Dynamics of the European demand for lignocellulosic (2G) ethanol:
An analysis of policy and leaming effects on market growth
Djerdj Horvat*, Christian Lerch*, Sven Wydra*
* Fraunhofer Institute for Systems and Innovation Research ISI
Breslauer Str. 48
76139 Karlsruhe, Germany
Phone: +49/721 6809-371
Email: djerdj.horvat@ isi.fraunhofer.de
A key approach to progress along the path to further sustainable development is the
continuous switch to second generation biofuels, which are derived from non-food re-
sources. One of the most promising biofuels is lignocellulosic (2G) ethanol. Neverthe-
less, due to the current absence of cost competiveness, the 2G ethanol market is strongly
dependent on policy support. Moreover, the production costs for 2G ethanol play a cru-
cial role for future cost competitiveness and thereby for market development. Hence, the
key question that arises is how the linkage between the policy-induced market growth
and the reduction of production costs can influence the dynamics for 2G ethanol de-
mand. To examine these dynamics and interaction in Europe over a forecasted 20 year
period, a simple System Dynamics model was developed as a part of a research project
funded by the European Union. The results of simulation runs show a positive impact of
learning effects resulting from build up and running of new capacities on 2G ethanol
price and thereby on 2G ethanol market demand. Moreover, it has been proven that
these effects are closely interlinked with the considered policy incentives in the form of
quota regulations according the Renewable Energy Directive (RED II).
System Dynamics, lignocellulosic (2G) ethanol, quota regulations, learning effects
1. Introduction
Currently biofuels contributed significantly to goals for renewable energy and ambitious
greenhouse gas emission reduction targets in Europe. But, first generation biofuels have
also been highly criticized because of food vs. fuel debates because of direct and indi-
rect change in land use (LUC and ILUC). Moreover, biofuel’s sustainability qualities
have been disputed (Timilsina/ Shrestha A. 2010). A key approach to the path to more
sustainability development is the continuous switch to second generation biofuels,
which are derived from non-food resources. One second generation biofuel is cellulosic
ethanol that may be produced from agricultural residues (e.g. straw and com stover),
other lignocellulosic raw materials (e.g. wood chips) or energy crops (miscanthus,
switchgrass, etc.) (EPure 2016).
Today, the global production of second generation (2G) ethanol is still very low, but
increasing, as several new 2G facilities have become operational in the last 3 years
(Bio-Tic 2015b; UNCTAD 2015). One full commercial plant is operative in the EU
(Beta Renewables in Italy), which accounts for somewhat less than 1% of the overall
ethanol production capacity in Europe (Philips et al. 2016). Still, significant technologi-
cal challenges in the build up of commercial plants occur. In particular, cost competi-
tiveness compared to 1G generation biofuels and fossil fuels has not been achieved yet,
as some production steps (e.g. pre-treatment of cellulosic), are still not optimized.
Because of absence of cost competiveness, the lignocellulosic bioethanol market is
strongly dependent on policy support, mainly on quota obligations for biofuels (Bio-Tic
2015). In 2015, the European Commission (EC) amended the Renewable Energy Direc-
tive and officially introduced a seven percent cap (=share of biofuels in total fuels) on
food based biofuels thus limiting future production of these first generation biofuels,
and introduced an indicative, non-binding 0.5% sub-target for second-generation of bio-
fuels (double counted towards the 10% renewable target in transport). However, these
indicative targets have not led to a strong market pull yet. At the end of 2016, the Euro-
pean Commission published a proposal for a new Renewable Energy Directive (RED
II), where an obligation of 3.6 % for 2G generation biofuels is envisaged. Currently,
uncertainties regarding future regulations still exist and so does the outlook for second
generation lignocellulosic ethanol in Europe (see e.g. OECD/FAO 2016; BioTic 2015;
Hirschnitz-G arbers/G osens (2015).
Even without reaching full cost competitiveness the production cost for lignocellulosic
ethanol will have an important role for market development, as it is more likely that
envisaged quotas will be actually transferred into national regulation and fulfilled or
slightly overachieved in reality. Once again, production costs are highly dependent on
the potential (policy-induced) market size. A significant amount of literature has studied
potential costs reduction from scale and learning effects and assumes significant reduc-
tions. This in tum may lead to convergence compared to 1G biofuels and fossil fuels in
around roughly 15-20 years (e.g. IEA-ESTAP, IRENA 2013; Daugaard et al. 2014;
Festel et al. 2015; Jonker et al. 2015).
