How long have we got to get our act together for renewable electricity?
By Phil Kreveld
One could be reassured by the leisurely acceptance with which the accelerating penetration of renewables connecting to the national grid, is accommodated. An AEMO updated Integrated Systems Plan (ISP) every two years and a plethora of talkfests, to which add the contributions of the state and territory ministers based on election cycles. All of that happens on a time scale of months and years. In reality, we require decision making on the basis of milliseconds.
The National Electricity Market (NEM) grid is an old fashioned, legacy, 50 cycles/second alternating current (AC) system—a composition formed by interconnection of erstwhile separate state grids. It is still governed by a mindset that much resembles the pre-2000 days. Privatisation and concomitant chasing out of engineering ‘cardigans’ in favour of new management types, saw refuge in much consulting work but very little electrical expertise needed by modern grids.
Renewable grids require millisecond control
A single cycle of electricity takes precisely 20 milliseconds; the renewable forms of generation incorporate technology operating on a time scale of 1 millisecond and even shorter times, requiring control systems on the same time basis—but guess what? Other than in the rarified academic engineering environment, no one seems worried about it.
One could be reassured by the ARENA projects such as investigations in battery-connected, voltage forming inverters—they will, we hope, ultimately replace those coal and gas-fired turbine-synchronous generator plant, but these efforts are at a tangent to the overall renewable integration challenge.
However, hope is of no value—neither is there time for relaxed experimentation. The pace of renewable penetration and retirement of conventional, synchronous generation is such, that we cannot sit on our hands. Projects like the CSIRO’s Control Room Of the Future (CROF) and the efforts of, for example, Monash University’s intelligent grid hub, have to be released and form part of the national integrated engineering plan, Australia so badly needs.
A national engineering plan for the grid
What precisely is a national integrated engineering plan—other than a ringing phrase?
Let’s dig in a little. Virtually 100% of renewable energy connects to the grid via grid-following inverters. The old-technology synchronous generators, provide the ‘commutation’ source for all these inverters. Commutation is tech-speak for switching the electronic switches of the inverters in synchrony with the AC voltage of the grid.
For those with knowledge of electrical engineering, the AC voltage, by means of Clarke-Park algorithms in the inverter control package and phase-locked loop circuitry provide very fast control of pulse-width modulation (PWM) of the inverter switches (generally gate-insulated MOSFET transistors) to furnish AC current to their connection point (PoC) in the grid—more on PoCs further on.
The long and the short of all this is that a supply of AC voltage is critically important—unfortunately this point is lost on energy ministers. Their vision is locked on 100% renewables—their political careers, they think, depend on realising this goal and in its pursuit they even encourage the closures of the one form stable AC voltage—old fashioned synchronous generators because they are driven by gas or steam turbines, the latter needing lignite or black coal.
The energy ministers have worked out, that irrespective of anything else, electricity must be transportable across the South-East of Australian grid (Western Australia is a somewhat separate case). It has all the appearance of a ‘cargo cult’! And it will definitely slow the connection speed of renewables—count on AEMO becoming more and more difficult in regard to connection approvals. And, yes investors will be less than impressed and seek opportunities elsewhere.
A drift towards a messy and expensive bandaid approach
To get back to time scales—the 1 millisecond time scale, because it ties in with the foregoing.
All those grid following inverters, mentioned above, respond to changes in AC voltage within a few milliseconds. Given that speed of response, control of these devices must operate at an even smaller time scale. That rules out manual control!
However, the entire Australian South-East grid operates essentially on manual control. It is a legacy from the past, appropriate for the days when only conventional synchronous generators supplied the nation’s electrical energy needs.
Even when eventually voltage forming inverters replace synchronous generators, the millisecond ‘control paradigm’ will be needed. This is serious! And we are not planning for this eventuality. It is the very fleshing out of the millisecond control paradigm that all our engineering attention must be focussed on.
By now, you will wonder what the role of AEMO is in all this. It is of course aware of the very uncertain future we are headed for. However, AEMO is the unloved child of two parents—state energy minister and network parents and although responsible for proper operation of the national grid, it constantly looks to its parents for approval—and they are hard to please. In fact, the parents between them are to blame for the treacle-like path being taken for the connection of renewables.
