TY  - GEN
N2  - This thesis presents an optimization model to simulate the global price formation of multiple commodities over multiple time periods. The model considers the connection of commodities through their production processes. The supply side maximizes its total profit taking account of the price-demand relationships of all products. The variables of this model are production quantities, transport quantities, storage quantities, and commodity prices. We apply the model to a part of the petrochemical market.
A large multi-commodity model requires many parameters. Moreover, the interpretation of the simulation results can become difficult. Therefore, this thesis focuses on the model and complexity reduction with respect to optimization models.
We propose a graph-theoretical approach to reveal the structure of large block-separable problems and to compare different decompositions into subproblems. The connections between primal and dual variables of a constrained optimization problem are represented on a hypergraph, which can be analyzed and beneficially partitioned using appropriate graph-theoretical methods. We show how different partitions of the hypergraph constitute different decompositions of the optimization problem. Furthermore, we address the approximation of subproblems.
The decomposition approach is adapted to the commodity market model. We formulate the subproblems for chosen sets of products and processes and present an algorithm for the automated identification of model components that are suited for an aggregation. The aggregation of components of the market model in terms of approximating subproblems is discussed from different points of view. Furthermore, we conduct sensitivity analyses within the overall problem and within subproblems. The numerical results of the application to a petrochemical market model reveal different possibilities of model reduction.
ID  - heidok16013
A1  - Kramer, Lilian
AV  - public
Y1  - 2013///
TI  - Modeling Price Formation in a Multi-Commodity Market - A Graph-Theoretical Decomposition Approach to Complexity Reduction
UR  - https://archiv.ub.uni-heidelberg.de/volltextserver/16013/
ER  -