LAMP

Learning-based Autonomic Management of Power Distribution

As society demands more renewable forms of energy and the management of energy grids must take place in the face of the emergence of multiple micro-producers, there is a growing need for autonomic management of power distribution.

LAMP will apply reinforcement learning techniques to the coordination of agent-managed micro-grids. In particular, LAMP will apply autonomic management techniques for inter-agent collaboration at multiple levels of the power distribution hierarchy.

A specific research focus of LAMP is the understanding of the limits of self-organising and learning based-techniques in an application domain subject to widely varying supply and demand and in which guaranteed quality-of-service agreements must still be maintained. As such, LAMP will draw significantly on Lero’s expertise in both self-organising systems and formal analysis.

Overall Objective

The principal objective of this work programme is to investigate the design of algorithms to support effective autonomic/self management of community-based micro grids that are capable of offering specific quality-of-service guarantees.

How will the question be addressed?

This work programme will investigate the use of decentralised, agent-based autonomic management techniques for optimisation of community-based micro grids. In particular, the work programme will take as its starting point the use of Distributed W-Learning (DWL) [1] (previously studied in urban traffic control scenarios) for the management of and undertake a simulation study of a community micro-grid.

The work programme will design extensions or adaptations to DWL to address issues arising from the study of DWL in community-based micro grids. For example, extensions may be needed to improve scalability or responsiveness, to react to changes in the available sensor/actuator infrastructure, or the network topology and, more generally, to ensure that bounds on performance can be achieved.

This work programme will be carried out in collaboration with Intel Labs Europe as part of a joint investigation of the use of agent-based techniques for management of micro grids.

References

[1] Ivana Dusparic and Vinny Cahill. Distributed W-Learning: Multi-policy optimization in self-organizing systems. In Proceedings of the 3rd IEEE International Conference on SelfAdaptive and Self-Organizing Systems (SASO 2009), pages 20-29, IEEE, September 2009.