Talk by William Johns (Dept of Mathematical Sciences, MSU)

1/30/2020  Wilson Hall 1-144  3:10-4:00pm


Electro-Magnetic Transients (EMT) models for fault tracking and real time simulation of electrical grids extensively use Multi-Input Multi-Output (MIMO) rational approximations. The current industry standard for MIMO rational approximations is the Vector Fitting algorithm. First, we will present a novel MIMO algorithm based upon the single input AAA algorithm to address some short comings of Vector Fitting. Second, this algorithm appears robust, but  we have to convert the result to a partial fraction form for use in EMT models. The challenges of this conversion include ill conditioned matrices, non linear root finding, and optimization for stable eigenvalue computations. Third, in order to use our algorithm in EMT simulations, we will need to guarantee that the poles are passive. We will present our ideas for approaching this potentially difficult problem. Finally, we want to minimize the number of poles used in the approximations, and thus minimize the running time of EMT models. Two possible approaches for this optimization will be discussed based upon a Rational Remez algorithm and AAK theory.

This presentation is a part of Ph.D  thesis proposal.