WRM optimization scenarios
To configure and drive and optimization scenarios
WaterWare manages Optimization Scenario OBJECTS
The OptimizationScenario OBJECT CLASS builds on
The OBJECT itself consists of
- a WRM base scenario
- a matching STREAM scenario
CONSTRAINTS and INSTRUMENTS are again implemented as OBJECTS,
their attributes as DESCRIPTORS so that eventually the rule-based
expert system can be used to generate values conditionally
(context sensitive, adaptively) to implement flexible heuristic search strategies.
- the header and META DATA including a description !
- scenario level information that provides the link to the WRM
(and STREAM) scenarios;
- a set of CONSTRAINTS
- a set of INSTRUMENTS
Generic INSTRUMENTS are managed in the
Water Technologies data base.
The top level scenario object shows a summary of these two sets.
CONSTRAINTS are used to
describe the OBJECTIVES or GOALS for the optimization scenarios;
the first step of the underlying multi-criteria method
is satisficing, i.e., the primary objective is to meet the constraints
to find one or more feasible solution.
The biggest advantage of this seemingly simple approach is that is makes
representation of multi-criteria multi-objective problems easy:
any number of CONSTRAINTS can be expressed in a problem oriented natural language
without any regard to preferences and trade-off, different constraints usually
corresponding to different stakeholders and interests.
The CONSTRAINTS define a desirable or at least acceptable region in model behaviour space:
they represent what the various stakeholders want or expect from the system in terms
of the model. CONSTRAINTS can be valid for the entire simulation period or
only some part of it (if they apply to seasonal requirements).
- Global Constraints
The global (default) CONSTRAINTS
are formulated in terms of:
- The concept name (the Descriptor display name from the systems knowledge base)
- strategy (minimize, maximize, minimize deviation from a norm)
- the nuumerical value the constraint is set at;
- the unit of measurement
- a tick box to toggle contraints on or off;
- at the results level:
- the number of trials that failed violating this constraint
- the reslut value for the current feasible solution
- the average, minimal and maximal values for the feasible subset and total set of trials.
- Sectoral and Node Specific Constraints
INSTRUMENTS define the control alternatives available to reach an optimum solution,
the define the search or decision space for the optimization.
Alternatives can define elements of the structure of the network
or the behavior of any or all nodes.
INSTRUMENTS can be
INSTRUMENTS can be defined for:
- Discrete, integer, or binary (selected or not, e.g., a new reservoir);
the models selects either on (1) or off (0).
- Continuous (i.e., they can be applied to some degree from 0 (not at all) to 100%
e.g., lining an irrigation canal: out of the 10 km any part or percentage
could be affected. The model selects an appropriate percentage of application.
- Mutually exclusive (can not be combined with certain other INSTRUMENTS);
- Additive or multiplicative (can be combined with other INSTRUMENTS);
INSTRUMENTS for individual DEMAND NODES are defined by:
- different SCOPE:
an INSTRUMENT can, in principle, affect
Please note: in the first installation, SCOPE is restricted to the
local or NODE level to simplify the implementation as well as scenario
- the entire basin (for example, overall policy or an education campaign)
- a specific sector (sectoral policy, incentives);
- a specific local NODE.
- NODE TYPE: At the local or NODE level, INSTRUMENTS can be defined for:
- START NODES, for example, inter basin transfer, desalination, or stronger pumps
for deeper well and well fields;
- RESERVOIR NODES, that can switch on or off individual/new reservoirs;
- DIVERSIONS that can be used to modify diversion rules by:
- modifying the diversion percentage split within a user defined range around the original value;
- modifying the diversion target flow (multiplier);
- modifyinh the downstream target flow (multiplier).
- DEMAND NODES that affect demand,
losses, and consumptive use for this NODE CLASS.
- Name and META DATA
- optional link to a Water Technology
from which its parameters (Descriptors)
get imported as defaults, that can be overloaded at the scenario level.
The Water Technology Object also provides a textual description of the technology.
- Any or all of the parameters listed below may be set to zero or 1.0 and thus,
depending on their nature, render them ineffective as the technology does NOT affect
the corresponding concept (e.g., consumptive use).
For each instance of SCOPE (a specific demand node) the data set includes:
- minimum application rate: the minimum level at which this technology can be applied,
can be non zero The interpretation is as follows: a given technology is either
applied not at all (0), or at least for N (say, 20%) of the corresponding node;
for example, a project lining an irrigation canal partly would have to cover at
least 20% to be feasible at all due to the fixed costs involved.
- maximum application rate: may be less than 100%; a technology may not
be realistically be applied to 100% of a demand node, for example,
water saving shower heads in a city.
The optimization will find the best value within this range.
For a discrete alternative (no scaling, can only be selected or nor), the
minimum as well as maximum are set 100%.
- demand multiplier: scales the demand, corresponding to the application rate;
- consumptive use multiplier: scales the consumptive use,
represents different efficiency of water use of the technology;
according to the application rate;
- conveyance loss multiplier :
- return flow loss multiplier:
- Cost factors FIXED (annualized investment) and VARIABLE
(operations and maintenance); these are meant to refer to a 100% application rate.
initially represented with a simple linear function of flow:
cost = FIXED * application rate + VARIABLE * reference_unit
where the ORM reference unit can be: per day, per m³ or by investment reference unit,
in the latter case to be scaled by the application rate.
The Alternative Generator, based on its STRATEGY, selects or generates new parameter values
that overload the original specification from the WRM base scenario)
and creates the (sample) input vector for a new model run within
the optimization framework.
The generation of alternatives is using a complex heuristic strategy.
Currently, however, only a simple Monte Carlo scheme is implemented.
A two-step proceudre is used:
- In a first step, we select which instruments should be applied;
this is based on a pseudo-random number in the intervall between 0.0 and 1.0
values below 0.5 lead to rejecting the instrument, values >e; 0.5 will select the instrument.
The instrument specific weight is used as a multiplier to increase
or decrease the probability of an instrument to be chosem by
multiplying the original random number.
- In the second step, the instrument configured:
another random number is generated, its value is mapped in the intervall between the minimum
and maximum application rate defined for the instrument,
the resulting application level is applied.
Combinations of INSTRUMENTS
For the same instance of SCOPE (the entire scenario, a sector, an individual node),
more than one instrument can be specified.
Some of these INSTRUMENTS can be combined (say, canal lining and an education program),
others are mutually exclusive (canal lining and a pipeline,
or different irrigation technologies for the same irrigation object).
For a given optimization scenario and its set of INSTRUMENTS,
an EXCLUSION MATRIX needs to be constructed that
specifies which INSTRUMENTS can jointly be considered for any SCOPE instance.