requirements using high blend. Regular users, on the other hand, are assumed to fill
80% of their fuel requirements using high blend under conditions of price parity. Movement
between occasional and regular users is driven by a long-term retail price differential between

the two products.

The distribution of high blend users provides a physical basis for ethanol usage among FFVs.
Logit functions are used to translate relative high-blend/gasoline retail prices into instantaneous
usage shares for both occasional and regular high-blend users. The distribution of occasional
and regular users is then applied to these usage shares. The resultant user-weighted usage
shares are multiplied against potential high blend fuel consumption in order to generate actual

high blend consumption within each region, as shown in Figure 24.

Occasional User High Blend Share ————> Desired High Blend Consumption
(Logit Function) From Occasional Users

High Blend
Gasoline Equivalent Price Occasional js Max High Blend Consumption
(from Pricing Module) High Blend Users From Occasional Users

Total Desired
Potential High Blend Consumption High Blend Consumption
(From Vehicle Module)
Gasoline Price

(from Pricing Module) \

Regular Max High Blend Consumption
High Blend Users From Regular Users
Regular User High Blend share Desired High Blend Consumption
(Logit Function) From Regular Users

Figure 24. Logic behind high-blend consumption

An Overview of the Biomass Scenario Model | August 2013 4l

Pricing and inventory module
The final module within the downstream sector accounts for ethanol pricing and inventory
dynamics. Pricing and inventory for butanol, which in the model forms a substitute for ethanol

in the low-blend market, are also captured here.

Ethanol inventory is aggregated across the entire supply chain within each region, allowing for
cross-regional movement of ethanol based upon regional surpluses or shortfalls within each

region (as shown in Figure 25).

There are several important features to this pricing and inventory structure. First, note the
three sources of regional ethanol production: the starch-to-ethanol module, the cellulose-to-
ethanol module, and the import module. Second, note the regional import/export structure

that facilitates cross-regional movement of ethanol. Third, note the single driver of ethanol

consumption, reflecting total ethanol demand from both low-blend (i.e. E10) and high-blend (i.

E85) uses. Finally, note the rich feedback that drives cross-regional movement of ethanol.

An Overview of the Biomass Scenario Model | August 2013

target regional inventory

regional inventory overage,

Cellulosic Ethanot
Production by R egion

Regional LOH Inventory regional EtOH consumption

tL 9

regional EtOH producton

det export by region expert] import_dstn import by region
than -

Production by Region

terR egion Trang
oo

total ~, NN

rolup desired import
across regions

desired inventpry adjustment

regional Al. w=

(ross regions

fra Desired Ethanol
Consumption

desired import by region

(otal desired regional EtOH adj

Figure 25. Downstream ethanol inventory dynamics

This cross-regional movement algorithm is relatively straightforward, and as described below:

¢ Calculate desired inventory adjustment in each region required to bring inventories to
desired levels (blue connections in Figure 25)

e Calculate the regional production/consumption gap as the difference between regional
production and consumption (green connections in Figure 25)

e Sum the inventory adjustment and production/consumption gap to arrive at overall
desired movement in ethanol by region

e Roll up total desired imports and exports across all regions.

An Overview of the Biomass Scenario Model | August 2013 43

e Limit total inter-regional movement to minimum of total desired imports/exports

e Allocate exports/imports in proportion to relative desired imports/exports

Pricing for ethanol is considered at multiple downstream points along the supply chain. Figure
26 provides an overview of the approach. Ethanol price is calculated at point of production, at
point of distribution, and at the pump. Supply/demand imbalances in the downstream supply
chain drive changes in price at point of production (see Appendix A). Transport and storage
costs, which vary based on distribution infrastructure within a region, are applied to the point
of production price in order to generate an ethanol point-of-distribution price. The price for
high-blend ethanol at the pump is determined as a weighted average of point-of-distribution
price and gasoline prices, based on a regression analysis of the two. Not shown in Figure 26,
but relevant to policy analysis, are multiple points along the supply chain where initiatives can
work to reduce costs and/or change price as perceived by producers, distributors, retailers, or

end users of ethanol or high blend.

Point of production price point of dstrbution price point of use price

change in price

pressure to change price,

trans port and storage costs
gasoline price from
Petroleum Industry Secto

production Inventary consumption

Figure 26. Simplified ethanol pricing structure

The pricing and inventory for butanol follows similar logic to that of ethanol, with some notable

exceptions:

An Overview of the Biomass Scenario Model | August 2013 44

e Assingle, national inventory is considered.

e In addition to its use in the low-blend oxygenate market, butanol can be consumed for
industrial uses.

e Pricing for butanol is captured at point of production only. There is neither a point-of-

distribution nor a point-of-use price for butanol.

The dynamics of butanol use and pricing center on the substitution of bio-butanol (produced
within the BSM cellulose to butanol module) for butanol produced by other means, and on the
substitution of butanol for ethanol in low-blend uses. These substitution dynamics are
determined by relative price considerations. To capture these two dynamics, logit formulations
are employed that translate relative prices into market shares. For industrial uses, the price of
bio-butanol competes against an assumed alternative price of $4/gallon (this value can be varied
as a scenario). For completion against ethanol, the endogenously-generated bio-butanol price is

compared against the price of ethanol.

Vehicle module

The primary purpose of the vehicle module in BSM is to provide inputs that represent potential
demand streams for ethanol and for gasoline, from “regular” vehicles and from FFVs. In order
to provide these inputs to the rest of the model, we have developed a highly simplified
accounting structure for vehicles of multiple types. Focusing on light duty vehicles, this vintaging
chain captures the cumulative impact of multiple scenarios around volume of new vehicles each
year, new-vehicle mix, new-vehicle efficiency, vehicle miles traveled, and vehicle mortality. The
model aggregates vehicles nationally. Regional population distributions are used to apportion

fuel consumption among the 10 USDA regions used by the model. In its operation, the module

An Overview of the Biomass Scenario Model | August 2013 45

applies age-specific survivorship estimates to vehicles as they vintage through the chain. The
model focuses on two distinct vehicle types (automobiles and light trucks) and 10 engine types
(gasoline, diesel, plug-in hybrid (PHEV), hydrogen, compressed natural gas, FFV, gas hybrid
electric, gas PHEV, bi-fuel, and other) within the light duty fleet. For each of these 20
combinations, a scenario for new-vehicle sales over time is accompanied by a scenario for new-
vehicle efficiency. The model dynamics track the implications of these new vehicle scenarios for

overall vehicle efficiency and resultant fuel demand, as shown in Figure 27.

