Forecast-Driven, Energy-Based, and Degradation-Aware Energy Management for a Photovoltaic-Integrated Hybrid UPS with a Dedicated Grid Battery Charger

This paper presents a forecast-plan/actual-replay energy management framework for a photovoltaic-integrated hybrid uninterruptible power supply. A day-ahead schedule is generated from forecasts and replayed using realized profiles. The formulation accounts for converter losses, LiFePO₄ battery degradation, reserve requirements, and terminal stored-energy constraints. Economic comparisons require terminal inventory comparability and critical validation compliance. Ten baseline 24 h scenarios cover positive and negative prices, photovoltaic forecast error, alternative objectives, and unannounced outages. Five additional cases examine prescribed PV overprediction and underprediction. Relative to a terminal-comparable grid-first UPS, system-level selected-objective cost reductions reached 15.818% in the positive-price PV case and 31.010% in the PV-rich outage case; these values do not isolate predictive scheduling or GA effects. In the five additional cases, selected-objective differences ranged from a 0.673% reduction to a 1.232% increase, despite preserved feasibility and complete load service in the primary runs. Null and control cases showed no artificial cycling. The non-compliant baseline case and additional diagnostic runs were excluded from economic inference. The results demonstrate traceable, validation-gated comparisons while showing that maintained operational feasibility does not guarantee an economic advantage under PV forecast mismatch.