LPV state-space identification via IO methods and efficient model order reduction in comparison with subspace methods

In this paper, we introduce a procedure for global identification of linear parameter-varying (LPV) discrete-time state-space (SS) models with a static, affine dependency structure in a computationally efficient way. The aim is to develop off-the-shelf LPV-SS estimation methods to make identification practically accessible. The benefits of identifying a computational straightforward LPV input-output (IO) model-that has an equivalent SS representation with static, affine dependency-is combined with an LPV-SS model order reduction. To increase practical relevance of the proposed scheme, in this paper, we present a computational attractive model order reduction scheme based on the LPV Ho-Kalman like realization scheme. We analyze the computational complexity and scalability of our method and compare its benefits to the PBSID_opt scheme. Two examples are provided to demonstrate that our introduced approach performs similar to PBSID_opt in a numerical example and outperforms the PBSID_opt on measurements of a real world system, the air-path system of a gasoline engine.