Smartphone-Based Real-Time Monitoring and Forecasting of Drinking Water Quality using LSTM and GRU in IoT Environment

Water quality plays an important role in human health. Contamination of drinking water resources causes waterborne diseases like diarrhoea and even some deadly diseases like cancer, kidney problems, etc. The mortality rate of waterborne diseases is increasing every day and most school children get affected to a great extent. Real-time monitoring of water quality of drinking water is a tedious process and most of the existing systems are not automated and can work only with human intervention. The proposed system makes use of the Internet of Things (IoT) for measuring water quality parameters and recurrent neural networks for analysing the data. An IoT kit using raspberry pi is developed and connected with a GPS module and proper sensors for measuring pH, temperature, nitrate, turbidity, and dissolved oxygen. The measured water quality data can be sent directly from raspberry pi to the database server or through the mobile application by QR code scanning. Recurrent Neural Network algorithms namely Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) are used for forecasting water quality. Results show that analysis made using GRU is much faster than LSTM, whereas prediction of LSTM is slightly more accurate than GRU. The data is categorized as poor, moderate, or good for drinking and it can be accessed using smartphones through mobile application. In general, the proposed system produces accurate results and can be implemented in schools and other drinking water resources.

Keywords: Gated Recurrent Unit, Internet of Things, Long Short-Term Memory, Recurrent Neural Networks, Water Quality Parameters.