This paper presents a simulation evolution and optimization approach for optimal configurations of a PV farm considering uncertain weather and dust soiling scenarios. Classic approaches employ the mean solar irradiance model which overlooks the risk that the produced power in real life may be far deviated from the expected value with a non-negligible probability. We handle such risk with three simulation optimization models based on analytics on benefit, cost, and risk. A large number of uncertain meteorological scenarios are generated by using Monte Carlo simulations. To find the optimal solutions to the proposed models, both mono-and multi-objective genetic algorithms are developed. We demonstrate the simulation with a real meteorology dataset. The results show the flexibility of our models which enable decision makers to perform fast, inexpensive, and non-disruptive simulations before implementing a real-world PV energy project.