Multivariate statistic validation of pH and ORP data as control inputs for biological nitrogen removal at full scale

Abstract

The objective of this study was to validate oxidation-reduction potential (ORP) and pH as input data for different advanced control strategies aimed at optimizing biological nitrogen removal under minimum aeration energy demand. For this purpose, a statistical multivariate projection approach was applied to different control inputs calculated from on-line ORP and pH data provided by several sensors installed in different locations of a full-scale plug-flow reactor, aiming to find the strongest correlations with the data provided by nitrogen-based sensors. It has been shown that pH and ORP data can be used as control inputs for optimizing the performance of continuous nitrification, SND, and denitrification processes. Specifically, the controllers were implemented based on the derivative signals from pH and ORP instead of on their absolute values. Multivariate projection methods have displayed and evidenced strong correlations of derivative pH and ORP data with the data obtained from nitrogen-based sensors. Moreover, pH and ORP derivative signals enhance the controller's resilience to sensor faults and data biases, as these signals are less affected by these issues compared to absolute signals and signal differences from different locations along the biological process.

Keywords: Biological nitrogen removal (BNR); Fuzzy-logic control; Low-cost sensors; Multivariate statistic methods; Partial least square (PLS); Principal component analysis (PCA).