A 2-d State Observer Using Particle Filter by Incorporation of Batch-to-Batch Dynamics

In batch processes with unmeasured states, state estimation problem is essential for control, monitoring and optimization of the process. In solving that problem, most state observers are inherently confined within a single batch. In this work, a 2-d state observer algorithm is designed taking into account the relations between adjacent batches in addition. First, A state-space model is introduced to characterize the state transitions over time and along the batch dimension as well. Then, an on-line alignment method that deals with the batch-to-batch shift problem is suggested. As in real world applications the environments are possibly be nonlinear and the process noise, measurement noise may be non-Gaussian, a 2-d particle filter method is presented, based on the 2-d state space model, to approximate the optimal solution of the Bayesian state estimation equations. The performance of the proposed state observer is evaluated by an application on a simulated chemical batch process.