摘要：针对目前风力发电机组（Wind Turbine，WT）噪声音调评估方法的不足，提出了相应的改进方法。首先，采用Gabor阶比分量提取技术从WT运行状态下采集的噪声数据中，提取出与WT本身音调相关的噪声信号，消除了用停机状态下背景噪声修正开机状态下的WT噪声所造成的误差；然后，利用基于转速的最优分割法将一分钟WT噪声信号划分为尽可能平稳的子段信号，避免了由非平稳信号FFT变换引起的频率混叠现象；最后，采用基于谱平面的音调判定法代替IEC标准中基于心理声学模型的音调估计法，消除了初始音调预测的误差并简化了音调辨识的过程。新疆达坂城的试验研究表明，提出的音调改进技术能够消除IEC中音调辨识方法的不确定性和不一致性，在WT的音调噪声评估中具有良好的应用前景。

Abstract: To overcome the shortages of IEC method of wind turbine noise tonality evaluating, an improved methodology is proposed. Firstly, the tonal signal generated from wind turbine (WT) is extracted from the captured noises in operating conditions, taking use of Gabor order component exaction technology. It is beneficial to eliminate the tonality evaluating errors caused by using background masking noises level of shutdown WT to correct that of running WT. Secondly, one minute WT noise was divided into some relatively stationary sub-segments, in order to suppress the frequency aliasing phenomenon in FFT analysis of non-stationary signal. Thirdly, the tonality identification method based on spectral flatness measure replaces that based on psychoacoustic model in the IEC standard. This replacement can eliminate the deviation brought by forecast of initial tone frequency and simplify the procedure of tonality identification to a great degree. The experiments in Danbancheng of Xinjiang province have verified that the improved methodology is able to eliminate the uncertainty and inconsistency of WT noise tonality evaluating with the IEC method, and has a good prospect in the application of WT noise tonality assessment.