Date of Award

2013

Degree Type

Thesis

Department

Chemical and Biomedical Engineering

First Advisor

Ungarala, Sridhar

Subject Headings

Chemical engineering -- Mathematical models, Dynamic programming, Nonlinear programming, Mathematical optimization, Chemical Engineering

Abstract

Moving Horizon Estimation(MHE) is a optimization based strategy to state estimation. It involves computation of arrival cost, a penalty term, based on the MHE cost function. Minimization of this arrival cost is done through various methods. All these methods use nonlinear programming optimization technique which gives the estimate. The main idea of MHE revolves around minimizing the estimation cost function. The cost function is dependent on prediction error computation from data and arrival cost summarization. The major issue that hampers the MHE is choosing the arrival cost for ensuring stability of the overall estimation and computational time. In order to attain this stability, this thesis incorporates dynamic programming algorithm to estimate MHE cost function. Dynamic programming is an algorithm for solving complex problems. The MHE cost function algorithm has been modied based on dynamic programming algorithm in order to ensure stability of the overall estimation. In order to apply this algorithm, a specic non-linear lter, particle lter is used for the initialization of MHE. The reason of using particle lter for initialization of MHE is due to fact that dynamic programming algorithm works on principle of samples and particle lter provides the samples. A comparison of mean squared error(MSE) using the nonlinear programming optimization and dynamic programming optimization is veried for the proposed theory of using dynamic programming algorithm in estimation of cost function

Download
COinS