A hybrid linear model for real-time optimization-based control of a combined sewer network has been developed to be used for the minimization of pollution during storm events. The model takes into account delays and attenuation in sewers together with piecewise linear approximations for flow over weirs, overflows in junctions and flow re-entering the network after overflows. Using the proposed model, an Optimal Control Problem (OCP) is formulated, which can be efficiently solved by means of a mixed integer linear or quadratic programming problem. The performance of a Model Predictive Control (MPC) strategy solving consecutive OCPs is assessed by means of closed-loop simulations using a physically-based complex model as virtual reality.