A Moving Horizon State Estimation algorithm applied to the Tennessee Eastman Benchmark Process

A moving horizon state estimation algorithm (MHE) is applied to the nonlinear and unstable Tennessee Eastman process, a well-known benchmark problem in the chemical process engineering community. The estimator fuses past measurements within a given time horizon and calculates the actual states in a maximum-likelihood fashion. The calculations are based on a first-principles process model. The arising least-squares optimization problem is solved numerically by the direct multiple shooting technique. Due to an intelligent real-time iteration approach, estimation CPU times lie between 0.5 and 4.5 seconds on a standard PC. Furthermore, an upper bound of the CPU time needed for one estimation task can be determined a priori. This makes the proposed MHE a true real-time algorithm for state estimation