Early fault detection of wind turbine yaw misalignment using Gaussian Mixture Copula Models on SCADA-derived power curves

Effective performance assessment and early detection of wind turbine failures are critical for maximizing energy generation and minimizing maintenance costs. In response to these challenges, this study proposes a novel approach for detecting yaw misalignment faults in wind turbines using Gaussian Copula, a statistical model. Yaw misalignment, a prevalent fault in wind turbines, typically results in a significant loss of power output due to yaw control errors. The proposed method leverages the power curve constructed from SCADA data, which delineates the relationship between wind speed and power output. By comparing this power generation data with reference data, variations in power output attributable to yaw errors can be detected. Significantly, the degree of yaw error correlates with the height level of data points on a Gaussian Copula. When the degree of yaw error surpasses a predefined threshold, the method flags operational anomalies, thus enabling real-time condition monitoring of wind turbines. This approach is not only robust compared to traditional methods but also boasts rapid detection capabilities. Specifically, the proposed algorithm can identify faults in the yaw drive as early as three hours after the initial anomalous fault, a response time that is markedly faster than many currently prevalent models. Consequently, this model facilitates early fault detection, thereby enhancing the decision-making process for turbine operators.