Data Driven Multiple Objective Optimization of AAO Process towards Wastewater Effluent Biological Toxicity Reduction

Wastewater treatment is crucial for environmental protection and sustainable development, addressing the challenges of purifying and discharging industrial and domestic wastewater. Wastewater treatment plants (WWTPs) are required to discharge effluents that comply with standards regulating pollutants, including organic compounds, nitrogen and phosphorus compounds, heavy metals, and other characteristic pollutants. However, traditional physicochemical indicators alone do not fully capture the comprehensive risk posed by wastewater. Even compliant effluents may contain low concentrations of pollutants or emerging contaminants, posing risks to the ecological environment and aquatic organisms. Therefore, controlling the ecological risk of WWTP effluents is crucial for maintaining the health of aquatic ecosystems.
 
The anaerobic-anoxic-oxic (AAO) biological wastewater treatment process is widely favored in WWTPs due to its operational convenience and high efficiency in pollutant removal. However, contemporary research in wastewater treatment technology primarily focuses on controlling specific pollutants, with fewer studies comprehensively addressing the biological toxicity of wastewater effluents. Current treatment processes face challenges in effectively mitigating the biological toxicity of effluents and often lack the necessary technologies to significantly enhance toxicity reduction levels. This prompts critical questions: How can we augment the efficiency of toxicity removal in AAO processes while ensuring compliance with stringent effluent standards?
 
Optimizing biological wastewater treatment processes is intricate and dynamic, influenced by diverse influent characteristics across different facilities and changes over time within the same facility. To address these challenges, the National Excellent Engineering Team of Industrial Wastewater Treatment Technology and Equipment, led by Prof. Hongqiang Ren, recognized the potential of machine learning (ML) to offer significant advantages in tackling the challenges of optimizing wastewater treatment process. Unlike traditional methods, machine learning methods do not rely on heuristic rules and fixed parameters, and can adapt to the dynamics and complexity of wastewater pollutants, which facilitates the design of a general process applicable to various scenarios.
 
In this study, incorporated information was compiled, including wastewater quality parameters, treatment process parameters, and wastewater biological toxicity. Such information was collected from 122 municipal WWTPs located in 26 provinces or municipalities across China, providing a detailed overview of the biological treatment processes employed in wastewater treatment. We fully leveraged knowledge and process data from the wastewater treatment process and established a two-step prediction scheme that enables accurate predictions of effluent quality parameters and toxicity reduction ratio under different influent water quality scenarios. Using various ML algorithms, we developed four prediction models for wastewater quality parameters and one model for toxicity reduction efficiency. These models accurately predicted effluent quality parameters and toxicity reduction ratios.
 
Additionally, we proposed a ML-based framework for the optimization of biological wastewater treatment processes, not only facilitating the identification of optimal unit combinations that adhere to fundamental effluent quality standards but also ensuring effective toxicity risk management under diverse influent water quality conditions. The method simulated bioreactor units in wastewater biological processes, predicting effluent quality and toxicity reduction efficiency. Through rigorous screening, we identified improved biological wastewater treatment unit combinations that met wastewater quality standards and achieved optimal toxicity reduction efficiency. The proposed framework recommended the optimal combinations of wastewater treatment unit processes that meet basic effluent discharge standards meanwhile maximizing toxicity reduction under varying influent water quality conditions.
  
In conclusion, this study serves as a proof-of-concept that artificial intelligence is a powerful tool for optimizing the anaerobic-anoxic anaerobic (AAO) process – the most common and classic wastewater biological treatment process, towards provision of safer water for discharge and reuse. Prediction models for effluent quality parameters and toxicity reduction efficiency were constructed and validated under different influent conditions. A machine learning-based optimization scheme was then developed for unit process selection and recombination, which successfully addressed the multi-objective optimization challenges of meeting effluent standards meanwhile maximizing the biological toxicity reduction. Leveraging ML and data driven approaches in optimizing the AAO biological wastewater treatment process provides new insights and practical solutions for the potential for significant advancements in wastewater treatment process with the, emphasizing the importance of comprehensive toxicity reduction and enhancing wastewater treatment process. The study provides a robust protocol for water practitioners and stakeholders for the design and operation of the next-generation wastewater treatment plants. This advancement is crucial for the design and upgrading of WWTPs, helping meet higher effluent standards aligned with evolving environmental regulations and societal expectations.