基于IMSGP-WEDI的水电机组故障预警方法

Fault warning method of a hydropower unit based on IMSGP-WEDI

水电机组故障预警指标对于机组早期故障预警时间影响较大,而当前预警指标多基于单传感器信息构建,且其特征信息单一,难以更全面地表征机组运行状态,针对此问题,提出了一种基于集成多传感器遗传规划(integrated multi-sensor genetic programming,IMSGP)与权重欧式距离指标(weighted euclidean distance index,WEDI)的水电机组故障预警方法。首先,将多传感器信号进行预处理,剔除干扰信息;然后从预处理后的信号中提取多元特征,构建原始预警特征集;接下来利用复合检测指数(composite detection index,CDI)进行特征选择,并利用IMSGP进行特征构造;最后结合主成分分析(principal component analysis,PCA)与欧式距离构建WEDI,判别机组异常状态。通过对水电机组实测数据的分析,证明了提出的方法可及时发现早期故障,实现故障预警。

The early warning of faults in hydropower units is greatly affected by warning indicators.However,these indicators are mostly based on single sensor signal and information,and therefore limited in their abilities to comprehensively characterize unit operating status.To address this issue,a method combining integrated multi-sensor genetic programming(IMSGP)and weighted euclidean distance index(WEDI)was proposed.First,multiple sensor signals are preprocessed to eliminate interference.Next,multivariate features are extracted from the preprocessed signals to construct the original warning feature set.Then,the composite detection index(CDI)is used for feature selection,and IMSGP is investigated for feature construction.Finally,principal component analysis(PCA)and the Euclidean distance are used to construct WEDI for identifying abnormal states of the unit.Through analysis of hydropower unit data,the ability of the proposed method on detecting early faults and achieving effective fault warning were verified.