The fast increase in wind power production has made predictive maintenance of wind turbines a key concern. Early identification of aberrant operating circumstances helps avert failure status, from which recovery is protracted. By way of monitoring the running nation of windmills using a data acquisition (SCADA) system and supervisory control, power loss may also be reduced and preservation efficiency can be accelerated. The majority of the widely used maintenance techniques were created using data mining and statistical analysis. Nevertheless, these strategies need advanced processing methods in addition to large amounts of data. To overcome the aforementioned difficulties, this work suggests a machine learning approach that includes thorough data preparation and batch size-specific hyperparameter tuning for aberrant early detection. The proposed method is able to predict the faults correctly 89% times. Lastly, a computer-based application for the fault classification and maintenance prediction of wind turbines uses the suggested machine learning technique.