For many people in the water supply business the concept of models operating on a daily basis in the control room is pure fiction; for others it is the “Holy Grail”, the ultimate development in modelling and system integration. How far have we come and what is truly possible? Let’s look at the facts.
1. What are the benefits of developing operational models ?
In general there are three potential uses for Network Models in the control room:
- Improving Operational Management of the network.
- Monitoring and improving Water Quality and Level of Service
- Emergency system management
Later we will consider the benefits of each of these uses in more detail, but we will firstly consider the current situation and the considerable work that will be required in order to be able to take full advantage of modelling in a control room environment.
2. The current situation
Many Water Providers are now using Geographical Information Systems (GIS) as the main source of asset information for their network. This gives an accurate representation of what the system comprises; but a modelling system can become the main source of information on how the supply system works, increasing staff effectiveness at managing the system by giving staff instant access to operational information. There are limits to what you can do with telemetry, personal experience, and going-and-looking. What you need is a model you can trust to give you reliable results when you are under pressure. But it needs to be up to date (from GIS), properly calibrated, and reflect current metered values.
So how many water providers have fully constructed and reliable strategic or detailed models covering both their Raw and Treated Water systems? In practice very few. Many have only partial coverage, and of those constructed models how many are regularly updated and maintained? Without this fundamental base the concept of models in the control room is indeed fiction.
So what are the essential stages in achieving this, and will the considerable investment be worthwhile? Many water providers are now asking these very questions. The way forward isn’t always clear, and the path is littered with potential bottlenecks and rough spots, not to mention the ever tightening financial controls being imposed in the drive for reduced costs. Can we as water system managers convince those that control the purse strings that the initial investment will reap far greater long-term financial benefits in the form of more efficient operation and better customer service?
Of course we have to raise the question of whether there is software available that can help us to meet this objective? For example, can any currently available software accurately model a possible combination of pressurised and non-pressurised flow, and give at least an acceptable representation of the effect of treatment works on the overall operation. We will address this issue later.
3. An example of an Operational Model
Let us consider the three main benefits of operational models stated above by using a hypothetical example:
Our example network is the supply and distribution system of a small Island. Water is supplied via a desalination plant and several borehole and surface water sources. Two Water Treatment Works receive the raw water and distribute it to separate High Level and Low Level zones. A total of 12 individually metered distribution zones are fed via a series of pressurised /gravity fed trunk mains. The High Level Zone comprises zones 1-7, zones 8-12 making up the Low Level Zone.
3.1. Benefit 1 - Improving Operational Management of the Network
3.1.1. Analysing System Performance
Performance of the raw and treated water networks can be assessed through a range of demand conditions. The modeller can assess:
- Is the system meeting operational targets?
- If not – why not. What can be done?
- If it is – can it be improved and/or operating costs reduced ?
How will the system perform when the demand grows and the system condition may have deteriorated?
Ongoing monitoring of system performance:
- Are you achieving the levels of service your customers expect?
- Are PRV settings suitable for current system state of development? Take account of or advantage of seasonal variation in demand.
What would be the best settings?
Assisting in minimising disruption due to planned maintenance and rehabilitation.
3.1.2. Holding Down Pumping Costs
An operational model can be used for estimating, monitoring and holding-down pumping costs. Where electricity providers operate a varying tariff system, it would be cost effective to make better use of storage and introduce selective pumping.
In our example network the impounding reservoirs may be a mixture of smaller reservoirs with large catchments (fast fill) and larger reservoirs with small catchments (slow fill). In order to preserve water stocks, the model could be used to develop an improved operating regime whereby sources are used selectively and water is pumped between the reservoirs. The pumping could also be optimised to reduce costs if varying tariffs are used.
3.2. Benefit 2 - Improving Water Quality and Level of Service 3.2.1. Raw Water Quality
A potential benefit of an operational model would come through management of water quality. Let us consider that a raw water system receives water from a number of different sources of variable water quality. Perhaps water supply comes from a number of impounding reservoirs, a desalination plant, and a number of boreholes. Monitoring devices would need to be installed at all sources and at the receiving treatment works.
Let us assume that some of the sources have a pH of up to 8.0 and the desalinated water has a pH of 8.5. This has to be reduced to about 6.4 in the treatment works. With the monitoring in place, if the model could track the movement and combination of water from the sources, then chemical dosing could be better controlled at considerable cost saving.
Perhaps some of the reservoirs produce water with a nitrate level that is higher than the target threshold. It may be more cost effective to desalinate water and mix it with the reservoir water, so reducing the overall nitrate level below the threshold, rather than removing nitrate. This provides an acceptable water quality whilst removing the need to dispose of the residue from nitrate removal. This saving often more than offsets the cost of desalination. Using a combination of monitoring flows and a model of the raw water network, the proportion of water taken from each source can be monitored and controlled. This will lead to better management of nitrate levels and optimal use of water resources.
