Algorithm for predicting water quality indicators in water bodies using a neural network

Clean and safe drinking water is a fundamental necessity for human health and well-being and a critical component in sustainable ecosystem development. In recent decades, water quality issues have become even more urgent due to population growth, industrial expansion and climate change.

A series of works by foreign researchers report results obtained by applying neural networks. There are studies confirming results of water quality prediction generated by neural networks to be quite valid.

In this research, we used Google Earth Pro, Microsoft Excel, water flow sensor based on Arduino UNO board with author's modification (tail feathering and built-in plugin for calculation of flow velocity), Python, Tensorflows keras2.2.0, Scikit-learn, Pandas libraries for machine learning and development of neural network architecture. In this study, two ANNs were combined to build a hybrid neural network model for predicting water quality indicators.

Neural network models offer unique opportunities to improve water resources management at various levels, ranging from local to global one. A key advantage of such models is a possibility to adapt them to specific conditions and requirements, which provides more accurate prediction and timely decision making under uncertainty. The relevance of the work is determined by application of neural networks for water quality prediction. This can improve systems for early warning about pollution, help optimize operational processes at water treatment plants and develop effective water management strategies.

In this research, an innovative hybrid neural network model has been developed for predicting water quality indicators. It is based on integrating deep convolutional neural network and bidirectional recurrent neural network, which consists of three functional parts.