A wind turbine fault identification using machine learning approach based on pigeon inspired optimizer

Wind turbines (WT) need regular maintenance, which should be carried out as soon as possible after installation. Additionally, due to unscheduled maintenance, wind farms may lose valuable time, resulting in a loss of revenue. Since the health of the turbine is constantly monitored, it is an easy task to discover faults early and schedule regular maintenance as soon as it becomes necessary. The supervisory control and data acquisition (SCADA) data used in this study was generated by an Irish wind farm. It is a well-known dataset that contains information on the performance of the turbines. The data is categorized into three main types: operational data, status data, and warning data. Feature selection is essential in developing machine learning methods. Irrelevant features in datasets have a negative impact on the algorithm's performance, and they also increase the durationof time required to train the machine learning model before it can be used. This work presents a feature selection technique based on pigeon inspired optimizer for the widely used SCADA data. The experimental findings demonstrate that the LSTM algorithmproduces the best results, with accuracy, precision, recall and F-score values of 98%, 95%, 98%, and 97%, respectively. As a result, the proposed model can be utilized to detect defects in wind turbines in the field.