To understand complex excitation energy transfer (EET) networks in photosynthetic systems, building a coarse-grained model is necessary to obtain a simplified representation. Here, we developed a systematic approach to produce coarse-grained models for photosynthetic systems by combining a minimum-cut method and a top-down clustering algorithm. The new approach was applied to investigate EET networks of three photosynthetic systems, and we demonstrate that our approach not only reproduces the population dynamics very well but also provides novel insights into the spatial-temporal EET dynamics in complex photosynthetic systems. The new approach could be a very powerful tool towards the elucidation of complex kinetic networks that is commonly encountered in Chemistry.