基于动态邻域指标重构的风电机组故障检测方法
WIND TURBINE FAULT DETECTION METHOD BASED ON DYNAMIC NEIGHBORHOOD INDEX RECONSTRUCTION
钱小毅 1孙天贺 1姜兴宇 1王宝石1
作者信息
- 1. 沈阳工程学院,辽宁省电网节能与控制重点实验室,沈阳 110136
- 折叠
摘要
复杂多变的运行工况是造成风电机组故障检测中误报和漏报现象的主要原因之一,为此,提出一种邻域规模与阈值动态协同的风电机组故障检测方法.定义阶梯近邻和状态分离度两种重构指标,以评估模型的故障分离能力,据此构建邻域规模的迭代修正策略.提出阈值静态分量与动态分量的融合方法,通过动态邻域规模和动态阈值减少由状态突变引起的误报和漏报.利用兆瓦级风电机组常见的10种故障进行仿真实验,验证所提方法可更有效分离正常状态与异常状态,进而降低误报率和漏报率.
Abstract
The changeable working condition is one of the main causes of false and missing alarms in wind turbine fault detection.For this purpose,a fault detection method based on dynamic coordination of neighborhood scale and threshold is proposed.The reconstruction indexes of stepped neighbor and state separation degree are defined to evaluate the ability of fault separation,on this basis,the iterative correction strategy of dynamic neighborhood scale is constructed.A fusion method of static component and dynamic component for threshold is proposed to suppress the false positives and missing positives caused by the abrupt change of working condition through dynamic neighborhood size and dynamic threshold.Ten common faults of megawatt wind turbines are used in the experiments and verified the proposed method can more effectively separate the abnormal state from the normal state and reduce the false alarm rate and missing false rate.
关键词
风电机组/故障检测/重构/复杂工况/动态邻域规模/动态阈值Key words
wind turbines/fault detection/reconstruction/complex working conditions/dynamic neighborhood size/dynamic threshold引用本文复制引用
出版年
2024
NETL
NSTL
万方数据