Using KD-Tree and DBSCAN to clean hydropower units monitoring data

To improve the poor quality of hydropower units monitoring data, a data cleaning approach using K-Dimensional Tree (KD-Tree) and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is proposed. Firstly, KD-Tree is constructed with the input data, and then DBSCAN is used to scan and complete the clustering of the nearest neighbour samples. After the clustering of the nearest neighbour samples, the noise points of hydropower units monitoring data are separated and removed. To validate the proposed approach, 1088 pieces of hydropower units monitoring data are selected, and 100 random data are inserted at the same time interval. The highest recognition accuracy of the proposed approach is 97.78%, and the lowest time consumption is 0.007 seconds. Compared with conventional data cleaning algorithms, the data cleaning effect of the proposed approach is improved, and the time consumption is significantly reduced.