Professional Documents
Culture Documents
A Multi Period Source Sink Mixed Integer Linear Progr - 2016 - Sustainable Produ
A Multi Period Source Sink Mixed Integer Linear Progr - 2016 - Sustainable Produ
Raymond R. Tan ∗
Chemical Engineering Department, De La Salle University, 2401 Taft Avenue, 0922 Manila, Philippines
A B S T R A C T
Biochar-based systems are a potentially effective means of large-scale carbon sequestration. Such systems rely on
carbonization of biomass into biochar, which can then be added to soil for the dual purpose of sequestering carbon
and improving fertility. When properly deployed, these systems can potentially achieve negative emissions through
the net transfer of carbon from the atmosphere into the ground. In this work, an optimization model is developed
to determine the allocation of biochar streams of different quality levels to various biochar sinks, which are farms
whose tolerance to impurities present in biochar are known a priori. The optimization model determines source–sink
allocation of biochar so as to minimize total system carbon footprint, while ensuring that soil quality parameters
for each sink are not exceeded. An illustrative case study is solved to demonstrate the use of the model.
Keywords: Biochar; Negative emissions technology; Carbon sequestration; Optimization; Mixed integer linear
programming; Source–sink model
c 2016 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
⃝
1. Introduction and from free biochar into soil carbon (via soil amendment).
The pyrolysis of biomass yields gas (syngas), liquid (bio-oil)
Negative emission technologies (NETs) may become neces- and solid fractions (biochar) in varying proportions, depend-
sary in order to reduce atmospheric carbon dioxide (CO2 ) ing on feedstock characteristics and processing conditions.
levels to safe levels during the course of the 21st Century. The biochar that is formed is a carbon-rich solid which is
Examples of NET options include direct air capture (DAC), in turn comprised of labile (degradable) and recalcitrant (un-
augmented ocean disposal (i.e., ocean liming or fertiliza- reactive) fractions. When biochar is applied to soil, the re-
tion), bioenergy with CO2 capture and storage (BECCS), and calcitrant fraction decomposes very slowly, typically with a
biochar application to soil (McLaren, 2012; McGlashan et al., half-life measured on the time scale of a few centuries;
2012). The latter option has traditionally been used as a soil thus, application of biochar can potentially result in semi-
amendment strategy, but is now recognized as a potentially permanent sequestration of carbon, which can prove useful
important agro-industrial strategy for climate change mitiga- for achieving medium-term reductions in CO2 levels in the
tion, because it results in the net transfer of carbon from the coming decades (Lehmann et al., 2011). Furthermore, biochar
atmosphere into the ground (Lehmann et al., 2011). This ben- can improve the quality of soil, thus improving crop yields
efit results from three main steps: the net transfer of car- or reducing the need for fertilizers. It has further been ar-
bon from atmospheric CO2 into biomass (via photosynthesis), gued by biochar proponents that additional climatic bene-
from biomass into biochar (via thermochemical conversion) fits can be achieved from reduced fertilizer demand, reduced
Sink Area (ha) Application Storage capacity (t) Limiting biochar Limiting biochar
dosage (t/ha) flowrate (t/y) quality (mg PAH/kg)
1 200 50 10,000 1000 25
2 500 20 10,000 1000 10
3 400 25 10,000 1000 5
4 1000 30 30,000 3000 2
each source, given fairly uniform biomass feedstock during its Table 3 – Transportation distances in km for source–sink
operating life, and without significant process retrofits, gives pairs in the case study.
