A method to estimate the process noise covariance for a certain class of nonlinear systems. (15th September 2019)

• Record Type: Journal Article
• Authors: Gliga, L.I.; Chafouk, H.; Popescu, D.; Lupu, C.
• Is Part Of: Mechanical systems and signal processing. Volume 131(2019)
• Page Start: 381
• Page End: 393
• Publication Date: 2019-09-15
• Subjects: Extended Kalman filter -- Adaptive -- Online covariance estimation -- Permanent magnet synchronous generator; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2019.05.054
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: A method to estimate the covariance of the process noise is presented. The procedure is independent of any type of state estimator. It can only be used for processes with an equal number of states and outputs. The output matrix of the model must be invertible, and the states must be observable. The method works if the process parameters are different from the ones of the model. Abstract: A new method to compute the covariance matrix of the process noise is presented in this paper. This procedure is shown in the context of an Extended Kalman Filter. However, it does not use any of the matrices of the filter and is therefore independent of it. The method uses a constant covariance matrix for the measurement noise and, at each iteration, it re-computes the values of the process noise covariance matrix. The proposed method and two other ones, selected from the literature, are tested to estimate the current generated by a Permanent Magnet Synchronous Generator. All three methods are tested in the context of the Extended Kalman Filter. The obtained results are compared and discussed to highlight the strengths and weaknesses of the proposed approach.