ProductionVerified Real deployment at a named or precisely described company, with a measured result from a primary source.
AI controls aeration and dosing in the plant's own wastewater treatment, cutting blower energy and chemicals (Sinopec Nanhua, Bayer Antwerp)
Added 2026-11-13 · Last updated 2026-11-13
Manufacturers in this case: Sinopec · Consorzio Aquarno · Bayer · all manufacturers using AI
Role
Utilities and environment (on-site wastewater treatment plant) · Typical users: utilities manager, wastewater treatment operator, environment or EHS manager, process engineer, plant energy lead
The problem
Many plants run their own wastewater treatment, and its biological stage needs air: blowers are often its largest electricity user, and too little air breaks the discharge permit while too much wastes energy and upsets the sludge. At Sinopec Nanhua Company (China), during production peaks two people were needed to adjust aeration: one in the control room watching four dissolved oxygen meters, the other carrying tools on a platform more than 3 metres above the basins to adjust valves by hand, at night and in the rain, with uneven aeration that wasted electricity and required extra chemicals. At Bayer's Antwerp site (Belgium), which treats 2 million cubic metres of wastewater a year, a highly acidic stream had always been fully neutralised with caustic soda before biological treatment, adding chemical costs and sludge.
How AI helps
At Sinopec Nanhua, an AI aeration system reads the dissolved oxygen signals of the basins in real time together with inflow and water quality data, calculates the best air flow for each blower and adjusts the valve openings remotely, instead of people monitoring and adjusting by hand. At Bayer Antwerp, the approach was model-based rather than machine learning: with the consultancy AM-Team, dynamic process models and computational fluid dynamics models calibrated on lab data were used as a virtual pilot to test a change (letting the biological process neutralise part of the acidic stream itself) and to design the mixing, before implementing it at full scale. Operators keep the final say and the permit limits stay the hard constraint.
Step by step
- Measure the starting point: blower kWh, air flow, dissolved oxygen per basin, chemical consumption, sludge volume and permit results for a few months.
- Check the instruments: dissolved oxygen probes, flow meters and valves must be reliable before any automatic control.
- Look at the history for overaeration (dissolved oxygen well above target) and for hours when dosing does not follow the load.
- Run the control or model in advisory mode first, with operators applying or refusing each recommendation.
- Switch to automatic control within limits agreed with the environment team, with alarms and a manual fallback.
- Report monthly on energy, chemicals, sludge and permit compliance, and keep the regulator informed if your permit requires it.
Tools
- AI aeration control system linked to dissolved oxygen meters and remote valve control: used at Sinopec Nanhua Company
- AM-Team dynamic process and CFD models: used at Bayer Antwerp
- DBS - Data Brain Services AI system: used at Consorzio Aquarno
- Other tools to evaluate: aeration and process control modules from your treatment equipment or blower supplier, advanced process control in your DCS or SCADA, digital twin services for wastewater
- To size the opportunity first: your SCADA history and any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below
In the AI tools index: AM-Team process models · all AI tools used by manufacturers
Result
Sinopec Nanhua Company (Sinopec News, the group's own newspaper): in service since 31 July 2025; blower air flow down 17%, more than 7,000 yuan a month saved on electricity and about 150,000 yuan a year on treatment chemicals; sludge settling ratio held in the target 30 to 40% range and effluent COD compliant; first use of this system in the Sinopec group. Bayer Antwerp (AM-Team project page): caustic soda (NaOH) for neutralisation down 40% and sludge production down 35%, including several hundred tons a year less dewatering chemicals; virtual piloting replaced onsite piloting and the change was implemented at full scale. Consorzio Aquarno (Italy), a mainly industrial treatment plant with more than 60 field sensors and over 250 lab parameters, announced an AI system with DBS - Data Brain Services to optimise compressors and chemical dosing, fully operational by July 2025; no results published yet.
As reported by the source; most figures are vendor or self-reported and not independently verified.
Human review: cautions
- Sinopec's figures come from the group's own newspaper and Bayer's from the consultancy's project page; neither is independently audited.
- Bayer's result comes from process and fluid dynamics models used as a virtual pilot, not from machine learning; it is included because it shows the same goal reached with modelling.
- Discharge permits come first: any automatic control needs limits, alarms and a manual fallback agreed with the environment team.
- Automatic aeration control fails with bad dissolved oxygen probes: plan their cleaning and calibration.
Try it yourself
You are a wastewater process engineer. I want to know if our on-site treatment plant wastes energy or chemicals. I will paste: (1) hourly SCADA data for the last one to three months [inflow, influent COD or load if measured, dissolved oxygen per basin, air flow per blower, blower kW, chemical dosing rates, pH]; (2) our effluent permit limits and lab results; (3) electricity and chemical prices. Do this: 1. Show how often dissolved oxygen was above and below our target, by basin and by hour of day, and the blower energy used in the hours above target. 2. Check whether air flow and dosing follow the inflow and load, and list the periods when they do not. 3. Estimate the possible savings in kWh and chemicals as a range, and show the calculation. 4. List the instrument problems visible in the data (flat signals, gaps, sudden jumps) that must be fixed before any automatic control. 5. Propose an advisory-mode trial: what to recommend to operators, what to measure, and which permit limits stop the trial. Never suggest anything that could break a permit limit, and mark every saving as an estimate. Use only the data I paste in; if something is missing, write 'not found' or list it as a question instead of guessing. Do not send, order, post or change anything in any system: produce drafts for my review and approval.
Want more like this? Join the newsletter
Related cases
- AI cuts plant energy cost: demand-based ventilation control, price-aware production plans and cross-plant energy benchmarking (Stellantis, Casuso Propellers, Jotun)
- Acoustic cameras find and size compressed air leaks while the line runs, so maintenance fixes the costly ones first (Schneider Electric, Altex/Toyota Industries)
- AI warns paper machine operators before a sheet break and points to the likely cause, so mills lose fewer hours to breaks and rethreading (Sappi, Georgia-Pacific, a Voith paper line)
- AI-guided welding robots see the joint and adapt to each part, so fabrication shops weld more with fewer scarce welders (Joe White Tank, Ganotec-Muga Fab, TYCROP)
Sources
- http://www.sinopecnews.com.cn/xnews/content/2025-09/08/content_7132693.html
- https://www.am-team.com/projects/how-smart-models-helped-to-optimize-bayers-industrial-wastewater-treatment
- https://consorzioaquarno.it/en/general/aquarno-is-the-first-predominantly-industrial-wastewater-treatment-plant-in-italy-managed-by-artificial-intelligence/
Weekly: 1–2 practical AI use cases for your plant
Free, one email a week. By subscribing you agree to receive the weekly newsletter (and its sponsor message) from AI Agents for Manufacturing. Every email includes an unsubscribe link, and you can unsubscribe anytime. See our Privacy Policy.