Smart Photonic Fibre Sensing and Digital Twin for Advanced Energy Storage Systems (SPAR)

Skein provide the AI and digital backbone of SPAR. We fuse embedded photonic fibre sensing with conventional BMS signals in a time-synchronised edge-to-cloud pipeline that supports schema-validated ingestion, efficient compression, device management, and secure role-based access. A physics-informed digital twin, combining coupled electro-thermal and degradation models with learned surrogates, assimilates live data using established filters such as EKF, UKF, and particle methods. The twin delivers robust SoC, SoH, and SoP estimates, internal state inference, and quantified uncertainty. On top of this, we run prognostics and health management: Bayesian anomaly detection, change-point analysis, probabilistic remaining-life estimation, and thermal-runaway risk scoring. The twin enables prescriptive optimisation so engineers can test operating strategies, including charge windows, cooling profiles, balancing, and cell quarantine, then deploy set-points through APIs to BMS, EMS, and SCADA. Models are packaged for constrained edge hardware and for cloud fleet analytics; dashboards present condition trends, alarms, and clear explanations using feature attributions and counterfactuals. The platform is production-grade, with CI/CD for models, a registry with versioning, drift detection, automatic recalibration, full audit trails, encryption in transit and at rest, and DPIA artefacts. We design for EU AI Act readiness with risk management, monitoring, and transparency controls. Simulation- and hardware-in-the-loop interfaces allow safe validation before rollout. The result is earlier fault detection, tighter uncertainty bounds on state estimates, longer service life, and higher round-trip efficiency, delivered as an integrated AI stack that makes SPAR’s photonic sensing actionable in the field.