Economic Aspects of Decision Making in Reliability Assessment of Water Distribution Networks - Pattern Recognition for Reliability Assessment of Water Distribution Networks - page 217

processing large number of scenarios, the network diagnostics tool (NDT) introduced in
Chapter 6, has been further upgraded with options that deal with calculations of investment,
operation and maintenance costs on annual basis. The reliability i.e. the resilience of each
network/demand scenario has therefore been assessed from the perspective of total costs. The
reliability measures that have been used in this assessment are the network buffer index ( NBI )
and network resilience ( I n ), introduced in Chapter 5.

8.7 PERFORMANCE OF OPTIMISED NETWORKS

As mentioned, the two initial networks have been manipulated with five topographic patterns
and three altitude ranges. Figures 8.9 and 8.10 give impression about the distribution of
pressures in GA-optimised i.e. initial layouts in each of the topographic patterns and the
range of nodal elevations taken as an example between 10 and 30 msl. As it can be seen, the
pressures in all cases are represented with contour lines of 20 to 50 mwc; the fraction of red
area in the second pattern shown on Figure 8.9 reflects the minimum pressure of 19.6 mwc.
Both figures show logical response to the particular topographic patterns: the areas of higher
pressure around the source and in low elevated areas, and the areas of low pressure more far
away of the source and around high elevated areas. Networks B have higher pressures, which
is the consequence of elevated tank selected to ensure the minimum pressure. The maximum
pressure amongst the A options is 53.6 mwc, while for the B options it is 71.2 mwc, both in
node N2 and in hilly terrain configuration.

Figures 8.11 to 8.15 show the relation between the reliability measures and total annual costs
taken as an average of all four investment options from Table 8.6. The initial layouts are
categorised per topography pattern and altitude range. The observations from the figures can
be summarised in the following bullets:
- In all the cases, the optimised networks of option A are more expensive than those of
option B . This is a direct consequence of the difference in supplying scheme in which the
pump and the maximum pipe diameter have to be designed on the maximum peak
demand, while in case of the tank existing in option B the design flow for the pump and
the maximum pipe diameter can be reduced.
- The costs of the networks in the flat topography shown in Figure 8.11 are the lowest of
all, as well as they are the least dependent on the selected altitude. In other four
topographic patterns the costs would generally grow with (significantly) wider altitude
range. This trend is more visible in option A than in B and has again the likely cause in
need of a bigger pump and maximum pipe diameter.
- There is a difference in the values of NBI that are consistently lower than those of I n . This
emerges from the origin of the two indices, which has been discussed in Chapter 5. The
network resilience, I n , is more of a head-driven index whilst the network buffer index,
NBI , is more responsive to the network connectivity and its resistance. Nonetheless, in
both cases, the values of indices depict rather low reliability, which is expected for GA-
optimised networks.
- The network resilience is lower in case of flat and slope-down configurations (Figures
8.11 and 8.13, respectively) which is the result of generally lower heads/pressures
occurring in those configurations. The other three terrain configurations inflict higher
pressures, be it from the lower nodal elevations, higher pumping heads and/or larger pipe
diameters. This additional buffer therefore makes those networks slightly more reliable.
- The NBI-values have erratic patterns, yet in much lower range. Similarities with I n are
visible in Figures 8.12, 8.14 and 8.15, but the correlations are generally weak.