Degradation-Aware Digital Shadow for Condition Monitoring of PV–BESS-Supported Cold Ironing in Smart Seaports

Abstract

Cold Ironing can substantially reduce emissions from berthed vessels, but it transfers large and highly variable electrical loads to port-side photovoltaic and battery assets whose condition changes during operation. Static monitoring references may therefore interpret normal ageing or environmentally induced losses as abnormal behaviour. This study develops a degradation-aware Digital Shadow for the condition monitoring of photovoltaic and Battery Energy Storage System assets supporting Cold Ironing. Hourly physics-based models with dynamic photovoltaic soiling and ageing, battery State of Charge, Equivalent Full Cycles and capacity fade define an evolving expected response, which is coupled with residual thresholds, data-quality checks and a three-hour persistence criterion. The framework is evaluated over a five-year simulation of the Port of Ancona with synthetic measurements. When the simulated plant and the Digital Shadow share the same models, the degradation-aware reference reduces false-positive rates from 4.31% to 0.364% for photovoltaic generation and from 0.204% to 0.077% for the battery. Across 200 simulated plants with independent parameter errors, the reduction persists but is smaller (median 5.04% versus 0.63% for photovoltaic generation) and is most sensitive to errors in the soiling model. At 20% underperformance, recall reaches 66.7% for photovoltaic generation and 67.2% for the battery; F1-scores improve mainly because false indications decrease rather than because recall increases. Measurement bias shorter than three hours and data losses of up to eight hours generate no degradation-aware indication, whereas longer bias is indicated in the same way as an asset deviation. The framework provides an interpretable condition-screening basis for maintenance prioritization in electrified smart ports, although the results are simulation-based and do not establish field accuracy or distinguish sensor faults from asset faults.