|
PROGRESS TO DATE
The first six workpackages have now been completed to a stage where both case
studies can be evaluated as part of Workpackage 7. Each application comprises
a dynamic, feedforward, near-optimal control system which is capable of being
demonstrated. In the case of Haifa A, the network comprises some 6 demand
zones, 9 storage tanks, 13 pumps and 1 operating valve. The corresponding
figures for the Valencia network are 6 demand zones, 2 storage tanks, 17
pumps and 10 operating valves. In the absence of Supervisory Control and Data
Acquisition (SCADA)
measurements, EPANET data have been used as a surrogate. At each SCADA
update, any errors between the current state of the network and that
previously predicted are grounded so as to minimise error accumulation.
Thereafter, a new operating strategy is generated for the period up to the
operating horizon. Although only the near-optimal control settings for the
current time step are implemented, having a strategy, say, the next 24 hours,
provides a degree of comfort that the tanks can be refilled to a prescribed
level by a specific time of the following day, at minimum pumping cost.
Obviously, if the actual demands are different from those forecasted, then
the control strategy will be amended at the next SCADA update to take account
of any discrepancies, as well as the possible errors in the predicted state
of the network. The control system for each case study has now been run for a
whole year.
PUBLICATIONS
As a matter of policy, very little effort has been made to publicize the
POWADIMA project, apart from this website. Initially, this was due to the
fact that public disclosure would have invalidated the patent application
which has now been filed. Even now, presentations at conferences etc. are
being treated with caution since the intention is to publish the entire
research project as a series of papers in the one journal, possibly as a
special edition. It is thought this will have a greater impact than separate
papers on different aspects in a variety of different journals. However, in
order to preserve this prospect, it is important that the main features are
not disclosed prematurely as conference papers etc. since the more
prestigious, refereed journals will not accept pre-published information.
|
