Systems Informatics and Analysis - Engineering and Science of Biomass Feedstock Production and Provision - page 210

famous “stage coach” problem. Application of the model to a case study of rice
straw led to the recommendation that cubing after chopping should be done on farm
only if the maximum transportation distance was 50 miles. Otherwise, cubing should
be done at the plant before direct combustion. Grado and Strauss [ 84 ] also proposed
an inventory control model using dynamic programming to optimize the production
of ethanol from woody biomass. The delivered cost of ethanol was $0.38 L −1 , which
was dominated by manufacturing (60 %) and harvest and shipment (18 %).
De Mol et al. [ 80 ] used a similar concept of network of nodes and arcs to develop
an optimization model. The simulation model in this work, as explained previously,
provided cost and energy consumption values as a function of time for a fi xed net-
work structure. In contrast, the optimization model was developed to select the opti-
mum network structure. The mixed integer linear programming (MILP) model
ignored the daily fl uctuations as well as biomass losses. A knapsack model was
developed to solve the optimization problem. Their work analyzed a number of pos-
sibilities, including mixed feedstocks (e.g., thinning and restwood, prunings, and
sewage sludge), multi-model transport (road, rail, and water), and pretreatment
(size reduction and drying). Their work highlighted the value of having a common
model structure and database.
Cundiff et al. [ 85 ], in an extension of their earlier work focused on harvesting
and baling [ 86 ], developed a linear programming model to optimize storage, load-
ing, and transportation of biomass. They also addressed the uncertain impacts of
weather by converting the problem into a two-stage problem with recourse. The
model was applied to study the production for a 5,600 Mg month −1 biorefi nery in
Virginia, and the total cost for the operations was about $14 to $19 dry Mg −1 . The
cost varied between these values depending on the exact scenario that was studied.
The transportation cost was $8 to $10 dry Mg −1 . Judd et al. [ 87 ] have recently pro-
posed another mathematical programming model that focused on the use of SSLs
with possible densifi cation (briquetting). The model optimized the location of SSLs
as well as the machinery infrastructure to be used at those SSLs. In particular, they
compared permanent and mobile loading equipment at these SSLs. They used GIS
to generate input data such as farm and potential SSL locations and distances for a
hypothetical plant in Gretna, Virginia. They concluded that densifi cation was not
justifi able for transportation distances less than 81 km.
The BioFeed optimization model has been developed using a philosophy similar
to that of Cundiff et al. [ 85 ]. BioFeed integrates the complete production and provi-
sion activities, including on-farm production, and optimizes the design and manage-
ment decisions [ 17 ]. It models a scenario in which many farms are producing
biomass feedstock for one or more regional biorefi neries and models the important
operations along this value chain. This includes harvesting, postharvest packing,
loading and unloading, on-farm or satellite storage, transportation, and preprocess-
ing, such as size reduction and densifi cation. In addition to using an optimization
approach, a unique feature about the model is the integration of design and manage-
ment decisions in a single framework. It is an MILP model, in which integer deci-
sions are typically machinery selection decisions while the continuous decisions are