This chapter has focused on and discusses the challenge of modeling reactive power-voltage response characteristics of renewable energy sources. A physics-informed deep learning approach is employed in this chapter to incorporate the influence of the grid model into the training process. Moreover, to improve the adaptivity of the approach to time-varying operational and weather conditions and reduce the reliance on sample volume, a knowledge graph-based approach is used to store, search, and utilize pre-trained deep learning models. The chapter presents the simulation studies on a modified IEEE 39-bus system and a 2486-bus real system and illustrates the efficacy of the introduced method.