AEMO does its best in investigating the effects of connecting more and more sources of energy using grid following inverters, be they solar farms or wind generators. Meanwhile, and totally out of its control, the mums and dads of Australia are installing rooftop solar which in aggregate already exceeds demand capacity of all coal, gas and hydro generation—all of it essentially uncontrolled! We will come back to the household generator!
Computer modelling must not be confused with real-time operations
AEMO is quite concerned about grid stability—in particular, voltage and frequency. So much so that it runs complicated computer-based studies to try to emulate the effects that inverters have on grid stability, this being essential for the operation of the inverters themselves—a kind of closing the loop.
It is something Daniel Westerman, the CEO, boasted about at last year’s COP26 Meeting in Glasgow. The modelling studies, electromagnetic transient (EMT) studies provide some sort of prediction on the effects inverters are having, ie., in the millisecond timescale region. But it is unable to provide any indication on how the grid system, consisting of some 3000 load and generator busbars might work in real-time.
The only way to find that out is to actually run the grid with only inverters or nearly so. When might that be? Interesting question, but this author believes that we won’t have the courage of our convictions and that therefore the path to the nirvana of 100% renewables will last a bunch of government terms—and attract fierce criticisms in the public domain.
A word about EMT; the term is waved about as if a mantra. In everyday language it is the recognition that electrical parameters based on 50 cycles/sec (root mean square parameters or RMS) are no longer the only game in town albeit still very usable in protection engineering for devices like circuit breakers and protection relays.
In actual fact there is in the world of inverters, not so much the phenomena of ‘transients; rather EMT is part of ‘business as usual’! Yet, EMT is treated as a kind of holy grail and AEMO would have us believe that EMT modelling is all that is needed to assure us that inverter connections ‘work’. But it is not to be believed—and in the tech bowels of the organisation it is not believed. There is already a lot of evidence of voltage, voltage angle, and frequency oscillation beyond fixing by inverter technicians in the field.
Earlier on the point of connection (PoC) of inverters was referred to, and modelling of EMT is carried out in theory at the PoC. And that is where models and reality part company. There are no effective models for the entire Australian South-Eastern grid that can take into account what the influences are of already connected inverters. Very likely every modelled PoC is assuming that at the far end there is a convenient infinite bus with constant voltage and frequency, ie, this in no way resembles reality. None of this is ever mentioned in the regular media—nor at the last House Standing Committee on Energy and the Environment (June 2021) where AEMO talked about ‘solving millisecond equations’ but without further details.
The Heisenberg Uncertainty Principle—applied to electric grids
In approximate terms, the Heisenberg principle states that particle position and velocity cannot simultaneously be determined with the same accuracy because we disturb a system by the act of observation.
This infuriating reality applies to the control of electric grids as well.
As mentioned we have choices of measurands and in deciding which to choose, we negatively affect the accuracy of other observations. Basically we face a choice between phasors and time-based phenomena, the latter including EMT.
Phasors are a mathematical fiction, basically describing timing differences in terms of angles, 180 degrees describing a timing difference of 10 milliseconds. However in making this conclusion, we are assuming our electrical system is ‘synchronised’, whereas if we were to use time as the measurement base, we cannot determine synchronicity to the same accuracy.
Of course nothing stops us from measuring parts of a large grid system in terms of time in some sections and with phasors in others.
We need to think through the control tasks—and therefore which of the two we want to use. The first harsh reality in using time is that light through vacuo and electricity in wire ‘wave guides’ travels at about the same speed. Therefore the idea of data streaming to central control locations is fraught.
We managed to control grids in the past by not having to stream data—old fashioned synchronous generators under more or less constant loading were able to communicate with each other over large electrical distances with timing delays of 5 to 10 microseconds and with end-to-end delays equal to a 100 microseconds in large grids. Those days are disappearing because of those nimble inverters.
