Online state and parameter estimation of a biological sequencing batch reactor using an advanced nonlinear Bayesian derivativeless filtering algorithm with Activated Sludge Model No. 1

To overcome the drawbacks of convensonal linear Kalman filter, nonlinear Bayesian derivativeless filtering algorithm, called unscented particle filter (UPF), is implemented for online estimation (or filtering) of the states and parameters in a sequencing batch reactor (SBR).
This UPF uses a dynamic Bayesian state-space model framework and does not require any model linearization procedure, and restrictive assumptions. The Activated Sludge Model No. 1 (ASM 1) with nonlinear biological kinetic models is used as a process model and formulated in a dynamic state-space model framework.
The UPF is implemented first time for recursively estimating the unknown states (microbial metabolisms) and parameters (biological oxygen demand (BOD) and ammonia) defined in ASM1 model in a sequencing batch
reactor (SBR). The performance and effectiveness of the proposed UPF, compared with the particle filter (PF), are demonstrated using the online data (dissolved oxygen (DO)) and offline measured data (BOD and ammonia)
obtained from the operation of a biological SBR. The excellent tracking result for the online measured DO concentration and the very good online estimation of unknown soluble BOD and ammonium (NH4-N) concentration
with a high level of accuracy are achieved by both UPF and PF. The performance and effectiveness of the proposed UPF, compared with the well-known particle filter (PF), are demonstrated that it can provide: (1) the flexible and robust recursive estimation framework with high levels of accuracy; and (2) a quantitative basis for probabilistic representation of unknown/unmeasurable states/parameters incorporating uncertainties to the model for highly nonlinear biological wastewater treatment processes.