by Chris Morris and Planning Engineer (Russ Schussler)
Technically, what are wind and solar doing to South Australia’s grid? And why is South Australia’s electricity so expensive?
This post provides details about the extra services and functions a grid provides and how the grid in Australia is being impacted by the increasing penetration of wind and solar generation. And how this makes Australia’s electric power so expensive.
Alternating Current (AC) power generation aims to produce an alternating voltage in the shape of sine wave, whereby the current has a similar wave at the same frequency. However, the current may not be timed in phase with the voltage. AC power can have three types of load, depending on how it affects the timing of the current to the voltage: resistive, inductive & capacitive. For a purely resistive load, the current and voltage are co-incident. However, this is not the normal case. Electric loads and grids have both inductive and capacitive elements. Inductive elements want to maintain constant current flows, while capacitive ones want to store charge (voltage). Active power is measured in Watts. With AC being a cycle, there is another value orthogonal to it, called reactive power, measured in units of Volt Amps Reactance or VARs. The diagram below explains the relationship.
Generators and inverters produce some mixture of “real” and “reactive” power. That is why them and transformers are rated in MVA, not MW like their prime movers are. Real power, in watts, is the form of electricity that powers equipment and does work. Reactive power, in VARs, is the energy supplied to create or be stored in electric or magnetic fields in and around electrical equipment. VARs are 90° out of phase with real power. Reactive power (inductance) is particularly important for equipment that relies on magnetic fields for the production of induced electric currents (e.g., motors, transformers, pumps, and air conditioning.) Negative reactive power (capacitance) seeks to slow voltage waves by acting as a store for charges. The balance between real and reactive power is adjusted to meet the needed load and grid requirements.
Transmission line elements both consume and produce reactive power. Under conditions of light loads, transmission lines are net producers, and at heavy loads, they are heavy consumers. Reactive power consumption by these loads tends to depress transmission voltage, while production of transmission voltage (by generators, transformers or synchronous condensers (syncons) or injection (from storage devices such as capacitors) tends to support voltage. Due to the characteristics of components on the grid , reactive power can be transmitted only over relatively short distances during heavy load conditions. If reactive power cannot be supplied promptly and in sufficient quantity, voltages decay, and in extreme cases a “voltage collapse” may result. If there is too much capacitance, voltages rise to excessive levels, damaging the insulation of equipment.
Worldwide, motors comprise about half the load on the grid. Air conditioning loads are almost exclusively motors. As well as needing the active power, motors produce significant reactive power. The distribution Network elements (local “low voltage (<66kV)” lines, transformers and switchyard components) have to absorb this reactive power, usually by increasing the voltage at the transformers. If the system can’t compensate for the reactive power or has no more capacity, then the voltage starts to drop, while the motors’ current and reactive power increases. This compounds the overvoltage problem – why you get brownout in the suburbs on those muggy hot afternoons.
Because of the lines and transformers in a grid, the VARs at each part, or even at either end of a single transmission line, can be significantly different. That means compromises or corrections need to be made throughout the distribution & transmission network. This is done by modifying generation output settings, changing transformer tap point settings and switching in or out capacitors, inductors and synchronous condensers installed at strategic locations. Grid operators monitor the various parameters, then adjust the settings as the generation source and loads change.
All this explanation and discussion about reactive power may seem esoteric and irrelevant. It isn’t. The large 2003 blackout in the USA/ Canada was caused because they weren’t managing reactive power properly because they had an inadequate system understanding.
Modern renewables are often sited a long way from where the load will be consumed, requiring long spur transmission lines. Consider the grid map showing current and proposed generators of New South Wales as an example – the load centres are on the east coast and the wind & solar generation a long way west. There will need to be long and uprated transmission lines to take the power east. The loading on those lines will be unpredictably variable. There will be times when they carry minimal current as well as times when they are at or near full rating. This variability requires the switchyards along the line to have expensive equipment to manage the VARs and stabilise the voltage. This extra switchyard equipment will be needed when there is little generation to provide the VARs. All of this additional switchyard compensation adds substantially to the cost.
There are problems with having transmission from generator to load by separate routes – effectively paths in parallel from an electricity flow viewpoint. There are also problems with running transmission routes in parallel. Because the impedance (resistance in an AC circuit) can vary significantly between the lines, circulating currents can occur that increase the line loading. The magnetic field lines around a transmission line can also induce currents in adjacent conductors when there are two circuits on the same towers. There are also additional risks associated with parallel circuits. When a low rating line constrains the capacity of a bigger line, the small line will reach its rating long before the other line approaches its rated capacity. This condition is called the spring washer effect. It can be corrected by using compensation devices in the switchyard, but this increases the cost and complexity of the system.
Frequency Control & Inertia
The frequency is the timing between wave cycles in an AC system 60Hz (a Hertz is one cycle per second) and 50Hz in most of the rest of the world. The frequency has to be the same across the whole grid – it is one of the things that defines it. A stable grid frequency is critical for effective operation. Thermal plants usually provide this by using governor control, whereby the frequency drives the plant output through a negative feedback device. The grid system operators may also run real time or short period dispatch, whereby the plant operators increase or decrease load over short time periods on grid operator’s instructions.
