摘要：Estimation of slowly varying model parameters/unmeasured disturbances is of paramount importance in process monitoring, fault diagnosis, model based advanced control and online optimization. The conventional approach to estimate drifting parameters is to model them as a random walk process and estimate them simultaneously with the states. However, this may lead to a poorly conditioned problem, where the tuning of the random walk model becomes a non-trivial exercise. Recently, Huang et al. (2010) has proposed a novel moving window weighted least squares parameter estimation approach, which is capable of simultaneous estimation of states and slowly drifting parameters/unmeasured disturbances. The slowly drifting parameters are assumed to remain constant in a time window in the immediate past and are estimated by solving a constrained minimization problem formulated over the window. In this work, the moving window parameter estimator of Huang et al. (2010) is recast as a moving window maximum likelihood (ML) estimator. It is assumed that the innovation sequence generated by the DAE-EKF is a Gaussian white noise process and further used to construct a likelihood function that treats the drifting model parameters as unknowns. This leads to a well conditioned problem where the only tuning parameter is the length of the moving window, which is much easier to select than selecting the covariance of the random walk model. Efficacy of the proposed ML formulation has been demonstrated by conducting simulation studies on an ideal reactive distillation system. Analysis of the simulation results reveals that the proposed moving window ML estimator is capable of tracking the drifting unmeasured parameter fairly accurately using only the tray temperature measurements.
