A case study investigation into the risk of fatigue in synchronous flywheel energy stores and ramifications for the design of inertia replacement systems

Flywheels are an attractive energy storage solution for many reasons; high turnaround efficiencies, long cycling lives and high “ramp-up” power rates have all been noted in the literature. Novel flywheel based hybrid energy storage systems have also been suggested by several authors which, due to the inherent partitioning of power sources in the system architecture, provide capacity for flywheels to deliver/receive energy over a comparatively large range of time scales and loading frequencies. Accommodating grid power fluctuations at the millisecond to second time scale is an ever growing problem that almost all grids undergoing de-carbonisation are facing. Synchronous flywheel energy storage systems have the attractive capability of being able to replace “real” (passively controlled) inertia with “real” inertia in a cheap and very robust manner. Flywheel design at the grid scale warrants careful consideration, as for static energy storage applications (i.e. those not used in transportation) the main driving factor is the reduction of manufacturing and material costs. It is paramount that material is used effectively, i.e. it is sufficiently stressed such that the flywheel is not oversized (and therefore expensive) while simultaneously guarding against the likelihood of catastrophic failure during service. Fatigue has the potential to be a serious life limiting mechanism due to fluctuating rotational speeds, however in depth analysis is lacking in the literature. The present work looks to quantify the severity of fatigue in flywheels which re-establish grid inertia by applying fatigue design methods (such as the rainflow cycle counting method and the generalised strain amplitude methods of Ince and Glinka for fatigue lifing) to loading scenarios that represent grid frequency fluctuations. Importantly flywheels are sized based on different limit stress criteria, thereby enabling differing levels of structural capacity usage between designs. For the realistic design cycles considered in the present work (representative of a large scale grid undergoing normal frequency fluctuations) all projected lives are extremely large, suggesting that fatigue is not a limiting factor and that any of the tested design methodologies is viable. Significant improvements in energy density and cost per unit of energy stored may however be achieved if elastic–perfectly-plastic (Tresca based) design criteria are implemented over simple strictly elastic variants. Neglecting containment costs for simplicity, improvements in energy density of ≈74% and cost per unit of energy stored of ≈290% are demonstrated to be achievable.