Loading...
Paper Type
Complete
Abstract
The quality of effluent in wastewater treatment plants (WWTPs) is crucial for protecting the environment, ensuring public health and meeting regulatory standards. We introduce a random forest (RF) model designed to predict the operational costs (OCI) and effluent quality (EQI). RF model offers comprehensive insights into the variables that impact the effectiveness and economic efficiency of WWTPs. To enhance transparency, our analysis includes an interpretative component that examines feature importances and SHAP values. This knowledge is needed to select key process variables for the design of a plant-wide controller for EQI and OCI optimization. Our model exhibits strong predictive performance for both variables, as demonstrated by high R2 values. The model's performance and generalizability were validated using data derived from the adjusted benchmark simulation model of the WWTP, achieving scores of 0.915 for EQI and 0.957 for OCI across test datasets, indicating a high level of model's predictive precision.
Paper Number
1166
Recommended Citation
Dominkovic, Lana; Hvala, Nadja; Vrečko, Darko; and Boshkoska, Biljana, "Efficiency and Explainability in Wastewater Treatment Plant: A Machine Learning Approach to Cost Management and Effluent Quality" (2024). AMCIS 2024 Proceedings. 2.
https://aisel.aisnet.org/amcis2024/dsa/dsa/2
Efficiency and Explainability in Wastewater Treatment Plant: A Machine Learning Approach to Cost Management and Effluent Quality
The quality of effluent in wastewater treatment plants (WWTPs) is crucial for protecting the environment, ensuring public health and meeting regulatory standards. We introduce a random forest (RF) model designed to predict the operational costs (OCI) and effluent quality (EQI). RF model offers comprehensive insights into the variables that impact the effectiveness and economic efficiency of WWTPs. To enhance transparency, our analysis includes an interpretative component that examines feature importances and SHAP values. This knowledge is needed to select key process variables for the design of a plant-wide controller for EQI and OCI optimization. Our model exhibits strong predictive performance for both variables, as demonstrated by high R2 values. The model's performance and generalizability were validated using data derived from the adjusted benchmark simulation model of the WWTP, achieving scores of 0.915 for EQI and 0.957 for OCI across test datasets, indicating a high level of model's predictive precision.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
SIGDSA