A Study on A Hybrid Water Quality Prediction Model Using Sequence to Sequence Learning Based LSTM And Machine Learning

Abstract

In the year 2022, around 99.4% of the population of South Korea is using tap water for drinking purposes. However, most of the sources are large seasonal turbidity and frequent algal blooms, which may pose a high risk to the treatment processes. Besides this, the facilities are getting aged and deteriorating day by day, raising concerns over WQ contamination and the need for stringent WQ management. The set point method is being applied in South Korea for monitoring WQ as well as pollution detection. This approach considers legally set WQ criteria for each process as threshold and sends out alerts for pollution when real-time WQ measurements exceed the threshold. This threshold base approach can trigger lots of false alarms. This study focuses on developing an efficient prediction strategy for WQ pollution active monitoring in South Korea with the help of LSTM Seq2Seq model. We further suggest hybrid WQ prediction model that integrates the LSTM Seq2Seq predictions with further machine learning models in order to improve the performance of the predictions. By comparing the short-term and long-term predictive performance of these models, we aim to develop an optimal WQ prediction model.