Entropy Snapshot Filtering for QR-Based Model Reduction of Transient Nonlinear Electro-Quasi-Static Simulations

摘要

Approximations of the entropy of a transient nonlinear electro-quasi-static field system induce an ordering of high-fidelity snapshots, in terms of their information content. The resulting ordering enables a direct reduction in the snapshot matrix, and hence, any orthonormalization process, such as the orthogonal-triangular decomposition QR, of a reduced set of vectors can be used, in order to obtain a reduced basis for model reduction. Numerical experiments for a 3-D surge arrester verify the benefits of the proposed method, both in terms of accuracy and computational effort.