摘要
针对电站数据库存在数据缺失的问题,提出了改进的模糊聚类缺失值填补算法,即支持向量回归与遗传算法优化的模糊聚类填补算法(SVR-OCSFCM)。对某600MW燃煤机组运行数据用支持向量机回归算法(SVR)、模糊聚类优化补全策略(OCS-FCM)与SVR-OCSFCM三种方法分别进行单属性和多属性缺失值填补实验,实验表明:同属性缺失时,算法性能随缺失率增加而降低,相同缺失率时填补性能随缺失属性的增加而降低;SVR-OCSFCM由支持向量回归和模糊聚类算法共同约束估计值,具有较好的准确性和有效性,缺失值填补性能优于SVR和OCS-FCM算法,且对多属性缺失数据填补具有较好的填补效果。
An improved fuzzy clustering algorithm for the missing data imputation was presented for the problem of data missing in database of the power plant,which is the fuzzy clustering fill algorithm that optimized by the support vector regression and genetic algorithm（SVR-OCSFCM）. The missing data imputation of a single attribute and multiple attributes for a 600 MW coal-fired unit was tested by using the support vector regression（SVR）,optimal completion strategy of fuzzy clustering algorithm and SVR-OCSFCM. The results show that the performance of the algorithm of the same deleted attribute degrades with the increase of the deletion rate,and the performance of the algorithm at the same deletion rate degrades with the increase of the deleted attribute. SVR-OCSFCM constrains the estimation value by the support vector regression and fuzzy clustering algorithm,which makes the system safer and more efficient. The performance of SVR-OCSFCM is better than that of SVR and OCS-FCM,which has a good filling effect on the imputation of multiple attribute missing data.
作者
李建强
赵凯
潘文凯
陈星旭
李世博
LI Jianqiang ZHAO Kai PAN Wenkai CHEN Xingxu LI Shibo(School of Energy Power and Mechanical Engineering, North China Electric Power University, Baoding 071003, China)
出处
《电力科学与工程》
2017年第1期43-48,共6页
Electric Power Science and Engineering
基金
中央高校基本科研业务费专项基金(916021007)
关键词
缺失值
数据填补
电站
模糊聚类
missing data
data imputation
power station
fuzzy clustering algorithm
