Résumé:
The energy transition and the digitalization of electrical infrastructure have led to the emergence of Smart Grids—intelligent networks that optimize energy production, distribution, and consumption. This thesis addresses the dual challenges of data indexing and prediction within Smart Grids. It presents a comprehensive review of indexing methods (time series indexing, RMI, autoencoders) and predictive models (regression, random forests, LSTM). A prototype system combining intelligent indexing and automatic forecasting using energy time series data was implemented. The experimental results demonstrate significant improvements in query response time and prediction accuracy, validating the proposed approach as an effective solution for managing massive and heterogeneous energy data in Smart Grid systems.
