Introduction
A couple of weeks ago, I wrote a brief post on our company website (Aurora Power Consulting) about EMT modelling, and it generated quite a bit of interest. This made me realise that unless you are directly involved in power system modelling and simulation, these issues can be a bit abstract and are a long way from the experience of most power engineers. Of course, there is no escaping the issue that power system modelling, and analysis is difficult and academic. The reason it is so important, is that power systems equipment is expensive and has long lead times, so modelling and analysis gives developers and network operators valuable insight into how a system will perform during steady state, fault and transient conditions in order to help keep the lights on.
As noted in the previous post, there is a general drive to move away from RMS (Root Mean Square) type simulation and use EMT (Electromagnetic Transient) modelling techniques to assess power systems within the industry. This is due to an actual (and sometimes perceived) shortfall within RMS modelling techniques. Whilst this may seem very academic it can have very real influences – the most recent case in the UK being the large amount of generation that unexpectedly tripped during the 2019 outage. This was in large due to problems with Fault Ride Through (FRT) capabilities of inverters and the associated control systems.
RMS vs EMT modelling is an interesting issue from Aurora’s point of view, as it brings many opportunities (we like power simulation modelling). However, it also brings a number of challenges and issues, both for us and our clients. This article is written from a U.K. / European context, but the general principles would also apply to practices in North America, Australia and many other parts of the world.
Modelling Approaches
Before discussing the details of RMS vs EMT studies and modelling, it is worth remembering a famous quote by a Statistician called George Box. “All models are wrong, but some are useful.” This is an important point, as a studies engineer should always remember that a model approximates reality, some models are good representation and others are less so. The suitability of the model and the analysis must match the level of understanding needed.
RMS models are very good for many cases, but they do have to simplify certain parts of the network representation, and this causes limitations. This means that certain studies can fall short in areas; specifically, RMS models and simulation techniques do not perform well in unbalanced, or fast transient conditions. Another possible limitation with RMS modelling techniques, is that as system strength falls (short circuit levels), the transient response of systems and control systems becomes more complex and may not be adequately captured in RMS simulations. EMT models are much more accurate, but they also have limitations and problems, and it is important to note that even an EMT model does not fully represent a real network. This is discussed in more detail later.
RMS Modelling
The first thing to be clear on with RMS modelling is the calculation method. The majority of RMS modelling is based on a balanced representation of the power system network and the calculation carries out all the calculations on an equivalent positive sequence basis. What this means, in simple terms, is that RMS simulations do not model individual phase values. Furthermore, RMS simulation techniques are based on a time step of (usually) 10ms, as this has been judged sufficient to capture general trends of system behaviour, so it will not capture any transient that is faster than this.
The reasons for an RMS based approach is that it is uses simpler formulas and is computationally efficient, and many of the traditional stability studies required tended to occur relatively slowly over a few seconds. This approach was adopted at a time before computers were as powerful as they are today, and simulation time was a major constraint. For smaller industrial systems this may not seem as important, but when operating at a transmission / country level, the computational time can be significant, running into several hours. Over the last few decades RMS simulations were found to be a pretty reasonable approximation of the overall system behaviour. A further key point to note is that for most steady state analysis, high frequency components (which RMS techniques do not catch) are not actually that important for loadflow analysis or simplified stability analysis.
To slightly complicate things a bit further, some software packages offer an unbalanced RMS calculation method. This is an interesting development, as this calculates individual phase values, during the calculation and therefore overcomes many of the traditional shortfalls of a positive sequence RMS calculation. However, this approach still, uses slightly simplified formulas, a slow time step, so it will not fully capture transient events. The debate then becomes if an unbalanced RMS simulation, with a faster time step are considered a reasonable approach to show FRT behaviour accurately enough.
What does all that mean? Positive sequence tools are very useful for providing analysis of large power systems if: a) the analysis does not need to capture fast transients and b) it does not need to consider unbalanced conditions. RMS models are excellent for modelling large power systems and carrying out analysis for power flows, and ‘slow’ transient behaviour, such as system stability. Unbalanced RMS simulations provide a good intermediate step that addresses some of the shortfalls, but fundamentally suffers the same underlying issue of not capturing fast transient correctly and using simplified formulas.
