Shall we use hardware sensor measurements or soft-sensor estimates? Case study in a full-scale WWTP

An array of data derived soft-sensors for estimating nitrate-nitrogen concentrations in the denitrifying post-filtration unit of a municipal wastewater treatment plant have been designed in our earlier work. The array was designed to back up the existing hardware instrumentation in the process unit. The importance of trustworthy information on nitrate-nitrogen concentrations is emphasized by their employment in control loops enhancing nitrogen removal. In this paper, we propose an automatic system for deciding whether the use of estimates or corresponding hardware measurements is preferable given the current condition in filters. The presented switching system is based on ranking the estimates and the measurements, and a preferability index that combines the assigned ranks. The ranking procedure employs a robust principal component analysis and statistics used as the measures of fit. The switching system was demonstrated to enable the employment of measurements and estimates in a complementary manner in operating the post-filtration unit.