Researchers from the School of Electronics and Computer Science (ECS) at the University of Southampton have devised a novel method for forming virtual power plants (VPPs) to provide renewable energy production in the UK. Small and distributed energy resources (DERs), such as wind farms and solar panels, have been appearing in greater numbers in the electricity supply network (Grid).
To ensure that energy demand is met without interruptions, the Grid requires power suppliers to provide an estimate of their production and the confidence in meeting it. This allows the Grid to choose the appropriate number of conventional generators needed to produce and supply energy whenever it is needed. The uncertainty and uncontrollability of renewable energy sources prevents individual DERs from profitably dealing with the Grid directly.
VPPs are fast emerging as a suitable means of integrating DERs into the Grid. They are formed via the aggregation of a large number of such DERs, enabling them to reach similar size and supply reliability as conventional power plants. The ECS researchers promote the formation of such ‘cooperative’ VPPs (CVPPs) using intelligent and multi-agent software systems. In particular, they designed a payment mechanism that encourages DERs to join CVPPs with large overall production.
By using a mathematical technique called proper scoring rules - a measure of the performance of an entity, be it person or machine, which repeatedly makes decisions under uncertainty) - intelligent software agents, representing the individual DERs, are motivated to report accurate estimates of their electricity production.
The researchers devised a scoring rules-based payment mechanism that motivates the provision of accurate predictions from the CVPPs - and in turn, the member DERs - which aids in the planning of the supply schedule at the Grid. The mechanism guarantees that DERs are rewarded for providing estimates that are both accurate and have a high confidence, ensuring that software agents are given credit for high probability estimates that are close to the realized ones.