The steady increase in Renewable Energy Sources plants has clear advantages on environmental impact. But it poses major challenges to electricity grid management, as we move away from a hierarchical model where few large power stations plan and generate energy, towards a new model seeing energy produced by non-programmable sources. Issues arise in ensuring service continuity.

We adopted a holistic and multidisciplinary approach by integrating different competence areas, ranging from engineering, mathematics, statistics, machine/deep learning and artificial intelligence. We used a Grey Box approach, combining models of electrical systems with predictive ones, exploiting Machine Learning and Deep Learning. We also developed algorithms to make accurate predictions and adapt both to the changes in the structure of the national electricity system and to the variations in production and load detected in the data. We evolved the model by adding associative algorithms to relate different plants to each other, optimisation algorithms to identify the actual load state of the electricity grid, simulation algorithms to assess performance trends and the number and optimal position of sentinel plants from which to extract real data to optimise grid load forecasting.