|
BACKGROUND
Losses of potable water from urban water-distribution networks can account
for up to forty percent of the amount put into supply. In part, this is due
to the unnecessarily-high pressures in the distribution network which result
from the inadequate levels of control currently available. If pressures in
the distribution network were kept as low as possible whilst still complying
with the statutory supply requirements (minimum pressure, adequate quantity
etc.), this would not only reduce leakages but also minimize pumping costs.
The problem is that water-distribution networks are large, complex and
subject to highly-variable demands. However, with the latest analytical and
optimization techniques available, it is possible to envisage an on-line
control system which dynamically responds to short-term fluctuations in
demand. In searching for the optimal combination of pump and valve settings,
consideration would be given to continuity of supply, minimum statutory
pressures and minimum accepted flow rates to avoid stagnation.
|
|
|
EXISTING STATE-OF-THE-ART
Hydraulic simulation models have been used for many years in the design of
water-distribution networks. All are based on solving the equations relating
to pipe flow and as such, are computationally time-consuming. Whereas in the
past, these models have largely been used on a "trial-and-error"
basis, more recently various methods of optimization have been introduced,
including genetic algorithms (GA). Even though GAs are very efficient,
computational times can be excessive but for design purposes , this would
probably be acceptable, particularly if it resulted in a significant saving
of construction costs.
Whilst a few hydraulic simulation models have been adapted for operational
purposes, such as anomaly detection, none have been used for optimal control.
However, various forms of mathematical programming have been used for pump
scheduling, including linear programming, dynamic programming, non linear
programming and decomposition-coordination methods. As a result, pump scheduling
is gradually being adopted by the water industry as a means of reducing
electricity costs. For Europe as a whole, annual pumping costs relating to
water distribution alone, must exceed a billion euro and therefore, even a
small percentage reduction represents a large monetary saving. It will, of
course, be appreciated that pump-scheduling does not equate to real-time
control since the former is limited to deriving a series of targets to
minimize pumping costs over the next 24 hours, assuming a fixed-demand
profile. Decisions relating to the control settings to achieve those targets
are left to the discretion of the operating staff who use their experience
and judgment on deciding where and when to turn pumps on/off and open/close
valves. For a complex network, it is almost certain that an objective control
system could identify a better overall solution than can be achieved by human
judgment alone.
|