At a practical level, concerns are present in relation to modelling of grid following and grid forming inverters, and HVDC links. These are all complex power system elements and the simplified representation in RMS has led to a few notable system events during faults where things have behaved in a way that was unexpected. At the moment Fault Ride Though studies and control stability are the key areas where RMS studies are falling short and need supplementing with EMT studies, but this is likely to develop as renewables becoming more dominant on the system.
EMT Modelling
EMT modelling tools are less commonly used that RMS techniques. This is for a variety of reasons, including more complex modelling requirements, simulation and computation times, limited software options and difficult user interfaces. EMT modelling was therefore generally limited to specialist studies such as insulation coordination, TRV or designing and analysing generators, motors, inverters, or transformers in detail.
What does an EMT simulation do that an RMS simulation does not? Fundamentally EMT analysis uses more complicated equations to represent each component in the power system, and this allows each phase to be calculated individually and solved in a much faster time step, allowing fast transient to be captured.
Even with modern advanced computing processors, modelling whole grid systems at an EMT level, is not widely done due to a) input data, b) practical considerations of processing power, c) ease of simulation and d) bug hunting and false results.
Input data: This perhaps the most challenging one in EMT simulations. As the calculation is more accurate it necessarily requires more detailed and accurate input information. This creates a lot of problems, particularly on large networks that have a lot of data and historical equipment that has been built up over many decades and information may not be available. The engineer building the EMT model, must have a very detailed knowledge of all aspects of theory in selecting and creating the model. EMT simulation packages have a much smaller ‘library’ of data and expect the user to have a deep understanding of the models it uses.
For a more complex case of an inverter-based generator, for example, an RMS model would typically treat it as a simplified impedance and a current source or voltage source with a series of control elements. In an EMT model, it would be necessary to model the power electronics, control systems, input transducers and so on in much more detail. A further problem that arises in EMT simulations is one of validation. EMT simulations are much more complex than RMS, so validation also becomes more complex. Inclusion, or exclusion of what would be minor details in an RMS study, can have a significant effect on the results. Depending on the study in question, this could include differences due to modelling of stray network capacitances, control system parameter, synchronous machine losses, choice of cable model etc.
Solution time: This is much easier to understand. A big model with lots of very detailed mathematical representations using short time steps will necessarily take longer to calculate. This moves simulation time for a few seconds, into minutes or hours. When you are running multiple simulations, and need to consider different scenarios and outage cases, this can become a problem. Some simulation packages have the ability to do parallel processing, to speed things up, and there are also techniques are available to ‘aggregate’ a model of a Solar PV / BESS / Wind farm into an equivalent, but fundamentally EMT models are longer to run.
Ease of simulation: This is perhaps one of the most interesting and challenging problems with EMT models. Most RMS simulation packages have very good tools for setting up lots of different study scenarios, operation cases, contingencies and so on. With a bit of scripting this can let you run through a large number of cases and get general trends and behaviours. EMT packages generally have either little, or no support for this kind of approach, and needs quite a lot of manual operation to set things up. This makes one of the fundamental problems of flexibility of analysis and repeatability of results a lot more challenging. Furthermore, EMT models cannot be easily initialised, to run steady state loadflows. This is because rotating machine initialisation parameters can be worked out in EMT and have to be back calculated from a loadflow solution elsewhere, so, that the machine terminal voltage and angle can be set in the EMT study. Some EMT packages handle this well as they are bundled with loadflow and RMS packages, others can use an external reference result via an interface module and others it needs to be done by hand.
Big hunting / false results: This is perhaps one of the most interesting and challenging areas. One of the most, if not the most important aspect of any modelling approach, is having confidence in the results. This means checking that you model is representing what you think it is representing, this is much harder than is often understood. Both RMS and EMT models will be different to reality, and the validation process of ensuring that the simulation is a close fit to reality is not an easy one, and EMT results can often be confusing, and need a high level of skill to interpret.
