Steffen Limmer, Takahiro Ishihara, Tobias Rodemann, "SimulationX Solver Setting Optimization via Automated Hyperparameter Tuning Approaches", ESI Forum 2019, 2019.

An ever-increasing complexity of technical systems requires sophisticated methods to optimize a larger number of design parameters under consideration of many objectives. For a broad class of problems evolutionary algorithms are the method of choice. Their main drawback is a huge computational effort since thousands or more simulation runs are required. It is therefore essential to reduce the simulation times as much as possible. One approach is to optimize the settings of the SimulationX internal solver. The task is therefore to find settings that minimize simulation run times for a group of similar simulation models or configurations, without leading to measurable differences in simulation results. We propose to use two well-known approaches for hyper-parameter tuning, SMAC and irace, to reduce simulation run times for the specific example of an energy management configuration optimization that was already presented on previous ESI-ITI events. After a general introduction to hyperparameter-tuning we will present results for our specific example and outline the specific challenges of tuning solver settings.