The inertia, provided by the rotating machinery of the generator, serves to slows the rate of change of frequency (RoCoF) . The slower the frequency changes occur, the less stress for the plant on governors. And as there is linkage, a small RoCoF in “normal” grid fluctuations will also stabilise the voltage and reactive power requirements.
However in recent times, there have been significant and rapid swings in the Australian grid frequency between the control limits, shown in the graph below. The gray region is the deadband of allowable frequencies where no intervening measures need be taken by the grid operators. When outside the deadband region, generators are supposed to be offering primary frequency response support – for underfrequency (load is greater than generation), grid operators increase the generators’ output. This is either done automatically or by dispatching plant to increase load. From the rate of frequency decline, calculations indicate there could have been at least 600MW shortfall in generation over the five minute dispatch process. The cause of the variability observed hasn’t been positively identified, but is likely to be uncontrolled solar generation. If that is the case, then it indicates that faster acting and more expensive frequency control services are needed.
Inertia on a renewables grid can be provided by synchronous condensers or by large battery banks with specialised electronics. Of course, the batteries have to have enough charge in them to function, so they are reserved for just that purpose and thus can’t be used for other purposes like general market dispatch. However, AEMO does not appear to believe that renewables and batteries are a substitute for the frequency response provided by synchronous units. To quote the latest available report:
“To comply with the requirements of IPFRR, Semi-Scheduled generators will typically need control of active power that allow for simultaneous MW curtailment, MW ramping, frequency response outside a relatively small frequency deadband, and ongoing variation in input energy. While such MW control capabilities do often exist in isolation, when they are tested simultaneously, and in an ongoing manner, software problems have often been found. This then requires further development, updates and testing to address, a process that has in some cases proven significantly more time consuming than initially expected.”
To manage the frequency balancing, the grid operates short period dispatches, mainly raising and lowering generation over short time periods: 5 minute, 60 seconds and 6 seconds are the main time periods. The time reflects how quickly the grid operator can respond to be up to the dispatch level. Because of frequency control problems that are occurring, Australia grid operators now proposing putting in place a 1 second response time dispatch. They go out to the market for bids from suppliers of these resources. This process is referred to as Frequency Control Ancillary Services (FCAS). Provision of FCAS is expensive for the NEM. In the first quarter of 2020, FCAS cost was about $110M, but this has dropped back to about $40M a quarter. In 2010, before there was significant wind or solar on the grid, the FCAS cost was about $2M a quarter. FCAS costs are operational charges to be paid by consumers for the privilege of the renewables penetration.
As an aside, for those that want more understanding of renewables and market grid operation, the commentary at WattClarity is worth following. They explain events well and use their showcased analysis software to provide detail and insight.
Resolving frequency control and inertia concerns does not resolve the imbalance between load and generation. This is where generation reserves come in. Reserves are plant capacity that is not actually generating, but is available to power into the grid within a short time period. By rebalancing the generation/ load, the frequency departures are resolved.
The reserves needed are usually sized to cover the loss of the largest supply item on the grid. This can be a circuit on a transmission line or the biggest single generation supply point. For South Australia, this is generally the rating of the Heywood interconnector when it is importing power into the state. Reserves can consist of large grid-following battery banks with sufficient charge in them to run for quarter of an hour or more. To reduce the risk of single point failure, the reserves should come from a variety of sources. Some of them may be units that are partially loaded and can ramp up their generation; others may be hydropower units acting as synchronous condensers, or they may be batteries.
To illustrate the problems with lack of reserves, South Australia power generation has been having major problems recently: the –9th , 16th and 23rd February in particular. On these days, the wind diminished early in the day, but the lost generation was covered by the grid solar and domestic solar. There were also gas turbines running to provide the inertia. As the afternoon went on and solar dropped, they needed to load first the gas turbines, then start the diesel engines. This is shown in the graph from Wattclarity below. Note this is only South Australian generation; it does not show flow across the interconnectors which at the last time interval was 677MW. Batteries provided very little power. It appears that despite batteries supposed to be there providing generation when there is a lull in renewables’ generation, it is there unused but providing FCAS reserves. This is probably because it gives the owners more income. But this is counter to why the batteries were supposed to have been installed. What is providing the backup generation, other than thermal plant which renewables and batteries are supposed to replace?
Part 1 and Part 2 gave examples of this duck curve problem . The Heywood link was running at capacity, but the grid declared a reserve shortage as there was not enough market bids to supply extra generation to cover the system need in the event of a potential trip of the interstate connector. Graphics at Box 6 provide a description of the system during this event. All this meant the price went very high, punishing the market buyers. And for those that claim negawatts and interruptible power is the answer, Part A6 on the 23rd event shows there was only 1MW available. As the Wattclarity author described it, “decidedly underwhelming”.
