Enhancing wind turbine gearbox reliability: a hy-brid deep learning approach for predictive maintenance

Авторы

Аннотация

This study explores the application of deep learning models for predictive maintenance of wind turbine gearboxes. We investigate Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and a novel Hybrid model combining neural networks with Random Forest classifiers. The models are trained on engineered features derived from wind turbine operational data. Results show that the Hybrid model achieves the highest accuracy (90%) in predicting gearbox failures, outperforming standalone CNN (83.33%) and LSTM (72.22%) models. This approach demonstrates significant potential for improving wind turbine reliability, reducing maintenance costs, and optimizing renewable energy production.