Pesticide metabolites: optimizing activated carbon for drinking water treatment
For several years, pesticide metabolites have become a primary concern for drinking water producers in France. Resulting from the degradation of plant protection products, these molecules are found in massive quantities in surface and groundwater.
French regulations set a strict quality limit of 0.1 µg/L for each pesticide and its metabolites in water intended for human consumption. For many water treatment plants, consistently meeting this threshold is a challenge: concentrations vary significantly and suddenly depending on weather, season, or agricultural practices. A plant can go from raw water that meets standards to a resource requiring emergency treatment in just a few hours.
Faced with this risk, treatment with powdered or granular activated carbon is the preferred solution. However, adjusting it presents a real dilemma for operators.
The activated carbon dilemma: health risks versus financial and environmental costs
Activated carbon is one of the most expensive reagents in a water treatment plant. While it is incredibly effective against micropollutants, determining the right dose to inject at the right time is a delicate balancing act.
Today, the vast majority of plants operate with reactive regulation based on periodic laboratory analyses. This time lag traps operators in a constant dilemma:
- Under-dosing: It creates a direct risk of health non-compliance, especially during rainy episodes that cause metabolite concentrations to skyrocket.
- Over-dosing: To play it safe, plants over-consume activated carbon. This leads to exorbitant operating costs and a disastrous carbon footprint (due to production, transport, and sludge treatment), which is disproportionate to the actual risk.
The Purecontrol solution: predictive control for activated carbon injection
To break this deadlock, Purecontrol offers a cutting-edge technological solution: a tool for predictive control of activated carbon injection using artificial intelligence. Proven in the field, the solution shifts drinking water plants from a reactive mode to a proactive one.
Our approach is based on a model built around three technological pillars:
1. A "virtual sensor" to forecast metabolite concentrations 6 hours ahead
No more waiting for lab results. Purecontrol models the evolution of metabolite concentrations (such as metolachlor ESA) in raw water in real time. By cross-referencing data on rainfall, UV levels, groundwater levels, pH, and nitrates, the algorithm generates an ultra-precise 6-hour forecast, recalibrated minute by minute.
2. The optimization engine: the perfect balance between quality and cost
Based on these forecasts, the AI continuously identifies the optimal dosing strategy. It balances the absolute health objective (meeting the 0.1 µg/L regulatory threshold) with economic and environmental goals (minimizing activated carbon consumption). The model even anticipates the natural loss of filter efficiency over time.
3. Real-time control of activated carbon injection
Injection setpoints are sent to the site's PLC minute by minute, adjusting the dosage to match real-world conditions. Acting as a true co-pilot for the operator, the solution relieves the mental burden of continuous monitoring and emergency manual adjustments, allowing them to focus on overall plant management. This ensures that cross-functional objectives are met consistently.
Proven field results: up to 15% savings on activated carbon
Deployed on the treatment lines of a major French drinking water plant in partnership with a leading national water utility and a committed local authority, the Purecontrol solution demonstrated its effectiveness spectacularly over a full year of operation:
- A 15% reduction in activated carbon consumption : The AI successfully smoothed and adjusted the dosage, drastically reducing it in summer and intelligently increasing it during winter peaks, whereas traditional regulation would have resulted in blind overconsumption.
- Flawless responsiveness: When the physical efficiency of the filters degraded in winter, the algorithm automatically detected the drop and compensated by increasing upstream injection, thereby avoiding non-compliance.
- Controlled water quality: Anticipating rainfall events allowed the plant to smooth its response, securing the production of drinking water.
"The value of the tool is that today we can compare in real time what we inject using Purecontrol versus the doses we would have used with traditional control. We can see that we are using lower doses with AI control. It’s a simple tool for the operator, with a safety feature that allows us to switch back to manual control if needed." - Jean-François Coupel,
Managing the pesticide metabolite crisis is no longer a matter of fate. Where traditional approaches struggle to keep pace with climatic and regulatory changes, artificial intelligence provides a concrete, robust, and easy-to-implement solution.