Each stage in the stock-flow chain represents a cohort of vehicles. Mortality flows remove
vehicles from the system; vehicles that survive to the end of the cohort’s time horizon are
moved to the next cohort in the sequence. Cohorts |-4 are each four years in duration.
Cohort 5 contains vehicles that are 16 or more years of age. (See Appendix C for details on

the BSM approach to aggregating vehicles into four-year-sized lumps).

The parallel pathway, shown in Figure 27, accounts for the efficiency of vehicles in each cohort.
Cohort-specific values for vehicles, efficiency, and vehicle miles traveled are used to calculate
cohort-specific potential fuel usage, which is then summed over all cohorts to calculate overall

potential fuel use.

An Overview of the Biomass Scenario Model | August 2013 46

vehicle influx logic

Y” suvive ra
Vid mortrate Suvive ate] Ee

Avg Vehicle Efficiency 2,

2 wtd mort rate

TVE 12

5 —

TVE 203

Total Vehicle
Efficiency]

Figure 27. Tracking vehicles and efficiency (2 cohorts)

The vehicle module is designed to facilitate exploration of the cumulative impact resulting from
changes in volume, mix, mortality, VMT, and efficiency. Structure in the model captures the
effects of changes in fuel prices, consumer attitudes, and other similar items. While we have not
provided an explicit representation of consumer choice mechanisms, in the vehicle module we
have created the potential to develop internally consistent scenario sets in which vehicle inputs
maintain a logical consistency with petroleum price scenarios. On the vehicleinflux side, levers
within the vehicle module enable use of (or departure from) Annual Energy Outlook (AEO)
Projections for inflow volume, inflow mix, and efficiency. Similarly it is possible to use or depart

from AEO mortality rates, and to use or modify AEO scenarios.

Oil industry sector

The oil industry sector in BSM is relatively simple, containing a single module that houses:

e Asset of scenarios used to determine crude oil prices

An Overview of the Biomass Scenario Model | August 2013 47

e Refinery product prices for diesel, jet fuel, and gasoline

e Algebraic relationships that translate crude oil prices, refinery product prices, and an
assumed refinery “drop-in point” for each infrastructure-compatible pathway into price
inputs for the different conversion modules and for the downstream pricing and
inventory module

e Accounting structure that captures petroleum displaced by diesel, jet fuel, and gasoline

produced by the different infrastructure-compatible pathways.

Import module

The import module is an exceedingly simple structure focused on the import of fuel ethanol
from outside U.S. borders based on relative price considerations. This structure compares the
ethanol point-of-production price generated within the downstream pricing-and-inventory
module against a threshold (including tariffs) that reflects the cost of bringing fuel ethanol into
the United States. As the price within the United States exceeds the threshold, an increasing
fraction of offshore production capacity is utilized. This simple structure enables analysis of
scenarios around tariff policies, cost reduction, and capacity growth for offshore ethanol

production facilities.

An Overview of the Biomass Scenario Model | August 2013 48

Interconnections among sectors

Figure 2 provided a high level overview of the sectors that comprise BSM, and the previous

discussion has given a detailed view into the modules that are found within each sector.

Another perspective on the system is given by the nature of the interconnections among the

different sectors. As shown in Table 3, the connections between sectors are relatively few in

number, typically consisting of price signals and supply/demand quantities.

Table 3. Inter-sector connections

Feedstock Conversion Import Oil Industry Downstream
From/To Supply &
Logistics
Feedstock « Feedstock
Supply & consumption
Logistics * Feedstock
price
(plantgate)
Conversion ¢ Feedstock ¢ Infrastructure- | ¢ Ethanol
demand compatible production
* Cost to price fuel © Butanol
ratios production by production
© Output pathway
capacity
Import ¢ Ethanol import
Oil Industry © Gasoline ¢ Module- © Gasoline point of
point of specific price distribution price
distribution input
price
Downstream e Ethanol point | ¢ Ethanol price
of production input
price
¢ Butanol point
of production
Price input
Vehicles ¢ Potential lo-
blend
consumption
from FFV

Potential lo blend
consumption
from non-FFV
Potential hi-blend
consumption
Potential gasoline
consumption

An Overview of the Biomass Scenario Model | August 2013

Data inputs

Multiple data inputs are required to run BSM, including agricultural cost and yield parameters
for the feedstock module, performance and learning parameters for the various conversion
modules, logit coefficients, petroleum prices, and adoption rates for new farm practices and for
dispensing station owners. Given the forward-looking nature of BSM, it is not surprising that
the availability and quality of input data is highly variable. In many instances, assumptions or

informed opinion were used to populate the parameter space, shown in Table 4.

An Overview of the Biomass Scenario Model | August 2013 50

Table 4. Summary of data inputs to BSM

Input data area

Source(s)

Comments

Crop production costs

ORNL/POLYSYS

Assumed constant over simulation time frame
Ongoing interaction with ORNL analysts

Energy price/crop
production price coupling

Pacey study (McNulty,
2010)

Price coupling factors derived from Pacey report

Yields

ORNL/POLYSYS

yield growth treated as assumption/scenario

Feedstock a USDA baseline (United Calibration done annually based on annual
Calibration data for ;
Supply & reduction: prices States Department of updates to baseline and updates to input data
Logistics P uP Agriculture) from ORNL
. ' Assumptions modified as needed as part of
Logit parameters Assumption aa
calibration process
eee mea INL/Biomass Logistics Structure and input data updated periodically to
vt Prep! 8 Model (BLM) reflect ongoing interaction with INL analysts
logistics
NREL design reports
PNL design reports NREL staff are assembling and vetting these data.
Performance and cost Analysis papers For some conversion options, formal analysis
data for different Expert opinion reports do not exist. We are in process of
conversion options Internal secondary vetting available data, developing assumptions,
analysis/interpolation and facilitating an expert review
Conversion’ Learning curve Assumption informed by
id Beck study (RW Beck, Sensitivity analysis planned
parameters 2010)
. - Plan sensitivity and robustness analysis around
Logit parameters Assumption
current logit parameters
Construction capacity Assumption
NPV of “other” option Assumption
— EIA (United States Energy | Data taken from “official” scenarios. Oil price
Oil price Information Agency), 5 :
fi shocks, other scenarios available to the system
Oil Industry arbitrary scenarios
r Treated as assumption but informed by design
Fuel mix Assumption
reports
Drop in points Assumption
Distribution Terminals by: | pia Developed from EIA data in 2008
region
Initial mix of terminals
with/without Assumption
infrastructure
Infrastructure acquisition .
Assumption
rate
NREL (Johnson &
Number, distribution of Melendez, 2007 draft) NACS provides a rich perspective on “other”
x) dispensing stations by NACS (National sales associated with dispensing stations in the
§ ownership, dispensing Association of spreadsheets that accompany the text of their
Z station economics Convenience Stores, annual report
E 2007)
¢ Initial repurposable
g pure: Assumption
S stations
ce Station adoption rates as i
z f(NPV) Assumption
a Logit parameters for fuel | Assumption

