Metaheuristic optimization methods for calibration of system dynamics models

System Dynamics models require calibration as part of the validation process. The available software tools only include one of the methods available for this end. But, the nature of this process requires sensitivity not only within a method, but with different methods. This study tests the effects of four optimisers (Genetic Algorithms, Simulated Annealing, Powell’s Algorithm, and a hybrid algorithm) in the calibration process of two system dynamics models. It was not possible to find an overall best optimizer algorithm due to three factors: model complexity, parameters for calibration and the measures (objective function) to evaluate the accuracy of the optimiser. Therefore, the choice of optimization method has an influence on the fit, limits, and representativeness of the model. More research is needed.