Standalone DC microgrids for EV charging stations are essential for the synergistic integration of energy storage technologies with smart grids and multi-energy complementary systems. However, stochastic dynamic EV charging loads in cold climates often overstress Battery Energy Storage Systems (BESS) integrated with dispatchable small/mini-hydro and renewables within DC microgrid, causing premature degradation of BESS. Conventional strategies struggle to coordinate these nonstationary demands effectively. This paper proposes a Two-Layer Learning Framework for storage-aware management, facilitating a low-carbon transformation of power networks. It integrates a machine learning layer for demand forecasting with a heuristic-based adaptive layer that dynamically adjusts control horizons based on real-time BESS constraints. By prioritizing renewable and hydro dispatch while restricting BESS to corrective support, the approach maintains DC-link stability and operational flexibility. Validation using Oslo’s cold-climate datasets demonstrates that the framework improves system reliability by 12.41% and sustainability by 24.65%. Notably, peak and cumulative BESS loading are reduced by 40% and 61%, respectively. Real-time hardware-in-the-loop (Typhoon HIL-604) validation confirms the framework’s efficacy for creating stable, resilient, and multi-energy standalone power infrastructures.