(85b) Globally Optimal Parameter Identification for ODE Models

In this work we present a new method of determining parameters from experimental data for systems described by ordinary differential equations. Except for simple cases where the ODE models can be analytically integrated, this problem is typically addressed using non-linear regression techniques. These methods, however, fail to guarantee the global optimality of the solution, and also fail to ascertain whether the proposed problem parameterization is appropriate for the available data. To address these shortcomings, we propose a novel method for parameter identification that is guaranteed to identify the global optimum of the non-linear regression problem and is also able to deliver ranges for the model parameters for which the proposed model can describe the available data within a predetermined level of accuracy. The method is illustrated in a case study involving a batch reactor model and associated data for the glucose-to-ethanol fermentation of the yeast Saccharomyces cerevisiae.