For an analysis of that convergence, a System Dynamics model is well suited (Vim-
merstedt et al. 2012; Barisa et al. 2015). Therefore, a System Dynamics model could be
developed which aims to analyze the following research question:
How does the linkage between the policy-induced market growth and the reduc-
tion in production costs resulting from learning effects influence the dynamics of
demand for 2G ethanol in Europe?
To analyze the dynamics of 2G ethanol demand based on interactions of policy incen-
tives and learning effects as result of production capacity changes, a System Dynamics
model was constructed for the ethanol value chain. The model represents work in pro-
gress in a research project funded by the European Union’s Horizon 2020 research and
innovation programme under grant agreement No 723687.
The paper is organized as follows. First, a short literature analysis was conducted to
identify the interactive effects of policy-incentives and learning effects on demand. The
findings are used in section 2 to construct a model containing stock and flows which
forms the basis of the System Dynamics model. Following that, in section 3 simulation
runs and tests are conducted to analyze the impacts of individual factors and to obtain
first insights into the system’s behaviour. The last section looks at conclusions which
can be drawn from these findings.
2. Policy-incentive and learning effects on 2G generation ethanol
demand
To develop a system structure which links policy-incentives and learning effects on 2G
ethanol demand, we built a simulation model with three subsystems. The subsystems
are interacting with each other via the key variables — “2G ethanol demand”, “process
costs” and “price change”. Figure 1 depicts a simplified structure of the System Dynam-
ics model with its three subsystems.
2G production —
‘monthly
2G enthanol
Inventory
Scale effect on
production costs
a
production
ana Price chanae=
ular} cllect
&%\ demand o9
Real ethanol 2G price
Figure 1: The simplified structure of the SD-model with three subsystems
The first subsystem (see figure 1) represents the initial demand of 2G ethanol which is
derived from the initial total fuel consumption based on the indicative targets presented
in the following table 1 (EC, 2016).
Year 2010 2015 2020 2025 2030
% of the total
fuel con- 0.01 0.11 0.25 0.45 0.5
sumption
Table 1: Indicative targets for 2G generation ethanol in Europe
As the production costs, and thereby the price for 2G generation ethanol, depend on the
potential (policy-induced) market size, we use the initial demand for 2G ethanol as a
starting point for the second subsystem, reflecting the demand on production capacities
(see figure 2). Furthermore, with increased production capacities, the process costs of
the ethanol production decrease.
The reason for this cost reduction is dynamic learning effects via technological process.
Learning effects have been discussed intensively for biomass-based innovations such as
lignocellulosic ethanol because of their rather low technological maturity compared to
established oil-based products (Ye et al. 2014; Festel et al. 2014). Potential aspects for
improvement regarding the production of lignocellulosic ethanol relate to a more effi-
cient organization of production and transportation processes, the use of advanced mate-
tials, lower costs of the enzymes for pre-treatment processes and lifetime prolongation
of catalysts (de Wit et al. 2010). Techno-economic literature for biomass technologies
has commonly accepted the experience curve approach to estimate the aggregated effect
of technological learning over future time periods. According to this concept, costs de-
cline by a fixed percentage amount with each doubling in cumulative production (De
Wit et al. 2010; Festel et al. 2014).
A noteworthy amount of literature has studied potential costs reduction from scale and
leaming effects and assumes significant reductions, which may lead to convergence
compared to 1G biofuels and fossil fuels in around roughly 15-20 years (e.g. IEA-
ESTAP, IRENA 2013; Daugaard et al. 2014; Festel et al. 2015; Jonker et al. 2015). For
modelling the learning effects, we calculated scaling effects (from 1% to 10%) in accor-
dance with the multiplication of the production capacities compared to the start demand
- from 1 time to 100 times. Because we started with a very low level of demand at the
initial time (2010), and this amount increases rapidly (e.g. IEA-ESTAP, IRENA 2013;
Daugaard et al. 2014; Festel et al. 2015; Jonker et al. 2015), we use low scale effects
(1%) for multiplications of up to 7 times. A delay of 6 months is also integrated into the
model, which indicates the time for real reduction of prices after a scale effect occurs.
