Efficient energy planning is a key feature for the future smart cities. The real-time optimization of the energy distribution and storage is the real added value for smart grid and cities. However, the available energy providers’ infrastructures are not able to predict real-time fluctuation of the energy demand, taking into account the new emerging energy demand behaviours in urban context with high density, fast moving people or new energy usages such as electric vehicles; moreover current energy forecast methods are not scalable enough to integrate, with low cost and effort, hardware elements able to estimate energy demand in real-time.

The solution proposed in this paper exploit heterogeneous data sources to forecast in real-time energy demands without requiring physical interventions on the energy providers’ infrastructures. The proposed approach is mainly based on the use of probabilistic models and it exploits geolocalized cellular network traffic as independent variable to estimate energy demand without observing the actual behaviour of the energy network.

Probabilistic models have been derived by using the cellular network data coming from the mobile network production environment of Vodafone Italy and the A2A’s grid in a real world case study in the city of Milan.