Precise Reliable, Process Control and Patience is Key to Success of Low DO Operation

Presentation Description: Abstract
This presentation demonstrates the implementation of advanced automation and machine learning (ML) control systems for optimizing wastewater treatment plant operations. Case studies from the City of Chico and Pomona Water Reclamation Facilities showcase the evolution from traditional Proportional-Integral-Derivative (PID) control to Model Predictive Control (MPC) and ultimately ML-based systems using DO/Nmaster™ and SRTmaster™ technologies.
The ML control system achieved exceptional ammonia control precision with 95% of readings within 1 mg/L of setpoint and a standard deviation of 0.3 mg/L, despite flow variations exceeding 100% during peak hours. This precision enabled DO setpoint reduction from over 2 mg/L to below 1 mg/L, resulting in 47.1% electrical energy savings ($200,000/year) and 900 metric tons of annual CO₂ reduction at Chico. The Pomona facility demonstrated over 60% energy savings during dry weather operations.
Comparative analysis revealed ML control achieved a mean absolute error of 0.06 mg/L (6%) versus PID's 0.2 mg/L (20%) for DO control. The system's multi-layer fault detection and predictive capabilities enabled unmanned operation 16 hours/day while maintaining 100% NPDES compliance. Implementation included automated SRT control, maintaining ±5% accuracy compared to ±25% with manual control, essential for robust low-DO operation and simultaneous nitrification-denitrification processes.