The development of electrical systems (e.g. electrical grids) as well as making inferences on their operation (e.g. different shares of electric vehicles simultaneously plugged in) usually involves performing many simulations. For example consider an electrical grid connecting customers with different parameters such as load types, electric demand or reaction times on price changes. One output of such simulations sought often are hints regarding the stability of the system. To identify such regions of stability, a procedure that is adequate at first sight is to tabulate each parameter with some pre-defined steps and perform simulations for all possible combinations. With a high number of parameters, the task becomes intractable. In this case, one currently tends to make overly simplifying assumptions, e.g., all participants are assumed to have the same parameter values.
The focus of our research is the development of algorithms performing the simulations that do not require unrealistic assumptions but still provide meaningful results, e.g., the area of stability. We seek areas that are as large as possible, but at the same time ‘simple’, so that we can summarize the results of such simulations in a way convenient to the researcher. For example, one may want to obtain the rectangular stability region in 2D space with mean and variance of the parameters on the axes. We also aim at minimizing the number of simulations through automated ‘wise’ choices of system parameters for each subsequent simulation.