The development of real-time decision-making AI virtual assistant systems requires semantic artefacts such as taxonomies, controlled vocabularies, and ontologies. These artefacts assist human operators in dealing with heterogeneous information. This paper presents Grid2Onto, an application ontology that leverages agent-oriented AI recommendations to aid power grid operators in solving future problems based on past observations stored in a knowledge database. The main contribution is a unified semantic model that formalises Grid2Op's output, a realistic simulation environment for electrical supervision. The proposed Grid2Onto ontology enhances a real-time power grid recommender system by automatically generating a knowledge graph and reasoning capabilities. This paper highlights the added value of the proposed ontology.