Abstract
This report summarises the work performed in the RAKETT‐project. The project goal is to reduce the calculation time of the FanSi model. The FanSi model is a power market
simulator for hydro dominated large‐scale power systems. The FanSi model formulates optimisation problems for the solution of the Economic Dispatch Problems.
The primary means for reducing calculation time is to apply decomposition techniques on deterministic Linear Programming problems from FanSi. Two methods for decomposition in space, Lagrangian Relaxation and Benders decomposition, were
studied. The computational experiments showed that results were not of high enough quality to be used with the FanSi model.
Benders decomposition was applied for decomposition in time. Three decomposition schemes were tested, one serial implementation and two schemes applying parallel
processing. One synchronous and one approximate asynchronous implementation. The asynchronous parallel implementation showed reduced calculation time and adequate result quality, but further study is recommended for use in FanSi.
A heuristic to relax the reservoir balance constraints was proposed and studied. The computational experiments showed a significant reduction in calculation time with minimal reduction in result quality.
simulator for hydro dominated large‐scale power systems. The FanSi model formulates optimisation problems for the solution of the Economic Dispatch Problems.
The primary means for reducing calculation time is to apply decomposition techniques on deterministic Linear Programming problems from FanSi. Two methods for decomposition in space, Lagrangian Relaxation and Benders decomposition, were
studied. The computational experiments showed that results were not of high enough quality to be used with the FanSi model.
Benders decomposition was applied for decomposition in time. Three decomposition schemes were tested, one serial implementation and two schemes applying parallel
processing. One synchronous and one approximate asynchronous implementation. The asynchronous parallel implementation showed reduced calculation time and adequate result quality, but further study is recommended for use in FanSi.
A heuristic to relax the reservoir balance constraints was proposed and studied. The computational experiments showed a significant reduction in calculation time with minimal reduction in result quality.
