Optimizing the Hyperparameters of a Mixed Integer Linear Programming Solver to Speed up Electric Vehicle Charging Control

摘要

Optimization of charging profiles for controlled charging of electric vehicles is commonly done via mixed integer linear programming. The runtime of the optimization can represent an issue for the practical use. However, by tuning the parameter setting of the employed solver, it is possible to speed up the optimization process. The present work evaluates two popular hyperparameter tuning tools - irace (iterated racing) and SMAC (sequential model-based algorithm configuration) - for the optimization of parameters of the SCIP (Solving Constraint Integer Programs) solver with the objective to speed up the solving process for four common variants of the electric vehicle charging scheduling problem. Based on the results, the most important solver parameters are identified. It is shown that by tuning a very limited number of parameters, speed-ups of 60% and more can be achieved.