It is often required to model an existing network for
1) the creation of a control system and optimization of the operation
2) the design of an extension or upgrade
3) the study of water quality and the investigation of ways to improve it.
In order for the model to be used for the intended job, it must simulate the actual mode of operation of the network with the accuracy required for each job. This is achieved by evaluating and calibrating the model.
Model Evaluation
The evaluation checks the results from solving the model for a normal, average charge, for which the network behavior is roughly familiar. Check if there are points with unexpected low or high pressures, tanks that do not stop filling, pumps that operate outside the permitted area or start and stop at intervals that are not actually observed.
These and other remarks that do not agree with the logical operation of the network are indications that some parts of the system have been misrepresented in the model. After correcting any errors and the resulting evaluation results, it is reasonable to adjust the model.
Model Calibration
Configuration requires data from network operation logs. The number of recordings to be made, the duration of each recording, and the state variables to be measured (pressure, flow, substance concentration) are directly related to the work for which the model is intended. In general, a record measures:
- the supply entering the network,
- the levels in the tanks,
- pressures and benefits at selected locations.
In cases where only pressures and flows are measured, it is common to form several charging scenarios with large outflows from taps so that the pressure differences recorded are relatively large. Such charges are applied for a short period of time, as needed to stabilize the effect of this outflow on the network.
Use of tracers to adjust the model
If the model will be used for control and optimization or quality study, then a more accurate hydraulic model is required, so the concentration of a tracer (eg fluoride) introduced into the network is measured. An inherent component of water can alternatively be used as a tracer if its concentration changes with its circulation time in the network (eg chlorine). The recordings last not less than one day, always depending on the time of propagation of the component in the network.
How is it regulated and what is regulated?
The same charge (scenario) with each recorded is entered into the model, which is resolved and the results compared to the recorded ones. The closer the results are to the values measured in the real network, the more accurate the model. The following is the setting, during which changes are made to the network parameters and its re-resolution, trying the resulting results to approach the real, measured conditions, to a satisfactory degree. Some network computing programs have sophisticated algorithms for configuring models. EPANET does not provide this capability, but it can accept the measurements made in the network and produce graphs that show how close the results of each solution are to the actual measurements.
The characteristic sizes of the model that are adjustable are the roughness of the pipes, the consumptions at the nodes and the initial concentrations of the quality parameters. The other parameters (length and diameter of pipes, height and coordinates of the nodes) are considered known (satisfactorily determined in advance). However, if the deviations from the actual conditions in the network are large and there are suspicions of wrong diameters, closed valves or pipes out of order, the model can be resolved with these changes to determine if they are valid and corrected.