THE professional use of plant modelling throughout all phases of plant control implementation could be used more widely in hydro power plants. Today’s modelling is based on simulation packages that use modern languages and established libraries of hydro plant components. The models of existing plants or new plants can help ensure in-depth knowledge of the plant’s behaviour is available directly from computers, aiding the control engineers who are working with plant analysis and control implementation.

Of course, the model based approach is meaningful only if reliable and cost effective tools for its development are available, and if the organisation can be convinced that the control implementation can be configured around the commonly accepted model of the plant.

Simulation tools and libraries


Cost effective modelling requires well-established simulation tools and the availability of model libraries.

Simulation tools should support the engineers developing and using models of the particular plant. As engineers are not
necessarily programmers, the package should be user-friendly and take care of all basic tasks of the modelling, e.g. efficiently solve differential equations, handle events and non-linearity, generate presentations of results and automatically generate model documentation.

Simulation tools should enhance library implementation, validation and maintenance accordingly. The author recommends that a professionally developed library be purchased – ad-hoc, home made developments, quite common today, will not ensure continuity in model improvement, validation and reusability.

Simulation package and language

The Modelica language simulation language, which is supported by the simulation tool Dymola, was developed recently by the non-profit Medelica Association to be used for modelling of physical processes, with particular emphasis on the easy exchange of models and model libraries. The reuse issue is solved by making the language well suited to an object-orientated approach when building models. The other main feature of Modelica is its non-causal approach to module interconnection, which allows bi-directional data flows and removes the need for pre-definition of inputs and outputs. That approach is most suitable when building up a model of a large system. It is no longer necessary to decide in advance what the inputs and outputs of the model must be; instead, the models can contain the constitutive relationships appropriate for the component, for example Ohm’s law. Object-oriented modelling allows behaviour of the lowest level expressed in terms of ordinary differential equations (ODE) without requiring it to be written on state space form. Instead, the models can be written using algebraic equations, the so-called differential algebraic equation system (DAE), without the restrictions against the presence of algebraic loops or implicit non-linear equations.

Detailed description of both Modelica and Dymola can be found in references 1 and 2.

Hydro plant library

The experience detailed here is based on the development initiated through a Master Thesis at Lund Institute of Technology (see Reference 3), and finalised through cooperation of Swedish partners – Modelon AB, Waplans AB and the author are responsible for the commercial version of the hydro plant library.

The models address medium scale plant dynamics (frequency domain up to 10Hz) and as such may be used both for plant control and for general analysis of the plant operation.

The following are the main features of the library:

Modularity – covers all systems of the hydro power plant shown schematically in Figure 1.

Full coverage of plant control and automation systems. New control schemes can easily be developed and tested.

Fully developed thermo-hydraulic basis of the water system models, of the basic structure of interconnected (lumped) volumes and connectors. Figure 2 illustrates this approach applied to the water conduit as penstock, surge tank and reservoir.

All system parameters are ‘physical’, e.g. waterways are given mainly through the tunnel’s penstock geometry.

Water properties are calculated as functions of the water’s state (pressure, temperature).

Models of generators and power systems are basically mechanical, which means modelling of active power only. This approach is good enough for modelling operating states of the plant; no-load operation, synchronisation to the main grid and load rejection. The grid can be changed from the stiff to local, allowing testing remote load/generation changes to decide suitable plant control.

Reservoir modules allow modelling of the interconnected waterways.

Costs of modelling

There are two main costs: the software tools and time spent on the model development. Costs of the initial investment in software tools, i.e. both simulation tool and library, vary today between h20,000 to h30,000 (US$24,000 to US$36,000), depending on the number of licensees and scope of the tools. Yearly costs of approximately 10-15% of the original purchase should be added for service agreements, software maintenance, updating etc.

The models are built in two stages; model configuration, followed by parameterisation of the model. The costs for the first stage depend on the skills of the model developers while the second stage depends on the availability of plant data. Assuming experienced developers and data are easily available, the model can be structured in only a few days, and dimensioned and verified in less than two weeks. Naturally, costs are also dictated by the project’s owner/s.

Model based organisation

Modelling can be either interwoven in the project organisation or added as an extra task. The ‘model based’ approach, can be characterised by the following:

The model is developed in the initial stages of the project.

• The model documents the plant to be controlled, and not only follows but copies the documentation.

The model should be developed initially by the specialists; the developers can be trained within the organisation or hired as consultants.

The project’s technical staff should be able to work on the model, and resolve various project issues that may occur.

The last point implies that even if model building remains specialised, the ability to use the models should be widespread in the project.

