In this article we will try to predict the end point (in terms of applied technology) of the energy transition. Getting directly to the point we now believe that:
– Directly connected spinning machines (generators and motors) on the grid are not required – but obviously they will continue to be with us for a long time to come.
– The ideal proportion of grid forming to grid following capacity (for stability) on the grid is about 15% to 85%, this compares with about 25% to 75% for synchronous generation to grid following inverters. Why the difference in proportions? Basically we don’t need inertia – in fact it makes control harder.
– The system strength rules in Australia do not currently encourage the technical needs of the system to get to these endpoints.
– The risks and benefits of a fully transformed electronic grid.
That was the short version of this article – for more details – read on.
A tale of Grid forming and Grid following
We have written an article on the differences between grid forming and grid following controls which you can find at here .
In summary:
– Grid following inverters behave like ideal current sources. Power and reactive power is controlled by changing the current and letting the system decide what the voltage should be.
– Grid forming inverters behave like ideal voltage sources. Power and reactive power is controlled by changing the voltage (in particular the voltage angle), and letting the system decide what the current should be.
These are the extremes – it’s also possible to consider mixtures of both strategies – the possibilities are virtually limitless, which can be a weakness as well as a strength.
With this new power system we are constructing we are like mariners exploring an unknown ocean with a well-constructed and provisioned ship but with no map or compass to guide us.
A new stability paradigm
The study of power system stability came to the fore after the 1965 North East blackout in the USA and Canada (affecting about 55 million people). It was a clear demonstration that as power systems increased in size, reliability generally improved, but the modes of failure became more complex and the potential for widespread cascading blackouts also increased.
Since then there have been many similar failures (see an incomplete list here here) many of which involved stability related issues.
Stability in traditional power systems is categorized of which the three categories most studied are:
– Transient stability – i.e. the potential of synchronous machines to lose synchronism with the power system.
– Oscillatory stability – this is typically a control system issue, feedback control systems may become unstable if they are tuned to operate too fast. There is thus always a balance to be struck between speed of response and propensity to instability.
– Voltage stability – this can arise if it is attempted to push too much current through a transmission link. Ohms law for AC systems is V = I Z, if I and Z is large, so is V, generators typically control the voltage at their terminals, but don’t control the voltage further away. If too much current flows, a voltage collapse will occur at a location remote from the generator.
These three issues have been of concern since we have had power systems, widespread blackouts typically involve one or several of these factors.
Transient stability is the most studied, faults on power systems cause rotating generating plant to increase in speed because faults interrupt the transfer of power. With nowhere to go the surplus power is used to speed up the rotating generator which governor control systems are too slow to reign in. When the fault is cleared, the generator could have sped up so much that it is out of synchronism with the power system.
This is known as pole slipping, and it can cause large voltage deviations on the power system and mechanical stress on the generator (shafts have been known to break due to sustained pole slipping). On large networks, several groups of generators can all exhibit similar behavior leading to widespread system failures.
Whilst synchronous machines and grid forming inverters can “pole slip”, in theory grid following inverters can’t. By definition they follow the voltage waveform and can rapidly track any changes in angle. Accordingly they don’t experience this type of instability. (There are subtleties we have glossed over – for a more complete discussion refer to the following presentation (warning it is over 50 minute long: here )
Synchronous generators typically experience pole slipping because they are accelerated by faults and their governor control systems are too slow to respond.
Grid forming inverters can also experience a form of pole slipping by not tracking changes in voltage phase angle quickly enough – but being electronic they can correct quickly. In summary a simplified ranking of transient behavior is:
– Synchronous generator – poor transient stability performance.
– Synchronous condenser – better than a generator (less acceleration during faults).
– Grid forming inverter – better transient performance than rotating machines (it can rapidly adjust to network transients).
– Grid following inverter – very good transient stability performance (in theory it should not experience transient instability at all).
In practice the characteristics of the power system and the various control system interactions can complicate things but the above ranking is a reasonable precis.
