Application of system identification techniques to simulation model abstraction

摘要

Abstract: This paper describes preliminary research into theapplicability of system identification techniques tosimulation model abstraction. Model abstraction enablesthe construction of a valid, low-resolution surrogateto a more detailed, high-resolution simulation model.When rapid, approximate results will suffice, we canalso apply system identification directly to actualsystem data, bypassing the simulation stage. Fournon-traditional system identification techniques arediscussed in relation to their ability to producelinear, time-invariant, state-space formulations ofmultivariable random systems. A simple example isprovided in which one of the techniques, Hidden MarkovModels, is used to identify the transitionprobabilities within a simulated Markov Chain. Theexample is used to illustrate the challenges in generalsimulation model abstraction caused by modeltransformation procedures, problem size, uncertainty,and computational complexity. At this stage, we can saythat the application of systems identification tosimulation model abstraction is promising, yetchallenging.!11