Search

Platforms based on digital twins are gaining widespread use in overseeing energy infrastructure, particularly within smart grid environments. Recent progress in artificial intelligence opens up new possibilities to build greater user confidence by strengthening predictive and prescriptive analysis capabilities and refining how users interact with these systems. This study introduces an innovative framework and architectural design for energy digital twins that combines data-driven techniques with knowledge-based elements. Machine learning-driven data integration and analysis are fused into a knowledge graph that captures asset condition information, forecasting results, and relevant domain expertise. This structure underpins a retrieval-augmented generation strategy, which powers an advanced conversational virtual assistant grounded in a large language model. The assistant delivers practical decision-making support for asset oversight and maintenance activities. Key parts of the suggested architecture have been aligned with readily available commercial tools to create a working prototype. This prototype was applied in a practical scenario involving a portion of the high-voltage energy network located in central Italy.