As wind and solar penetration increases, grid operators face growing challenges in managing uncertainty and maintaining balance between supply and demand. A new two-stage robust optimization framework addresses this challenge by determining economically efficient capacity mixes for distributed generation (DG), battery energy storage systems (BESS), and hydrogen storage under multiple sources of uncertainty.
The methodology first derives flexibility requirements—both upward and downward—from feasible net-load intervals derived from wind, PV, and load forecasts. These requirements are then embedded within a min-max-min planning architecture: capacity decisions are made before uncertainty manifests, an adversarial uncertainty set identifies worst-case scenarios, and operational strategies are optimized in response. This approach avoids overbuilding resources while maintaining reliability across a realistic range of operating conditions.
The solution algorithm employs column-and-constraint generation with KKT-based subproblem reformulation, enabling practical computation for large-scale problems. Validation used hourly synchronized forecast and actual data from a provincial grid in Northwest China, with uncertainty radii calibrated to empirical 95th percentile forecast errors.
Key findings show optimal capacity allocations of 2,472 MW distributed generation, 852 MW battery storage, and 446 MW hydrogen storage. Compared to conventional deterministic planning, the robust solution increases DG capacity by 13.8% and battery capacity by 11.2%, with total cost increasing 16.4%—a modest premium for enhanced reliability. Notably, this robust approach reduces costs 3.7% versus full-box robustness methods, demonstrating superior computational efficiency.
Operational patterns reveal complementary roles: batteries absorb short-duration fluctuations efficiently, while hydrogen storage provides multi-hour or longer duration support. This insight aligns with physical characteristics of each technology and suggests portfolio approaches outperform single-storage solutions in high-renewable grids. The framework offers grid planners a practical decision-support tool for capacity planning under realistic uncertainty.
