Gallium Nitride high electron mobility transistors (GaN HEMTs) enable control circuits operating in extreme environments, but their bias point instability induced by thermal drift degrades system robustness. This paper proposes a temperature-adaptive empirical model to address this challenge. The im-proved Angelov formulation incorporates parameters η and β to correct self-heating effect-induced errors in knee voltage transitions and high-current regions, achieving 46.58% RMSE reduction and exceeding 99.75% fitting ac-curacy. Temperature parameterization extends the model to a tri-variable (Vgs, Vds, T) framework validated from 25 °C to 300 °C. By reconstructing full I-V characteristics at arbitrary temperatures, the model enables dynamic bias point optimization for control circuits, with experimental verification showing less than 8.2% deviation in saturation regions. This approach provides a foundation for reliable decision-making in high-temperature aerospace and industrial systems.