Due to their numerous advantages, combined cycle power plants (CCPPs) have become an important technology for power generation. The part played by CCPPs on the power generation market includes frequent start-up/shutdown operations which must be optimized by means of innovative strategies, such as model-based methods. The availability of dedicated languages and libraries for physical system modeling facilitates the implementation of accurate large-scale models. However, the efficient resolution of optimization problems based on such models requires powerful algorithms, which impose a series of constraints on the model formulation, in particular, that of the lack of discontinuity source. This paper presents a method to transform a CCPP physical model designed for simulation in an optimization-oriented model, which can be further used with efficient algorithms to improve start-up performances. The methodology is based on the reformulation of the plant initially represented with discontinuity sources into a smooth model, by using continuous approximations of the Heaviside function. The validation results demonstrate the model consistency and emphasize the fact that it is accurate enough to be used for optimization and control purposes.