3.2.2. Treated Water Quality
If water quality is monitored at strategic points throughout the distribution system, these data can be used to update the on-line water quality model calibration. The model will then provide an accurate estimate of the quality of the water supplied to every customer in the distribution network. The model can then be used to improve quality of water delivered to the customer.
3.3. Benefit 3 - Emergency System Management
A well calibrated and trusted model is an effective tool for helping to effectively manage emergency situations. The Operator can use the model to confirm the extent of the emergency and to develop and test the most appropriate action to take to minimise disruption to customers. Evaluating scenarios on the model is considerably more efficient than sending operatives out to change valve and pump status and trying to work out the effectiveness of those system changes. In a short period of time the operator can establish the extent of the emergency, the customers that have been and will be affected by the incident and can develop an effective plan of action.
Of course this reactive approach to emergency system operation may be satisfactory, but strategic use of the operational models for contingency planning could be an effective tool in reducing response time and improving staff performance in difficult high pressure situations. Knowing you are well prepared to deal with events encourages rational thinking, reduces stress and demonstrates a level of professionalism that can be beneficial when the media spotlight is turned on an organisation in the event of an emergency. The operational model can be used to establish suitable action to take in the event of such situations as:
- Fire flow
- Major burst
- Terrorist attack / Security risk
- Pollution incident
- Loss of impounding reservoir
- Loss of service reservoir
- Loss of treatment works
4. Further benefits of developing operational models
Of course these three main benefits can easily justify the development of operational models in many cases. However there are a number of other benefits that should not be overlooked, as they can also reap rich rewards.
The ultimate aim is to produce a single model that combines the raw water system with the treated water network. Output from this model would be available to both distribution and treatment managers, giving a common view of what is, after all, essentially a single system. These two parts of the company rarely communicate well in many organisations, but this would help them co-operate effectively.
Linking different systems such as Modelling software, SCADA, Asset Management Systems, Billing data etc. is often complex and certainly requires teamwork, co-ordination and above all a shared goal to make it all work. Unfortunately these qualities are sometimes in short supply particularly when staff turnover is high and the team faces other uncertainties. However shared success often brings better working relationships and this can benefit other areas of the organisations work.
5. What is the level of detail required, and are existing models good enough?
The level of detail required for an “Operational Model” is wholly dependent on the tasks for which it will be employed. A simple strategic model of trunk mains with meters to report all in and outflows will suffice if only an overview of system operation is required. This overview is already available in the control room in many cases using a combination of mimic boards and SCADA & Telemetry system reporting devices, a strategic model would be of little value to this situation. The opposite of this is a model that wholly encompasses raw and treated water networks through the treatment works, down to properties linked to the billing system, bringing with it a number of possibilities, but also requiring significant time and capital investment. In reality something between these extremes would be a both sensible and attainable for most water providers.
The addition of flow measuring devices, pressure and water quality monitors at suitable locations opens up the model for more detailed analysis and a number of more advanced uses. This of course also places further demands on the model construction and calibration, requiring considerably more investment and time, but provides potential for an on-line, real-time, level of service performance monitoring system as well as an operational and planning tool.
Many water providers would have to undertake a comprehensive meter installation and/or telemetry installation programme, and a model building and calibrating exercise to meet even the more basic of these requirements. Others may have fairly advanced metering and telemetry or SCADA systems, but would require work to co-ordinate these systems sufficiently to provide on-line frequently updated data for model operation and predictive studies.
If the model is to go a stage further and be capable of operating in an optimising capacity making use of predictive algorithms, then both current and historical data need to be readily available, and regular maintenance of the monitoring devices is essential to ensure accurate data is constantly available. This is particularly important if the model will be used to directly operate control structures such as motorised valves or pumps based on current conditions and predicted system changes. Building in sufficient safeguards to prevent non-existing or erroneous data from being used in the model is a further problem for consideration. The model should certainly be fitted with “a safety net” in the form of alarms and calls for user intervention, as system failures can be catastrophic, and even if they are minor and irritating the operators will soon lose confidence in the model, and it will rapidly become “another obsolete bit of unnecessary equipment stopping me from doing my job properly”.
6. Who should operate the model – operations staff or dedicated modellers?
To be a truly operational tool the completed model should be used daily by operations staff, who will require training and motivation to perform this task. The skills of dedicated modelling staff will be required for model construction and calibration. Subsequent links to telemetry, SCADA, GIS, Asset and billing systems where required will also require significant staff input from specialist staff, and a good deal of co-operation and good will between all parties.
