Introduction
Overview
Development Philosophy
Definition
of System Components & Configuration
Input of Technical & Cost
Data
Data Units
Model & Data Storage
Simulation Module
Simulation Modes
Simulation Outputs
Firm Yeild Deterimation
Optimisation of Long-Term Operating Policies
Program Details and Usage
Applications
Availability
Click Here For Aquarius Presentation
Outline Description of Program AQUARIUS

Introduction top
Rising water demands, increasing fuel costs, and the threat of greater flow variability due to climate change, are further heightening appreciation of water as a scarce and valuable resource. It is thus vital that water resource and hydro-thermal power systems are operated as efficiently as possible.
Benefits from the conjunctive use of water sources and different forms of electricity generation have led to the development of a range of computer programs for simulating and optimising system operation over both long and short time scales. However, computational constraints have often dictated that significant simplifications be made when simulating and optimising policies designed to minimise long-term operating costs while satisfying specified supply reliability constraints. Continuing advances in computer technology now make it feasible to create generalised software which eliminates the need for many such simplifications, enable models to be rapidly constructed and modified, whilst providing greater transparency of the decision making processes.
Overview top
Program AQUARIUS has been developed by PWSC to optimise and simulate the operation of water and power supply systems. It has been specifically designed to provide:
Development Philosophy top
AQUARIUS builds on experience gained from the successful application of PWSC's earlier programs for simulating water resource and hydro-thermal generation systems (Program SYSIM), and for simulating and optimising the operation of large scale water resource/water supply systems (Program MOSPA). Initially constructed as FORTRAN 'batch' programs, SYSIM and MOSPA incorporated logical rules to allocate resources in each simulation time step, rather than formal optimisation algorithms, so as to contain computation times when simulating complex systems over long hydrological sequences. Subsequently, Windows based GUI's were developed for both programs in order to simplify the entry and modification of input data, and to provide comprehensive facilities for the screen and hard copy display of results in tabular, graphical and 'mimic' diagram form.
The impetus for developing AQUARIUS arose from a number of considerations and observations associated with the evolution of water supply and power systems, including:
The scope and design of AQUARIUS has also been influenced by continuing advances in computing power and disk storage, which now enable the incorporation of analytical methods not previously compatible with detailed and lengthy simulations of complex system operation.
At the same time, the development of programming languages such as Microsoft's Visual Basic now enables 'engineering' type programs to be written directly within a Windows compatible environment while having execution times comparable with versions written in 'scientific' languages such as FORTRAN. An additional advantage of the Visual Basic language is that it allows direct read and write access to common database structures. Such considerations led to the following key features being incorporated within AQUARIUS:
Simulations can be carried out at with a level of detail appropriate to the available data and application objectives, while the long-term optimisation methodology ensures that the resulting operating policies are 'practical' in terms of complying with constraints on system operation.
Use of a variable dimension system so that computer memory requirements can be tailored to the size of the system being modelled and remove limits on the number of components that can be accommodated.
Definition of System Components and Configuration top
AQUARIUS enables the 'on-screen' construction of a simulation model for a given water resource, water supply and electricity supply system. The user selects, from a menu, the component type to be added to the current diagram and positions the component by 'clicking' the mouse. AQUARIUS currently allows the user to select from the following Component Types.
Water Resource/Supply System Components
Reservoirs, Flow Points, River Reaches, River Abstractions,
Pumping Stations, Transfer Aqueducts, Supply Aqueducts,
Water Demand Areas, Water Sources (e.g. groundwater
or bulk supplies) and Treatment Works.
Electricity Supply System Components
Hydroelectric Plants, Thermal Plants, Wind Power
Plants, Pumped Storage Units (i.e. reversible turbines),
Transmission Lines, Transmission Nodes, and Electricity
Demand Areas.
Transmission lines, transfer aqueducts, supply aqueducts and river reaches are 'connected' to appropriate component types by firstly clicking the mouse on the 'from' component, and then dragging the cursor to the 'to' component and clicking again. In-built logic restricts the component types that can be connected; for example, a transmission line can only go 'from' a hydro, thermal or wind power plant 'to' a transmission node or electricity demand area, 'from' a transmission node 'to' another transmission node or electricity demand area, or between two electricity demand areas.
The position of individual symbols and their assigned names can be adjusted using 'drag and drop' to make the diagram visually acceptable and reflect actual topography. While moving symbols only alters the appearance of the diagram, deleting components or changing connections will alter the logic of the model. A specimen model diagram, created using AQUARIUS for a demonstration system, is reproduced on the front page of this description.
Input
of Technical and Cost Data top
With
the exception of stream flow, wind speed and demand
time series, all input data to AQUARIUS is entered
via a collection of 'on-screen' templates which
vary with Component Type. These enable both initial
data entry and the modification of existing values.
The appropriate template is accessed from either
a menu or by 'right-clicking' a component on the
model diagram. An example of an input screen for
entering reservoir data is shown as Figure 1

