To try and establish the effects
that one or more reservoirs would have had on the current system we wrote a
deterministic network flow model. This model took into account newly
proposed enhancements to the scheme and looked at the effects that they
would have had historically on the reliability of the scheme. The model
determined a historical optimum release schedule from any reservoirs placed
within the scheme. Generation
benefits are also modelled to determine any ill or beneficial effects and
overall economic benefit.
The actual model is not displayed
here, the diagram below shows a simple network flow model that is similar
to what was applied to the RDR.
The model was written in Python 2.7 using the open source PuLP modelling language
and solved using CPLEX
12.1. A full explanation of how to run and edit the model, along with
detailed annotated code can be found in the accompanying user guide.
The model was
only solved for a period of two years due to memory restrictions. When
solving for a two year period we had to take into account the behaviour of
the model at the end of the period. We firstly stepped the model through
one year at a time so that each season was optimised twice, once as the
first season and once as the second season. The levels of each reservoir
were noted and also the dual price of the maximum storage constraints, this
showed that in each of the ten seasons there was a “reset point” where
storage was consistently full. Knowing this meant that when running the
model for different storage configurations we could start with full storage
and also constrain the model to end with full storage.