Research On Abnormal Detection Method Of Wind Turbine Operation State Based On Supprot Vector Machine Regression

Nowadays,the total capacity of wind turbine assembly machines in China and the running time of wind turbines continue to increase.It is self-evident that the importance of accurately assessing the operational status of wind turbines is obvious.The traditional equipment operation status is divided into normal and faults,but for a special group of wind turbine generators,there is an abnormal condition that does not meet the normal standards and does not reach the fault alarm.In this paper,the following researches are conducted on the detection method of abnormal operation status of such wind turbines:The operating conditions of wind turbines are complex and changeable.A single anomaly detection model is not sufficient to accurately evaluate the operating status of wind turbines.the wind turbine operating conditions are determined by multiple factors such as wind speed and output power.The artificial classification method is not sufficient to meet the accuracy requirements.This paper proposes an unsupervised self-organizing map clustering method to cluster the wind turbine operating data using the unsupervised self-organizing map clustering method,analyzes the wind turbine operating conditions,and then models and evaluates the wind turbine operating status.The clustered sample data of working conditions presents features of small samples and multi-dimensions.The support vector machine regression method with good learning effect for this type of data model is used to model each type of working condition class sample and to use normal operation.The state model detects abnormal data.At present,the research on how to define the abnormal operating state of the wind turbine from the data of different field detections has been separated from the historical operating data of the wind turbine,ignoring the impact of long-term operation of the equipment on the parameter changes.This paper proposes a sliding window statistical analysis method of single variable and comprehensive state for this deficiency,combining the threshold setting with the current healthy operating state of the wind turbine,which is more suitable for real-time monitoring of wind turbines.Simulation and examples are used to verify the validity of the algorithm model and successfully applied to the wind turbine abnormality detection system.Finally,the algorithm model is compared with nonlinear state estimation methods,and the self-organizing map is combined with the support vector machine regression.The detection method is more suitable for the abnormal detection of wind turbine operation status.