An Overview of the Biomass Scenario Model | August 2013

use

Vehicle influx, miles
traveled, miles per gallon

EIA/NEMS

Ethanol price at point of
distribution

Assumed

Assumed values for storage and transport
applied to endogenous point of production price

High blend point of use
price

NREL/Lexidyne regression

Regression of available data provides weighting
factors for point of use price

Import

Capacity, price threshold
for import, learning curve

Assumed

Values used to calibrate against observed data

for fuel ethanol imports

parameter

Analysis infrastructure

As has been outlined above, the BSM is a robust model that has undergone rigorous testing,
validation and refining by the BSM team, and was built to explore multiple facets of the biofuels
supply chain and its numerous drivers, bottlenecks, and system interactions. The model was
designed to be a comprehensive, agile tool that would allow U.S. Department of Energy (DOE)
to perform quick-turnaround analyses in response to evolving policy, scenario, and research
questions. In order to quickly perform multiple runs of the model with different scenario inputs
and to always be able to review runs that were made historically, it was important to set up a
framework for storing all inputs and outputs of the model for all runs made for important

analyses.

We use modern software-engineering methodologies to maintain model quality and enable
flexibility and responsiveness in response to analysis requirements that evolve as new bioenergy
issues gain interest from stakeholders. An open-source configuration management and version
control system, named Subversion, is used to track changes in the BSM model, documentation,
and other project-related files. Documentation and metadata for variables are embedded
directly in the STELLA model. Input data are stored, raw data sources are archived, and
provenance/pedigree metadata is tracked within a relational database: furthermore, input data

sets are processed within that database. Multidimensional data analysis, statistics, and

An Overview of the Biomass Scenario Model | August 2013 52

visualization tools are linked to the database in an architecture that allows for the automated
“refresh” of visualizations and analyses when new scenarios are run. This database-centric
approach makes it easy to develop and package “scenario libraries” for stakeholder use, as

shown in Figure 28.

~ Build Numbers = ~~
Zotero

Figure 28. Computing infrastructure for BSM.

The aforementioned computing infrastructure supports a high-throughput analysis process that
is outlined in Figure 29. In particular, it enables a “design-of-experiments” approach for
simulation studies that involve complex combinations of policy scenarios, sensitivity analysis,

and uncertainty quantification. The automation of simulation studies involves retrieving input

An Overview of the Biomass Scenario Model | August 2013 53

parameters from the database, running STELLA models in “batch mode”, and then storing
output into the database. To further enhance the approach, the BSM source files can be copied
and run on multiple machines at once to quickly make thousands of runs simultaneously. The
required output variables are specified in the database, and the values for these variables are
taken from STELLA for the specified simulations and transferred to the database. In this way,
we have a central system where any team member can re-create any past scenario—either by
viewing the previous runs or finding the correct model on the model repository. The outputs
can then be imported into any graphics software to visualize the simulation results, analyze
trends, and develop insights. Using these techniques, we have been able to analyze and compare
thousands of BSM runs with little effort, completing analyses that have been included in over I5

internal analysis reports, 8 external publications, and 10 forthcoming publications.

Modeling
Build Numbers
Inputs
Variables

Quality

Control

Exploratory Visualization
Simulation 4 Analysis bs nd Reporting

Data
Preparation

Data sources,
Provenance, ELT

Requirements/ Needs Analysis

Scenario LY oe
Design yo 2
a“
Runs ea
Studies _ oa a“
a Lo
~~ —
ee —_ oo

Figure 29. High level overview of the BSM analysis process.

An Overview of the Biomass Scenario Model | August 2013 54

Analysis, scenario development, and insights‘

Specific policy-relevant scenarios or past scenarios can be used to drive BSM simulations,
though BSM is not limited to scenario analysis. Under a specified scenario, BSM can be used to
track the hypothetical development of the biofuels industry given the deployment of new
technologies within various elements of the supply chain and the reaction of the investment
community to those technologies and given the competing oil market, vehicle demand for
biofuels, and various government policies over an extended timeframe. Note, however, that
high-level models such as the BSM are not typically used to generate precise estimates but

rather to:

¢ analyze and evaluate alternate policies
e generate highly cost-effective scenarios
e identify high-impact levers and bottlenecks

e focus discussion among policymakers, analysts, and stakeholders.

When BSM output includes unexpected system behaviors, modeling assumptions—particularly
the behavioral aspects of decision making and the adequacy of the representation of feedback—
need careful reexamination to distinguish potential insights from model limitations. The model
itself often indicates what assumptions need the most scrutiny; hence, it helps define the

research and learning agenda.

‘ Analysis efforts using BSM have been ongoing since early in 2010, initially using an earlier version of the model
that focused on ethanol from cellulose and starch crops. Beginning in the fall of 2011, the BSM team has been
designing experiments, creating scenarios, and conducting analyses using the current version of the model which
includes both ethanol and infrastructure-compatible fuels.

An Overview of the Biomass Scenario Model | August 2013 55

Although BSM inputs can be altered to include any combination of policies, al analysis efforts
included establishing a “reference policy case” to which subsequent scenarios could be
compared. The BSM reference policy case includes moderate incentives for ethanol production
and a 50 cent per gallon gasoline tax (which could be interpreted as a “carbon” or emissions
tax in dollars per ton of carbon dioxide). Policies are phased out in a staged manner, with the
policies involving grants for capital equipment or loan guarantees ending earlier and the policies

involving volumetric subsidies phasing out anywhere from 2020 to 2050. Each of the policies

included in the reference case is based on historical precedence or future plausibility.

As BSM functionality increased and research questions from stakeholders became more
sophisticated, we created an expanded list of scenarios to be easily incorporated, tested, and
analyzed. These scenarios are not intended to be prescriptive or comprehensive, but instead
represent an extended backdrop of cases against which policies can be tested and possible
industry evolution can be explored. The scenario library approach has proven to be quite agile
and useful from an analysis perspective, and the BSM team expects this functionality to be

augmented and expanded as part of the project’s ongoing development efforts.

