Novel Machine-Learning-Based Stall Delay Correction Model for Improving Blade Element Momentum Analysis in Wind Turbine Performance Prediction

Abstract

Wind turbine blades experience excessive load due to inaccuracies in the prediction of aerodynamic loads by conventional methods during design, leading to structural failure. The blade element momentum (BEM) method is possibly the oldest and best-known design tool for evaluating the aerodynamic performance of wind turbine blades due to its simplicity and short processing time. As the turbine rotates, the aerofoil lift coefficient enhances, notably in the rotor’s inboard section, relative to the value predicted by 2D experimentation or computational fluid dynamics (CFD) for the identical angle of attack; this is induced by centrifugal pumping action and the Coriolis force, thus delaying the occurrence of stall. This rotational effect is regarded as having a significant influence on the rotor blade’s aerodynamic performance, which the BEM method does not capture, as it depends on 2D aerofoil characteristics. Correction models derived from the traditional hard computing mathematical method are used in the BEM predictions to take into account stall delay. Unfortunately, it has been observed from the earlier literature that these models either utterly fail or inaccurately predict the enhancement in lift coefficient due to stall delay. Consequently, this paper proposes a novel stall delay correction model based on the soft computing technique known as symbolic regression for high-level precise aerodynamic performance prediction by the BEM process. In complement to the correction model for the lift coefficient, a preliminary correction model for the drag coefficient is also suggested. The model is engendered from the disparity in 3D and 2D aerofoil coefficients over the blade length for different wind speeds for the NREL Phase VI turbine. The proposed model’s accuracy is evaluated by validating the 3D aerofoil coefficients computed from the experimental results of a second wind turbine known as the MEXICO rotor.

Computational investigation and validation of new MEXICO experiment

“…The validated models developed here will be employed in the near future to test and develop the diffuser-augmented wind turbine concept. Also recently, sophisticated Machine Learning (ML) algorithms have been employed successfully to reduce uncertainties in computational predictions, and they show promising potential to even improve the existing modelling tools (Gajendran et al, 2023; Purohit et al, 2022; Syed Ahmed Kabir et al, 2022). This will also be explored in the near future.…”

Machine Learning-Based Approach to Wind Turbine Wake Prediction under Yawed Conditions

“…Wind energy, specifically in the context of wake modeling, stands as a conspicuous beneficiary of this technological influx. Although conventional ap-proaches like the Blade Element Momentum (BEM) method and Computational Fluid Dynamics (CFD) provide valuable insights, their limitations in accurately capturing complex aerodynamic behaviors or high computational cost have led researchers to consider data-driven approaches [12].…”

“…Leveraging machine learning techniques, such as symbolic regression, allows for the generation of models that can accurately predict wake behavior under various operating conditions, including yaw misalignment [11]. Particularly, symbolic regression provides the added advantage of producing transparent and interpretable models, a significant departure from the 'black-box' nature of many conventional machine learning methods [12]. Consequently, the development and refinement of machine learning-based symbolic regression models for predicting yawed wind turbine wakes represent a promising and emerging field of research.…”

On the influence of twist and taper of HAWT blades on the rotational augmentation phenomenon: the NREL Phase VI – Phase II comparison

“…3 (bottom) the sectional distribution of the chord to radius ratio (c/r) for the two blades shows that the single sections have quite similar c/r values. The chord to radius ratio is widely recognized as an important parameter in the rotational augmentation models [2,8,15,16,17,18,20]. Indeed, most of 3D correction models are based on the chord to radius ratio.…”

“…They concluded that the lack of generality of stall delay models is mainly due to the assumptions about the physics of the centrifugal pumping. By means of a Machine-Learning-Based methodology, Kabir et al [17] implemented a new stall delay model. They found that the proposed model led to a good prediction of the lift coefficient while the drag coefficients were over-predicted.…”