K. M. van Schagen

A fully adaptive forecasting model for short-term drinking water demand

For the optimal control of a water supply system, a short-term water demand forecast is necessary. We developed a model that forecasts the water demand for the next 48 h with 15-min time steps. The model uses measured water demands and static calendar data as single input. Based on this input, the model fully adaptively derives day factors and daily demand patterns for the seven days of the week, and for a configurable number of deviant day types. Although not using weather data as input, the model is able to identify occasional extra water demand in the evening during fair weather periods, and to adjust the forecast accordingly. The model was tested on datasets containing six years of water demand data in six different areas in the central and Southern part of Netherlands. The areas have all the same moderate weather conditions, and vary in size from very large (950,000 inhabitants) to small (2400 inhabitants). The mean absolute percentage error (MAPE) for the 24-h forecasts varied between 1.44 and 5.12%, and for the 15-min time step forecasts between 3.35 and 10.44%. The model is easy to implement, fully adaptive and accurate, which makes it suitable for application in real time control.

Good modelling practice in drinking water treatment, applied to Weesperkarspel plant of Waternet

Good modelling practice increases the credibility and impact of the information and insight that modelling aims to generate. It is known to be crucial for model acceptance and it is a necessity to amass a long-term, systematic thorough knowledge base for both science and decision making. This paper shows how ten steps in model development and evaluation can also be applied to numerical modelling of drinking water treatment, using models of drinking water treatment processes of the Weesperkarspel treatment plant of Waternet. The Weesperkarspel plant consists of ozonation, pellet softening, biological activated carbon filtration and slow sand filtration. For the different processes models were developed that were used for operational improvements. The modelling resulted in new insights and knowledge about the treatment processes and improved operation of the processes. From scenario studies for the pellet softening it was concluded that chemical dosing can be diminished when by-pass ratio is increased and that pellet size can be controlled by measuring the difference in pressure guaranteeing fluidisation of the pellet bed. In addition, ozone dosage can be optimised by modelling ozone exposure, bromate formation and biologically degradable natural organic matter (NOM) under varying influent water quality.

Model-Based Dosing Control of a Pellet Softening Reactor

Abstract Abstract The control of a drinking-water treatment plant aims to produce the correct quantity of water, with a constant quality. Achieving constant water quality is not an obvious task, since the online water-quality measurements and possible control actions are limited. Applying model-based control improves disturbance rejection and online process optimisation. For the softening process step, the integral control scheme is shown with multiple controllers for different time scales and process detail. The dosing control is elaborated and verified using simulation experiments. The control is implemented and tested in the pilot plant of Weesperkarspel (Amsterdam). It shows that in the case of accurate state estimation, quick changes in setpoint can be tracked.

Cascaded parameter estimation for a water treatment plant using particle filters

Advanced online control of drinking water treatment plants requires reliable models. These models in general involve temperature-dependent, uncertain parameters, which can only be measured in laboratory conditions. We propose to estimate these parameters online, using the available pH quality measurements. Since the pH measurements are a nonlinear combination of the systems states, a particle filter is used. Thanks to the cascaded nature of the plant, the estimation is also performed in a cascaded setting. The performance is evaluated both for simulated and real-world data. Results indicate that the filter can be effectively used to improve the model accuracy.

Modeling and predictive control of pellet reactors for water softening

Abstract A nonlinear chemical/physical dynamic model in the form of partial differential equations was adopted and further developed to serve as a basis for model predictive control of a pellet reactor for drinking water softening. The model was calibrated using full-scale process measurements. A linear predictive controller based on a lineralization of the model has been designed to achieve the desired hardness of the effluent water through cost-effective operation of the reactor. This controller has been extensively validated in nonlinear simulations. The results are promising and the strategy found by the predictive controller leads to a smoother operation compared to the currently used heuristic controller.