The wind energy industry is growing every year, being one of the most prominent energies in the world. The reduction of operating and maintenance costs through the implementation of new condition monitoring systems is essential to ensure the competitiveness, reliability, and maintainability of wind turbines. This work introduces a novel methodology for wind turbine monitoring based on acoustic analysis of rotatory elements from the nacelle using an unmanned aerial vehicle carrying a data acquisition sensor. The approach is based on pattern recognition of acoustic signals associated with faults that modify the typical noise produced in healthy conditions. The extraction of reliable data from the acoustic dataset is performed through a signal processing methodology using wavelet transforms for the detection of patterns associated with different scenarios. The validation of the methodology is developed with a case study in an operating wind turbine combined with the simulation of rotatory faults. K-Nearest Neighbor is applied to validate the results, demonstrating the reliability and suitability of acoustic monitoring for fault detection in offshore wind turbines.