Application of Maximum Likelihood Method to Boiler System Identification

Abstract

A maximum likelihood (ML) method is applied to a boiler system identification. The mathematical model used in this paper is a discrete-time, singleinput and single-output (SISO), constant, linear system excited by an “innovation” process. Since the ML identification is reduced to a nonlinear optimization problem with equality constraints, the Davidon's conjugate gradient method is employed for numerical solutions. By using the given input/output data, the dynamics of the governor/steam pressure and the governor/steam temperature relations are identified as an SISO system, respectively. AIC and a test based on the innovation process are also applied for selecting an appropriate order of the assumed model.