Fault diagnosis and forecasting contribute significantly to the reduction of operating and maintenance associated costs, as well as to improve the resilience of wind turbine systems. Different from the existing fault diagnosis approaches using monitored vibration and acoustic data from the auxiliary equipment, this research presents a novel fault diagnosis and forecasting approach underpinned by a support vector regression model using data obtained by the supervisory control and data acquisition system (SCADA) of wind turbines (WT). To operate, the extraction of fault diagnosis features is conducted by measuring SCADA parameters. After that, confidence intervals are set up to guide the fault diagnosis implemented by the support vector regression (SVR) model. With the employment of confidence intervals as the performance indicators, an SVR-based fault detecting approach is then developed. Based on the WT SCADA data and the SVR model, a fault diagnosis strategy for large-scale doubly-fed wind turbine systems is investigated. A case study including a one-year monitoring SCADA data collected from a wind farm in Southern China is employed to validate the proposed methodology and demonstrate how it works. Results indicate that the proposed strategy can support the troubleshooting of wind turbine systems with high precision and effective response.

Given the prejudicial environmental effects of fossil-fuel based energy production, renewable energy sources can contribute significantly to the sustainability of human society. As a clean, cost effective and inexhaustible renewable energy source, wind energy harvesting has found a wide application to replace conventional energy productions. However, concerns have been raised over the noise generated by turbine operating, which is helpful in fault diagnose but primarily identified for its adverse effects on the local ecosystems. Therefore, noise monitoring and separation is essential in wind turbine deployment. Recent developments in condition monitoring provide a solution for turbine noise and vibration analysis. However, the major component, aerodynamic noise is often distorted in modulation, which consequently affects the condition monitoring. This study is conducted to explore a novel approach to extract low-frequency elements from the aerodynamic noise background, and to improve the efficiency of online monitoring. A framework built on the spline envelope method and improved local mean decomposition has been developed for low-frequency noise extraction, and a case study with real near-field noises generated by a mountain-located wind turbine was employed to validate the proposed approach. Results indicate successful extractions with high resolution and efficiency. Findings of this research are also expected to further support the fault diagnosis and the improvement in condition monitoring of turbine systems.

Local oscillatory-characteristic decomposition (LOD) is a relatively new self-adaptive time-frequency analysis methodology. The method, based on local oscillatory characteristics of the signal itself uses three mathematical operations such as differential, coordinate domain transform, and piecewise linear transform to decompose the multi-component signal into a series of mono-oscillation components (MOCs), which is very suitable for processing multi-component signals. However, in the LOD method, the computational efficiency and real-time processing performance of the algorithm can be significantly improved by the use of piecewise linear transformation, but the MOC component lacks smoothness, resulting in distortion. In order to overcome the disadvantages mentioned above, the rational spline function that spline shape can be adjusted and controlled is introduced into the LOD method instead of the piecewise linear transformation, and the rational spline-local oscillatory-characteristic decomposition (RS-LOD) method is proposed in this paper. Based on the detailed illustration of the principle of RS-LOD method, the RS-LOD, LOD, and empirical mode decomposition (EMD) are compared and analyzed by simulation signals. The results show that the RS-LOD method can significantly improve the problem of poor smoothness of the MOC component in the original LOD method. Moreover, the RS-LOD method is applied to the fault feature extraction of rotating machinery for the multi-component modulation characteristics of rotating machinery fault vibration signals. The analysis results of the rolling bearing and fan gearbox fault vibration signals show that the RS-LOD method can effectively extract the fault feature of the rotating mechanical vibration signals.

The rolling bearing and gear fault features are generally shown as modulation characteristics of their vibration signals. The empirical envelope (EE) method is an accordingly common demodulation method. However, the EE method has the defects of over- and undershoot, which may lead to demodulation error. According to this, an envelope optimization algorithm -- empirical optimal envelope (EOE) is introduced into the EE method, and an improved empirical envelope (IEE) method is obtained to calculate the instantaneous amplitude and instantaneous frequency of mono-component modulation signal. Furthermore, aiming at the actual measured mechanical vibration signal has multi-component modulation feature, the IEE method is combined with an adaptive signal decomposition method -- local oscillatory characteristic decomposition (LOD) proposed by the author, thereby a new multi-component signal demodulation method based on LOD and IEE is proposed. The proposed method is compared with Hilbert transform (HT) and Teager energy operator (TEO) demodulation methods by the simulation signal and actual measured mechanical vibration signal. The results show that the demodulation effects including edge effects, negative frequency, over- and undershoot of the proposed method are significantly improved and can extract the rolling bearing and gear fault feature information clearly.