A Techno-Economic Analysis of Lithium-Ion and Vanadium Redox Flow Batteries for Behind-the-Meter Commercial/Industrial Applications with a Focus on Achievable Efficiency and Degradation Rates.

This thesis concerns vanadium redox flow batteries (VRFB), and whether their posited
advantages over the more commercially advanced lithium-ion battery (LIB) can translate
to improved economic outcomes in realistic use-cases.

The key advantage of the VRFB is increased lifetime; the energy storage medium (and
major cost component) is simply two solutions of vanadium at differing oxidation states
hence here is no scope for the myriad permanent degradation mechanisms that exist in
LIB. As such, over a project lifetime VRFB will potentially have lower economic and
environmental costs than LIB. A second posited advantage of the VRFB was the low
incremental cost of storage duration, allowing longer durations to be more cost competitive.

However, VRFB are disadvantaged by lower round-trip efficiency and a higher power
capacity cost due to the relatively complex power generating apparatus.

In this thesis, bottom up cost modelling for a state of the art VRFB predicted that
following cost reductions in LIB over the last 5 years, the cost of incremental usable
duration would now be very similar for the two technologies, negating one of the posited
benefits.

For the full cost-benefit analysis, it was hence important to rigorously define the
use-cases and resulting cycle rates. The chosen case study was a commercial/industrial
facility in South California. This region is a very promising market for stationary electrical
storage, and as such was considered an arena in which VRFB and LIB are likely to compete
in the near future.

In order to thoroughly explore the thesis two differing archetypal use-case were formulated.
In use-case A, the battery was called upon to reduce the electricity bill at the facility
by time-shifting power imports to cheaper hours, reducing the peak power consumption
each month, and generate revenue by providing spinning reserve and frequency response
to the grid operator. The objective was strictly economic; to maximise the net present
value of a ten year project.

In use-case B, the battery was deployed in conjunction with a PV array in order to
achieve self-sufficiency in power. In this case the self-sufficiency objective is in competition
with the economic objective (to minimise the levelised cost of electricity), hence a multi-objective
optimisation was used to size the battery and PV array.

An important contribution made by this thesis was the incorporation of detailed
degradation models for both VRFB and LIB. For VRFB, previous case studies had
assumed zero degradation, whereas in practice regular intervention is required to avoid
electrolyte imbalance. For LIB, similar case studies had employed models attributing all
degradation to cycling, whereas continual temperature dependent aging is also important.
The latter was modelled in this work.

A novel mixed integer-quadratic programming (MIQP) method was introduced that
allowed the VRFB operation to be optimised while accounting for the considerable variation
in efficiency with power input/output. This is an improvement over previous VRFB case
studies where a constant efficiency is assumed. In use-case A this resulted in the discovery
of an energy saving strategy whereby the charging was performed at moderate power in
order to track the peak efficiency as closely as possible. In a further novel contribution,
this model was used to demonstrate the benefit of operating multiple VRFB modules as
an ensemble. The benefit arises when a low load must be covered, and some modules may
be idled to reduce parasitic losses.

In use-case A, it was concluded that VRFB may compete with LIB under certain
scenarios at 4 h duration, although the most profitable system is a shorter duration LIB.
Both were predicted to break even at 6 h duration when current long duration storage
incentives were included.

For use-case B, both systems were predicted to achieve a SSR of 0.95 at under
¢21.5kW−1 h−1. Although the costs overlap depending on the scenario, VRFB were
estimated to be more likely to be cheaper up to 0.9 SSR, above which reducing cycle rates
favoured LIB. This level of self-sufficiency called for a usable duration of 6 h - 7.5 h. An
important finding for project developers is hence that 6 h would be a sensible duration for
both LIB and VRFB systems as this would cover both use cases effectively.

Another novel contribution of this work to estimate the benefit of a hybrid LIB/VRFB
system, the hypothesis being that the LIB could be used to cover the less frequent high
charge/discharge power events. In use-case B this had the hypothesised effect of increasing
the LIB lifetime, but there was negligible predicted effect on the overall levelised cost of
electricity.

Lastly, a number of important findings were made relating to practical operation of
both LIB and VRFB, which should be of interest to asset owners. Firstly, in use-case A, it
is unlikely that bidding for regulation provision would be feasible alongside demand charge
reduction, as performing the former can result in a loss in the latter. Maintenance timing
was predicted to be important for VRFB in use-case A where available revenue varies
seasonally, and the capacity should be replenished prior to the peak revenue periods of
the summer months. For LIB, it was predicted that managing state of charge will prolong
life considerably in use-case B, and climactic variations across Southern California may
strongly affect lifetime in both cases.