EMWD

Eastern Municipal Water District (EMWD) serves nearly one million people across 558 square miles in Riverside County, California, operating four reclamation facilities that collectively treat approximately 48 million gallons of wastewater daily. Aeration — supplying dissolved oxygen to biodegrade organics and oxidize ammonia — accounts for nearly half of the district's total electricity spend. Traditional PID control systems could not keep pace with continuously shifting flow rates and ammonia loads, requiring daily manual operator interventions and creating a persistent lag between adjustments and measurable results. Setting dissolved oxygen targets too high wasted energy; setting them too low triggered ammonia bleed-through and drove up chemical disinfection costs.

EMWD partnered with Rockwell Automation to pilot a predictive machine learning control system at its San Jacinto Valley reclamation facility, which processes 7 million gallons per day. Rockwell Automation data scientists identified wastewater flow rate and ammonia load as the primary drivers of dissolved oxygen demand, then built a predictive ML model to anticipate DO requirements before deviations occurred — replacing reactive feedback loops with forward-looking adjustment. The AI software was remotely programmed and deployed on an Allen-Bradley CompactLogix 5480 controller integrated directly into EMWD's existing plant automation network. Two dedicated processor cores run the AI software continuously, adjusting air valve positions in real time as conditions change. EMWD ran one month of standard PID baseline followed by one month of AI-enabled control to generate a direct performance comparison before committing to broader deployment.

The AI-enabled system reduced airflow delivered to the aeration basin by as much as 31% compared to the traditional PID baseline. Combined energy savings are estimated at 2,330 kWh of electricity per day, broken down as follows:

• 960 kWh/day from tighter dissolved oxygen control (~$42,000 in annual savings)
• 1,370 kWh/day from the ability to lower DO setpoints more aggressively (~$60,000 annually)
• Total projected savings: over $100,000 per year

Qualitative outcomes included fewer incidents of ammonia bleed-through, reduced disinfectant consumption, improved effluent quality, and complete elimination of the daily manual operator interventions that previously burdened plant staff.

• Predictive ML outperforms reactive PID control in high-variability environments — fluctuating wastewater loads are an ideal fit for AI-driven optimization.
• Deploying AI on existing control hardware rather than replacing infrastructure dramatically reduces cost and implementation friction.
• A structured A/B pilot (one month baseline, one month AI) generates credible comparative data to justify broader rollout.
• Energy savings compound: tighter control enables more aggressive setpoints, unlocking efficiency gains beyond the initial improvement.
• Downstream process benefits — reduced chemical use, better effluent quality — often emerge as secondary wins from optimized aeration control.