Table 5. Scenarios in current BSM scenario library

Scenario Name Scenario objective/constraints | Strategy employed

|: Minimal Policy Starch until 2012 Apply minimal subsidies and policies

Provide support for ethanol only; analogous to BM

2: Ethanol Only Ethanol pathways only reference case

All pathways in order to produce | Allow all fuel types equal access to generous scenario

Sh EqualAccess 36 billion gallons/year by 2031 subsidies

To maximize growth restricted to | Target most promising technology and withhold

AL Outpitilociised $10 billion per year subsidy access from other pathways

Design subsidy timeline to enable take-off of multiple
fuel pathways by staggering start and end dates based
on pathway progress and potential

To maximize pathways restricted

5: Pathway Diversity | $10 billion per year

An Overview of the Biomass Scenario Model | August 2013 56

Scenario I: Minimum policy

The minimum policy scenario (Table 5, Figure 30) includes only a $0.45 price subsidy at the
point of production for starch ethanol that expires in 2012; it does not have any additional
subsidies directed towards renewable fuel production. Without government intervention in the
form of renewable fuel subsidies and given the oil price assumptions used in the model, neither
the cellulosic ethanol nor infrastructure-compatible fuel industries gain industrial momentum,
and thus fail to “take off’ to any significant extent. Starch ethanol is able to satisfy the market
for oxygenate in gasoline. The declining demand for oxygenate over the BSM time period is
attributable to the overall decline in gasoline demand as more fuel-efficient vehicles enter the
market in response increased Corporate Average Fuel Economy (CAFE) standards.

Scenario 2: Ethanol only

The ethanol only policy, (Table 5, Figure 30) applies all subsidies to the renewable ethanol
industry exclusively. With all cellulosic feedstock available to the ethanol industry, the cellulosic
ethanol industry is able to reach nearly 9 billion gallons in annual production (bgy). Annual
spending peaks at $6 billion (aside from the initial starch subsidy.)

Scenario 3: Equal access

The renewable fuel standard (RFS2) mandates that 36 billion gallons of renewable liquid
transportation fuels will be in the market place by the year 2022; the annual RFS2 volumes are
allowed to be adjusted by the U.S. Environmental Protection Agency (EPA) based installed
capacity and the amount of fuel demanded (US EPA 2011). The equal access scenario is
designed to mimic pathway agnostic policies such as the RFS. In this scenario, subsidies are set

to levels that spur renewable fuel output to RFS2 levels (i.e., 36 bgy). The results of this

An Overview of the Biomass Scenario Model | August 2013 57

scenario need to be viewed in the context of the initial settings and assumptions regarding the
industrial maturity of the infrastructure-compatible fuel technologies examined.

Scenario 4: Output-focused

Scenario 4 focuses subsidies on one pathway, i.e., fast pyrolysis. Fixed capital investment
subsidies and loan guarantees were limited to fast pyrolysis and not available to other fuel
pathways (Figure 30). Spending was limited to $10 billion per year, and after the expiration of
the starch ethanol price subsidy at the end of 2012, total subsidies reached a peak of only $5.3
billion in 2023. As the output for fast pyrolysis grows in later years, the fast pyrolysis subsidies
grow as well because of the price subsidy on each gallon of fuel. The exposure to loan
guarantees is not counted in the total subsidy figure. Infrastructure-compatible fuels — almost
exclusively fast pyrolysis — contributed 34.1 billion gallons to the total 51.4 billion gallons

produced in 2030.

Scenario 5: Pathway diversity

In the “pathway diversity” scenario, we explored the possibility of promoting pathway
production diversity by launching four different technologies to produce volumes of significant
output (over | billion gallons) with a total annual budget of $10 billion. After selecting the four
most competitive technologies through preliminary analysis (Fischer-Tropsch, fast pyrolysis,
fermentation, and methanol-to-gasoline), different subsidy amounts, start times, and durations
for each technology were applied in order to achieve output levels spread most evenly across
the technologies. The staggered start times and durations increase the attractiveness of
technologies with promising mature commercial-plant techno-economic parameters but starting

with limited industry maturity and experience. By the time pathways are ready to take off

An Overview of the Biomass Scenario Model | August 2013 58

(around 2023), technologies are on more even footing, allowing for greater pathway diversity
than the other scenarios. Industry output for infrastructure-compatible fuels reaches the 3
-billion-gallon volumetric limit just before 2023, at which time heavier “startup” subsidy values
switch to lower “background” subsidy values (as indicated in Figure 30). Annual production
reaches 34.9 billion gallons of renewable fuels per year in 2030 and reaches a peak of $8.9
billion of spending in year 2023. Of the total production in 2030, fast pyrolysis, Fischer-Tropsch,
fermentation, and methanol-to-gasoline produce 5.7 billion, 5.3 billion, |.3 billion, and 5.5 billion

gallons, respectively.

Scenario insights

BSM simulations based on these scenario libraries have provided a wide range of insights to the
Project team and policy-makers alike. Potential policies designed to accelerate the development
and sustainability of the biofuels industry can be easily tested across these embedded scenarios
under a wide range of assumptions regarding the magnitude, duration, and sequencing of
various policy interventions. Subsequent sensitivity studies on important model parameters are
used to quantify the responsiveness of various key BSM output metrics to policy initiatives,
alone or in combination, in these different scenario cases. This rigorous testing has built
confidence in the robustness of BSM, as well as informed key insights into the nature of the

evolution of the biomass-to-biofuels supply chain.

Insight 1: Momentum in the infrastructure-compatible fuels industry causes hierarchical

d h

ition for feed: ks, thus r ing | market share.

An Overview of the Biomass Scenario Model | August 2013 59

The existence of three renewable-fuel industries creates an interesting hierarchical competition:
the infrastructure-compatible fuels industry competes with the cellulosic ethanol industry for
feedstocks, while the starch ethanol industry competes with the cellulosic ethanol industry for
market share (both low- and high-blends). In the minimal policy scenario, the unsupported
biofuels industry produces only 12 billion gallons of starch ethanol output in the year 2030; the
cellulosic ethanol industry produces only about 60 million gallons of ethanol in the year 2030.
Starch ethanol is well established and continues to provide oxygenate for gasoline and high-
blend ethanol gasoline (E85), accounting for most, if not all, of the market for ethanol. Without
government intervention, the starch industry does not face competition from the cellulosic

ethanol industry and is able to meet all the ethanol demand.

