摘要
To ensure the efficient operation and timely maintenance of wind turbines, thereby enhancing energy security, it is critical to monitor the operational status of wind turbines and promptly identify abnormal conditions. This process relies on data collected over time by turbine sensors, including measurements such as current, voltage, temperature, and vibration signals. However, in practical applications, data from normal and abnormal conditions often exhibit an imbalance in quantity, posing challenges to traditional anomaly detection methods. Additionally, sensor data inherently contains temporal information, making the effective extraction of time-dependent features another key challenge. To address these issues, this paper proposes an anomaly detection method for wind turbine operations based on an improved Auxiliary Classifier Generative Adversarial Network. The proposed approach first employs the latent features of the training samples to augment the dataset and subsequently utilizes a Long Short-Term Memory network discriminator to extract temporal features from the samples for classification. This process directly outputs the anomaly detection results for test samples. To validate the effectiveness of the proposed method, this study uses a wind turbine blade icing dataset obtained from a Supervisory Control and Data Acquisition system. The proposed method is compared with other commonly used anomaly detection approaches. The validation and comparison results demonstrate that the proposed method achieves the lowest false alarm and missed detection rates on the blade icing dataset, underscoring its superior performance in wind turbine anomaly detection.
To ensure the efficient operation and timely maintenance of wind turbines, thereby enhancing energy security, it is critical to monitor the operational status of wind turbines and promptly identify abnormal conditions. This process relies on data collected over time by turbine sensors, including measurements such as current, voltage, temperature, and vibration signals. However, in practical applications, data from normal and abnormal conditions often exhibit an imbalance in quantity, posing challenges to traditional anomaly detection methods. Additionally, sensor data inherently contains temporal information, making the effective extraction of time-dependent features another key challenge. To address these issues, this paper proposes an anomaly detection method for wind turbine operations based on an improved Auxiliary Classifier Generative Adversarial Network. The proposed approach first employs the latent features of the training samples to augment the dataset and subsequently utilizes a Long Short-Term Memory network discriminator to extract temporal features from the samples for classification. This process directly outputs the anomaly detection results for test samples. To validate the effectiveness of the proposed method, this study uses a wind turbine blade icing dataset obtained from a Supervisory Control and Data Acquisition system. The proposed method is compared with other commonly used anomaly detection approaches. The validation and comparison results demonstrate that the proposed method achieves the lowest false alarm and missed detection rates on the blade icing dataset, underscoring its superior performance in wind turbine anomaly detection.
作者
Xiangyan Meng
Jiyu Zeng
Zuquan Zhang
Peng Luo
Lin Yang
Xiangyan Meng;Jiyu Zeng;Zuquan Zhang;Peng Luo;Lin Yang(Hunan Provincial Key Laboratory of Health Maintenance for Mechanical Equipment, Hunan University of Science and Technology, Xiangtan, China;Collaborative Innovation Centre for Rail Transport, Hunan University of Technology, Zhuzhou, China;NO. 95072 Unit of PLA, Nanning, China)
