Coordination of Distributed Energy Storage Under Spatial and Temporal Data Asymmetry

We consider the problem of controlling distributed storage in a network with renewable distributed generation to minimize operational cost while satisfying power quality constraints. We assume control is distributed between a global controller (GC) and local controllers (LCs) located at the nodes with storage units. Each LC has access to its most recent net load data and runs at every time step while the GC has delayed data due to smart meter infrastructure or communication network delay, hence runs less frequently. We describe three control schemes: 1) direct storage controller (DSC) in which the GC computes the storage control signals for an upcoming window and the LCs directly use these signals; 2) net load following controller (NLFC) in which the GC computes a net load profile for each node and each LC tries to track its set profile; and 3) nodal slack controller (NSC) in which the GC computes upper and lower bounds on the net load at each node and the LC optimizes the local control action constrained by these bounds. We use a radial network with real load data to compare the performance of these schemes based on arbitrage profit and maximum solar penetration relative to a perfect foresight controller. We find that NSC and NLFC increase the supported maximum solar penetration to 29% and 40%, respectively, as compared to 10% for DSC. Moreover, NSC is able to capture 90.7% of the available arbitrage profits which is significantly higher than that achievable with NLFC and DSC.