When subsidies are applied exclusively to ethanol (with an emphasis on cellulosic ethanol), the
hierarchy between starch and cellulosic ethanol is salient. To prevent significant industrial
bottlenecks and encourage market penetration, downstream infrastructure subsidies are
critical. When cellulosic ethanol subsidies are high, the industry takes market share away from
the starch ethanol industry, but the latter is able to recover in the long-term because of its
maturity. Although this case subsidizes cellulosic ethanol heavily relative to all other pathways,
ultimately growth is restricted by limited market for ethanol (described above) as starch
ethanol and cellulosic ethanol compete for the same market. In the competition for market
share, the minimal policy and ethanol only scenarios confirm that cellulosic ethanol is able to

compete with starch only with sufficient subsidies in the developing years.

Providing subsidies for all pathways in the RFS2 scenario, the infrastructure-compatible fuels

industry is able to outbid cellulosic ethanol, driving up feedstock costs, which disadvantages the

An Overview of the Biomass Scenario Model | August 2013 60

cellulosic ethanol industry compared to the starch ethanol industry. Additionally, unlike ethanol,
the infrastructure-compatible fuels are modeled with potentially unlimited demand and no
interference (bottlenecks) from lack of downstream infrastructure. The net effect of these
factors is a 59% reduction in cellulosic output by the end of the simulation (i.e., 2030) relative

to its peak output, and a rapidly growing infrastructure-compatible fuels industry.

Insight 2: RFS2 volumes are achievable in 2030 with heavy startup subsidies.

Under the ethanol only scenario (Table 5), a total of 35.9 bgy of renewable fuels is produced in
the year 2030. Infrastructure-compatible fuels contribute over half of this amount (18.7 billion
gallons), while the starch- and cellulosic-based ethanol industries cumulative comprise 17.2 bgy.
The RFS2 timeline is shown to be impractical because the high-blend ethanol market applies
pressure on the system (i.e., the high-blend ethanol market is not large enough), and the
infrastructure-compatible fuels industry is not mature enough to produce 36 billion gallons in
2022. Reaching this level of production requires investment in the form of start-up subsidies,
particularly fixed capital investment subsidies and loan guarantees for commercial-scale facilities.
Total annual subsidies for the industry peak at a demanding $34.2 billion in the year 2024. The
start-up fixed capital investment and loan guarantee subsidies (especially commercial-scale) are
more effective at quickly building the industry than other subsidies but are far more costly than

the others.

An Overview of the Biomass Scenario Model | August 2013 6l

Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5
FS2

Minimal Policy Ethanol Only RI Output-focused Diverse pathways
e Ba 3 Ba 1 billion 2
2 ge 2 35 slonstartup &
Pathway | 3 8a 8 && int 8
Cellulosic — rr 0
0.6 0.6 6 — 0
0
Starch Ethanol 0.45 = 0.45
15
All Ethanol ae
“es
#| 32
2 me
5 Be
£3
36
§
E
&
1 billion fung! 0.3 billion fungible
HII Pointof production {s/sation} MERI Loan for Pioneer [%6] jelgallon fuel gaan startup
-d Capital Investment (FCI) Loan for Commercial [9%] startup limit limit for FCL

for Pioneer [9%] Downstream Distribution and storage [$/gallon] Commercial

I Ft for Commercial [%] Downstream Point of use [$/gallon]

Figure 30. Subsidy summary for scenarios | through 5. "Startup" value refers to values left of the annotated limit
line(s) for the technology, background values refer to values right of each respective line; thickness of the duration
bar indicates relative magnitudes of subsidies. Start and end years can be mapped to the year on the x-axis.

An Overview of the Biomass Scenario Model | August 2013 62

Insight 3: Production levels can exceed RFS2 levels if subsidies promote the most

economically attractive pathway

Even though the volumetric output of Scenario 4 was higher than Scenario 3 (RFS2), the
spending in Scenario 4 was less than that of Scenario 3 (RFS2). In Scenario 3, most years had
annual spending on subsidies exceeding $10 billion, with the highest year at $34 billion. While
Scenario 3 applied the same startup fixed capital investment (FCI) subsidies to all infrastructure-
compatible fuel pathways, Scenario 4 saves the most favorable startup FCI subsidies for fast
pyrolysis. Because these FCI subsidies end up being directed toward fast pyrolysis, the subsidies
are more efficient in promoting take off of that pathway than if subsidies are spread to different
pathways or if subsidies are directed to a pathway that is not as economically attractive. As a
result, even though a $10 billion annual subsidy was allowed, subsidies in Scenario 4 never

surpassed $6 billion in any year after 2012.

Fast pyrolysis receives FCI subsidies for pioneer and commercial plants in 2012-14. After that
time, cellulosic ethanol becomes a more attractive investment because it starts with more
learning at the pilot and demonstration levels and builds on that head start. Only when this
initial wave of fast pyrolysis plants is built and generates its own learning do fast pyrolysis plants
again become attractive investments. Additional fast pyrolysis commercial plants are built after
2017 without a subsidy. The initial subsidies in 2012-14 are enough to set in motion the learning
necessary to make commercial plants an attractive investment without additional FCI subsidies.

The threshold volume of 0.3 billion gallons for the FCI subsidy for commercial plants is reached

An Overview of the Biomass Scenario Model | August 2013 63

in 2017. As a result, after 2014, fast pyrolysis is supported by only the price subsidy and by loan

guarantees.

Although additional money was available to spend in each year, greater spending on subsidies
did not result in substantially more output. Spending more money on subsidies toward fast
pyrolysis ends up subsidizing plants that would have come online without subsidies and/or
result in more output but at a rate of spending above $10 billion per year. Additional subsidies
available to pathways other than fast pyrolysis have little effect. By restricting subsidies to fast

pyrolysis, it becomes more mature and locks out other pathways.

Though it appears most economically efficient, relying on this single pathway presents nontrivial
technology risks. Relying on a pathway with unfavorable long-term economics could result in
less volumetric output. In the BSM, the most economically attractive pathway is obvious based
on the available input data. In reality, the consequences of choosing a less than ideal pathway

may not be evident until several years after a policy decision.

Insight 4: Technologies with favorable long-term economic cost structures can succeed if
supported by targeted subsidies.

BSM simulations have shown that technological “lock-in” is likely to occur. Fischer-Tropsch has

the highest initial level of maturity among the infrastructure-compatible fuels; its initial settings
for pilot-scale and demo-scale maturity are higher than or equal to all other pathways (with the

exception of the starch and cellulosic ethanol pathways). However, the mature commercial

plant economics of fast pyrolysis are better than that of Fischer-Tropsch, based on the available

An Overview of the Biomass Scenario Model | August 2013 64

process designs. In Scenario 4, the maturity of fast pyrolysis has to increase in order to make

the investment look more attractive and to prevent technology lock out from Fischer-Tropsch.

