Intelligent FA/FNA alternating strategy for nitrite-oxidizing bacteria inhibition: Data-driven prediction and process...

Area of Science:

• Environmental Engineering
• Wastewater Treatment Technologies
• Biotechnology

Background:

• Partial nitrification (PN) is crucial for wastewater treatment, but inhibiting nitrite-oxidizing bacteria (NOB) traditionally relies on manual monitoring.
• Manual assessment of nitrite accumulation rate (NAR) hinders automation of the free ammonia (FA) and free nitrous acid (FNA) inhibition strategy.

Purpose of the Study:

• To develop an intelligent control system for automated, real-time switching of FA/FNA inhibition strategies.
• To enable real-time control of inhibition concentration for stable NOB suppression in PN processes.

Main Methods:

• Compared support vector machine (SVM) and multilayer perceptron (MLP) for classification and regression tasks.
• Utilized SVM for strategy alternation based on NAR thresholds and MLP for real-time FNA concentration control via nitrite prediction.
• Trained and tested models on a 180-day experimental dataset from a continuous flow PN reactor.

Main Results:

• Optimized SVM achieved 91.67% classification accuracy; MLP showed R² of 0.96 and RMSE of 53.16 for nitrite prediction.
• Real-time control demonstrated SVM classification accuracy of 97.5% and MLP R² of 0.83 (RMSE 0.04) for FNA concentration.
• The system maintained NAR above 95.55% and ammonium removal efficiency above 52.99%.

Conclusions:

• The intelligent control system successfully automates FA/FNA alternating inhibition and accurately controls FNA concentration.
• This automation simplifies operations, ensuring stable and efficient PN process performance.
• The developed system promotes wider application of the PN/A process in wastewater treatment.