Application of LPCF model based on ARIMA model to prediction of water quality change in water supply system

Abstract

Currently, various water quality parameters (WQPs) are monitored for real-time contamination warning (CW) in the water supply system (WSS) of South Korea. If the measured values of WQP exceed the threshold value, CWs are issued. However, the U.S. Environmental Protection Agency (EPA) reported the following limitations of the CW system based on these thresholds. First, irregular and sudden hydraulic changes in WSS caused by pump or valve malfunction may cause measurement error of the WQP sensors, which may cause nuisance and unnecessary false-positive alarms. Second, in the case of long-term outflow of micropollutants, WQPs change is slightly within the thresholds, which causes a serious monitoring error of false-negatives that cannot be detected even in actual contamination. Therefore, the U.S. EPA applied a linear prediction-correction filter (LPCF) model for real-time CW, which is based on the autoregressive (AR) model. The main purpose of this study is to develop a CW technique to be applied to WSSs in South Korea. For the development of a real-time CW technique, the LPCF model was applied with reference to previous research of the U.S. EPA. However, the time series of the WQP observed in WSSs often does not satisfy stationarity even though they are important fundamental assumptions of the AR model. Therefore, in this study, we developed an LPCF model by applying the autoregressive integrated moving average model considering nonstationary WQPs.