An Intelligent Dosing Algorithm Model for Wastewater Treatment Plant

Abstract

The precise coagulation add-in in the wastewater process treatment is key for efficient contamination removal. However, the complexity of the coagulant chemical theory and affected by many factors (turbidity, pH, conductivity, flow rate, etc.) that it is difficult to determine the optimal dosage. The traditional method in the production process, such as PID controller had a bad adaptability on the complex systems and high performance required systems due to its inefficient parameter coordination, and it has a large time delay, difficult to achieve precise control. Excessive dosage will lead to waste and cost-waste, insufficient dosage could not guarantee the quality of effluent water. In this research study, we proposed an intelligent precisely dosing prediction algorithm based on LightGBM, using the characteristics of the influent water quality parameters PH, turbidity, electrical conductivity and flow rate to predict the dosage of coagulant. Perform experiments based on the actual data collected from the sewage treatment plant. Compared to experimental results with the optimal dosage solution, it demonstrated that the proposed approach could predict the dosage more accurate, resulting in intelligent and precise dosing add-in in water treatment process.

Exploring potential dual-stage attention based recurrent neural network machine learning application for dosage prediction in intelligent municipal management

“…Xu 3 employed eight ML models to predict the chemical dosage WWTP and compared the performance of each model. Fang 19 utilized the light gradient boosting machine algorithms to forecast dosage in WWPT, and relatively good results were achieved. Wang 20 compared principal component regression (PCR), support vector regression (SVR) and long short-term memory (LSTM) neural networks to build predictive models, and the results show that the LSTM algorithm outperforms both PCR and SVR.…”

Using a supervised machine learning approach to predict water quality at the Gaza wastewater treatment plant

“…These algorithms speed up the model training without compromising prediction accuracy or increasing memory loss. 48 However, GBDT's iteration process can result in slow training speed and large memory consumption. The chosen evaluation metrics for assessing the model's performance were set to include 'l1' and 'l2', enabling a comprehensive assessment of model accuracy and robustness.…”