Figure 1 : Input data Screen for Reservoir Characteristics
System level data input to AQUARIUS includes :
Input operating cost and penalty cost data used by AQUARIUS to calculate the least-cost of meeting specified electricity and water demands within each simulation time step include:
Technical and demand data that can be considered during each time step of the simulation include:
Daily time series of electricity demands, water demands and stream flows can be input to the simulation from individual disk files, the names of which are user defined as part of the model data. As previously noted, user defined seasonal profiles can be associated with a large number of input 'daily' values so that these can be varied either as a function of calendar month or 'day-of-week'.
Data Units top
The water volume unit used within AQUARIUS is the Megalitre (Ml), with flow rates in Ml per day (Ml/d). However, water related data can optionally be displayed and edited in Cubic Metre units i.e. with volumes in million cubic metres (Mm3) and flows in m3/s. The power unit used is the Megawatt (MW), and the energy unit is MegaWattHours (MWh).
Model and Data Storage top
With the exception of individual time series data files, all information relating to a particular AQUARIUS Model is stored in a single comma delimited disk file, and automatically given the attribute *.mdl. Thus any changes made to an existing Model, including those simply data related, can be saved under a different file name for later retrieval using AQUARIUS. At the same time such files may be archived, so as provide a permanent record of particular program run inputs for, say, regulatory auditing.
Simulation Module top
The steps performed during an AQUARIUS simulation are shown in flowchart form as Figure 2.
The start and end of the simulation period are set by the user, with the duration only limited by the coincident period for which any input time series are available. The minimum simulation time step is one day, but simulations can also be undertaken using calendar weekly or monthly time intervals.

Figure
2 : AQUARIUS -
Outline Flowchart of Simulation Module
In all cases, the outputs ('dispatch') of any power generation plants and transmission line flows are optimised across a number of user defined load blocks, corresponding to the definition of imposed electricity demands. The duration of each load block is defined on a system basis with the total duration summing to 24 hours.
Within each simulation time step, AQUARIUS minimises the total cost of satisfying the imposed electricity and water related demands using PWSC's proprietary Linear Programming (LP) algorithm.
The LP input matrix is automatically constructed on the basis of the specified model components and topography (linkages), and the associated physical and economic data, including imposed constraints. To avoid the occurrence of 'infeasible solutions', all electricity and water demands, including minimum river flows and reservoir compensation releases, are subject to (user specified) penalty costs which are applied in the case that a demand cannot be satisfied.
Unit benefit costs can be assigned to electricity supplies (by load block), water supplies, reservoir compensation and minimum flow satisfaction. This can be used for allocating supplies of energy or water in times of unavoidable shortage.
In order to minimise execution times, the basic input matrix is created for the first simulation time step and only those elements which change between time steps are modified. For similar reasons, the maximum dimensions used by AQUARIUS e.g the maximum number of reservoirs or transmission lines etc., can be adjusted by the user so as to minimise computer memory requirements.
Within the LP formulation reservoir releases can be assigned costs in accordance with input (long-term) Stored Water Values as a function of month and reservoir content, and such releases are optimised taking account of (downstream) incremental inflows within the river system. Hydro plant capacities are calculated as a function of the available flow and a MW per m3/s conversion factor. This conversion factor can be fixed or given as a function of flow and available generating head, while the latter may in turn be dependent on an upstream reservoir level.
Simulation Modes top
The following different types of simulation can be performed using AQUARIUS.
Simulation
Outputs top
'Real Time' graphs
As a simulation proceeds AQUARIUS can simultaneously display up to 15 'real time' screen graphs to show the behaviour of selected system components, as indicated below :
As shown in Figure 3 below, a summary of simulation results to date is also provided, and includes: Stored Water Cost ($), Daily Direct Cost ($) & Average Daily Direct Cost [to date] ($), Daily Penalty Cost ($) & Average Daily Penalty Cost [to date] ($), System Storage (Ml) & Lowest System Storage [to date] (Ml), Water Supply Deficit [to date] (Ml) & Electricity Supply Deficit [to date] (MWh).

