Principal Component Analysis of Water Pipe Flow Data

In this study the Principal Component Analysis (PCA) technique was applied to the night flow data collected from a water distribution system in South Korea. The principal components of the observed flow data were determined, and the scores of the data and the model residual errors were used to calculate the statistics T2 Hotelling and Distance to Model (DMOD) which represent severe and moderate outliers for the PCA model, respectively. The changes in the dates which have the outliers obtained from the PCA using various analysis periods were analyzed. As a result, the best flow data size for the PCA of a District Metered Area was determined.