Modeling control and simulation of a drinking water treatment plant

Clean water is an important prerequisite for human health. In this thesis, the water treatment for drinking water production is investigated with a focus on the compact unit as a specific water treatment plant. First, control problems related to the sub-processes of clarification and sterilization are identified and discussed. Then, a model of the sedimentation process is obtained based on real measurement data. In this thesis, a neural network model is chosen, since the validation of analytical models did not lead to satisfactory results. Together with the model, a control method for the sedimentation process is proposed. The control architecture is feedforward control in combination with feedback control. The feedforward controller is realized as a neural network, that is obtained from existing measurement data. The feedback controller is realized as a fuzzy logic controller based on expert knowledge of the sedimentation process. Simulations of the control system with real input data show that the control architecture is suitable for the control of the sedimentation process. In addition, our discussion points out that the only modifications in order to implement the proposed control method are a water quality sensor and a variable speed pump. With these modifications, an un-experienced water treatment plant operator can be replaced by an automatic control system.