摘要
针对风电机组的实时状态监测问题,提出了一种基于增量式相对熵的残差分析方法。首先,通过分析滑动窗口数据特点,推导了适用于实时运算的增量式相对熵的计算公式,其时间复杂度为O(1),要低于常规计算方法的O(n)。接下来,提出了一种基于数据驱动和正常行为建模的风电机组实时状态监测方法,并将增量式相对熵作为实时残差分析的指标。最后,用某2 MW风电机组的实际齿轮箱故障数据为算例,验证了该方法的有效性。结果表明,相对熵残差分析能够至少提前8~10 d实现故障预警,优于常规的统计量;增量式相对熵的计算时间仅相当于常规方法的0.4%~1.9%,在实时性上有显著优势。
Aiming at the problem of real-time condition monitoring of wind turbine,a real-time monitoring method based on incremental relative entropy was proposed.Firstly,based on the analysis of the characteristics of sliding window data,the formula of incremental relative entropy which is suitable for real-time calculation was derived.And the time complexity of incremental relative entropy is O(1),which is lower than O(n)of conventional calculation method.Next,a real-time wind turbine condition monitoring method based on data-driven and normal behavior modeling was proposed and the incremental relative entropy was used as the index of real-time residual analysis.The effectiveness of the proposed method was verified by the actual gearbox fault data of a 2 MW wind turbine.The results show that relative entropy residual analysis can realize fault warning at least 8~10 days in advance,which is better than the conventional statistics.The calculation time of incremental relative entropy was only 0.4%~1.9%of the conventional calculation method,which has significant advantages in real-time performance.
作者
王梓齐
张书瑶
刘长良
Wang Ziqi;Zhang Shuyao;Liu Changliang(School of Control and Computer Engineering,North China Electrie Power University,Baoding 071003,China;State Key Laboratory of Altermate Electrical Power System with Renewable Energy Sources,North China Electric Power University,Beijing 102206,China)
出处
《电子测量与仪器学报》
CSCD
北大核心
2020年第12期125-132,共8页
Journal of Electronic Measurement and Instrumentation
基金
北京市自然科学基金(4182061)
中央高校基本科研业务费(2020JG006,2020MS117)资助项目。
关键词
增量式相对熵
风电机组
状态监测
残差分析
incremental relative entropy
wind turbine
condition monitoring
residual analysis
