摘要
大型发电机的故障诊断对电网安全和经济运行具有重大影响,但由于实际操作中的故障数据较少,且预警指标单一无法满足汽轮发电机故障诊断的要求。因此,提出一种基于堆叠自动编码器的汽轮发电机多指标早期故障预警与诊断模型。首先,根据多种相关性方法对数据进行降维处理,采用堆叠自动编码器来学习降维后数据之间的深层映射关系,并提取出重构误差;在此基础上,建立包括参数与测点温度的静态阈值,重构误差的自适应动态阈值,温度波动差值的动态阈值和电流及有功功率上升速率阈值的多指标综合故障预警与诊断模型;然后建立故障征兆–参数关联合集,将超过阈值的数据结合实际值与预测值的残差值,完成故障的具体诊断。最后,以京能集团河北涿州电厂以及山西某热电厂350MW汽轮发电机实际数据为例,验证所提出的基于堆叠自动编码器的汽轮发电机多指标故障诊断模型的有效性,算例分析表明,所提出的基于堆叠自动编码器的汽轮发电机多指标故障诊断模型能够提前2~10h预警故障并诊断出故障类型,为汽轮发电机安全稳定运行提供保障。
The fault diagnosis of large generators imposes a significant impact on the safety and economic operation of the power grid.However,due to the lack of fault data in actual operation and the single early warning index,it cannot meet the requirement of steam turbine generator fault diagnosis.Therefore,this paper proposed a multi-index early fault warning and diagnosis model for steam turbo generators based on stacked autoencoder.First,the dimensionality reduction processing was made on the data according to a variety of correlation methods.Then,a stacked autoencoder model was applied to learn the mapping relationship between the data based on dimensionality reduction to extract the reconstruction error.A multi-index comprehensive fault warning and diagnosis model was established,including the static threshold of parameters and temperature of measuring points,the adaptive dynamic threshold of reconstruction error,the dynamic threshold of temperature fluctuation difference,and the threshold of current and active power rise rate.The data exceeding the threshold value was combined with the fault symptom parameter correlation set and the residual value of the actual value and the predicted value to complete the specific fault diagnosis.Finally,the effectiveness of the proposed multi-index fault diagnosis model is verified by the actual data of a 350MW steam turbo generator in Hebei Zhuozhou power plant of Jingneng group and a thermal power plant in Shanxi.The example analysis shows that the proposed multi-index fault diagnosis model of the steam turbo generator based on stackable autoencoder can warn the fault and diagnose the fault type 2~10 hours in advance,which guarantees the safe and stable operation of steam turbo generators.
作者
叶林
葛鸥翔
郭永红
梅东升
毛永清
王斌
路朋
戴斌华
YE Lin;GE Ouxiang;GUO Yonghong;MEI Dongsheng;MAO Yongqing;WANG Bin;LU Peng;DAI Binhua(College of Information and Electrical Engineering,China Agricultural University,Haidian District,Beijing 100083,China;Beijing Jingneng Energy Technology Research Co.,Ltd.,Chaoyang District,Beijing 100020,China;Hebei Zhuozhou Jingyuan Thermal Power Co.,Ltd.,Zhuozhou 072750,Hebei Province,China)
出处
《中国电机工程学报》
EI
CSCD
北大核心
2022年第10期3656-3669,共14页
Proceedings of the CSEE
基金
国家自然科学基金项目(51977213)
北京能源集团有限责任公司科技项目(1180KJ201902)。