Abstract: Weak-grid regions generally suffer from fragile backbone power grids, congested power transmission channels, and conflicting constraints between local clean energy consumption and power supply stability. To address these problems, this paper constructs a hybrid hydro-wind-solar-hydrogen collaborative system, in which hydropower serves as a continuous regulation resource and green hydrogen acts as a long-cycle energy carrier. Seasonal-trend decomposition and variational mode decomposition are employed to extract the spectral characteristics of source-load unbalanced power. Based on the known operating states of hydrogen energy storage, the scale of mixed-integer programming is effectively reduced to improve computational efficiency. On this basis, a bi-level capacity optimization configuration model is established with the objective of minimizing the full-life-cycle comprehensive cost. The outer layer optimizes the capacity of system equipment, while the inner layer verifies the annual operational trajectory and the 168-hour continuous off-grid autonomous operation performance. Under identical hydrological, meteorological, load, and installed power capacity conditions, three typical schemes are compared, including a three-level electricity-hydrogen structure with hydropower merely as an energy supplier, a four-level structure without feature extraction, and a four-level structure with pre-determined hydrogen operation strategies. Simulation results demonstrate that the introduction of the hydropower low-frequency regulation channel reduces the seasonal hydrogen energy storage capacity from 420 MW·h to 280 MW·h and improves the system self-balance rate to 0.88. Compared with the conventional three-level structure, the proposed scheme decreases the comprehensive cost by 22% and shortens the relative solution time to 0.35. Moreover, during the 168-hour extreme low-power-output window, the reservoir capacity, state of charge of energy storage, and state of hydrogen storage remain within safe operation boundaries. The proposed source-storage collaborative optimization method can significantly enhance the clean energy accommodation capability and operational autonomy of weak-grid systems, providing a theoretical basis and technical reference for the planning and design of integrated hydro-wind-solar-hydrogen energy systems in remote weak-grid areas.