Enhancing Wind Turbine Reliability through Intelligent Fault Prediction Techniques

Abstract

Wind turbines (WTs) stand as one of the main sources of renewable energy, playing a crucial role in achieving sustainability objectives and increasing the proportion of renewable energy in the global energy matrix. Nevertheless, WTs are often exposed to various types of stresses during their operation in external environments. This scenario negatively affects the operation of WTs, accelerating their aging and leading to critical failures. Consequently, the costs associated with operation and maintenance (O&M) actions increase, while the financial appeal of such power sources diminishes. To address these issues, WT condition monitoring techniques have become indispensable, aiming to detect failure patterns prior to the occurrence of the failure event. However, most of the papers found in literature focus on forecasting critical failures based on the detection of incipient failures. The primary drawback of this approach lies in the fact that there are no viable models that allows to infer the evolution of a incipient failure into a critical one. In this paper, a novel methodology is developed for predicting the time until critical failure occurrence. This methodology relies on simple machine learning (ML) methods that are fed with WT’s supervisory control and data acquisition (SCADA) system data, eliminating the need for complex sensor hardware. It is expected that this method will provide a valuable tool for energy companies to optimize their O&M processes.