基于加权AAKR算法的发电设备状态预警技术研究  

Study of Power Equipment Condition Early Warning Technology based on Weighted AAKR Algorithm

作　　者：刘双白[1] 朱龙飞 仇晓智 周卫庆 LIU Shuang-bai;ZHU Long-fei;QIU Xiao-zhi;ZHOU Wei-qing(North China Electric Power Research Institute Co.Ltd.,Beijing 100045,China;Beijing Jingneng Power Co.Ltd.,Beijing 100025,China;School of Energy and Power Engineering,Nanjing Institute of Technology,Nanjing 211167,China)

机构地区：[1]华北电力科学研究院有限责任公司,北京100045 [2]北京京能电力股份有限公司,北京100025 [3]南京工程学院能源与动力工程学院,江苏南京211167

出　　处：《热能动力工程》2020年第7期235-241,共7页Journal of Engineering for Thermal Energy and Power

摘　　要：针对发电设备的状态监测问题,引入加权AAKR算法建立了设备的状态估计模型,通过四重交叉验证学习机制对模型进行优化,并提出了完整的实时状态预警方案。为提高历史状态矩阵的信息量,提出了主参数等间隔划分结合多参数聚类的历史存储矩阵构建方法。以某600 MW机组的送、引风机为例,采集实际运行数据进行验证计算,结果表明:基于加权AAKR算法的状态估计模型在正常状态下能够对监测参数进行准确的估计,在部分测点失效的情况下模型不仅能够及时给出预警信号,而且能够抑制异常值的残差污染,维持稳定的估计跟踪能力。For the condition monitoring for power generation equipment,the state estimation model for the equipment is established by introducing weighted AAKR algorithm which is optimized through a four-fold cross-validation learning mechanism,and a real-time status early warning scheme is also proposed.In order to increase the amount of information of the historical state matrix,a method for constructing historical storage matrix combing equal interval partitioning of main parameters with multi-parameter clustering is proposed.With a supply fan and induced draft fan of a 600 MW unit as an example,the real operation data of monitoring parameters was collected for verification calculation.The results show that the state estimation model based on weighted AAKR algorithm has the ability to accurately estimate monitoring parameters under normal conditions.In the case of partial measurement failure,the model can not only provide early warning signals in time,but also suppress the residual pollution of abnormal values and maintain stable estimation tracking.

关 键 词：状态预警 AAKR 状态建模 机组设备 状态矩阵 

分 类 号：TM621[电气工程—电力系统及自动化]