In Scenario 5, subsidy policies are crafted to avoid lock out by any one pathway. To overcome
lock out, subsidies target learning through pioneer plants and do not include commercial plants.
By staggering policy start times and varying durations of subsidies according to maturity, the
other technologies have a chance to build experience. Limiting more mature or economically
attractive technology subsidies to begin after the volumetric threshold is reached allows the
other technologies to also develop. Rather than pouring extra subsidies into a relatively mature
technology, this approach provides the minimum subsidies needed for a more mature
technology (such as Fischer-Tropsch) to develop on a commercial scale, while providing the
others with the extra support they need to accelerate their experience levels. This approach
allows the successful take-off of four technologies while also approaching RFS2 production

levels in 2030.

However, though heavy subsidies may help overcome initial maturity differences, they are not
necessarily sufficient in overcoming differences in long-term economic cost structures.
Technology, such as fermentation (which is as commercially mature as the other pathways by
the end of the simulation), may need support beyond the subsidies exercised in this analysis

order to reach greater production levels.

Current and potential use of BSM
The BSM has provided an invaluable tool for the Bioenergy Technology Office of the DOE for

gaining intuition around the biomass-to-biofuels supply chain, and the insights detailed herein

An Overview of the Biomass Scenario Model | August 2013 65

only begin to address the impact the project has had in building understanding around several
key industry dynamics. The BSM has also been utilized in collaboration with other parties, such
as the EPA. Although to date the model has been mainly used by the DOE and other
governmental agencies, it has the potential to be highly useful to many different stakeholders

and across a wide range of analysis areas within the biofuels industry, as shown in Table 6.

Table 6. Potential collaborations for ongoing BSM use

‘ Climate | Suppl Biomass | Polic' R&D | Ener; Trade | Region-
Stakeholder iAnalysis Area Change Cire! Yield Y Secure specific
EPA x x
USDA x x x x
DOD x
Oil C x x
Biofuels Companies x x
Think Tanks x x x x x x x x
Foreign Governments/ x x
Organizations
Universities x x x x x x x x

Concluding remarks

The Biomass Scenario Model provides a rich representation of the supply chain associated with
the production of biofuels. By integrating feedstock production and logistics, multiple
conversion options, and market dynamics for butanol, fuel ethanol, and infrastructure-
compatible fuels (gasoline, diesel, jet fuel), the model serves as a vehicle for exploring the
mechanisms by which the biofuels industry might develop beyond its current state. By providing
an operational structure that reflects both the physics and economics of the system, BSM is a
tool for building understanding around initiatives that seek to stimulate sustained development
of the industry. And by representing the system of interactions simply and transparently, the
model sheds light on gaps in the data as well as areas where understanding of system structure

is in need of enrichment. Analyses of the BBM—both as standalone modules and in integrated

An Overview of the Biomass Scenario Model | August 2013 66

form—have underwritten powerful insights about the nature of the biomass-to-biofuels supply

chain and of the nature of policy initiatives required to stimulate industry take off.

Over the course of this project, the BSM working team has developed numerous internal
reports and briefing documents, which are housed on the project repository at (bsm.nrel.gov).
The References section of this paper also highlights several publications based on the BSM

project.

An Overview of the Biomass Scenario Model | August 2013 67

Appendix A: Pricing within BSM
In BSM, an endogenous pricing mechanism is an essential component of the structure that

underwrites industry development. The model incorporates endogenous pricing structures for

e Each of the commodity crops (corn, wheat, cotton, small grains, soy)
e Hay (regional markets)

e Cellulosic feedstocks (regional markets)

e Ethanol

e Butanol

Price mechanisms within BSM can be viewed as central components of an economic control
system. Each price signal evolves in response to the interplay of the forces of supply and
demand. As production, consumption, and inventories change over time, price responds to
imbalances. Prices, in turn, play a critical role in the investment, allocation and utilization
decisions of producers of agricultural products and of biofuels. They also play a critical role in

the fuel use decisions for butanol and for high-ethanol-blend fuels.

In developing the pricing structure used in BSM, we were mindful of multiple design
constraints. First, the pricing mechanism needed to be simple so as to be understandable to a
broad audience of model users. Second, the structure needed to be sophisticated, in order to
not generate spurious dynamics. A simplistic pricing formulation can lead to steady-state error
in controlled quantities or can become trapped in unrealistic states in response to extreme
condition tests. Finally, the pricing mechanism needed to be flexible enough to support real

world circumstances such as market initiation and scale-up.

An Overview of the Biomass Scenario Model | August 2013 68

The basic feedback relationships of the BSM pricing mechanism are shown in Figure 31.

Price

pressure to (3)
change price qj

Inventory
Relative to
Target %
(-) bi Required
(-) Inventory Inventory
Production Ps
Production Relative to ;
Consumption Consumption

(-)

Figure 31. Stylized view of feedbacks in BSM pricing mechanism

In this simple diagram, price works to balance production and consumption and to balance
inventory against desired or target levels. Production/consumption imbalances create pressure
to change price, as do imbalances between inventory and target inventory (which, in turn,
depends on consumption). In order to accumulate or integrate pressure over time, price must
be represented as a stock. The representation of price as a stock, in conjunction with pressure
from inventory, results in oscillatory tendencies in the system; oscillations are dampened by the

presence of feedback connections around production, consumption, and price.

Figure 32 shows output from a simplified model of pricing/inventory/producer/consumer
dynamics, which uses the basic pricing structure found in BSM. The test shows the equilibrium-

seeking tendencies of the structure, in response to a 10% shift in product demand.

An Overview of the Biomass Scenario Model | August 2013 69

@ 1: supplyrelative to demand 2: inventory relative to target 3: Price
1804

1104

Pl: 1 3 a —
2

60 120 180 240
Page 1 Time

Figure 32. Response of pricing system to 10% step-increase in demand

In BSM pricing, the mechanisms that connect production, consumption, and inventory to
fractional change in price are significantly more detailed. The structural arrangement shown in

Figure 33 is used to determine dynamic prices of several products throughout BSM, including:

e Ethanol at point of production
¢ Commodity crops

e Hay

¢ Cellulosic feedstocks

¢ Butanol

An Overview of the Biomass Scenario Model | August 2013 70

price input from inventoryindex

price input from prodn consn (alee mica epee

logistic price response cue from inventory

curve from prodn consn

spread in frac chg

spread in frac chg from inventory

from prodn consn, frac ch price
of).
frac chgfpiice prodn cons frac chg AS price inv
offset in frac chg change Wptice

offset in frac chg from inv

from podn consn

Piice

Figure 33. Detail of BSM generic pricing structure

The algorithm associated with this structure uses a bit of sophisticated math, but is relatively
straightforward. It begins by calculating the price input—either from inventory or from
production relative to consumption—as a distance from equilibrium in doublings or doublings.
When the ratio is |, the input is at its equilibrium value. When it is 2, it is one doubling away
from equilibrium. When it is 0.5, it is one halving away from equilibrium. To capture this

distance simply, the model uses logarithm functions as illustrated in Figure 33.

