Machine learning framework for optimization of flocculation process of water treatment

• Autor(a) principal: Bankole, Abayomi Oluwatobiloba
• Data de Publicação: 2023
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://hdl.handle.net/11449/244140
• topic: Floc length evolution; Flocculation; Machine learning; Neural network; Smart water treatment; Evolução do comprimento dos flocos; Floculação; Aprendizado de máquina; Rede neural; Tratamento inteligente de água

The implementation of machine learning (ML) model that could improve both the effectiveness and sustainability of water treatment system is a major problem in the water sector, with the optimization of flocculation process being a major setback. In this study, we have developed the first ML model for floc length evolution monitoring and a framework for its potential adoption in large-scale water treatment. Artificial Neural Network (ANN) and Long-Short Term Memory (LSTM) models, and traditional time series model; Auto Regressive Integrated Moving Average (ARIMA) were explored to predict floc length evolution data that was obtained through non-intrusive image analysis from a jar test batch assay and model the orthokinetic process. Batch assay data of two velocity gradient (Gf 20 sec-1 and 60 sec-1) and flocculation time of 3hrs were partitioned into 5 bins for floc length range 0.27 – 3.5 mm and upscaled using linear method. Results showed that ARIMA model is not suitable for predicting number of flocs with a negative test accuracy (R2). ANN recorded R2 of 0.86 – 1.0 for training and 0.84 – 0.99 for testing, across Gf 20 sec-1 and Gf 60 sec-1. LSTM model has the best prediction accuracy of 98 – 100% for Gf 20 sec-1 and perfect prediction of number of flocs across all bins and Gfs. Our study has proven that the developed framework can be replicated in large scale water treatment and will promote application of smart technology in large-scale water/wastewater treatment.