Work

Bayesian Neural Network based analysis of supervised control and data acquisition of Horizontal Axis Wind Turbine for fault detection

Year: 2025

Type: article

Abstract: This paper demonstrates the Bayesian Neural Network-based Supervisory Control and Data Acquisition data for fault detection in Horizontal Axis Wind Turbines. Traditional fault detection methods, such as vibration analysis and thermal imaging, often suffer from predictive accuracy and real-time application limitations, leading to costly maintenance and decreased operational efficiency. These systems continuously monitor turbine performance, generating extensive data sets that can be used for early fault detection. However, the complexity and variability of the data present significant challenges for effective analysis, followed by prediction with minimal false alarms. Bayesian Neural Networks address these issues by incorporating uncertainty into the predictive model, providing more reliable fault detection, even in noisy environments. This paper leverages operational data from wind turbines to train and validate the neural network model, focusing on critical faults such as generator overheating and excitation errors. The proposed approach offers superior performance compared to traditional neural networks, particularly in its ability to handle uncertainties in the data. This improves the reliability of real-time fault diagnosis, resulting in more accurate predictions, reduced false positives, and decision-making for improved maintenance. Furthermore, the study compares the performance of Bayesian Neural Networks with other machine learning models, e.g., Support Vector Machines, K-Nearest Neighbors, Random Forests, and XGBoost, using evaluation metrics such as precision, recall, and F1-score. The proposed model achieved the highest accuracy, outperforming other algorithms. These results highlight the potential of Bayesian Neural Networks to revolutionize fault detection systems in the wind energy sector, offering a more reliable and scalable solution for predictive maintenance and enhancing the overall efficiency and sustainability of wind turbine operations.

Source: Wind energy and engineering research.

DOI: 10.1016/j.weer.2025.100016

Authors: Gaurav Jyoti Gogoi, Sourav Das, Arunasis Chakraborty

Institutions: Indian Institute of Technology Guwahati, Indian Institute of Technology Hyderabad

Language: English

Topic: Machine Fault Diagnosis Techniques

Subfield: Control and Systems Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Affordable and clean energy

Open Access status: gold

Estimated APC paid: $2,400