Second, the price input processed through a logistics function to generate a well-behaved

response curve. Price input and logistics calculations are shown in Figure 34.

smaponan cove «poe mest

Price input =
IF ratio >O

THEN logn(ratio) / logn(2)
ELSEO

Response curve =
1/(14€XP(price_input))

Figure 34. Illustrative price input and response curve calculations

An Overview of the Biomass Scenario Model | August 2013 71

The third step in this algorithm is to scale the response curve by shifting its intercept to (0,0)

and setting its asymptotes to desired maximum and minimum fractional changes in price. Finally,
the total fractional change in price is calculated as the sum of fractional changes from inventory
and production/consumption, and the result is applied to the price to generate a total fractional

change in price.

In BSM, this generic pricing structure is applied to multiple market situations, with context-

specific details (beyond the scope of this paper) applying to specific fuel markets.

Appendix B: Logit as allocation mechanism within BSM

In BSM, logit functions are a mechanism for allocating resources among multiple competing
uses. Detailed discussion of the logit function can be found in a variety of texts and articles
dealing with consumer choice. For example, Train (Train, 2003) provides a thorough
introduction to the logit, generalized extreme value, and a wide range of other approaches. The
logit function expresses the likelihood P of choosing alternative i from the set of j alternatives
given an observed utility of x. A simple form of the logit is shown below:

_elk;+ Bx)

Zles* 9)
The parameter k reflects unobserved or unexplained utility, while the parameter B is a scaling
factor. The logit function has several desirable characteristics. Among them:

e It can be interpreted in terms of the utility associated with alternatives within a set of

choice.

An Overview of the Biomass Scenario Model | August 2013 72

e The sum of probabilities across all choices is |
« There is a sigmoid relationship between utility and the resultant probability, which is

beneficial under extreme conditions

The typical interpretation of the logit formulation, in the context of consumer choice, is the
probability of choosing a particular alternative. In BSM, this probabilistic interpretation is
applied to a population of actors (for example, farmers, investors in conversion facilities,
consumers as they are deciding to fuel their vehicles) in order to generate an aggregate

allocation of land use, investment, or fuel use.

Logit formulations can be found throughout BSM, as summarized in Table 7.

An Overview of the Biomass Scenario Model | August 2013 73

Table 7. Uses of logit formulation throughout BSM

Module Usage Dynamic Inputs Notes
Crop land allocation
¢ Commodity crops
o With/without residues For crop land, nested logit
Perennial cellulosic crop function is used to allocate
© Hay Per-acre grower among broad groups (e.g.,
Feedstock Supply Pasture land allocation payment for commodity vs perennial

As pasture

As pasture harvested as
cellulosic feedstock

Perennial cellulosic energy crop

respective uses.

cellulosic vs hay) and then
among different
commodity crops

Conversion and
Relative
Attractiveness

Allocation of facility construction
resources among alternate pioneer
and commercial scale conversion
pathways in different regions

NPV of respective
conversion
pathways

Nested logit function is
used to allocate
construction capacity
among different
conversion platforms (e.g.,
fast pyrolysis) and then
among different regions

Pricing and Inventory

Displacement of ethanol by butanol
in lo-blend mixes

Butanol, ethanol

(Downstream) Displacement of non-bio-butanol in prices

industrial market

Allocation of fuel sales between hi- Price of gasoline
Fuel Use

blend and gasoline

Price of high-blend

Appendix C: Aggregation of age classes in the vehicle module

The current version of the vehicle module, like the other modules within BSM, reflects design

tradeoffs between the competing pressures of detail

realism” and usability. It is conceptually

straightforward to create a model containing great detail around vehicle type, regional

distribution, and age distribution of vehicles. Unfortunately, the computational overhead

required to simulate this detail would quickly become unmanageable. In an earlier version of

BSM which incorporated this detail, we were required to run the vehicle module separately

from the rest of the model and then import fuel demand scenarios separately.

An Overview of the Biomass Scenario Model | August 2013

In the current version of the model, we have reduced computational overhead significantly by
aggregating age distribution of vehicles. We represent vehicle vintages using 5 distinct cohorts.
Each cohort represents the 4 years of vehicle life. Within each cohort, each year vehicles are

scrapped or they get older. Those vehicles that survive to the end of a cohort are transferred

to the next cohort. A portion of the vehicle aging logic is shown in Figure 35.

vehicle influx logic

le Efficlency 1 vid morte suvve mie 1
ane Vi wtd mortrat V 2.wtd mort rate

Figure 35. Structure of vehicle vintaging

This structure aggregates together vehicles of multiple ages, and it is important to provide a
reasonable estimate of the distribution of vehicles within each cohort. To do so, we consider
the age-specific survival rates within each cohort, using these to derive an approximation of

distribution of vehicles across the cohort:

Let S, = survival rate for year n in cohort, 0 <= S, <=1, S, = 1

D, = fraction of cohort population in year n

Dj ='Syf (Sp SotSp*SotS; ® Sy St SF S5*'S) D
D, = Sy*S, 1 (Sy + So*S, + So*S, *S, + S,*S,* S, * $5)

D, = Sy*S, * 5,1 (Sp + So®S, + Sp*S, * S, + S,*S,*S, * S,)

0S, * Sp * S31 ( Sp + SoS; + So*S, * Sp + So*S,* Sy * Ss)

An Overview of the Biomass Scenario Model | August 2013 75

Age-specific survival rates are then applied to this distribution of vehicles in order to calculate
distribution-weighted age-specific mortality rates, which are then summed and applied to the
number of vehicles in the cohort to generate a mortality flow. The survival rate for the last year
in the cohort is applied to the appropriate distribution, in order to generate movement of

vehicles to the next cohort.

Figure 36 compares the transient response of a single 4 year cohort of the BSM vintaging
structure against a simple one-stock structure and against a more disaggregated structure with
4 |-year cohorts. For both systems, yearly survival rates are set to 50%. In the test, both
systems are initialized at zero. Inflow to each system is set to 100 initially; the inflow steps

down to 50 at time 10.

