Recently, the fast frequency response (FFR) service by large-scale battery energy storage systems (BESSs) has been successfully proved to arrest the frequency excursion during an unexpected power outage. However, the adequate frequency response relies on the proper evaluation of the contingency reserve of BESSs. The BESS FFR reserve is commonly managed under fixed contracts, ignoring the various response characteristics of different BESSs and their coexisting interactions. This paper proposes a new methodology based on dynamic grid response and various BESS response characteristics to optimise the FFR reserves and prevent the frequency from breaching the under-frequency load shedding (UFLS) thresholds. The superiority of the proposed method is demonstrated to manage three large-scale BESSs operating simultaneously in an Australian power grid under high renewable penetration scenarios. Further, the proposed method can identify the remaining battery power and energy reserve to be safely utilised for other grid services (e.g., energy arbitrage). The results can provide valuable insights for integrating FFR into the conventional ancillary services and techno-effective management of multiple BESSs.