THESIS
2015
xi, 99 pages : illustrations ; 30 cm
Abstract
This thesis studies two topics related to distributionally robust optimization in the
areas of transportation and supply chain management.
The first part investigates the problem of location of quick charging stations to
support long-distance trips for Electric Vehicles. Electric Vehicles (EVs) are gaining
popularity with surging fuel costs and concerns on the environment. With
the limited technology on battery capacity, it is necessary to locate enough quick
charging stations along highways to ensure long-distance travel. We consider a location
problem for quick charging stations for EVs and focus on the attractiveness
of the different charging locations for travelers. A traveler who chooses to stop en
route to recharge the EV incurs monetary (for electricity and charging serv...[
Read more ]
This thesis studies two topics related to distributionally robust optimization in the
areas of transportation and supply chain management.
The first part investigates the problem of location of quick charging stations to
support long-distance trips for Electric Vehicles. Electric Vehicles (EVs) are gaining
popularity with surging fuel costs and concerns on the environment. With
the limited technology on battery capacity, it is necessary to locate enough quick
charging stations along highways to ensure long-distance travel. We consider a location
problem for quick charging stations for EVs and focus on the attractiveness
of the different charging locations for travelers. A traveler who chooses to stop en
route to recharge the EV incurs monetary (for electricity and charging service) and
inconvenience costs, which vary by location and travelers’ preferences. A traveler
chooses to use charging locations that minimize total trip cost while ensuring
completion of the trip. We prove that this charging location selection problem is
mathematically equivalent to a bounded inventory replenishment problem in the
production planning literature, which can be reformulated as a shortest path problem
on an augmented network. By doing so, we formulate the problem for the service
provider as a linear problem with uncertain and ambiguous coefficients. The
problem is equivalent to completely positive programming (CPP) and the travelers’
preferences can be reflected by the persistency value which is the optimal solution
for the CPP problem. We also consider a two-stage stochastic model to identify the
proper locations to minimize the total construction costs and the trip costs.
The second part investigates the problem of sourcing and supplier selection
under yield uncertainty and ambiguity. Supplier selection is a typical problem for manufacturing. With a proper selection of the suppliers, companies can reduce the
cost and thus improve the profit margin. Over the years, researchers and practitioners
tend to focus on consolidated sourcing with the benefit of economies of
scales and qualify inspection. Yet, due to the uncertainty of the sourcing, yield etc.,
risk pooling effect has been concerned for the main effect when referring to the
sourcing problem. As the yield is an unknown factor to the manufacturer, selecting
only one of the suppliers available seems not a good solution. Companies tend
to diversify the sourcing portfolio. However, selecting too many number of suppliers
will also have a downside on building long-run relationship with the suppliers
and quality monitoring and controlling. In our work, we assume only the knowledge
of the means and covariance matrix of yield for suppliers, and consider the
worst-case bounds, out of all distributions with the given moments, on the expected
costs. We consider to select suppliers to source a single product. To extend
the model, we also consider the ordering strategy to source complimentary components
to assemble a product where each component has one unreliable supplier
available. In addition, we consider sourcing strategy for an assembling production
line, with each component facing a number of unreliable suppliers.
Post a Comment