By Phil Kreveld
Data centres have to ignore unserved energy statistics because national and regional long-term planning targets for unserved energy (0.002% being AEMO’s required national standard), are not a guarantee of continuous, uninterrupted power.
So, when the connections of data centres are seen as a ‘growing problem’, it’s not their energy requirements per se but their effect on voltage, and even frequency stability when they switch over to their own ‘uninterruptible supply systems’, often within milliseconds of detecting an anomaly in electricity supply possibly leading to grid collapse.
Data centres cannot tolerate any interruption to electrical power. That includes momentary voltage dips—just in case they are the harbingers of worse to come. Weak grids are subject to voltage dips—and there’s the rub.
Related article: Blue Mountains residents score win against data centre
One way of washing one’s hands of the problem is to get data centres to run independently of the grid, 24/7. i.e., get your own electricity supply. Some might, but others might well decide to take their business elsewhere. However, data centres want to be close to their customers so as to minimise signal latency. And customers reside in large population centres, already served by electrical energy.
Therefore, their first port of call is the local electricity supply which they beef up within their premises by batteries with very limited storage and stand-by diesel generators. The national government’s wishy-washy approach is for data centres to ‘contribute to the national energy needs’ by storing large amounts of energy in battery energy storage systems.
It totally sails around the real problem but it sounds good enough for public consumption. And, it is of course heading in the direction of ‘get your own electricity, and stop bothering us’.
Our grid is weak because it is a radial one, characterised by long distances of hundreds of kilometres between generators and centres of energy consumption (loads). By contrast, high population density countries have meshed grids with many interconnected generators serving electrical loads.
Ours is a natural outcome of having just about everyone living along the coast—and just connecting it all up as one long string—5600km from Port Augusta to Far North Queensland. Although not directly related to grid strength—an impedance measure—the decline in inertia from 130MWs in 2016 to 50MWs today isn’t helping frequency stability, voltage stability decline being the primary factor of decreasing grid strength.
Understanding the factors of voltage instability caused by data centres is not difficult. Voltage at the data centre’s connection lags the voltage at the generator’s end as the result of power demanded by the facility. Weak grids, i.e., those with high impedance because of their length, exhibit a large voltage phase difference (lag) between generator and load, and one which increases with increasing power demanded by the load.
Asudden decrease in power demand, e.g., caused by a data centre ‘going off line’ because it has detected a ‘wobble’, is a bit like letting go of a rubber band previously stretched to the limit. The oscillations in voltage angle can be such that neighbouring solar inverters ‘lose the plot’ and switch off, resulting in wild power swings thus making AEMO’s stability maintenance task impossible.
Wild power swings are harmful to frequency stability. The loss in inertia, as mentioned above, exacerbates frequency instability. The hope is that sufficient ‘virtual inertia’ provided by battery-supporting grid forming inverters will replace loss of rotational inertia provided by synchronous generators. Hope is not the same as confidence, and the employment of synchronous condensers at data centre connection points would contain voltage angle swings thus also heading off power swings as mentioned above.
If nothing else, worrying about data centre connections—a part of modern economies—demonstrates that fundamentally our national grids are inadequate for a modern economy. Wide employment of synchronous condensers will help in strengthening our weak grids. Some of the responsibility might be offloaded to data centres as a condition for connection, but that doesn’t hide the basic problem.
Related article: US AI firm to build Australia’s biggest data centre in SA
We are adding to it with remote energy zones connected by long transmission lines, and distribution networks stressed to their limits in power handling capacity. Time is not on our side, and a ‘national expertise in sitting on our hands’ is frustrating investment plans—and not just for data centres.
There appears no alternative but for the taxpayers of Australia to knuckle down to a national grid-strengthening exercise.