The same general process would occur for over-frequency situation, but for this circumstance generation would be shed or batteries switched to charge. That is a process a lot easier to manage.
Grid operation is more than just keeping the load and generating balanced and providing a stable frequency. Grid capabilities must be able to cover its fundamental operating parameters like frequency, voltage and reactive power when things go wrong. Examples would be when a lightning strike occurs, or a transmission line failure or a generation plant trip. The grid needs both inertia and fast acting reserves as backup. Protection systems take time to sense and then react to changes in frequency. Circuit breakers typically take 3-10 cycles to fully disconnect once they receive the trip signal, but these are only the last steps in the chain. The protective electronic sensors have to sense that the frequency or line impedence is outside its allowable range, and then activate the switching relays and the switches take time to break the circuit. The gap between allowable and load shedding is a fraction of a Hertz. To give time for the protection and reserves to function, the frequency (and voltage) have to decay as slowly as possible, which is where the (RoCoF) becomes important. There is a relationship between RoCoF, inertia and load lost as a fraction of total load, but even a change of 1 Hertz decline a second is considered risky.
That is why when major events happen, inertia to slow the RoCoF is so critical. South Australia went black in 2016 because a lot of wind generators tripped off, the major interconnector got overloaded and tripped, then there was not enough inertia, and the frequency collapsed so fast that system load shedding schemes couldn’t operate and everything protectively shut down.
Because of the blackout, AEMC now takes the problem of low inertia very seriously. They instituted rules setting inertia requirements for each state, particularly South Australia. AEMO used the rules to set minimum inertia levels. This means that gas turbines are required to be on, while wind or solar is taken off the grid, which overrides the merit order bid stack. South Australia is now replacing some of the inertia previously supplied by gas turbines with synchronous condensers and grid forming battery banks, using reserved sections of the giant batteries. This has added further to the cost of operation.
AEMC has been struggling with the problems of frequency management in a renewables dominated grid. They are looking at tight requirements for frequency management, RoCoF and related issues. AEMO will then use those to set rules for compliance by each state’s grid . Whatever the final rules and requirements are determined to be, they will increase costs to the power user, both in capital and operating costs.
Automated load shedding is usually managed at the distribution network level. These schemes normally work on under-frequency trigger setpoints. This is achieved by tripping complete feeder circuits from substations. There will be target load reductions (typically 10 to 20%) and these typically occur in two separate stages.
As might be expected, problems are occurring in South Australia associated with the automatic under-frequency trip levels When there is significant uncontrolled domestic solar on the grid, then there is very little grid-supplied load. This means that the underfrequency load shedding will have little effect stopping frequency decline. AEMO have put limits on the Heywood interconnector well under the cables’ rating in certain conditions to counter this risk. The SA Network company is also looking for solutions to these concerns. There are major potential problems during the daily duck curve rampings. The grid can’t protect itself if a significant event occurs while the power is having to be rapidly increased. AEMO has recognised this risk but there are no measures put in place yet. Previous experience suggests that it will likely take another major outage event before a solution is found.
System functions during frequency excursions
The graph below of an idealised frequency excursion caused by a generation trip shows how it all fits together. The section labelled ‘dynamic response’ is normal frequency control behaviour. After the incident, the slope of the RoCoF is a function of size of load loss and the inertia. The higher the inertia, the shallower the RoCoF slope. When the frequency is dropping, the generation frequency control and fast acting reserves put extra power into the grid. The frequency nadir means the input is enough to stabilise things. The higher the RoCoF, the lower the nadir.
After the nadir, frequency slowly recovers as the slower acting reserves come in. Often, there is no steady state section, but there is enough excess generation from the reserves and existing generation picking up load to bring the frequency back to the normal value.
The above graph shows the “ideal” grid behaviour to a loss of generation. In reality, the behaviour can be more like the graph below, which plots the number of the Queensland coal stations tripping in a cascade. Here the frequency got so low that under frequency load shedding occurred.
As part of the Queensland failure, the voltage there dropped so the reactive power went up, which tripped switchyard compensation. The current inrush flows in the interconnector got so high that it tripped, islanding Queensland for 15 seconds until things were stable and balanced enough to reconnect automatically.
Queensland was “saved” by the high grid inertia. If it had of been renewables powering the grid, and a similar cascading failure had occurred, it is likely that it would have been a collapse to blackness.
The above is a simplified explanation of what is needed for reliable grid operation. Proponents of renewable energy do not want to discuss concerns of this sort, particularly the costs involved. When forced to address these issues, they rely on magical thinking, advocating for technologies that either do not yet exist or have not yet been proven to work reliably on a grid. The known solutions are expensive, but the renewable sector doesn’t want to pay for them – their mantra remains that renewables are cheaper than fossil fuels so the others should pay for them – hiding the expense. Add in the costs from the needed system support requirements described above, then renewables are significantly more expensive (and less reliable) than conventional generation. The extra costs of renewables support are being paid for a deteriorating quality of electricity supply. That is why there is a new industry adage –
Cheap renewables are very expensive.