Structural Identifiability Analysis of Steady-State Induction Machine Models

Many mathematical models have been developed to describe the dynamic behaviour of induction machines and have been utilized in induction machines parameter identification. In some cases, model parameters may not be uniquely estimated, regardless of the used algorithm and the quality and quantity of the used measurements. This non-identifiability is related to the structure of the model itself. In this paper, the structural identifiability of three commonly used steady-state induction machine models (the standard T-model, the inverse Γ-model and the Γ-model) is investigated. Such analysis deals with the uniqueness of the solution for the unknown model parameters and is, therefore a prerequisite for induction machine parameter identification. Two structural identifiability techniques, the transfer function and bond graph, are reviewed and applied for testing the identifiability of the three models. The results show the importance of identifiability analysis before performing parameter identification. Structural identifiability investigation confirms the non-identifiability of the T-model and, on the other hand, the global identifiability of both the inverse rand rmodels.