Soft Sensors Development for Model Based Control of Groundwater Treatment Plants

The drinking water in the Netherlands has been aimed to achieve high quality of water along with a low production cost. In-order to produce such high quality of treated water, constant monitoring of the process variables at the end of each treatment step is recommended. These processes are monitored and operated by motivated and skilled operators and process technologists, which leads to an operator dependent, subjective, variable and possibly sub-optimal operation of the treatment plants. Furthermore, the extensive automation of the treatment plants reduces the possible operator attention to the individual process units. The use of mathematical or black-box process models might solve these problems. This thesis thus focusses on the development of soft sensors i.e models that can simulate the rapid sand filter (one of the drinking water processes) process and thus be able to estimate/predict the output variables of interest, such that optimal control schemes could be developed to control the quality of the produced water. Before an appropriate model can be designed, it is necessary to analyse the drinking water treatment processes of the pilot plant. In general the treatment processes are robust, but ignoring the typical process behaviour can hamper the optimal performance. One of the readily available mathematical/white-box model is assessed, only to conclude that assessed white-box model has few limitations hampering the main objective of the research, thus motivating us to use black-box modelling approach instead. Using black-box approach,the water quality parameters to be estimated are determined using subspace and parametric estimation schemes. The strategy for implementing these system identification schemes have been investigated. These techniques have been applied to experimental data collected for the rapid sand filter process and a good fit to the observed dynamics is obtained. Models obtained through these techniques have been discussed and compared using various validation tests.

Subject

embedded systems
soft sensors
water treatment
systems & control

To reference this document use:

http://resolver.tudelft.nl/uuid:3cfb667e-c109-4877-a634-0c83eeeb45d3

Embargo date

2012-05-23

Part of collection

Student theses

Document type

master thesis

Rights

Files

PDF

mscfinal.pdf

2.92 MB

Close viewer