Pre-Filtering SCADA Data for Enhanced Machine Learning-Based Multivariate Power Estimation in Wind Turbines DOI Creative Commons
Bubin Wang, Bin Zhou, Denghao Zhu

и другие.

Journal of Marine Science and Engineering, Год журнала: 2025, Номер 13(3), С. 410 - 410

Опубликована: Фев. 22, 2025

Data generated during the shutdown or start-up processes of wind turbines, particularly in complex conditions such as offshore environments, often accumulate low-wind-speed region, leading to reduced multivariate power estimation accuracy. Therefore, developing efficient filtering methods is crucial improving data quality and model performance. This paper proposes a novel method that integrates control strategies variable-speed, variable-pitch maximum-power point tracking (MPPT) pitch angle control, with statistical distribution characteristics derived from supervisory acquisition (SCADA). First, thresholds for rotor speed are determined based on SCADA distribution, effect visualized. Subsequently, sliding window technique employed secondary confirmation potential outliers, enabling further anomaly detection (AD). Finally, performance validated using two turbine datasets machine learning algorithms, results compared without filtering. The demonstrate proposed significantly enhances accuracy estimation, proving its effectiveness turbines operating diverse environments.

Язык: Английский

A Multivariate Machine Learning Approach for the Prediction of Wind Turbine Blade Structural Dynamics DOI Creative Commons
Amr Ismaiel

Applied System Innovation, Год журнала: 2025, Номер 8(1), С. 12 - 12

Опубликована: Янв. 16, 2025

Wind turbine blade structural dynamics are crucial in the design phase. Blade deflections and loads can affect weight of rotor as well power performance a wind if extremely high. Predictions turbine’s lead to informative decisions on optimizing blade. In this work, multivariate machine learning (ML) approach is used predict blade’s based flow conditions control actions turbine. Three different datasets were generated using OpenFAST software tool for three turbulence classes. Various ML algorithms trained at tip root edgewise flapwise directions. The models tested generalization model conditions. A one dataset with classes then outputs other two datasets. random forest algorithm gave best accuracy predicting it was for, predictions found be higher direction both load deflection outputs. direction, could data an around 99% over 75%. While only prediction above 95% all

Язык: Английский

Процитировано

0
Pre-Filtering SCADA Data for Enhanced Machine Learning-Based Multivariate Power Estimation in Wind Turbines DOI Creative Commons
Bubin Wang, Bin Zhou, Denghao Zhu

и другие.

Journal of Marine Science and Engineering, Год журнала: 2025, Номер 13(3), С. 410 - 410

Опубликована: Фев. 22, 2025

Data generated during the shutdown or start-up processes of wind turbines, particularly in complex conditions such as offshore environments, often accumulate low-wind-speed region, leading to reduced multivariate power estimation accuracy. Therefore, developing efficient filtering methods is crucial improving data quality and model performance. This paper proposes a novel method that integrates control strategies variable-speed, variable-pitch maximum-power point tracking (MPPT) pitch angle control, with statistical distribution characteristics derived from supervisory acquisition (SCADA). First, thresholds for rotor speed are determined based on SCADA distribution, effect visualized. Subsequently, sliding window technique employed secondary confirmation potential outliers, enabling further anomaly detection (AD). Finally, performance validated using two turbine datasets machine learning algorithms, results compared without filtering. The demonstrate proposed significantly enhances accuracy estimation, proving its effectiveness turbines operating diverse environments.

Язык: Английский

Процитировано

0