Smart grids

Traditional architecture

The traditional electrical grid was built around a relatively simple architecture: a small number of large power plants generated electricity and transmitted it towards consumers.

High renewable energy penetration changes this picture.

Solar panels, batteries, electric vehicles and small generators introduce large numbers of controllable devices throughout the network. A future electricity grid therefore looks increasingly like a network of intelligent, interacting subsystems.

Smart grids

Smart grids are the main architecture supporting this transition. They combine local generation, storage, and loads and can operate independently from the main grid. Their basic physical requirement is seemingly simple:

Generation = Consumption + Losses.

But maintaining this balance while keeping voltages within safe limits, respecting converter constraints, distributing power appropriately and allowing devices to connect and disconnect is a challenging control problem.

My early research focused particularly on DC microgrids, which are attractive for integrating batteries, photovoltaic generation and many modern electronic loads.

Voltage regulation

One of the main challenges in DC microgrids is maintaining stable voltages [1] despite variations in generation and consumption. When a load changes or a generator adjusts its power output, currents redistribute throughout the network, potentially causing voltage fluctuations. Since all units are electrically interconnected, the actions of one converter inevitably influence its neighbours.

We develop decentralized voltage controllers [2] that regulate each generation unit using primarily local information, without relying on a central coordinator. By exploiting passivity theory, we establish conditions under which the individual controllers collectively guarantee stability of the entire microgrid. 

Plug-and-play

An important feature of these controllers is plug-and-play operation [3]. When a generation unit is added or removed [4], only a small number of neighbouring controllers need to be updated; the complete network does not need to be redesigned. For several of these designs, the stabilizing controller parameters can be characterized through explicit inequalities [1] involving local electrical quantities.

Nonlinear models

We investigate more realistic nonlinear load [1]. In particular, constant-power loads can have a destabilizing effect on DC networks. Moreover, medium-voltage smart grid [2] models are ineherently nonlinear, and require nonlinear control techniques to ensure voltage stability. Using passivity-based arguments, we derived conditions under which networks containing these loads, dynamic power lines and distributed generators remain stable, while maintaining a decentralized and plug-and-play control architecture. 

Higher-level decision making

Primary controllers operate at a fast time scale and keep local voltages stable. At a slower level, however, the network must also decide who should provide how much power. We therefore studied secondary control strategies [5] that adjust voltage references while accounting for actuator limitations, electrical losses and the changing total load. We compared centralized, distributed, leader-based [4], and cluster-based ways of solving this coordination problem.