3:45 PM - 4:00 PM
[MZZ40-02] Evaluating Sound Energy Loss and Frequency Characteristics in Wind Farm Using Acoustic Method
Keywords:wind power , sound, frequency
Renewable energy has been gaining attention due to recent environmental concerns and limited energy resources available. Wind power is among the source of domestically produced energy with high economic efficiency and significant power generation potential on land and offshore. However, many issues remain to be addressed for promoting wind energy generation, including noise mitigation, equipment maintenance, and energy conversion loss. This study aims to utilize sound as a valuable information source to quantify energy loss due to sound, and to investigate temporal changes in frequency characteristics of sound that help detect equipment failures.
We acquired data from large wind turbines (GE 1.5MW) No. 5 and 6 at the Hibikinada wind farm in Kitakyushu, Fukuoka Prefecture, Japan, using a sound level meter, linear PCM recorder, anemometer, thermometer, and barometer. We performed frequency analyses using conventional FFT and continuous wavelet analysis to obtain temporal changes in the detected sound frequencies. We also computed the sound energy from time-series sound pressure data, corrected wind speed, pressure, and temperature, and compared it with estimated changes in wind energy.
The wavelet analysis identified sound sources such as 240 Hz near the ground-level transformer, aerodynamic noise from the blades (500-1,500 Hz), and mechanical noise originating from the nacelle (~1,500 Hz). These detected frequency bands are consistent with those reported at a same-model turbine with a compact microphone array (Ramachandran et al., 2014). The sound energy was estimated to be less than 18.4 W, accounting for only 0.008% of wind energy. We also found that wind turbine No. 6 had a more significant energy loss at ~1500 Hz than wind turbine No.5.
The results indicate that the sound generated by wind turbines, which could be problematic as noise, is nearly negligible in terms of energy loss. Furthermore, our method can successfully acquire the temporal changes of the sound in a given frequency band generated from wind turbines. Our results also reveal that frequency characteristics can depend on measurement conditions and turbines, even for the exact model specification. We suggest that real-time detection of the operation conditions and equipment failures of wind turbines will be possible, with further continuous data acquisition using our method in various weather and site conditions. Our study will advance a better understanding of the influences of weather and environmental conditions on wind power generation as well as energy loss and equipment failure detection.
We acquired data from large wind turbines (GE 1.5MW) No. 5 and 6 at the Hibikinada wind farm in Kitakyushu, Fukuoka Prefecture, Japan, using a sound level meter, linear PCM recorder, anemometer, thermometer, and barometer. We performed frequency analyses using conventional FFT and continuous wavelet analysis to obtain temporal changes in the detected sound frequencies. We also computed the sound energy from time-series sound pressure data, corrected wind speed, pressure, and temperature, and compared it with estimated changes in wind energy.
The wavelet analysis identified sound sources such as 240 Hz near the ground-level transformer, aerodynamic noise from the blades (500-1,500 Hz), and mechanical noise originating from the nacelle (~1,500 Hz). These detected frequency bands are consistent with those reported at a same-model turbine with a compact microphone array (Ramachandran et al., 2014). The sound energy was estimated to be less than 18.4 W, accounting for only 0.008% of wind energy. We also found that wind turbine No. 6 had a more significant energy loss at ~1500 Hz than wind turbine No.5.
The results indicate that the sound generated by wind turbines, which could be problematic as noise, is nearly negligible in terms of energy loss. Furthermore, our method can successfully acquire the temporal changes of the sound in a given frequency band generated from wind turbines. Our results also reveal that frequency characteristics can depend on measurement conditions and turbines, even for the exact model specification. We suggest that real-time detection of the operation conditions and equipment failures of wind turbines will be possible, with further continuous data acquisition using our method in various weather and site conditions. Our study will advance a better understanding of the influences of weather and environmental conditions on wind power generation as well as energy loss and equipment failure detection.