Current trends of soft sensor applications to monitoring pollutant removal efficiency in wastewater treatment

Soft sensors offer a way to estimate wastewater treatment outcomes that are difficult to measure continuously. This review synthesizes 17 studies that predicted pollutant removal, focusing primarily on domestic and municipal wastewater treatment and including selected studies of other wastewater types for methodological comparison. We examined removal targets, model inputs, modeling approaches, validation practices, and reported performance. The studies addressed nutrients, organic matter, microbial indicators, viruses, and other pollutants. Most predicted a continuous removal outcome, while others classified whether a removal threshold or treatment endpoint had been reached. Input selection and modeling strategies varied substantially, and comparisons within individual studies did not identify a consistently superior algorithm. Differences in targets, datasets, performance metrics, and validation designs also prevent direct ranking of results across studies. Although the reported models show promise for specific monitoring tasks, evidence of performance across independent operating periods or treatment systems remains limited. Some models also rely on inputs that cannot be measured at the frequency needed for continuous prediction. Future studies should define removal outcomes consistently, report when and how inputs are measured, and evaluate models under changing conditions and at independent sites.