Electricity-Purchase Optimization for PV-Storage Microgrids Based on Historical Error Scenarios and Rolling Correction

This study investigates electricity-purchase optimization for a PV-storage microgrid under changing photovoltaic output, load and prices, with storage coupling adjacent intervals. It establishes a unified ten-minute physical and accounting framework and compares four information settings using deterministic linear programming with an MILP cross-check, historical-error scenario programming, finite-state dynamic programming for rolling correction, and scenario programming under time-varying prices. The deterministic schedule costs 35,126.95 CNY, 26.90% below the no-storage baseline. Over 334 days, the historical-error scenario setting costs 14,731,085.32 CNY. Rolling updates at 06:00, 12:00, and 18:00 reduce cost from 16,143,244.46 to 13,905,968.07 CNY (13.86%) and emergency purchases by 81.55%. Under variable prices, intraday updating cuts emergency purchases by 38.69% but raises total cost slightly (0.06%). These results show that the shared framework supports physically consistent comparisons, rolling feedback can mitigate forecast errors, and price updates should be assessed alongside adjustment costs and storage state.