Comparison of Several Filtering Approaches on Water Treatment Processes

Abstract

This paper addresses the state estimation problem of a bioreactor in wastewater treatment processes. The state variables of this process are the concentrations of the organic pollutants and of the bacteria inside the bioreactor. A specific growth rate function is used to describe the variation of the bacteria concentration when the amount of pollutants increases. This rate can also represent the speed of the biological degradation of the pollutants. Most research work in this field uses only deterministic models that do not conveniently account for uncertainties. These models are often obtained using several simplifications during the modeling procedure such as neglecting the measurement noises. In this paper, we consider stochastic models and study the state estimation problem using three approaches: the Extended Kalman filter, the Unscented Kalman filter and the Particle filter. These methods are adapted to the models in study and compared to understand which is the most adequate for this type of processes considering their slow evolution, discrete time measurements and high-intensity noises. Further, we also apply a Multiple Model Adaptive method which adapts the filters to the correct growth rate type. This method is also used to automatically choose the most efficient estimation method for this type of biological processes.