a consistent contaminant level in its product, such that Qik1 =
Qik2 . . . = QikT . Parameter ψ is used in the model to quantify Source Sink
1 2 3 4
the extent to which the decision-maker is willing to risk soil
contamination; its value is zero for a completely risk averse 1 70 60 50 120
decision-maker (i.e., if no level of soil contamination can be 2 80 60 40 50
3 100 80 80 40
tolerated), and unity for a risk-taking decision-maker who
is willing to let soil contaminant levels reach the prescribed
physical limit. Finally, limits are imposed on all flowrates (Kuppusamy et al., 2016). Table 1 shows the source data for the
between source–sink pairs: three biochar plants; note that Sources 1 and 2 are operational
throughout the entire 10-year planning horizon, while Source
0 ≤ xijp ≤ bij SU
ip
∀i, j, p (8)
3 only begins to operate in the third year. The characteristics
bij ∈ {0, 1} ∀i, j (9) of the sinks are given in Table 2. The storage capacity of each
farm or tract of land is based on the amount of biochar that
where bij is a binary variable signifying the existence or
can be added until the soil is saturated; it is further assumed
non-existence of a stream of biochar from source i to sink
that the annual limit to the rate of biochar application is just
j. As with other applications in PI literature, the presence
one-tenth of the storage limit, based on the specified 10-year
of these binary variables allows case-specific constraints
time horizon. The final column of Table 2 gives the maximum
on network topology to be added to the model (Poplewski
level of PAH in biochar that can be safely added to the soil
et al., 2010). Examples of such constraints include limits on
(note that 1 mg/kg = 1 g/t); in practice, this limit is based on
network complexity, specification of forbidden source–sink
background levels of PAH already present at the site, as well as
combinations, etc., and in many cases such considerations
the characteristics of the crop to be cultivated at the farm. The
can be formulated as additional linear constraints. The
distances between the biochar sources and sinks are given in
given objective function and constraints constitute an MILP
Table 3. It is assumed that the carbon footprint of transporting
model. Solution of such models for problems of practical size
biochar by truck is 0.1 kg CO2 /t/km (Foo et al., 2013). The
generally presents no significant computational challenges
MILP model corresponding to this problem was implemented
using branch-and-bound solvers embedded in currently
using the commercial optimization software LINGO 13.0 and
available commercial software (including solvers embedded
solved with negligible CPU time using a laptop with 8.00 GB
in common spreadsheet applications such as Microsoft Excel).
RAM, i7-3540MCPU and a 64-bit operating system running on
The use of this MILP model as a planning tool is illustrated in
Windows 8 Pro.
the next section.
Table 4 – Optimal source–sink network for the baseline scenario of the case study (biochar
flowrates in t/y).
Table 9 – Optimal source–sink network for ψ = 0.6 and N = 2 (biochar flowrates in t/y).
Table 10 – Optimal source–sink network for ψ = 0.6 and N = 1 (biochar flowrates in t/y).
network topology may be simplified by adding the following contamination tradeoff to be calibrated, while the inclusion
constraint: of integer variables allows network topology to be controlled
on a case-to-case basis. The MILP model proposed here can
Σi bij ≤ N ∀j (10)
thus serve as a high-level planning tool to determine the
where N is the limit to the number of biochar sources that are benefits that potentially accrue from a large scale, biochar-
allowed to be linked to any given sink. Solving the model for based carbon sequestration project. Future work can focus on
N = 2 gives the network shown in Table 9, for which the CO2 detailed operational extensions of this basic framework, us-
sequestration is 122,196 t. Note that the resulting network ing the model developed here as the core to which new fea-
simplification reduces the extent of carbon sequestration by tures (i.e., parameters, variables and constraints) are added;
a mere fraction of 1%. On the other hand, using N = 1 multiple-objective extensions can also be developed to incor-
gives the drastically simplified network shown in Table 10; porate economic aspects (Vochozka et al., 2016) as well as var-
this solution sequesters 114,958 t of CO2 , which represents a ious supply chain sustainability metrics used in the literature
reduction of about 6% relative to the unrestricted network in (Ahi et al., 2016). Alternative modes of long-term biochar stor-
Table 6. age other than direct soil amendment can also be explored.
Furthermore, the model should eventually be tested on pilot-
scale biochar demonstration projects.
5. Conclusions