Emitter Based Algorithm for Pressure-Driven Demand Calculations of Water Distribution Networks - Pattern Recognition for Reliability Assessment of Water Distribution Networks - page 63

differences in the minimum (negative) pressure would result from the position of the node
compared to the failed pipe, which would not quite affect the value of available demand. The
better equality of the maximum pressures is like in all previous cases a consequence of
upstream location of these nodes, towards the failed pipe.

3.8 CONCLUSIONS

The discussion presented in this chapter assesses four ways of using EPANET emitters in
running pressure-demand driven simulations, namely:
- The basic mode where emitters are used for all the pressure below the threshold, allowing
negative nodal pressures and demands (coded as 'pdd1').
- The mode where all pipes connecting negative pressure nodes are disconnected ('pdd2').
- The mode where all negative demands will be set to zero, allowing negative pressures as
the result of extreme topographic conditions ('pdd3').
- The mode where lower pressure threshold has been introduced for pipe disconnection or
alternative setting of negative nodal demands to zero ('pdd4').

The results of these four modalities have been compared with the PDD version of EPANET
developed by Pathirana (2010), as option 'pdd5'. The analyses have primarily focused to the
boundary conditions that could yield peculiar results and have briefly explored a few simple
options of bringing them closer to the reality without massive programming interventions.
For this purpose, an algorithm has been developed using the EPANET toolkit functions in
C++ programming environment. Following is the list of conclusions based on the results of
analysed network layouts:
- Negative nodal pressures in the PDD mode of hydraulic calculation will occur only in the
cases of really extreme topographic conditions and in normal situations even the basic
PDD mode ('pdd1') should be yielding satisfactory results.
- In cases of negative nodal pressures, setting the corresponding nodal demands to zero
('pdd3') yields more stable computation, but avoids consideration of negative pressures
both in demand- and non-demand nodes potentially leading to the reduction of conveying
capacity i.e. further demand reduction.
- Network reconfiguration as the result of pipe disconnection, be it a result of negative
pressures ('pdd2') or the pressures below the lower threshold ('pdd4'), offers more
conservative and likely more accurate results but the simplified algorithm can create
convergence problems if integrity of the network layout has been severely affected. To
arrive at more robust algorithm, additional methodology is necessary that can identify the
network 'islands' without connection to any of the sources.

The final conclusion of this study is that in case of calculations of large networks, the 'pdd1'
and 'pdd3' approaches are likely to create fewer troubles than the 'pdd2' and/or 'pdd4'
approach, although in extreme topographic conditions those may also yield less accurate
results. Nevertheless, the concept of upper and lower pressure threshold looks valid and
should be further tested on more robust PDD algorithm.

The remaining but also important general issue is the one of calculation of emitter
coefficients based on the upper pressure threshold, which in itself is an approximation. The
model calibration in low-pressure conditions is commonly a problem. Nevertheless,
reasonably estimated PDD-thresholds make the presented approach of emitter calculations
the best possible simulation of reality.