a
“ ~
fe
2G enthanol - Se
2G ethanol lnyentory 2G ethanol \
Ler sales production
ae A |
gf / |
we / ,
YL S Scale effect on
of / production costs
art de }
get f
- /
ethanol |. Ps | we
a Process costs 2
~ Increase of | Reduction of
= process costs
process costs
Figure 2: Effect of 2G ethanol market price on demand
In the third subsystem we modelled the effect of changed production costs on 2G etha-
nol demand (see figure 3). Following the study of IEA-ESTAP and IRENA (2013) we
calculated production costs as a sum of process costs (42%), energy costs (16%) and
feedstock costs (42%). To simplify the model considering only the effects of process
costs on real price and demand, we calculate production costs as individual variable of
the real 2G ethanol price.
-~ Overall > a
2G :
ethanol Price change
demand w , eilest “f
/ i seal pe \ _ .
A R { Real ethanol 2G price _‘Inereas
/ | a process costs
| \ /
{ \ {
\ | \ \
4 Initial ethanol 2G price\, Pieastiea’ f
\ Policy driven demand of 2G ethanol costs 4
—— oe fh 2
— \
b Ske op
\ ~~. Energy costs
Feedstock costs
Figure 3: Effect of changed production costs on ethanol 2G price and overall demand
Assuming price elasticity = 1, the model calculates the change of the 2G demand based
on the difference between the initial and the calculated real price as result of the de-
creased process costs. As the customers do not simultaneously react to price changes, a
delay of two months is integrated in the model to capture this effect.
3. First results of the ethanol demand dynamics
After validation tests of our System Dynamics model, in which we verified the structure
of the model and validate it through extreme conditions experiments and sensitivity
tests, the model is ready for first system behaviour analysis for policy making. For a
dynamic analysis and a deeper understanding of the system’s behaviour, various simula-
tion runs and tests were conducted. Some exemplary runs are shown in the following
and interpreted to generate the first dynamic hypotheses.
Figure 4 shows the dynamics of the 1G (Line 1) and 2G (Line 2) ethanol demand in the
time period of 240 Months starting from January 2010. For the 2G ethanol we simulate
under the assumption of an indicative target — following the Renewable Energy Direc-
tive (RED II) - that the share of 2G in total fuel consumption in 2030 will be 0.5%.
11,000
8,250
q
g 5,500
5
2,750
015 3 pp | pp | tp 2
0 24 48 72 96 120 144 168 192 216 240
Time (Month)
1G ethanol demand ——-1———4 2G ethanol demand 2——2——
Figure 4: 1G and 2G ethanol demand between 2010 and 20301
The runs in figure 4 show 1G ethanol demand increases between 2015 and 2020, up toa
level of approximately 10.500 million litres. This boost results from a higher share of
1G ethanol of the total fuel consumption in 2015, switching from 4% to 6%. After
reaching its peak in the year of 2020, the 1G demand slowly decreases again, because
the share remains the same, but the total fuel demand decreases from 2020 until 2030.
Moreover, at around 2025 the 2G demand starts to increase slowly, reaching a level of
600 million litres in 2030. This derives from boosting the share of 2G ethanol from
0.25% in 2020 to 0.45% in 2025 and hence, it is almost doubled (compare tab. 1). But
due to this very slow increase it seems necessary to generate further simulation runs,
showing the potential effects resulting from an additional initial share.
1 By policy-driven share of the 2G ethanol demand 0.5 % of the total fuel consumption (Base run)
0 24 48 72 96 120 144 168 192 216 240
Time (Month)
Policy-driven demand of 2G ethanol + t t 4 4
Overall demand of 2G ethanol 2 2 2 2 2 2
we
Figure 5: 2G ethanol demand by policy-driven share in total fuel consumption 0.5 % in
2030 (Base run)
In order to gain a deeper insight into the dynamic behaviour of the 2G ethanol market
and its interactive links to learning effects and policy incentives, we analyzed various
demand scenarios for Europe. Therefore we modified the share of 2G ethanol in 2030
from 0.5% to 3.6 % and compared the policy-driven demand and the overall demand of
2G ethanol. The policy-driven demand describes the amount of litres generated by the
legal share of 2G ethanol of the total fuel demand. In contrast, the overall demand in-
cludes the legal share plus the additional consumption of 2G ethanol. While the policy-
driven demand is given by law, the additional demand depends on the market price.
