Smart River Water Quality and Level Monitoring: a Hybrid Neural Network Approach

River water plays an important role in every metropolitan society; it contributes significantly to the production of agricultural products and hence to the economy of a country besides offering many other benefits. Therefore, river water monitoring is necessary though difficult. The goal of this research is to create a quantitative technique for assessing the water quality state of the Indian rivers in the southern part of India. Water test samples were obtained at three distinct places along the Kaveri River for this study. The water level information was retrieved from the photos using a hybrid method (a combination of convolutional neural network and long short-term memory network) called CNN-LSMN. The level points were measured using the field camera placed in the test locations. The following six typical metrics were used to assess the water quality: turbidity, temperature, pH, TDS, conductivity, and total hardness. In this study, the water quality index (WQI) of the modified National Sanitation Foundation (NSF) was used to determine the quality of water. Furthermore, the flower pollination optimization method was used to optimise the critical water quality indicators. Standard performance metrics were used to compare the performance of the proposed approach with that of the existing techniques. Upon comparing the performance of the suggested CNN-LSMN in terms of performance measures, it was found that the detection accuracy had improved and it was 4.62%. The proposed technique in this study was found to be beneficial for precisely estimating the water level and quality of the rivers.

I. Introduction

River water quality monitoring (WQM) is a timevariability, uncertainty, and a highly non-linear problem. It has two parts, namely observed states and an online measuring state. Internet of Things (IoT) based WQM has been proposed by Hanifah and Supangkat; different sensors (turbidity sensor, rain sensor, flow sensor, salinity sensor, temperature sensors, and pH sensors) have been interfaced with a controller and k-mean clustering has been used for data classification. A comparison study has been done and the result shows the superiority of the method [1]. To recognize the location of WQM stations in the river system is proposed by Ilker et al. Water contamination has a significant impact on how different water parameters are calculated. Four most water pollutant includes trash, parasites, bacteria, and chemicals [2]. Large amount of domestic and industrial waste water is delivered to the river. The sewage system and the wastewater treatment facility comprise an interconnected wastewater system. Non-linear model is used for the biological process. WQM was done using soft measurement based on interval observer and it can give accurate estimation [3].