Optimal parameters selection for simulated annealing with limited computational effort

Simulated annealing (SA) is a stochastic global search approach, which is of the ability to escape from local minima with deteriorative accepted probability and has been successfully applied to many difficult combinatorial and numerical optimization problems. But it is well known that the performance of SA highly depends on its parameters, especially the annealing schedule. Traditionally, the parameters of SA are determined empirically or by trial and error. In this paper, the determination of optimal SA parameters with limited computation effort is viewed as a stochastic problem, and then a systematic procedure based on ordinal optimization (OO) and optimal computing budget allocation (OCBA) is applied to select the most reasonable parameter combination. Simulation results based on flow shop scheduling benchmarks demonstrate the effectiveness.