The penetration of new types of devices, such as domestic storage and electric vehicles, offers increasing flexibility on demand side. This will bring both new opportunities and challenges to the operation of power systems. The aim of this thesis is to design novel distributed control strategies for large scale coordination of flexible devices. To this end, flexible devices are modelled as self-interested rational agents that aim at minimizing their individual costs in response to the broadcast price signals. This thesis mainly consists of three parts, considering that the price signals can be designed in different forms, and that flexible devices could operate in different markets (e.g. energy markets, and integrated energy and reserve markets).

The first part presents a multi-agent framework for the coordination of large populations of micro-storage devices in energy markets, under the assumption that the electricity price is some monotone increasing function of total power demand. The second part extends the work of the first part through taking into account the topology of power networks: the proposed modelling framework envisages heterogeneous groups of loads that operate at different buses, connected by transmission lines of limited capacity. The locational marginal prices of electricity are used as price signals, which are different in general for each bus and calculated through an optimal power flow problem. In the framework of the third part, it is envisioned that micro-storage devices and electric vehicles participate in an integrated energy-reserve market, and that they can contribute to the provision of reserve by being available to reduce their power consumption. These flexible devices autonomously schedule their operation in response to two kinds of price signals - the locational marginal prices of energy and reserve.

Iterative schemes for the coordination of the flexible devices are presented in the three parts. It is proved that the proposed coordination schemes can ensure the convergence to stable market configurations, characterized as aggregative equilibria at which each device cannot further reduce its cost by unilaterally changing its power profile. Distributed implementations of these proposed control strategies are discussed, and their performance is evaluated in simulations on large scale power systems.