Consequently, the higher the learning rate, the lower the market price and hence, the
higher the overall demand. This demonstrates how the legal share affects an impact on
the overall demand, too. The following figures show the dynamics of the 2G ethanol
demand, as the result of policy incentives, and the overall 2G ethanol demand, as the
result of learning effects resulting in an additional 2G ethanol demand.
Figure 5 represents the base run in which an initial share of 2G ethanol in total fuel con-
sumption of 0.5 % was used. As the graph illustrates, the “Overall Demand” (line 2) is
growing faster than the “Policy-driven Demand” from 2016. In 2030 it reaches a de-
mand capacity of approximately 600 million litres and hence, almost 100 million litres
more than the initial demand. Moreover we observe the “gap” between real demand and
initial demand is opening over time. These effects result from an accelerated price de-
crease, leading to a faster reduction of production costs and thus, a higher demand over
the years. This loop is accelerating over time, which results in the demand gap further
increasing over time. This loop and hence this gap, is influenced by the share of 2G
ethanol reached in 2030. Consequently the question arises, how does a higher initial
share of 2G ethanol (e.g. 1.5% or 3.6%, following the Renewable Energy Directive
(RED II)) influence the real demand of 2G ethanol?
2,000
1,500
1,000
Million Litres
500
0 24 48 $72 96 120 144 168 192 216 240
Time (Month)
Policy-driven demand of 2G ethanol t t t 4 4
Overall demand of 2G ethanol 2 2 2 2 2 2
an
Figure 6: 2G ethanol demand by policy-driven share in total fuel consumption 1.5 % in
2030
As expected the figures show that the leaming effects have a stronger impact on the
overall demand, if the initial share of 2G in total fuel consumption is increasing. Figure
6 shows the run for an initial share of 1.5% and figure 7 for 3.6%. As the diagrams
show, the higher the initial share of 2G ethanol, the faster the gap between the policy-
driven demand and the overall demand is growing. This derives due to a lower price
level, leading to additional demand beyond the policy-driven demand. While the gap
holds a value of 100 million litres for an initial share of 0.5%, it is 350 million litres for
1.5%, and approximately 850 million litres for 3.6%. Finally, figure 8 illustrates the
development of the gap between initial demand and real demand over time (comparison
between Fig. 5, 6 and 7, see also Fig. A-1).
6,000
4,500
Million Litres
w
rod
S
S
24 48 72 96 120 144 168 192 216 240
Time (Month)
Policy-driven demand of 2G ethanol t t t t t +
Overall demand of 2G ethanol = = = = -: =
Figure 7: 2G ethanol demand by policy-driven share in total fuel consumption 3.6 % in
2030
1,000
750
:
= 500
250
eee ees
24 48 72 96 120 144 168 192 216 240
Time (Month)
Demand gap by initial share of 2G demand in total fuel % 4 4 4 4 4 4 f
Demand gap by initial share of 2G demand in total fuel consumption 1.5% —2 5 2 3 2 2 2
Demand gap by initial share of 2G demand in total fuel consumption 3.6% 3 3 3 3
Figure 8: Gap between policy-driven and overall 2G ethanol demand for various shares
(0.5%, 1.5%, 3.6%)
11
4. Conclusion
This paper takes the first step towards analyzing the interactive effects of policy incen-
tives and learning effects on 2G ethanol demand. Therefore a System Dynamics model
was first developed, which reproduces the system structure in a very simplified way, but
is able to provide initial insights into the system’s behaviour. By means of simulated
tests it was possible to develop the first hypothesis about the dynamics of 2G ethanol
demand. Through the simulations we can show that the learning effects influence posi-
tively the reduction of process costs and the 2G ethanol market price, thereby influenc-
ing the 2G demand positively. This effect is enhanced by reinforcing policy incentives
in terms of indicative targets (Renewable Energy Directive (RED II)).
