关键词: 风电机组, 长短期记忆网络, 风险态势, 聚类, 数据采集与监控

Abstract: In view of the difficulty of accurately reflecting the state of wind turbines in the traditional static evaluation of wind turbines,by considering the historical record,current state and operation trend of the wind turbine Supervisory Control and Data Acquisition(SCADA) system and other multi-period information,the situation prediction of wind turbines was conducted based on risk theory in the absence of support from the historical fault data.The long short-term memory network was used to construct the short-term prediction model of active power,and the prediction residual of active power was used to quantify risk state seriousness.The fuzzy C-means algorithm was used to construct the outlier model of unit risk state severity and to classify the risk state severity level.The current risk state severity level and potential trend of wind turbines were predicted based on the Markov chain state transition model.The 2 MW wind turbines of a wind farm in Xinjiang were taken as an example to verify the rationality and validity of the proposed method.

Key words: wind turbine, long short-term memory network, risk situation, clustering, supervisory control and data acquisition