Stochastic resonance (SR) is an effective tool to enhance weak signal by utilizing noise to reach a certain synergistic effect, which has been widely studied in the field of weak signal detection. Currently, using SR to enhance the weak fault feature of wind turbine faces two challenges: First, it is difficult for SR to select the optimal system parameters, while the traditional adaptive method based on SNR needs to predict the precise frequency of the target signal. Second, the wind turbine load changes frequently, making the vibration and noise large. As a result, the traditional SR cannot effectively highlight the target fault feature by inducing a stable resonance phenomenon at the target frequency. To improve the ability of SR to enhance the weak fault feature of wind turbine under strong noise, this paper proposes an adaptive fractional SR method based on weighted correctional signal-to-noise ratio (WCSNR). Firstly, the proposed method considers the adiabatic approximation applicable condition in the SR system and combines characteristics of the expected output signal to construct the WCSNR evaluation index to quantify the system output response, so that the system can adaptively obtain optimal parameters without predicting the accurate frequency of the target signal. Then, the fractional-order theory is applied to the SR system to overcome the shortcoming that the integer-order SR cannot induce stable resonance phenomenon at the target frequency when enhancing the fault feature of wind turbine, and use WCSNR to search for the optimal fractional order to further enhance the weak fault characteristics. Simulation and engineering actual data analysis results verify the effectiveness and superiority of the proposed method in the fault feature enhancement of wind turbine. The analysis results show that compared with the traditional SR method, the method proposed in this paper can more effectively reduce the interference of background noise and accurately enhance the weak fault feature.

随机共振（Stochastic Resonance，SR）是利用噪声达到一定的协同效应来增强微弱信号的有效工具，在微弱信号检测领域得到了广泛的研究。目前，利用SR增强风机弱故障特征面临两个挑战：一是SR难以选择最优的系统参数，而传统的基于SNR的自适应方法需要预测目标信号的精确频率。二是风机负荷变化频繁，振动噪声大。因此，传统的 SR 无法通过在目标频率处诱导稳定的谐振现象来有效突出目标故障特征。为提高 SR 增强强噪声下风力机弱故障特性的能力，本文提出了一种基于加权校正信噪比（WCSNR）的自适应分数SR方法。首先，该方法考虑SR系统中绝热逼近的适用条件，结合期望输出信号的特点，构建WCSNR评价指标来量化系统输出响应，从而使系统在不预测准确频率的情况下自适应地获得最优参数。的目标信号。然后，将分数阶理论应用于SR系统，以克服整数阶SR在增强风力机故障特征时无法在目标频率诱导稳定谐振现象的缺点，并利用WCSNR寻找最优分数阶以进一步增强弱断层特性。仿真和工程实际数据分析结果验证了所提方法在风机故障特征增强中的有效性和优越性。分析结果表明，与传统的SR方法相比，本文提出的方法能更有效地降低背景噪声的干扰，准确增强弱故障特征。