The Role of Sensitivity Analysis in Stress Testing Real Time Models of Power Systems Controllers
Behshad Mohebali, Karl Schoder, Mark J. Stanovich, Michael Mischa Steurer, Gordon Erlebacher · 2021
Real-time Hardware-in-the-Loop (HIL) simulation has been a major step in design, development, and implementation of new technologies in the field of power systems. The distinct advantage of HIL simulations is their ability to incorporate external controllers or power hardware into the simulation using control or power interfaces. Due to their real-time characteristics, these simulations cannot be accelerated by using more powerful hardware, which makes the comprehensive evaluation of a given device-under-test (DUT) more challenging. Usually the simulation scenarios (mainly events and parameter values) are orchestrated by experts to observe the behavior of the DUT under stress by observing a set of performance metrics. The issue with this approach is that it cannot push the DUT to its absolute limit. Doing so requires a systematic approach to determine and manipulate a usually vast number of model parameters to check all possible scenarios and identify the ones that push the metrics beyond their acceptable range. However, not all model variables have the same influence on the performance metrics. Here, we use a Power Generation Module (PGM) model found on next generation naval vessels as a case study to implement fractional factorial design as a factor screening approach on two metrics defined for the PGM. The results confirm that a very small subset of the model variables have the majority of influence on the performance metrics. By knowing which variables are more influential on the metrics, the dimension of the search space, and therefore the computational burden of the analysis, can be drastically reduced.