Abstract
IEC 61800-9-2 Edition 2 defines a standardised framework for evaluating losses in complete drive modules (CDMs) and power drive systems (PDSs). Although suitable for efficiency classification, the predictive accuracy of its reference assumptions for individual industrial converters remains insufficiently quantified. This paper evaluates the IEC loss model using experimental data from a commercial 37 kW two-level voltage-source frequency converter comprising a diode-front-end rectifier, DC-link and IGBT inverter. Three approaches are compared: a strict IEC implementation using the prescribed 4000 Hz switching frequency, the same model using the measured switching frequency of approximately 2400 Hz, and an adaptive-parameter IEC model incorporating measured operating-point values for switching frequency, power factor, DC-link voltage and modulation index. The strict IEC model overestimated losses, yielding an average deviation of 15.3%. Replacing only the switching-frequency assumption reduced the deviation to 6.61%, demonstrating the strong influence of this parameter. The adaptive-parameter IEC model achieved the lowest mean absolute error (31.5 W) and the highest coefficient of determination (R² = 0.979). The results show that switching frequency is a dominant source of prediction error and that limited parameter adaptation substantially improves agreement with measured converter losses.