Surfactant-Polymer (SP) flooding has become an attractive Enhanced Oil Recovery (EOR) method. Defining chemical concentrations, chemical types and an injection schedule, according to geological features of a reservoir and well pattern, is key to making decisions for reservoir management. In this paper, we introduce an innovative approach for EOR optimization under geological uncertainty by integrating a reservoir geological property modelling and a robust optimizer. Multiple reservoir realizations are generated automatically by geology-driven modeling software and sent directly to an optimizer to analyze the effect of single or multi-parameters on objective functions such as cumulative oil production and net present value (NPV). Clay minerals play an important role in chemical flooding, but it is rarely included in the reservoir simulation. In this study, the distribution and proportion of clay are investigated in terms of facies and its relationship with porosity and permeability for a sandstone reservoir. Different facies and petrophysical properties are geostatiscally generated in a geologic manner that significantly improves the quality of history matching and optimization processes. It is found that SP flooding has the highest oil recovery factor in comparison with waterflooding, polymer flooding and surfactant flooding, and it demonstrates good performances even in high clay content reservoirs. The optimal formulation of SP and polymer slugs and injection schedule were proposed. The effect of clay content in cumulative oil and NPV were addressed, in which the more clay content is the lower NPVs obtain. A comprehensive geological uncertainty analysis has been performed for: (1) facies distribution only; (2) facies distribution and proportion. The results indicated that NPV uncertainty is less than 2.25% for (1) and about 4.18% to 5.68% for (2). The proposed optimization approach could be effectively applied to tertiary EOR techniques in various reservoir conditions under geological uncertainty. By integrating geological software, reservoir simulator and robust optimizer, it serves as a powerful tool for design and optimization of these processes. SP flooding is definitely a complicated process, therefore, an innovative modeling and optimization approach for SP flooding described in this paper is needed to improve the prediction of process performance.