Decision Support Modeling for Supply Chain Planning

Creating an efficient supply chain requires many decisions, such as the amount of feedstock that should be transported by truck or rail and the location of storage yards.

Research Question

What are the physical and operational characteristics of a supply chain that can deliver feedstock to a biofuel processing facility in a low cost, reliable, and environmentally sound manner?   This is an important research question because biomass harvesting, handling, and transportation costs represent the vast majority of the costs associated with the production of bio-ethanol, and there are concerns about the environmental sustainability of these supply chain activities.  Since a supply chain does not currently exist, computer models are needed to evaluate potential alternatives and provide support to decision makers.  Some decisions to be made are how much feedstock should be transported by truck or by rail, where should storage yards be located, and what inventory strategy should be followed to minimize feedstock delivery shortages during periods when seasonal road restrictions are in place.



The first step in this project is to develop a conceptual model that captures all the key elements of the biomass feedstock supply chain, including transportation infrastructure and equipment, storage facilities, and information exchange.  To limit the scope of this study, the conceptual model considers only transportation of feedstock from the roadside landing to the biofuel processing plant, although it incorporates information on harvesting and handing costs in the forest.   The next step is to develop two operational models—one using a simulation approach, and the other using an optimization approach.  The optimization model essentially answers the question, “What’s best?”, but it requires a significant degree of simplification.  The simulation model, on the other hand, can better address the real-world complexities of supply chains, but it is cumbersome to use to answer some of the questions decision makers may ask.  To overcome these limitations, the two models are being designed to operate in tandem, with the optimization model guiding strategic decisions, and the simulation model able to evaluate decision and policy outcomes in more detail.

The final step in this study is to apply the simulation and optimization models in an integrated manner to identify key parts of the supply chain where improved knowledge or changes in system capacity would have the largest effects on feedstock cost, reliability, and overall environmental impacts.  The models may also be used to evaluate a number of “what if” scenarios, including policy changes, economic swings, or severe climatic events. Throughout all phases of this project, strong integration with the other projects is needed in order to make use of the best available data and potentially guide further data collection.


Progress and Results

Simulation and optimization models have been constructed using professional modeling software. The optimization model supports long-term, strategic decision making by prescribing “optimal” (as determined by least cost, lowest emissions, or lowest energy use) harvesting and transportation plans for up to a 20-year planning horizon. Users can adjust the constraints on timber availability in order to match their assumptions about landowner participation in the biomass feedstock supply chain.  Baseline results from optimization indicate that the least cost solution calls for harvesting from areas close to the facility in the first 6-10 years, and then extending the harvest zone to areas farther away (up to 150 miles) in later years. However, this strategy leads to transportation cost increases of up to 30% in years 16-20, and also potential disruption of local workforces. The user is advised to consider adjusting constraints to represent socioeconomic objectives.


The annual decisions prescribed by the optimization model can be scaled to weekly harvesting and transportation plans for input to the simulation model. Unlike the optimization model, which assumes perfect knowledge of future conditions (i.e., timber availability and transportation availability), the simulation model considers uncertainties that can affect the supply chain on a daily basis.  Primarily, the simulation model represents the randomness of spring break-up timing, which can affect supply chain reliability as inventories are drawn down during spring thaw load restrictions. This model is recommended for analysis of trade-offs and risks associated with weekly to annual operational decisions.  For illustration, scenarios with reduced truck availability, reduced rail use, and reduced storage yard capacities are simulated to predict potential impacts on feedstock reliability and cost.  In general, cost impacts could not be fully analyzed because no assumptions are made regarding the cost of “emergency wood” or the cost of an unplanned facility shutdown.  However, the reliability of meeting facility demand is shown to be quite sensitive to each of these three factors, with the probability of a feedstock shortage roughly doubling due to either a 5% reduction in harvest volume, a 90% reduction in rail use, or a 40% reduction in storage yard capacity.  Simulation results also illustrate the trade-off between reliability and log age, e.g., increasing reliability from 94% to 97% requires larger inventories, with average log age increasing by about 10-15 days (to ages of 50-55 days) just prior to spring breakup.

The accompanying presentation slides illustrate model interaction, inputs, and outputs.

Final Project Report:

Feedstock Supply Chain Models for the Frontier Renewable Resources Biofuel Facility in Kinross, MI 
See also: Appendix G