The design of Renewable Energy Systems (RES) involves the simultaneous consideration of technical, economic, and regulatory constraints. Traditional approaches often lack the flexibility required to efficiently explore the full range of possible configurations. Generative Adversarial Networks (GANs), as generative models, offer promising potential to automate this task. However, their purely data-driven nature makes them poorly suited to highly constrained domains such as RES. This paper proposes a hybrid methodology that combines the generative power of GANs with the rigorous structuring of knowledge through an energy domain ontology. The formalized ontology guides the generation of RES configurations by embedding domain specific rules, functional relationships, and physical constraints. The GAN is conditioned by this semantic information, ensuring that the generated configurations are more coherent and relevant. This work lays the foundation for a knowledge guided approach to RES configuration generation. It opens promising perspectives for automated, interpretable, and domain compliant design, pending future validation through concrete case studies.