A hydro informatics system for real-time optimal operation of a surface water treatment plant

Abstract

The need for clean drinking water in the sub-Saharan Africa is on a rapid increase due to the fact that the existing surface water bodies are facing a problem of pollution resulting from both urban and industrial developments which are on a rise in most African countries. Water treatment plants have been adapted to avail the people with water that is clean and safe for drinking and other related uses, but as a result of increasing pollution of surface water sources, these plants are facing challenges related to operation. This project, therefore, was inclined towards optimizing the operation of surface water treatment plants through the use of a real time hydro informatics system to effectively monitor the parameters right from the surface water source to the final treatment unit. Deep learning and Arima models were developed for the forecasting of available water and raw water parameters, for which the deep learning models had better performance based on their high coefficient of determination values and were adopted. Also, three tools were used in developing the prediction models of the effluent water unit parameter outputs. These included multilinear regression, deep learning and support vector machines. The deep learning modals had better prediction accuracy based on their slightly higher values of coefficient of determination and hence were adopted during the optimization. The optimization was carried out using the genetic algorithm tool of MATLAB and the results indicated better parameter control when compared with a real-world system. An Arduino based IoT system was also developed which was used for system monitoring and control, and this proved to be an easier and better way of treatment system monitoring and parameter control. Key terms: Surface water treatment plant, Hydro-informatics, models, genetic algorithms, optimization, IoT