Voltage stability and oscillatory stability issues can be experienced by rotating machines and inverter connected plant in various ways.
Voltage stability is an inherent characteristic of the transmission system – which generators assist with via injecting reactive power, but if the transmission is inadequate, no amount of reactive power will solve the issue.
Control system instability has now emerged as THE problem to solve. It always was an issue but it has now emerged as the main area of concern with the faster systems we now have available. So much so it deserves its own section.
Why is control so problematic on power systems?
All automated control relies on feedback of some type. For single input, single output systems this is relatively straightforward – as an everyday example consider driving a car. Drivers use their eyes to steer the car and avoid obstacles. The feedback control loop consists of sensing the environment (using the drivers eyes), moving the steering wheel which changes the cars direction, and if the direction is wrong, apply corrections via the steering wheel.
So far, so good. However, consider what would happen if all the car’s occupants (say four people) all had their own steering wheel which is used to steer the car –say they each had control over their own wheel, able change both direction and speed independently of their co-occupants.
This could easily be a recipe for disaster particularly if the driving styles or awareness of the road of each “driver” are in conflict with each other. Add in the other controls (i.e. accelerator, brake, etc.) and the potential for conflicting commands to the car multiplies up very quickly.
This is similar to how a power system operates. In this metaphor the drivers are the generators, the car is the power system.
Specifically, on the NEM there are about four hundred registered generators –many of which have multiple sub-generators. Over short time periods (seconds) all of them could be trying to control the power system frequency, and all of them are trying to control their local voltage which is linked to varying degrees with voltages everywhere else on the system.
Over longer time frames (minutes) the generators are following dispatch commands in response to network constraints and market bid pricing. Over very short time periods they could be reacting to system transients caused by faults or switching.
It is possible to safely drive a car with several drivers if they all agree, or at least don’t disagree too much. By analogy the power system works with similar principals. Using techniques such as control system droop, tuning of response etc. the various generator control systems are set up to “not disagree too much”.
Sometimes these approaches fail and it can take a lot of effort to disentangle the mess that ensues. The various grid requirements defined in schedule 5 of the rules are meant to head off these issues, but unfortunately in many cases they have been the cause, or at least been unhelpful.
For example:
– The rules specify reactive power requirements regardless of what the local system actually requires for voltage control. Over supply of reactive power makes voltage instabilities more likely, undersupply could lead to voltage collapse.
- – The rules specify rapid response to system transients – which as a side effect makes control system instabilities more likely.
There is also an issue around transparency of information. Owners of new plant, their consultants and equipment suppliers are flying blind to a large degree, being reliant on information from the network companies and the system operator which is incomplete.
Equipment suppliers are also often unreasonably protective of their IP, which places unreasonable reliance on encrypted software models. How does the airline industry deal with this issue? If a plane crash needs to be investigated – would the investigators be required to rely on encrypted models of the aircraft?
The challenges discussed above are real and need to be addressed, however we also have some legacy thinking that needs to be jettisoned. One of these is inertia – which will be discussed next, the other is fault current.
Inertia – who needs it?
There is a lot of concern being expressed that as we transform to an electronic grid we will lose the inertia from spinning machines. A moment’s reflection should dispel this, a more detailed discussion is here.
The quick explanation follows: traditional inertia uses rotating masses (mainly the generators) to store energy. If the system slows down (i.e. the frequency declines) the energy transfers from the rotating mass to the system via the electromagnetic coupling of the system. The converse happens if the system speeds up. This is automatic and uncontrolled in much the same way that a car slows down when going up a hill.
Traditionally this was useful because generator governor control systems (which control the rotational speed of a generator) are relatively slow to respond.
However, we now have fast acting electronic control systems. As long as these are connected to a power source (wind, solar, battery) – they are able to shuffle energy to and from the system as needed and in a controlled way. This enables the system frequency to be accurately controlled without the disadvantage that comes with uncontrolled energy storage in rotating masses.
