Sponsored by: Senter
This project, called Promicit (Process Modeling and Integrated Control of Water Treatment), is a cooperation between Amsterdam Water Supply, ABB, DHV and the Delft University of Technology - Department of Civil Engineering and Delft Center for Systems and Control.
Currently, water treatment plants are primarily controlled by experienced operators, based on laboratory and on-line measurement of water quality parameters. The goal and the challenge of this project is to develop models of the complete water purification plant, by using first principles (chemical, biological and physical) as well as novel gray-box, data-driven techniques. Based on these models, an automatic control system will be designed for integrated, plant-wide control of the entire process chain. This system can be used both for on-line process control and for decision support. The main treatment steps considered in this project are ozonation, softening and biological active carbon filtration. It is expected that by using advanced modeling and control techniques, water supply companies will gain more insight in the operation principles of the plant, improve monitoring, prediction and control of the processes and thus will consistently produce high-quality drinking water.
In a first pilot study, a model-based predictive controller for the softening process stage (consisting of several pellet reactor operating in parallel, see Figure 16) was developed. This process was selected as it is relatively independent of the other treatment steps, the chemical and physical principles are well understood and a sufficient number of sensors and actuators are available. The softening controller should maintain the desired water hardness and at the same time minimize the super-saturation of calcium in order to prevent calcium deposits in later water treatment steps.
First, a model was developed within the Stimela environment under Matlab and Simulink. Sensitivity analysis has been conducted on this model to identify the most important parameters and to compute the uncertainty bounds in the predicted outputs. The parameters of the model were optimized by using process data. A hierarchical control structure was proposed to comply with the requirement of constant effluent hardness and minimal super-saturation. A supervisory control level is responsible for determining optimal water flow and reactor effluent hardness set-points. Local controllers take care of the actual control of the individual softening reactors. The performance of the local controller was evaluated in four different scenarios and it was compared with the controller currently used in the plant.
Next: Model based optimization of emulsification Up: Industrial process control Previous: Artificial intelligence for the control
Last modified: 24 March 2005, 10:16 UTC
||Search Site map|