Electricity networks exhibit multiple organisational scales, for example: local or large-scale production and decentralised consumption. Since the instantaneous supply of electricity must always meet constantly changing demand, operation of an electric power system involves a complex process of forecasting production and demand.
We will apply the results obtained in Work Package 1 about large-scale control to build efficient controls for electrical networks, using a model that incorporates the following information: the activity state of a customer; the particular bus to which a customer is connected; the external state information, such as time of day and weather; and the controllability of the load, i.e. the state of the load with respect to the service curve constraint.
Using our methodology and tools, we will define and evaluate algorithms to distribute the control signals in quasi-real-time (e.g. response times < 1 sec) while taking into account:
- electrical constraints (voltage, power flows, ampacities),
- the topology of the distribution network; and
- the presence of storage within the distribution network.
We will use mean-field analysis to provide heuristics for good suboptimal policies, in particular to account for spatial aspects and for systems incorporating electric vehicles.