摘要：AbstractIn this article, we propose an optimization based formulation for design of optimal inputs for multivariate systems. A well designed identification experiment can generate good quality models while incurring significant costs. In a typical process plant, the nominal policy is to operate the plant at or near constraints to achieve economic benefits. In this work, we quantify the cost of the experiment carried out on such systems in terms of the deviation from the nominal operational policy. The objective is to minimize the cost incurred during the experiment without violating the operational constraints while guaranteeing model quality. The proposed economics based optimization formulation is non-convex and hence we present a twostep iterative algorithm. The inputs are realized as white noise sequence filtered through an M-tap multivariate FIR filter. The filter coeffcients are obtained by the spectral factorization. A detailed simulation study is presented to illustrate the proposed approach.
关键词：KeywordsSystem identificationconvex optimizationinput designmultivariate system