Figure 3 : AQUARIUS -
Example of 'Real-Time' Screen Graphics Output
Run Time Log
At the start of each AQUARIUS run a modelname.log file is created to which run details and summary results are written. The dates and quantities of any water or electricity supply shortages and failures to meet compensation or minimum flow requirements are also logged.
Database Outputs
For each time step of the simulation, information on the inputs, outputs and performance of each component included in an AQUARIUS model, and for the overall system, are stored in a model specific Microsoft ACCESS© 2000 format database. Tables are also included for Drought Conditions and Time Dependent Licences :
A new database ‘modelname_SIM.mdb’ is created at the start of a simulation and summary records are automatically produced showing average values for the simulated period. All items in the database can be viewed directly from AQUARIUS, with facilities to step forwards and backwards by record, day, week, month and year. An example is shown as Figure 4.

Figure 4 : AQUARIUS -
Database Contents View Facility
Time Series Plots
Any combination of time series values stored in the database can be plotted and viewed on screen or printed with high definition. Values are automatically summed to provide weekly, monthly or annual time series, or to give average values over the total simulation period. Line graphs (Figure 5), vertical bar charts, area graphs (Figure 6), or pie charts can be produced. The user is able to 'scroll through' values over the simulated period, and a 'hot hit' facility is included so that any plot point can be identified in terms of the component name and associated value. The Graphics Server© facility employed gives the user control over the content and appearance of any graph so that they can be tailored for inclusion in reports. Graphs can be output to disk file in ‘bmp’ or ‘wmf’ formats.

Figure 5 : AQUARIUS -
Example of Line Graph Using Stored Database Values

Figure 6 : AQUARIUS -
Example of Stacked Area Graph Using Stored Database Values
Selected time series can also be viewed in tabulated
form and output as *.CSV files, for direct input
into spreadsheet programs such as Microsoft EXCEL©,
for further analysis.
Mimic Diagram Displays
It is often instructive to study and display system operation in a spatial perspective, and AQUARIUS provides facilities for this using so-called ‘mimic’ diagrams. Such diagrams can be constructed interactively based on information contained in the model file, and the display location of each model component can be adjusted using ‘drag and drop’. Location, size, colour and other information is stored in *.MIM files for future retrieval.
Mimic diagrams can be used to display a wide range of values recorded in the simulation results database. For example, reservoir values that can be shown include: average % active storage to date, average content to date, compensation release, end content, end % active storage, evaporation, minimum % active storage to date, minimum storage to date and spill. Marginal costs can be shown at water and electricity demands and at junction points and transmission nodes.
Flows in Ml/d or m3/s through individual aqueducts and river reaches can be shown, together with any losses. For power plants, transmission lines and transmission nodes, energy flows in GWh or power flows in MW in each load block can be displayed. Total system values are automatically shown below the main diagram, as shown in Figure 7 below.

Figure 7 : AQUARIUS -
Mimic Diagram Display of Simulation Results for Example System
Values corresponding to each simulation time-step can be displayed as well as average values for the simulated period. The former facility is particularly helpful when developing new AQUARIUS models, while the latter can be used to provide high definition graphical outputs for inclusion in reports.
Load Dispatch Details
When modelling power supply systems, it is often instructive to inspect the way in which each individual generation plant is dispatched in order to meet the electricity demands imposed. AQUARIUS therefore includes a facility to graphically display the output from each hydro, thermal and wind power plant in each load block for any simulation time step.
As shown in Figure 8, a ‘step through’ facility is provided which enables the user to rapidly locate a time step of particular interest and, as with all graphical outputs, high definition printer output can be produced.

