How well can inaccurate sensors quantify and improve the performance of a fleet of on-site wastewater treatment plants?

On-site wastewater treatment plants (OSTs) are widely seen as a stopgap solution, mainly because of a lack of monitoring and the resulting unreliable treatment performance. To address this concern, low maintenance, but inaccurate soft sensors are emerging. However, the impact of this inaccuracy on the treatment performance of entire fleets of OSTs has not been quantified. We develop a stochastic model to estimate these performances. In the modelled case soft sensors with a 70% accuracy improve the treatment performance from 66% (percentage of time functional) to 98%. Soft sensors optimised for specificity (true negative rate) improve the system performance, while such optimised for sensitivity (true positive rate) quantify the treatment performance more accurately. Based on this new insight we suggest to build two soft sensors with the same data input in practical settings: one soft sensor geared towards high specificity, for maintenance scheduling, and one geared towards high sensitivity, for fleet performance quantification. The findings suggest that inaccurate sensors in combination with an appropriate alarm management have the potential to largely improve the treatment performance of a fleet of OSTs. We present a management strategy to reduce undetected failures drastically and thereby diminish negative impacts on environmental and human health.