Offshore wind turbine harmonic models are increasingly required over wider frequency ranges, at finer frequency resolutions, and across more operating conditions. While these requirements aim to improve the reliability of grid studies, they do not necessarily ensure that a delivered model preserves the relevant converter behaviour or predicts quantities that are comparable with standardized power-quality measurements. This paper examines the construction and validation of frequency-domain wind turbine harmonic models from the perspective of the original equipment manufacturer.

The work combines field measurements, reprocessed emission records, analytical considerations, and high-fidelity software-in-the-loop studies using production-representative control software. Alternative methods for deriving the frequency-dependent turbine impedance and Norton-equivalent source are assessed, including analytical modelling, controlled simulation, and measurement-based identification. Particular attention is given to the distinction between the intrinsic Norton source and the measured terminal current, which also contains the turbine response to background voltage distortion.

Model validation is considered at three levels. First, impedance scans with different frequency grids are compared to determine when narrow-band control features are lost through coarse sampling or interpolation. Second, narrow-band model predictions are processed using equivalent spectral grouping, time aggregation, and percentile calculations to enable comparison with IEC 61400-21-1 power-quality reports. Third, the additional model fidelity required when extending the frequency range from 75–2500 Hz into the 0.1 Hz–9 kHz range is assessed, including low-frequency control coupling, sequence interactions, switching-related emissions, and measurement-bandwidth limitations.

The results demonstrate that model resolution, measurement resolution, and harmonic grouping are distinct concepts and should not be specified interchangeably. A 1 Hz model cannot be validated directly against grouped P95 measurements without applying a consistent measurement-equivalent observation process. The paper concludes that future TSO requirements should impose a transparent burden of validation while remaining method-neutral regarding model construction. Application-specific acceptance criteria, independent validation data, and automated quality checks can provide stronger confidence than prescriptive data density alone, while avoiding unnecessary modelling and disclosure burdens for manufacturers.