Context
Accurate forecasting of the load on Milan’s electricity grid is necessary to anticipate critical situations and support the development of a local flexibility market.
Challenge
Urban distribution networks in Milan face increasing pressure to accurately anticipate grid load and avoid emerging risks, particularly as flexibility mechanisms begin to play a larger role in system balancing.
AI Solution
Machine learning (ML) and statistical forecasting models analyze historical consumption data and meteorological variables, such as temperature and solar radiation, as well as calendar information. Based on these inputs, the system anticipates the load required by each asset. Activation is then managed through operations research algorithms, which identify the most efficient sequence.
Impact & evidence
The solution delivered forecast accuracy improvement as opposed to traditional methods. It also assisted with early detection of network congestion and improved effectiveness of flexibility activation, network reliability and operational efficiency.
Implementation / Adoption
Implementation of the solution requires availability of 15-minute MV network measurements (voltage, current), access to historical load data, weather forecasts, and calendar data along with suitable data platforms, forecasting tools, and integration with control room and flexibility systems. Deployment of the solution involved data integration (network, weather, calendar), development of AI forecasting models (short, very short, long term), operational validation and integration with flexibility service activation, supported by Italian NRA funding.
Data
The solution requires MV network measurements (voltage, current – 15-minute resolution), historical electrical load data, weather forecasts (temperature, solar radiation) along with calendar and seasonality data.
Future enhancements
Planned future enhancements to the solution include extension to other geographical areas, deeper integration with local flexibility markets, continuous improvement of AI forecasting models and enhanced real-time decision support for control rooms.