$@ +: Aggregated 4 Year Cohort 2: Disaggregated 1 Year Cohorts
1 ] 20
: sme
5] 1004 ~~, 2
_——
he
2 r r
0 6.00 12.00 18.00 24.00
Page 1 Time

Figure 36. Comparison of BSM and disaggregated vehicle cohorts.

An Overview of the Biomass Scenario Model | August 2013 76

References
Anderson, S. T. (August 2009). The Demand for Ethanol as a Gasoline Substitute, Michigan State

University, East Lansing, MI

Bush, B., M. Duffy, D. Sandor & Peterson, S. (2008). Using System Dynamics to Model the
Transition to Biofuels in the United States: Preprint, National Renewable Energy Laboratory,

Golden, CO

Energy Information Administration. (May 2010). Annual Energy Outlook 2010, In: Forecasts &

Analyses, 22.02.2011, Available from http://www.eia.doe.gov/oiaf/archive/aeo | 0/index.html

Hax, A. C., & Majluf, N. S. (1982). Competitive Cost Dynamics: The Experience Curve.

Interfaces, 50-61.

Interagency Agricultural Projections Committee. (2007). USDA Agricultural Projections to 2016.

U.S. Department of Agriculture

isee systems. (2010). STELLA: Systems Thinking for Education and Research Software,
22.02.2011, Available from

http://www.iseesystems.com/softwares/Education/StellaSoftware.aspx

Johnson, C., & Melendez, M. (2007 draft). E85 Retail Business Case. National Renewable Energy

Laboratory.

Johnson, C., & Melendez, M. (2007 draft). E85 Retail Business Case. National Renewable Energy

Laboratory.

An Overview of the Biomass Scenario Model | August 2013 77

Johnson, C., & Melendez, M. (2007 draft). E85 Retail Business Case. National Renewable Energy

Laboratory.

McNulty, M. S. (2010). Energy Price-Biofuel Production Cost Coupling Analysis. Boulder, CO: Pacey

Economics Group.

Merrow, E. W. (1983). Cost Growth in New Process Facilities. Rand Corporation.

National Association of Convenience Stores. (2007). Annual Report.

Newes, E., Inman, D., & Bush, B. Understanding the Developing Cellulosic Biofuels Industry
through Dynamic Modeling. In M. e. dos Santos Bernardes, Economic Effects of Biofuel Production.

InTech Open Access.

Office of the Biomass Program and Energy Efficiency and Renewable Energy. (March 2008).

Multi-Year Program Plan March 2008, U.S. Department of Energy, Washington, D.C.

RW Beck. (2010). Biorefinery Learning Curve Analysis. Report for NREL Subcontract LCI-9-88660-

Ol.

Sterman, J. (2000). Business Dynamics. Irwin/McGraw-Hill.

United States Department of Agriculture. (n.d.). Agricultural Baseline Projections. Retrieved from

ERS/USDA Briefing Room: http://www.ers.usda.gov/Briefing/Baseline/

United States Energy Information Agency. (n.d.). EIA Analysis and Projections. Retrieved from

http://205.254.135.24/analysis/projection-data.cfm#annualproj

An Overview of the Biomass Scenario Model | August 2013 78

The University of Tennessee. (n.d.). Agricultural Policy Analysis Center Research Tools -
POLYSYS, In: Agricultural Policy Analysis Center - The University of Tennessee, 22.02.2011, Available

from http://www.agpolicy.org/polysys.html

U.S. Department of Agriculture. (February 2011). Conservation Reserve Program, In:
Conservation Programs, 22.02.2011, Available from

http://www. fsa.usda.gov/FSA/webapp?area=home&subject=copr&topic=crp

U.S. Environmental Protection Agency. (December 2010). Renewable Fuels: Regulations &
Standards, In: Fuels and Fuel Additives, 22.02.2011, Available from

http://www.epa.gov/otaq/fuels/renewablefuels/regulations.htm

U.S. Government. (2007). Energy Independence and Security Act of 2007

Vimmerstedt, L., Bush, B., & Peterson, S. (2012). Ethanol Distribution, Dispensing, and Use:
Analysis of a Portion of the Biomass-to-Biofuels Supply Chain Using System Dynamics. PLOS

ONE.

An Overview of the Biomass Scenario Model | August 2013 79

Metadata

Resource Type:: Document
Description:: Biofuels are promoted in the United States through aggressive legislation as one part of an overall strategy to lessen dependence on imported energy as well as to reduce the emissions of greenhouse gases. Meeting mandated volumetric targets has prompted substantial funding for biofuels research, much of it focused on producing ethanol and other fuel types from biomass feedstocks. A variety of incentive programs (including subsidies, fixed capital investment grants, loan guarantees, vehicle choice credits, and aggressive corporate average fuel economy standards)have been developed, but their short-and long-term ramifications are not well known. This paper describes the Biomass Scenario Model, a system dynamics model developed under the support of the U.S. Department of Energy as the result of a multi-year project at the National Renewable Energy Laboratory. The model represents multiple pathways leading to the production of fuel ethanol as well as advanced biofuels such as biomass-based gasoline, diesel, jet fuel, and butanol). This paper details the BSM system dynamics architecture, the design of the supporting database infrastructure, the associated “scenario libraries” used in model runs, as well as key insights resulting from BSM simulations and analyses.
Rights:: CC BY-NC-SA 4.0
Date Uploaded:: March 17, 2026

Using these materials

Access:: The archives are open to the public and anyone is welcome to visit and view the collections.

Collection restrictions:: Access to this collection is unrestricted unless otherwide denoted.
Collection terms of access:: https://creativecommons.org/licenses/by/4.0/

About the Archives

Access options

Ask an Archivist

Ask a question or schedule an individualized meeting to discuss archival materials and potential research needs.

Ask a Question Book a Meet

Schedule a Visit

Archival materials can be viewed in-person in our reading room. We recommend making an appointment to ensure materials are available when you arrive.

Schedule a Visit Hours

Collection

System Dynamic Society Records, 1978-2016

Collection ID	ua435
Total components	7460
Items available online	6307
Last indexed	2026-06-29

Schedule a Visit

Collections

Peterson, Steve with Brian Bush, Emily Newes, Danny Inman, David Hsu, Laura Vimmerstedt and Dana Stright "An Overview of the Biomass Scenario Model", 2013 July 21 - 2013 July 25

Online content

Metadata

Using these materials

About the Archives

Access options

Ask an Archivist

Schedule a Visit