Model-based optimization has been widely used to exploit economical and environmental potentials of industrial processes. The quality of the model is crucial to the applicability of the optimization results. In part 1 of our paper [Faber, R.; Li, P.; Wozny, G. Ind. Eng. Chem. Res. 2003, 42 (23), 5850-5860], we proposed a sequential parameter estimation approach to solving large-scale errors-in-variables estimation problems with multiple data sets. In this part (part 2) of the paper, we present the application of the approach to an industrial coke-oven-gas purification process. The model of the reactive absorption/desorption process is composed of a complex, highly nonlinear equation system with 420 state variables per absorption/desorption unit. The parameters are estimated based on 15 experimental data sets. With the estimated parameters, a significant improvement in model accuracy is achieved in comparison to the results from a previous simulation study.