State detection in the biological water treatment process

Abstract

<div lang="en" class="abs"><p class="abs_header">Abstract</p><p>This study introduces a theoretical bioprocess of ideally stirred Chemostat. Chemostat models give an insight to real-life bioprocess systems, in particular biological water treatment. The studied process is very nonlinear due to inhibition of desired reactions by high substrate concentration. The aim of this study is to demonstrate the possibilities of modelling and state detection of this system. This report gives some basic knowledge about bioprocesses and reviews the latest work done for modelling and control of biological waste water treatment. Data-based modelling and state detection are studied thoroughly. Also a couple of control designs are proposed.</p><p>The theoretical bioprocess is described by a set of differential equations. Steady-state and dynamic models were both derived from the differential equations. The steady-state model was used to obtain knowledge about system’s behaviour and the dynamic model to generate data for data-based modelling. The dynamic model was also used in testing of developed models and control designs. Two separate operating points were identified from the data and first order models were identified for both operating points. Based on the knowledge from the steady-state model and data analysis a rulebase was defined. The rulebase and operating point models were integrated to form a Takagi-Sugeno-type fuzzy process model. Testing with the dynamic model showed that the developed fuzzy model performed well. The fuzzy model with slight changes was also used in state detection with good results.</p><p>Two control designs were developed. One utilizes the recycling of the process output back to the reactor while the other is based on the developed process model. The prior control design is of feedback type and the latter is of feed-forward type. Both control designs performed well while tested with the dynamic process model.</p></div>