中文摘要：

时间序列的相似性挖掘是水电机组故障诊断的重要方法,本文提出一种基于频率模糊贴近度的时间序列相似性的数据挖掘方法,用来解决水电机组故障诊断中振摆特征曲线的相似性比较问题。该方法将复杂的时域问题转化为频域问题,通过模糊贴近来度量时间序列之间的距离,刻画出数据时间序列的相似程度。该算法应用到大峡水电站二号机组的故障诊断中,结果表明,该方法能够对故障做出准确判别,分离各种故障类型。由于需要存储的数据比较少,速度快,非常适于水电机组故障诊断中大规模图形序列挖掘。

英文摘要：

Time-series similarity mining is an important method for vibration fault diagnosis of hydropower units.In this paper,a new data mining algorithm based on the frequency fuzzy lattice close-degree is presented to compare the similarity of two vibration characteristic curves.This complicated problem in time domain is transformed into frequency domain.The distance of two time series is measured by the fuzzy lattice close-degree and it is used to describe the similarity degree of the series.This algorithm is applied to fault diagnosis for unit No.2 of the Daxia hydropower station.Results show that this method can help accurately identifying the faults and their categories.Since the data storage needed is relatively small and fast,the method is especially suited to large-scale graphic series mining of faults diagnosis for hydropower units.