Autonomous Electrical Infrastructure Scheduling for Data Centres Integrating On-Site Hydrogen Fuel Cells and Battery Energy Storage Systems

Abstract

Rising computational demand has made data centre electricity consumption a significant driver of grid stress and operational carbon intensity, motivating on-site generation and storage architectures that reduce reliance on utility electricity during peak and contingency periods. This paper synthesizes thirteen studies of data centre microgrids, hydrogen energy systems (HES), and battery energy storage systems (BESS), and proposes an autonomous electrical infrastructure scheduling framework that integrates these assets with grid and renewable energy inputs. Three themes emerge from the reviewed literature: 1) The complementarity of the duration and response-time properties between batteries and hydrogen; 2) The increasing importance of reinforcement learning and mixed-integer optimization of real-time dispatch; and 3) The on-going economic challenge of electrolyzer and fuel cell capital costs compared to conventional backup generation. The formulation of a mixed integer scheduling model is provided that includes the power balance, battery SOC, and hydrogen storage level constraints, as well as the electrolysis, fuel cell reconversion and cost of dispatch - considering degradation. Four backup configurations are compared, showing that hybrid battery–hydrogen coordination can reduce levelized operating cost by approximately thirty-five to forty-five percent relative to hydrogen-only backup while providing a longer backup duration than batteries alone can achieve. These results indicate that autonomous, forecast-driven scheduling—treating hydrogen as long-duration storage and batteries as a short-duration buffer—offers a technically viable and increasingly cost-effective route to resilient, lower-carbon data centre power infrastructure.