Presentations

Abstract

Floating LiDAR systems (FLS) provide cost-effective, rapid offshore wind measurements, but platform motion introduces artificial fluctuations in horizontal wind speed (HWS) that inflate measured turbulence intensity (TI). Two correction families dominate today: machine-learning (ML) and deterministic (kinematic). ML approaches can operate on low-resolution (e.g., 10-minute) data at low run-time cost once trained, but require robust, generalisable training across diverse environmental conditions. Deterministic methods combine high-frequency motion and line-of-sight (LOS) data to remove kinematic contamination, offering physics-based performance across sites—but typically demand high-rate (≈50 Hz) streams that are costly to transmit over satellite during remote campaigns.  We review current market methodologies and introduce a pragmatic “middle-ground” deterministic approach that uses low-frequency motion data (1–4 Hz) instead of 50 Hz. This low-frequency motion correction (LF-MC) enables generalisability and near-real-time correction within realistic telemetry and power budgets.  Both LF-MC and a high-frequency motion correction (HF-MC) are applied to a six-month validation campaign. We discuss implementation challenges and mitigations from a systems-engineering perspective—time-stamping, synchronisation, resampling, LOS vector reconstruction, and QC integration. Performance is evaluated against recent Carbon Trust updates, including the multi-regression HWS KPI, and the TI KPI roadmap binned by wind speed.  Results show substantial KPI improvements when applying accessible, generalisable kinematic equations to motion-corrupted data, enabled by targeted design changes to the study FLS. LF-MC materially improves HWS and TI KPIs versus uncorrected data and approaches HF-MC performance in many bins, while HF-MC provides the strongest overall metrics when high-rate data are available for post-processing.