This thesis presents methods for the aggregation, scheduling and dispatch of
fleets of heterogeneous
storage units. We start with real-time control, where we present policies that are unambiguously
optimal, regardless of the future; in terms of time-to-failure, energy-not-served and
flexibility. In
the first instance, this optimality is achieved by restricting the operating regime: to unidirectional
operation with no cross-charging among devices. We present policies in continuous and discrete
time, both for nominal operation and recovery between events.
Once dispatch is guaranteed to maximise
flexibility, it is coupled with a matching aggregate
representation for scheduling ahead of real-time. As an exact characterisation of the capabilities
of the
fleet, this can also be used as a binary check on request feasibility, or to optimise ancillary
service specification or perform comparisons among
fleets. We then present a generic aggregation
framework based on the aforementioned scheduling and dispatch routines, for application to coupled
discharge-recovery operation. This is scalable, in the sense that it can be applied to a hierarchical
aggregation structure and incorporates a complexity-reducing toolkit using a virtual battery
fleet.
Moreover, the framework inherits maximum discharging capabilities and guaranteed feasibility of
round-trip operation, with the additional property of minimum recovery time.
Later on, we justify the restrictions to the operating regime, by showing that these achieve
optimality under the majority of ancillary services procured by system operators today. We then
relax these conditions, and instead place specific restrictions on the variation of device parameters
across the
fleet. This provides an alternative means to precipitate set-theoretic optima, and we
present a policy that achieves optimality in this new setting. An algorithm is provided, through
which the user can implement any of the presented dispatch policies in discrete time systems or
simulation studies.