In steady state simulations it is easy to ‘eyeball’ the results and see if they make sense or not, with RMS simulations this is harder and needs an experienced studies engineer to assess the results. With EMT studies this is much harder again, and it is very easy to end up with misleading results, either showing a problem when there isn’t one, or not showing a problem when one exists. The skill and experience level for an engineer carrying out such studies needs to be far higher than those for a simple RMS study. The danger with EMT simulations is it can lead to a false sense of complacency and trust in results, that may not be accurate, as there are assumptions and simplifications in the modelling approach that may not be apparent.
Discussion and Conclusion
A good example of RMS vs EMT studies, can be seen in considering Grid Code compliance studies, as required by many system Transmission System Operators (TSO). Most TSOs require a suite of studies to be supplied looking at steady state loadflow, short circuit events, harmonics, frequency response, voltage control and so on; these are all either steady state studies to specific standards or ‘slow’ dynamic studies,a dn cannot usually be done easily (if at all) on an EMT package – EMT packages are too powerful, and cannot simplify the calculations down enough. Fault Ride Through (FRT) studies are fasters, have a number of transients, require unbalanced analysis and consider operation of elements like PLLs, so are much better suited to EMT studies, although many would contend that a well configured unbalanced RMS could produce accurate enough results, if backed up manufacturer validated Hardware in the Loop (HIL) testing.
The TSO requirements for EMT models, thus creates a practical problem as this then requires two different simulation models. An RMS model for most simulations, and an EMT model for detailed analysis of FRT cases. Some packages can handle both, but many cannot. In the case of the U.K., the RMS simulations need to be done in DIgSILENT Powerfactory but the EMT simulations need to be done in PSCAD. There are some reasons for this, as while Powerfactory is powerful, PSCAD is a more established modelling environment for EMT studies. This therefore duplicates effort and slows the whole compliance process down.
I hope this short article has been useful and explained the difference between RMS and EMT simulations. In simple terms RMS simulations, are powerful tools and well suited for the majority or large power system analysis cases, however they do have limitations as they use simplified formulas, only consider the positive phase sequence, and use large time steps. EMT studies are very powerful and address many RMS shortfalls, but they are much harder to use, as they require much more input data, require more complex initialisation. They require a lot more knowledge and experience in interpreting the results. EMT simulations are more accurate, but they are not always better, it very much depends on the type of analysis need.
Remember that the key is understanding the limits of the software and your own knowledge, and the needs of the simulation being requested. For many studies, a skilled and experienced studies engineer might be able to provide a higher level of analysis in an RMS environment than a less experienced engineer using a more powerful EMT model.
Future
What does the future hold for both RMS and EMT modelling and simulation. Nobody really knows, as the pace of change and deployment of renewables is almost outpacing development in the software, and there is a massive shortage of power engineers with the right level of skills. Here are a few predictions:
- Grid Codes start to converge on requiring FRT, LVRT and HVRT simulations to be carried out on EMT simulation packages.
- EMT simulations will become the dominant method for carrying out all routine power system analysis.
- More established ways will be found of ‘simplifying and standardising input’ to EMT models.
- Aggregation of embedded generating facilities into smaller simplified equivalent models will become a standard practice.
- Existing RMS providers will have to expand their offering to include EMT modules or analysis methods.
- EMT packages will undergo a big shift in usability, and include much better functionality for handling multiple cases, scenarios, and reporting.
- Real Time simulation packages will start converging with EMT simulation packages and possibly the real time packages will be offered as a basic EMT model for normal computers, and with an enhanced model for real time modelling.
- Standardised models for inverter configuration and control will become more common (such as the existing WECC models) and standards will need to be developed in a similar manner to Exciters and IEEE 412.5.
References
These are some references and links that might be useful:
CIGRE: TB 727 Modelling Of Inverter Based Grids For Power Systems
CIGRE: TB 881 Electromagnetic transient simulation models for large-scale system impact studies in power systems having a high penetration of inverter-connected generation
NERC: Technical Report: Beyond Positive Sequence RMS Simulations for High DER Penetration Conditions, October 20221 PSCAD
NREL: Final Technical Report: Stabilizing the Power System in 2035 and Beyond
NREL: Open-Source PSCAD Grid-Following and Grid-Forming Inverters and a Benchmark for Zero-Inertia Power System Simulations
