Data-driven fault detection in wind turbine systems using support vector machines and artificial neural networks

Efficiency and reliability of wind turbine systems are very important in the production of clean sustainable energy sources. Malfunctioning of sensors, actuators, or other system components may affect system functionality and increase maintenance costs. This study presents a data-driven approach to fault detection in wind turbine systems using Artificial Intelligence (AI) techniques, including Support Vector Machines (SVMs) and Artificial Neural Networks (ANNs), with Random Forest (RF) included as an additional baseline for comparison. A comprehensive benchmark model of a variable-speed wind turbine system is implemented using the FAST/Simulink environment under realistic turbulent wind conditions. The key system variables, namely generator speed, generated power, and blade pitch angles, were monitored. Multiple fault scenarios are introduced to evaluate system performance. A five-fold blocked temporal cross-validation strategy was used to evaluate the generalization performance of the models while preserving the temporal ordering of the simulated data. Simulation results show that fault-detection performance depends strongly on the fault type, with SVM and ANN generally providing stronger performance than RF, while ANN shows advantages for some nonlinear and challenging fault patterns, particularly in terms of recall and robustness under noisy conditions. This research highlights the effectiveness of AI-based, data-driven approaches in enhancing fault detection capabilities and improving the reliability of wind turbine systems.