Figure 8 : AQUARIUS -
Graphical Presentation of Optimised Load Dispatch
Details
A ‘hot hit’ facility is also provided so as to identify the output from each plant in each load block.
Simulation Summary Tables
A summary of simulation results can also be viewed in the form of tables for selected component types. Currently such tables are generated for Reservoirs, Hydro Plants, Thermal Plants, Water Demands and Electricity Demands.
Firm Yield Determination top
Planners
and regulators frequently request an assessment
of the maximum water or electricity supplies that
can be met by a system, based on given flow sequences
and supply reliability criteria. For water resource/supply
systems such estimates are sometimes referred to
as 'firm yields' or 'deployable outputs'. For hydro-thermal
power generation systems they are analogous to
'firm energy' assessment if a single load block
is employed in the simulation.
AQUARIUS incorporates a search procedure for identifying
the demand multiplier consistent with satisfying
such criteria, and also permits the user to select
the individual water and electricity demands to
be subjected to the multiplier. A key attribute
of the approach is that the assessment takes into
account the constraints on system operation incorporated
within the AQUARIUS simulation model.
Optimisation of Long-Term Operating Policies top
Within each AQUARIUS simulation time step, the quantity of water to be released from a reservoir can take account of one or more of the following types of operating rule.
For all such rules the critical reservoir contents can be expressed in terms of weekly or monthly values and, if required, be subject to linear interpolation. The frequency of inspection can also be set to daily, weekly or monthly, so as to reflect operational practicalities.
AQUARIUS enables each type of rule to be optimised with PWSC's proprietary Policy Iteration Stochastic Dynamic Programming (DP) algorithm, which has been used to optimise the long-term operation of major hydro-thermal power generation systems1 and for optimising operation of the largest and most complex water resource/supply system in the UK2. The way in which this algorithm is used in conjunction with the simulation module is illustrated in Figure 9.
1.
Wyatt T. & Robinson P.E.
'Medium-Term Operating Policies for Hydro-Thermal
Systems'Modern Power Systems, January 1988.
2. Walker S., Walsh P.D. & Wyatt T. ' Derivation and Application of Medium-Term
Operating Policies for the Northern Command Zone System of North West Water (UK) IAHS
Baltimore Symposium, May 1989.

Figure 9 : AQUARIUS -
Integrated Use of Simulation and Optimisation Modules
Operating policy optimisation data and results are stored in a Microsoft ACCESS© 2000 format database, which is automatically assigned the file name modelname_OPT.MDB. Contents of this database can be inspected using the facility illustrated in Figure 4.
Experience shows that application of this fully integrated simulation/optimisation technique results in practical operating policies with potential savings of more than 10% of average long-term operating costs when compared with more conventional methods of operation. Stored Water Value rules can be particularly effective for balancing releases in multiple reservoir systems, and provide information often required for short-term optimisation models.
An example of a two regime weekly Water Value rule optimised using AQUARIUS is shown as Figure 10 below.

Figure 10 : AQUARIUS -
Optimised Two Regime Water Value Rule
Program
Details and Usage top
Program AQUARIUS is written
in Microsoft Visual Basic© (Version 6), and has been tested running
under the Microsoft Windows 95, 98, NT4, 2000 & XP
operating systems. All code, including that associated
with the proprietary Linear and Dynamic Programming
algorithms, has been written by PWSC, thereby eliminating
reliance on any third-party suppliers.
The execution times required for simulating the performance
of a water or power supply system using AQUARIUS will depend upon a number of factors, including:
In practice the level of detail used for a particular simulation will vary with the application. Thus, for planning studies it may be sufficient to employ, say, three load blocks, whereas for optimising 'day ahead' system operation the use of 12 or even 24 load blocks might be appropriate.
AQUARIUS allows users to switch between daily, weekly or monthly time steps and, in this way, investigate the sensitivity of the simulation results to the time interval employed. Hence, while single simulations might be made with a daily time step, it may be acceptable if those made within the yield search or long-term operating policy optimisation process employ a weekly or even monthly interval.
When optimising long-term operating policies a further determinant of execution time is the number of alternative operating regimes used when producing the input data for the optimisation process, since a separate simulation is performed for each.
Applications top
AQUARIUS is equally applicable to planning and operation studies as for use in optimising long and short term operation, and models have been built representing the following systems.
Availability top
AQUARIUS is available for purchase by individual
utilities, subject to standard software protection
measures being installed. Usage by consultants and
international agencies on a project-by-project basis
is by negotiation. For further details on AQUARIUS and other software for optimising the development
and operation of hydro-thermal and multi-purpose
water resource systems, please contact us.