The shape of new control systems will include fast local control at generator busbars (collection points of a number of remote energy zone PoCs) and phasors elsewhere, where there is phasor-based control hardware, already in place.
In some cases individual PoC controls may be required. In simple terms the basic control task is the maintenance of constancy of timing differences throughout the grid (inter alia, this is essentially the same as constant frequency) and constancy of voltage levels.
For time-based (EMT) observations, we are basically interested in whether we have smooth sinusoid voltage and current waveforms or not. Rather than streaming that data, and clogging up the communication systems, filtering of measured quantities can be applied, with local judgement of ‘acceptable/not acceptable’ limits.
Comparisons between digital filter-extracted, close to ideal sine waves and for example, actual peak amplitudes through sample and hold circuits in combination with fuzzy logic utilising ‘if then’ statements, can be applied locally. For example the local comparison evaluation can be used to reduce power output (grid current of grid-following inverters) and/or apply a varying time constant to outer control loops of inverters.
Phasors can be employed at synchronous generator busbars, out-of-step and power oscillation blocking relays, at thyristor-controlled series capacitors, var compensators, static compensators, etc. Worryingly phasors, although they can be updated rapidly through high sampling speeds, require a minimum 20 millisecond interval in order to extract the fundamental phasor. They are therefore suitable for low frequency voltage, current and power oscillation.
A valid question therefore is the general structure of grid control with very large penetration of inverters. As mentioned further down, grid stability control will have to sit on top of energy dispatch.
Currently we have the bull by the tail with the AEMC authorising ridiculously complicated rules to avoid technology solutions. A sensible new control system will have central cores situated at large battery backed voltage forming inverters and gas/hydro/steam synchronous generators. These cores will use phasor based measurement (PMU’s) and may have limited streaming to the CROF centre. Grid following inverter sources, elsewhere, will only stream alarms, also ending up in the CROF centre. Their alarms will override dispatch instruction of the AEMO Dispatch engine.
Obviously a centrally based, integrated approach to grid design and generator rules is urgently required. The current ‘free for all’ is a recipe for disaster.
The elephant in the room—uncontrolled distribution grids
Before discussing controls appropriate for inverter-dominated grids, let’s look at the elephant in the room—distribution grids.
Talk to distribution engineers and their only real concern is voltage control in the medium and low voltage portions. As far as AEMO is concerned, the latest thing is fast frequency control ancillary services (FFCAS) to be provided by aggregation of participating rooftops. How exactly that might work in practice is not clear, nor to what extent it will ameliorate frequency instability. But there is blindness when it comes to the really serious problem of under frequency load shedding relays in distribution grids.
Without very serious planning for forms of dynamic restraint probably requiring forms of fuzzy logic, or even AI, FFCAS might not work. Under frequency load shedding will have to be based on dynamic restraint envelopes, which are likely to be complicated by virtue of having to share the restraint and control of hundreds of thousands solar inverters in typical low voltage networks. This nettle awaits to be grasped!
New forms of control
The control side of the renewable grid poses major challenges, initially requiring a mixture of synchronous controls such as damping through excitation control and new, programmable limit parameters for out-of-step and power oscillation blocking relays, as well as variable time-base, millisecond, tens and hundreds of millisecond controls applicable to inverters.
The combination of very fast (sub-millisecond control loops) with typically 10 millisecond loops for inverters and their integration with governors and excitation controls of synchronous generators in second-timescale is an unexplored field in terms of its real-time exploitation in large, meshed networks, let alone in basically linear grids with large voltage angle differences as in Australia. Thus AI will ultimately be involved, resulting in major challenges to cyber security. In addition, the new control mechanisms will have to be integrated with the national dispatch engine under the control of AEMO. It is highly unlikely that the current regimes of FCAS will be compatible. In other words, ensuring the stability of the grid will have to take precedence over commercial operations as presently conducted by AEMO.
We are not so much engaged in a difficult experiment with the Australian South-Eastern grid, but rather we are, through not grasping the technical challenges in an organised manner, allowing a drift to an uncertain set of outcomes with very difficult to predict costs in order of assuring a reasonable level of service to the electrical energy consumers of Australia.