An Innovative Approach for Optimal Time Delay Estimation in System Identification

Abstract

In this study, a novel strategy for time delay estimation (TDE) is proposed. The method, OTDE-Gen, uses a genetic algorithm to find the best time delay, so that the identified model presents the best performance. OTDE-Gen also allows to evaluate different model structures and distinct orders for processes of any dimension. A brief review of classical TDE techniques is presented, and the TDE problem is further analyzed from a System Identification perspective, through examples that illustrate and quantify how the time delay affects the model performance, and also how the identification setup (signal-to-noise ratio, model structure and order) affects the choice of the best time delay. The OTDE-Gen and the classical TDE methods are tested in a benchmark and performance comparisons are carried out. © 2017, Brazilian Society for Automatics--SBA.