Online Monitoring System for Water Quality Based on Machine Learning Algorithms

Abstract

The international attempt to develop sustainable, intelligent water delivery systems faces considerable challenges due to the urbanisation that has characterised modern cities. In urban planning, the quality of the water we use for aquaculture and we consume is becoming more crucial due to the environmental changes. The main focus of urban water quality control has historically been the physical, chemical, and biological categories of quality indicators. However, occurrences like widespread infections have increased in frequency in many big cities due to the biological indications' inevitability, worsening the threat to the fishes and public's health. We begin this task by outlining the problem at hand, going over its technical challenges, and outlining open research issues. Then, we suggest a potential solution by creating a methodology for risk assessment specific to the urban water distribution system.  Using the indicator data we acquired from industrial activity, we can track changes in water quality and spot potential threats. We provide an Adaptive Frequencies Analysis (Adp-FA) approach that uses the frequency domain data of indicators for their internal linkages and individual prediction in order to produce results that can be explained. We also examine the approach's indicator, geographic, and temporal scalability. As a part of a larger study, we use data sets of industrial quality collected in four Norwegian cities (Oslo, Bergen, Strmmen, and Alesund). We put the proposed method to the test, comparing its spectrogram, prediction precision, and time commitment to better established AI techniques like the Artificial Neural Network, Random Forest, CNN, and LSTM. The results demonstrate that our strategy outperforms alternatives in most respects. Risk prediction for water quality are possible.