Representative-Day Selection for PV-Storage Lifespan Estimation Using Prior SOC and DOD Binning Features

Representative-day methods reduce the computational burden of energy storage dispatch and planning. Conventional clustering criteria typically preserve operating scenarios relevant to dispatch economics, whereas lifespan evaluation requires days that better represent battery degradation. Days selected only from load, photovoltaic generation, and electricity-price profiles may therefore miss cycling structures that determine degradation. We propose a representative-day selection method based on prior battery-behavior features for lifespan estimation in photovoltaic-storage systems. The method first clusters daily scenario profiles to maintain scenario coverage. Within each cluster, it constructs a capacity-aware prior state-of-charge sequence from net-load variations and applies rainflow counting to obtain depth-of-discharge bins. The method then standardizes and covariance-whitens the first- and second-order moments of this feature vector. A global moment-matching problem selects one real representative day from each cluster. Case studies using multiple annual user-side datasets show that the proposed method reduces lifespan estimation errors compared with the baseline. The method also retains a clear computational advantage over full-year daily dispatch, supporting repeated calculations for capacity searches and economic assessment.