Location Awareness in Multi-Agent Control of Distributed Energy Resources

Abstract

The integration of Distributed Energy Resource (DER) technologies such as heat pumps, electric vehicles and small-scale generation into the electricity grid at the household level is limited by technical constraints. This work argues that location is an important aspect for the control and integration of DER and that network topology can inferred without the use of a centralised network model. It addresses DER integration challenges by presenting a novel approach that uses a decentralised multi-agent system where equipment controllers learn and use their location within the low-voltage section of the power system.

Models of electrical networks exhibiting technical constraints were developed. Through theoretical analysis and real network data collection, various sources of location data were identified and new geographical and electrical techniques were developed for deriving network topology using Global Positioning System (GPS) and 24-hour voltage logs. The multi-agent system paradigm and societal structures were examined as an approach to a multi-stakeholder domain and congregations were used as an aid to decentralisation in a non-hierarchical, non-market-based approach. Through formal description of the agent attitude INTEND2, the novel technique of Intention Transfer was applied to an agent congregation to provide an opt-in, collaborative system.

Test facilities for multi-agent systems were developed and culminated in a new embedded controller test platform that integrated a real-time dynamic electrical network simulator to provide a full-feedback system integrated with control hardware. Finally, a multi-agent control system was developed and implemented that used location data in providing demand-side response to a voltage excursion, with the goals of improving power quality, reducing generator disconnections, and deferring network reinforcement.

The resulting communicating and self-organising energy agent community, as demonstrated on a unique hardware-in-the-loop platform, provides an application model and test facility to inspire agent-based, location-aware smart grid applications across the power systems domain.