Abstract
Grid-connected energy communities on weak or capacity-limited feeders increasingly consider hybrid PVbattery-hydrogen systems, yet planning studies often size assets against annual balances while leaving the grid connection as an effectively unconstrained buffer. Here, we develop a boundary-centred planning architecture that couples long-horizon investment decisions to schedule-feasible operation. A genetic algorithm selects component capacities, inverter rating, and contracted import/export power, while a rolling-horizon MILP (model predictive control) resolves tariff-coupled dispatch with bounded daily load shifting. Using a public-sector community on the island of Formentera (Spain), we benchmark reactive load-following against predictive operation under 0 to 8% flexible demand. Predictive coordination reshapes both, sizing and energy flows by converting temporal foresight into "capacity value" at the point of common coupling. Relative to load-following, the optimiser tightens the contracted exchange envelope from ±78 kW to ±37 kW and reduces the inverter rating from 120 kW to 77 kW, while increasing PV utilisation and reducing grid import reliance to roughly 9-11% of demand. These gains are not driven by indiscriminate oversizing, but by differentiated storage roles: batteries absorb diurnal variability and preserve headroom, while hydrogen buffers longer-duration mismatches. Economically, modest flexibility amplifies the effect, reducing net present cost by 9% and lowering levelised electricity cost from 0.243 to 0.218 €/kWh. Together, the results show that hybrid community value is anchored in enforceable boundary control, and that limited behavioural adaptation can substitute for contracted capacity in weak-grid planning.