The ongoing maintenance procedures should be established and responsibilities clearly defined.
7. Is the software currently available to achieve this task ?
A number of software providers will claim to have “seamless” links to other systems, some will even claim to have models in the control room already, but are these systems really as good as their claims, and are they one off bespoke applications or transferable “off the shelf” technology that can affordably be used elsewhere?
One of the problems of course is that GIS, SCADA, Telemetry, Asset Management and Billing systems will generally have been purchased by different parts of the organisation, and often the last consideration was “will this system readily link to all the others that we have or may conceivably purchase ?”. Does it meet my requirements and budget ? is the usual criteria. This means that either the modelling software has to be flexible enough to receive input from these various sources, or some intermediate files are required as a central repository for the data. There is no one size fits all solution.
Another important factor is the strength of the modelling engine, and its ability to model accurately complex operational controls (Programmable Logic Controllers in particular) and real pumps and valves. An operational model has to be just that, operational. It is not good enough to produce modelling fixes that may have been acceptable for initial model calibration. The Operator has to be sure that the model is truly representative under all conditions, otherwise there will be an uncertainty in their mind that at best will result in them losing confidence in the model and at worst could cause a serious breakdown in the system affecting a potentially large number of customers. This is where many current modelling packages fall down. Buyers beware; operational models simply must model the reality.
My advice would be to talk to a number of software vendors, ask them to prove that their system can integrate effectively with the other relevant systems, and test the capability of the model to accurately represent the most complex controls currently in the system (make some up if you are not sure).
8. Developing an Operational Model
Let’s consider our hypothetical project and the tasks required to construct an operational model for this typical Island supply system and integrate this into the control room as a key tool in the management of the supply and distribution systems.
8.1. Phase 1
8.1.1. 1a: Build Raw Water Network Model
Construct a network model of all the links between raw water sources (boreholes, reservoirs, desalination plant and the treatment works). This will be a model that balances throughput, with forced flows. The model would probably be calibrated using historical data where available.
Required data: GIS maps of the water network and digital background maps for display, a suitable log of historical data: one week of flow data, including output from the water sources and input (or output if that is all that is available) from the treatment works. The data should be available digitally in files readable on a PC.
8.1.2. 1b: Connect metered data to the model
It is essential that the raw water network is comprehensively metered, and all meters should be digital, reporting back via telemetry to control panels in the control room automatically.
These logged flows would then be transferred into the raw water model. Instead of historical data, the model will now be driven by actual flows in the network. It is essential to iron-out data timing issues and to get the model running with real data.
8.2. Phase 2
8.2.1. 2a: Expand the model to include the treated water trunk mains
Expand the model into the treated water network, down to strategic meter points. Again it is essential that the treated water network is comprehensively metered, and all meters should be digital, reporting back to the control room via telemetry.
8.2.2. 2b: Add calibrated distribution networks to the model Add all available calibrated distribution models to the model, thus expanding the model beyond a strategic level to provide detailed distribution system information. Any critical control points in the distribution system such as Booster Stations, Motorised Control Valves and PRVs can be linked via telemetry to the control room, and hence provide detailed operational data to the model.
8.3. 2c: Produce a fully integrated operational model The integrated model will be driven by the metered flows from the water sources and by demand estimates on the distribution network. It is essential to provide a means of regularly updating demand on the integrated network.
8.4. Phase 3
8.4.1. Make the operational model available at all relevant sites
Up to this stage, we have assumed that a single model would be running in the headquarters building. The water provider should consider the operational needs of each of the main sites, and make available either the full ability to run operational scenarios on the model, or to view the central model runs and results. It may be necessary to provide further access to run models or purely to view models. Having invested significantly in the development of an operational model it is desirable to allow widespread but controlled access to this valuable source of data. It is effectively a real time representation of the system that could be used for a wide range of uses, such as contingency planning and developing emergency operating scenarios, future development planning, optimising system operation and reducing costs etc.
9. Are Operational Models Reality or just a Pipe Dream ?
Developing an operational model, linking it with other company systems and integrating it into the control room environment is not an easy task. It requires capital investment and a considerable amount of staff time. It also requires teamwork, enthusiasm and a shared goal to make it happen. If all these things are in place the rewards are potentially high, with significant opportunity to improve system operation, reduce costs and most importantly monitor and improve customer service.
For organisations who do not make this management and financial investment, control room models will remain a “pipe dream” or a “Holy Grail”, at least in the foreseeable future. But for other organisations, who are already in good shape and making progress along the modelling systems path, this target is achievable, and models in their control rooms in the next five years is a realistic prospect.