CORMSIS Seminar - QUASAR: A General-Purpose Solver for Multistage Stochastic Optimization Event
- Time:
- 16:00 - 17:00
- Date:
- 21 April 2016
- Venue:
- Room 3041 Building 2, Southampton Business School
For more information regarding this event, please email Dr Yuan Huang at yuan.huang@soton.ac.uk .
Event details
Abstract: This talk introduces QUASAR, a general-purpose solver for stochastic optimization, which combines approximate dynamic programming with scenario reduction techniques to tackle real-world decision problems under uncertainty. The solver handles stochastic-dynamic programming problems in a fashion similar to solvers for linear programming but allows model coefficients to be modeled as stochastic processes. Embedded into a mathematical workbench, QUASAR provides users with an interface for algebraic modeling, stochastic input model estimation, as well as simulation output analysis. The basic functionality of the solver and the interface is demonstrated for a simple stochastic-dynamic optimization problem. The talk finishes with presenting results of a case study that demonstrates the usage of QUASAR for medium-term planning of an Austrian pumped-hydro storage system.
Speaker information
Dr Nils Loehndorf,Vienna University of Economics and Business,Bio: Nils' academic background is in operations research. He holds a master’s degree from the University of Mannheim and a PhD from the University of Vienna. Nils taught several undergraduate and graduate level courses in operations management and operations research. Currently, he is teaching a graduate course on data mining and decision support systems. Nils’ current research revolves around the economics of energy storage. By being able to buffer renewable energy supplies, energy storage is becoming a key component of the future energy supply chain. His research aims at developing models and methods suitable for efficient storage operation, valuation, and investment. Further research interests include everything that is related to computational stochastic optimization - a relatively new field that brings together concepts from applied probability, machine learning, mathematical programming, and stochastic simulation.