Turbine Data cleansing based on conditional depth convolution generation confrontation

With the advantages of fast response speed and high processing capacity, Big data technology provides great convenience for the operation data analysis of power plant thermal equipment. A large amount of operation data will be generated during the operation of steam turbine, but data loss, discontinuity, abnormality and other situations will occur due to various reasons. Through Big data cleaning algorithm, data content can be optimized, data application accuracy can be improved, and subsequent data processing can be facilitated. In this paper, the C-DCGAN model is used to take the turbine boundary conditions as the input condition information, and the trained discriminator and generator are used to clean the Data cleansing. It can effectively eliminate the Outlier of the unit data to be cleaned and the data filling of the missing data after the value elimination. While ensuring the consistency of the data volume before and after cleaning, the data change trend of the filled complete data set is relatively flat and more consistent with the actual unit data distribution trend.