• AIMD
• decentralized control
• forecast
• loading margin
• machine learning
• plug-in electric vehicles
• power grid simulations
• road traffic simulations
• smart cities
• solar power
• stochastic control
• wind power

Nowadays we are living a change of the electric system paradigm: the penetration level of energy generation from renewable sources is increasing, as well as the world's sale of electric vehicles, that are estimated to play a relevant role in future power systems. These new forms of power generation and consumption may pose challenges to the electric system if not properly handled; on the other hand, if they are correctly exploited and managed, they may represent a very valuable resource for the system. The present research aims at developing tools and management strategies able to facilitate the penetration of renewable energies, and to fully take advantage of the electric vehicles fleets, as mobile storage units, improving also the robustness and the resilience of the power system. The first part of this manuscript presents a forecasting model of future regional power generation from solar and wind sources, represented by a pure data-driven method, based on machine learning algorithms. While generally the power research community focuses on power forecasting at the level of single plants, in a short future horizon of time, this work is about aggregated macro-area power generation (i.e., the so called “market zones” of Italian Country, in a territory of size from 24x10^3 to 119x10^3 km^2), with a future horizon of interest up to 15 days ahead. Real data are used to validate the proposed forecasting methodology on a test set of several months. The second part of this thesis is about the electric mobility management tools, and it is composed by two different works. The first one is the study about the impact of the charge of plug-in electric vehicles on the dynamic response of power systems. Contextually, an efficient solution to control electric vehicles chargers is defined: it dynamically allocates the available power in an optimized way, preserving the stability of the network. The proposed approach is based on an Additive-Increase-Multiplicative-Decrease (AIMD) stochastic decentralized control strategy to efficiently and seamlessly manage the charge of a high number of PEVs with little communication efforts. A modified version of the New England network is utilized to validate the proposed control through a variety of scenarios and control setups. The second work about electric mobility is rather focused on the allocation of electric vehicles looking for public chargers in an urban context. More precisely, it presents a stochastic decentralized algorithm to recommend the most convenient charging station to plug-in electric vehicles that need charging. Different utility functions are used to describe the possibly different priorities of drivers, such as the preference to minimize charging costs, charging times, or the distance between them and the stations. In this work, the notion of charging station is generalized to include the possibility of supplying other loads in addition to the electric vehicles, of exploiting locally generated from renewable sources, and of providing ancillary services to the outer grid. Extensive simulations performed with the mobility simulator SUMO in realistic city-wide networks have been conducted to illustrate how the proposed vehicles assignment procedure works in practice, and to validate its performance. Finally, in the last part of this thesis, there is a brief description of the Web App tool developed in i-EM, for an optimal allocation of new charging points in an urban context. The proposed dashboard is an up and running web app, aimed at providing optimal charging infrastructures, based on key information of the territory, such as the density of the population, the presence of shopping centers, and the topology of the distribution grid; the algorithmic core is based on a deterministic procedure.