Neural network modeling for real-time water quality assessment

In recent decades, water quality problems have become even more pressing due to population growth, industrial expansion, and climate change. A number of studies by foreign researchers have shown the results of applying neural networks. There are studies confirming the reliability of water quality prediction results generated by neural networks. During the work, OpenAI Earth Pro, Microsoft Excel, a water flow sensor based on the Arduino UNO board with author’s modifications (tail feathers and a built-in plugin for calculating flow velocity), Python, Tensorflows Keras 2.2.0, Scikit-learn, Pandas libraries for machine learning and developing the neural network architecture were used. Two neural network models were combined to build a hybrid neural network model for predicting water quality parameters in the research. Neural network models provide unique opportunities to improve water resource management at various levels, from local to global. One of the key advantages of such models is the ability to adapt to specific conditions and requirements, providing more accurate predictions and timely decision-making in the face of uncertainty. The relevance of the work is due to the application of neural networks for predicting water quality can contribute to improving the early warning system for pollution, optimizing operational processes at water treatment plants, and developing effective strategies for water resource management. During the research, an innovative hybrid neural network model for predicting water quality parameters was developed, based on the integration of a deep convolutional neural network and a bidirectional recurrent neural network, which consists of three functional parts.