The ‘added task’ approach is probably the main reason for the slow progress in the general usage of plant modelling. In this frame of mind, modelling is seen as an extra task normally done late in the project, often as a last resort if something goes wrong; it is not surprising that management often counts modelling as ‘extra expenses’.

Modelling in all stages of a project

Verification of requirements

The main features of the power plant are well known when the control systems is designed. Parameters of the dam, reservoir and water ways are decided and already under construction. Turbines are selected and grid characteristics are well known. So a model of the plant, which can already be prepared at this stage (Figure 3), is probably not entirely thorough but is good enough to be used for:

Estimation of the dynamic properties of the water system, including advanced aspects as e.g. interference of the turbine units using common manifold.

Tests of the new plant in the existing grid. This information will set system objectives of the control system.

Estimation of allowable/necessary start and stop rates of the guide vanes. This information will set requirements on the guide vane servo systems, including main data for dimensioning servo oil supply.

• Identification of objectives for the water level control (influence of the upstream plants), or for other schemes of the joint control of plant units.

A base for deep knowledge of the plant dynamics.

Models of the plant can prove invaluable for writing control system specifications, by allowing requirements to be evaluated, verified and demonstrated. And of course, if faults in specifications are identified and rectified at the modelling stage, expensive repair work down the line can be avoided.

Benchmarking and verifications

Having models for specifications will allow the evaluation of control schemes/components proposed by the supplier. The model will become a common platform for benchmarking offered solutions.

Benchmarking of the supplier may be done during contract negotiations or as the first stage of the contracted project.
The plant model can now be complemented with models of the supplier’s control system. This is an important stage, as it allows an early evaluation of the plant’s operation under the proposed control scheme. Profit will be made through the efficient selection of the supplier and the provision of an easily accessible test bench ensuring early detection of malfunctions. Additional time and costs can be curbed by resolving technical conflicts and responsibilities of the project parties, assuming that project parties agree on model validity.

Costs of modelling cover the time for running various test cases on the model in this stage. Definition, execution and documentation of a single test case will take 8-16hrs, assuming that the plant model is already available. As control models may add some costs it is advisable to ensure the supplier’s cooperation by adding a requirement for model based benchmarking in the contract.

Commissioning and operation

It is a common knowledge that nearly all control implementation projects lose profits during plant commissioning. The following are the main reasons for the prolonged commissioning, along with possible solutions:

Faulty planning of commissioning tests and procedures: using/complementing models from the previous stages will allow the running of all planned tests on the models, reducing the risk of unexpected developments during testing and allowing revision of redundant testing.

Commissioning engineers unfamiliar with the plant: it is quite common that the commissioning engineers learn about the plant for the first time only when they arrive on site. Allowing them better preparation would shorten their expensive time on site. It is highly probable that control experts will not be required to attend commissioning as all presetting of control parameters can be done on the model.

• Conflicts between the project parties: all plants when initially operated are prone to getting into unexpected situations. Such situations are often difficult to analyse and investigate; it takes time and can lead to conflicts between the involved parties. The plant model will not only allow better understanding of data acquired from the plant, but allow the running of operational scenarios to prove various assumptions.

The model of the plant should remain part of the plant documentation throughout the its lifetime. Planning of tests for plant maintenance, training of the staff and analysis of accidents can be enhanced in the same way as presented above for the commissioning. As a result, profit should be made through reduction of the commissioning time and shortening off-load time.

Cost comparison

Considering an investment of up to h60,000 (US$72,000) to cover the purchase of software and basic training, is the cost of utilities modelling really justified?

Let us assume the following cost comparison between a traditional and model enhanced control modernisation of a 50MW turbine unit:

Project initiation and specifications – work time unchanged as traditional project time will be reduced but time added for the model building.

Benchmarking and verification – a probable 20-30% work time reduction corresponding to 200 working hours.

Commissioning and operation – 50% time reduction is plausible corresponding to 100 work hours of the commissioning team.

Costs of lost production correspond, at the energy price of h0.02/kWh (US$0.02/kWh) at 50MWh, to h1000/h (US$1200/h). Assuming prevention of one 24h stand-off period during the first five years of plant operation, the total is h24,000 (US$29,000).

Using modelling could mean a cost reduction of approximately h55,000 (US$66,000), i.e. the initial investment repaid in one project only. Yet still the perception is that the main advantages of modelling lay in risk reduction. Faults in specifications can lead to several months of plant stand-still and huge costs for redesign of the wrongly implemented equipment. But on the evidence here insurers are not the only benefactors of the modelling. Admittedly the changes in project routines suggested here may be difficult to realise. But can the water power industry really afford to ignore the potential for increased profit and decreased risk that utilities modelling may offer?

Author Info:

Jan Tuszynski is with Datavoice. Email: