基于DCS的汽轮发电机转子绕组匝间绝缘早期故障预警研究  

Research on early fault warning of interturn insulation in turbogenerator rotor windings based on DCS

作　　者：刘君 王晋宏 凌旺 刘文伟 牛志强 LIU Jun;WANG Jinhong;LING Wang;LIU Wenwei;NIU Zhiqiang(Shanxi Jingyu Power Generation Co.,Ltd.,Shuozhou 037200,China)

机构地区：[1]山西京玉发电有限责任公司,山西朔州037200

出　　处：《中国高新科技》2023年第18期41-43,共3页

基　　金：北京能源集团有限责任公司科技计划项目“发电机转子匝间绝缘在线监测与故障预警关键技术及其系统应用”(DL202207)。

摘　　要：文章针对大型汽轮发电机转子绕组匝间绝缘早期缺陷预警问题开展研究,提出了一种汽轮发电机转子绕组绝缘状态评估与缺陷预警模型。首先基于机组DCS系统实时监控数据采用多层感知机(Multi-layer Perceptron,MLP)构建发电机励磁电流预测模型,为提高模型泛化能力和预测精度,通过双MLP抽取DCS实时监控数据的时间和电磁空间特征,并结合残差连接、层标准化等技术实现时空信息的融合。考虑到DCS数据噪声较多易导致预测残差波动较大,采用Savitzky-Golay滤波分析法对预测结果进行平滑处理,提高模型的稳定性。最后,基于残差越限法对转子绕组的匝间绝缘早期故障进行预警。实例分析结果表明,所提出的模型具有较高的灵敏度和准确性。The paper conducts research on the early warning of inter turn insulation defects in large turbogenerator rotor windings,and proposes a model for evaluating the insulation status and defect warning of turbogenerator rotor windings.Firstly,based on the real-time data of the DCS,a generator excitation current prediction model is given based on multi-layer perceptron(MLP),dual MLP structure is used to extract the temporal and electromagnetic spatial characteristics of the DCS data to improve the model’s generalization ability and prediction accuracy,and techniques such as residual connection and layer standardization are combined to achieve the fusion of spatiotemporal information.Considering that the high level of DCS data noise can easily lead to significantfluctuations in prediction residuals,the Savitzky Golayfiltering analysis method is also used to smooth the prediction results and improve the stability of the model.Finally,based on the residual exceeding limit method,early warning of inter turn insulation faults in rotor windings is given.The example analysis results show that the proposed model has high sensitivity and accuracy.

关 键 词：DCS数据 多层感知器 转子绕组匝间短路 故障预警 

分 类 号：TM311[电气工程—电机]