PLS-based soft-sensor to predict ammonium concentration evolution in hollow fibre membrane contactors for nitrogen recovery

Hollow fibre membrane contactors (HFMC) have emerged as a promising technology for nitrogen-recovery that can be implemented in wastewater treatment plants (WWTPs) to promote circular economy. In this process, a hydrophobic membrane allows the transference of free-ammonia across the hollow fibres. During its operation, the ammonium concentration decreases, and real-time measurements would be of great value for process monitoring, optimization and control. Ammonium probes exist, but they are expensive and present noticeably maintenance costs. In this work, results from eight N-recovery experiments performed at different pH values using real supernatant of a full-scale anaerobic digester were analysed in terms of the time-evolution profiles of pH and total ammonium nitrogen (TAN). The pH revealed to carry relevant information related to the TAN concentration, as it decreased in the feed solution due to free ammonia stripping. The pH is an inexpensive-to-measure process variable that can be routinely acquired in any WWTP. Therefore, a data-driven soft-sensor has been developed. It uses the pH, its derivative, and the pH increments after each reagent dosing as input signals, to estimate the TAN concentration via PLS. An extended PLS-model incorporating interaction terms, quadratic and cubic forms of the three input variables improved the TAN concentration estimation. The developed soft-sensor was able to accurately reproduce the evolution of TAN concentration (in the range 0–1000 mgNH4+-N/L with R2 > 0.97 and RMSE <40 mg/L) during the HFMC process operation, thus making it possible to monitor the process as well as enabling future development of different control and optimization strategies.