Simulations of large power systems have been carried out by the NREL which indicate that grid forming inverters can adequately provide a frequency stable voltage waveform if they make up about 15% of the generation capacity of the grid ( the remaining generation capacity being grid following).
If synchronous condensers were used instead of grid forming inverters the capacity rises to about 30%. Why the difference? Synchronous condensers use uncontrolled inertia to help maintain system frequency. The fact that inertia is uncontrolled makes stable frequency control a bit more difficult.
This seems to be the reason why simulations indicate that the minimum ratio of grid forming to grid following is less than the ratio of synchronous condensers to grid forming capacity.
Australia’s NEM system strength rules and registration practices
Australia’s NEM has recently made rules which act to encourage grid forming over grid following installations. These are based on an approach confusingly developed from system fault level (basically fault current contribution from grid following is considered negative compared to other devices on the system – which for real fault currents is physically impossible) which creates a pseudo-measure of where we might need grid forming devices.
Overlaying this is a system of system strength nodes (specific locations on the grid) to which a non-transparent worked out system of monetary penalties are attached. This has created a strong motivation to roll out grid forming technology even in situations where this type of control is not appropriate in order to avoid punitive system strength charges, and to avoid some areas of the network altogether.
This system appears to have been established for commercial reasons – the engineering rational is dubious at best. As discussed above there are advantages in having a mixture of grid forming and grid following inverters because they have complementary control weaknesses and strengths.
The current rules do not consider this – they appear to have adopted an overly simplistic grid forming = good, grid following = bad philosophy. In part this could be because we currently have too many grid following installations – but as various researchers are showing – we don’t need to flood the system with grid forming or synchronous condensers.
The registration process is a process which has issues with engineering legitimacy. In effect a lot of effort gets expended trying to tune control systems to be as responsive as possible (at the risk of instability), to attempt to prove models agree perfectly with the installed hardware, and to overbuild various electrical requirements such as reactive power capability and harmonic filtering.
It is perplexing that this process is often on the critical path of projects – issues such as site selection, community engagement, supply chain, business case development and equipment design/procurement/logistics seemingly being relatively easy by comparison.
Despite the effort put into system modelling, things still go wrong. However when the reasons for this are closely examined it is sometimes the paucity or inaccuracy of information about the network which is the main contributing factor.
The network companies and AEMO have invested a lot in network modelling but much of this is activity is hidden from the wider industry. When transmission companies carry out wide area network studies – it usually results in a yes/no or a request for a change in settings. There is no shared report or analysis commentary which casts light on the reasoning.
This is a very undesirable state of affairs, which needs to change if we are to guide the grid to the desired outcome.
We believe the grid we should be targeting will be a mixture of grid forming and grid following inverters with occasional use of rotating machines. The existing system strength rules on the NEM (the WEM so far has not followed this path) do not encourage this outcome and the registration process is focusing on issues to the detriment of more important factors.
A grid built on silicon and software – what could go wrong?
Halfway through the writing of this tome – a major software failure occurred in the form of the “crowdstrike” update. In common with many other organisations – we temporarily suffered the so called “blue screen of death”.
This was nothing more than an inconvenience, but if it had occurred on the generation assets across the NEM it could have easily resulted in a nationwide blackout which might have taken weeks to recover from. This is the risk we need to guard against as we install more and more electronic control systems based mainly on software.
However, the fact is we are already running this risk. Even the most archaic coal fired power plant on the system in all likelihood has a sophisticated control system installed within the last decade or so. In common with the laptops that failed on Friday 19th July 2024, these systems are vulnerable to software glitches and external attack.
To ensure a robust system going forward system should be decentralized and avoid the possibility of common failure modes.
As solar farms, wind farms and batteries proliferate, the system is naturally becoming more and more decentralised. The other step is to ensure that common software and hardware is avoided so as to ensure that when failures occur they are limited in extent.
Nemo Bourbaki has a PHD in electrical and electronics engineering and is a keen observer of the NEM and WEM markets. This article was originally published on their LinkedIn page and is reprinted with permission of the author. You can read the original version here.