However, this dynamic analysis is just the first step based on simplified systemic struc-
tures and is still ongoing research. The production costs, for example, include a couple
of other factors that have not been regarded so far. Moreover, feedback influences on
feedstock may arise in the case of large scale production, as residues and straw might
become scarce. Furthermore, land use changes and/or feedstock price increases may
occur and affect production. Summing up, there are still a few loops, restrictions, re-
source limitations and political decisions that are not included in this first model so far.
Nevertheless, these effects may take place in a more detailed System Dynamics model
in the future.
12
References
Barisa A, Romagnoli F, Blumberga A, Blumberga, D. 2015. Future biodiesel policy
designs and consumption patterns in Latvia: a system dynamics model. Journal of
Cleaner Production 88, 71-82.
Bio-Tic 2015. Overcoming hurdles for innovation in industrial biotechnology in Europe
- BIO-TIC Market Roadmap.
Daugaard T, Mutti LA, Wright MM, Brown RC, Componation P. 2015. Learning rates
and their impacts on the optimal capacities and production costs of biorefineries.
Biofuels, Bioproducts and Biorefining, 9(1), 82-94.
De Wit M, Junginger M, Lensink S, Londo M, Faaij A. 2010. Competition between
biofuels: Modeling technological learning and cost reductions over time. Biomass
and bioenergy, 34(2), 203-217.
EC (European Commission) 2016. Proposal for a directive of the European Parliament
and the Council of Europe on the promotion of the use of energy from renewable
sources (recast);
https://ec.europa.eu/energy/sites/ener/files/documents/1_en_act_partl_v7_1.pdf
ePURE 2015. European renewable Ethanol — Enabling Innovation and Sustainable De-
velopment. State of the Industry 2015. Brussels.
Festel G, Wirmseher M, Rammer C, Boles E, Bellof M. 2014. Modelling production
cost scenarios for biofuels and fossil fuels in Europe. Journal of Cleaner Produc-
tion, 66, 242-253.
Festel G, Wirmseher M, Rammer C. 2014. Scaling and learning effects of biofuels con-
version technologies. Energy Technology, 2(7), 612-617.
Hirschnitz-Garber M, Gosens J. 2015. Producing bio-ethanol from residues and wastes -
A technology with enormous potential in need of further research and devel-
opment. Policy Brief No. 2, November 2015. Sweden.
IEA-ESTAP, IRENA 2013. Production of Liquid Biofuels, Technology Brief.
Jonker JGG, Van Der Hilst F, Junginger HM, Cavalett O, Chagas MF, Faaij APC. 2015.
Outlook for ethanol production costs in Brazil up to 2030, for different biomass
crops and industrial technologies. Applied Energy, 147, 593-610.
OECD/FAO 2016. OECD-FAO Agricultural Outlook 2016-2025. OECD Publishing.
Paris.
Timilsina GR, Shrestha A. 2010. Biofuels: markets, targets and impacts. The World
Bank Working Paper 5364.
13
Vimmerstedt LJ, 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 7 (5), e35082.
UNCTAD 2015. Second Generation biofuel Markets: State of Play, Trade and Devel-
oping Country Perspectives. UNCTAD/DITC/2015/8. Switzerland.
Ye F, Paulson N, Khanna M. 2014. Technology uncertainty and learning by doing in the
cellulosic biofuel investment. Agricultural & Applied Economics Association's
2014 AAEA Annual Meeting. Anais. Minneapolis, MN.
Appendix
800
600 Le Base simulation run:
E Policy-driven share of
Cee i 2G ethanol demand in
3 total fuel demand
200 0.5%
Million Litres
2,000 Z
1,500 Second simulation run:
Policy-driven share of
= 2G ethanol demand in
total fuel demand
500 [oa 15 %
Third simulation run:
Policy-driven share of
3,000 = 2G ethanol demand in
total fuel demand
1,500 3.6%
faba gg eee Ty | | | |
Poe 4 72 «96 ~=«120 ane 168 192 216 240
‘Time (Month)
of 2G ethanol 33 33g
Oven demand of 2G ethanol
Million Litres
Figure A-1: 2G ethanol Demand by Initial Share of 2G Ethanol in Total Fuel Con-
sumption for 0.5%, 1.5% and 3.6% in 2030
