Machine Learning-Based Tools for Wind Turbine Acoustic Monitoring

Giuseppe Ciaburro,Amelia Trematerra,Virginia Puyana-Romero,Gino Iannace

doi:10.3390/app11146488

Abstract

The identification and separation of sound sources has always been a difficult problem for acoustic technicians to tackle. This is due to the considerable complexity of a sound that is made up of many contributions at different frequencies. Each sound has a specific frequency spectrum, but when many sounds overlap it becomes difficult to discriminate between the different contributions. In this case, it can be extremely useful to have a tool that is capable of identifying the operating conditions of an acoustic source. In this study, measurements were made of the noise emitted by a wind turbine in the vicinity of a sensitive receptor. To identify the operating conditions of the wind turbine, average spectral levels in one-third octave bands were used. A model based on a support vector machine (SVM) was developed for the detection of the operating conditions of the wind turbine; then a model based on an artificial neural network was used to compare the performance of both models. The high precision returned by the simulation models supports the adoption of these tools as a support for the acoustic characterization of noise in environments close to wind turbines.

Highlights

IntroductionThe global energy supply has been characterized by the development of various technologies defined as renewable, which make it possible to obtain energy in a clean and inexhaustible way by exploiting the wind, solar, hydroelectric, and geothermal resources present on Earth
The different operating conditions have been highlighted in the following paragraphs to better appreciate the differences in the measured levels from a visual point of view
The operating conditions with the turbines off, returned the following values: 24.9, 40.7, 50.8, and 36.7 dBA. These results show that the extreme operating conditions recorded noise values conditioned by the wind speed, demonstrating that the contribution of the wind to the background noise became significant under certain circumstances

Summary

Introduction

The global energy supply has been characterized by the development of various technologies defined as renewable, which make it possible to obtain energy in a clean and inexhaustible way by exploiting the wind, solar, hydroelectric, and geothermal resources present on Earth. The term “renewable energies” refers to energies that can be regenerated in a short time compared to human history [1]. The sources of these forms of energy are called renewable energy resources. Wind has represented a form of local energy that can be transformed, with examples such as mills and sailing boats illustrating that humans had already recognized its great significance in 2000 BC [2,3]

Methods

Results

Conclusion