Evaluating exponential power coefficient models for wind turbines: accuracy, complexity, and system relevance

Abstract

Abstract This work presents a calibration and qualification framework for exponential aerodynamic power coefficient models, designed to enhance predictive accuracy and interpretability across diverse wind turbine control regimes. Unlike conventional aerodynamic fitting, the proposed framework integrates nonlinear regression, model selection using the corrected Akaike Information Criterion (AICc), and residual-based statistical validation to systematically evaluate model parsimony and reliability. The methodology is applied to three turbine configurations representing different operational strategies: a constant-speed pitch-regulated turbine (MOD-2), a variable-speed pitch-regulated turbine (NREL 5 MW), and a stall-regulated turbine (NREL Phase VI). Results show that while conventional exponential forms adequately capture modern variable-speed machines, significant trade-offs emerge in stall-regulated regimes between aerodynamic fidelity and rotor-level power accuracy. This divergence underscores the need to reconsider aerodynamic model selection in hybrid-performance applications such as wind farm digital twins and predictive maintenance. By linking aerodynamic modeling with system-level integration, the framework establishes a statistically grounded foundation for scalable wind energy forecasting tools. The proposed approach supports robust energy yield estimation, facilitates the digitalization of wind infrastructure, and advances the development of control strategies for a decarbonized and distributed power grid.