The European goal of climate neutrality by 2050 requires a significant increase in energy production from renewable sources. According to the National Integrated Energy and Climate Plan (PNIEC), Italy is expected to see a 250% increase in solar capacity and a 190% increase in wind capacity by 2030, compared to 2019 levels. However, high levels of renewable generation can threaten the security and stability of electrical grids due to the inherent variability of these sources. This leads to a higher need for flexible energy reserves to address imbalances and increases uncertainty in the energy market. The variability of renewable sources affects various time scales, but knowing future production and demand profiles can help manage reserve requirements and enable timely actions in several areas, including grid stability management, storage control, and energy market bidding.

To address this, at the end of the three-year period, RSE aims to implement a predictive system for photovoltaic production across different time scales and spatial areas. Planning operations require hourly forecasts extending several days ahead, while congestion resolution and intraday markets demand accurate 15-minute forecasts for a few hours ahead.

The production forecasting described here was performed using forecasts from various meteorological numerical models as inputs to appropriate post-processing systems. These systems use different algorithms based on the forecast type and time horizon, such as Analog Ensemble, Support Vector Machine, Random Forest, ARIMAX, and neural networks. Along with the forecast, a confidence band is provided, which is useful for assessing the flexibility needed to manage imbalances. The plan also includes developing a demand forecasting system for market areas and at the national level. A national demand forecasting system, HELPME, which uses Analog Ensemble and Support Vector Machine models with a one-day horizon, has been complemented by an ARIMAX-based system for 15-minute forecasts with a six-hour horizon for all market areas. Additionally, a predictive system for national gas demand based on Random Forest, using not only reanalysis temperature data but also forecasts from numerical models, is also a goal for the three-year period, and an initial version has been implemented this year.