Robust estimation of fractional models in the frequency domain using set membership methods

In this paper, the usual definition of Grünwald–Letnikov fractional derivative is first extended to interval derivatives in order to deal with uncertainties in the differentiation orders. The Laplace transform of interval derivatives is computed and its monotonicity is studied in the frequency domain. Next, the main objectives of this paper are presented as the implementation of three methods for set membership parameters estimation of fractional differentiation models based on complex frequency data. The first one uses a rectangular inclusion function with rectangle sides corresponding to real and imaginary parts of the complex frequency response; the second one uses a polar inclusion function and the gain/phase representation; the third one uses a circular inclusion function with disk representation. Each inclusion function introduces pessimism differently. It is shown that all three approaches are complementary and that the results can be merged to obtain a smaller feasible solution set. The proposed methods can be applied to estimate parameters of certain/uncertain linear time variant/invariant systems.

► Fractional derivatives and integrals extended to intervals.
► Set membership estimation expanded to fractional models based on complex frequency data.
► Robust system identification algorithms developed for guaranteed solutions.
► Rectangular, polar, and circular inclusion functions used to compute certain/uncertain models.
► Results merged to obtain a more compact solution set.