In engineering, as in life, we all have moments when something that is troubling us suddenly crystallizes, the fog lifts to reveal an insight. So it was after several months of grappling with issues in the Frequency Control Ancillary Services (FCAS) market when I started to examine events in detail and study the performance of the power system.
On 1 Nov 2015, South Australia separated due to a communication protection error during a prior line outage, and the frequency control in the 35 minute islanded period was extremely poor. It was unlike any past separation event; it triggered under frequency load shedding for the loss of 150 MW of import. After a brief recovery to 50 Hz, the frequency meandered between 49.85 and 50.6 Hz before finally getting close enough to resynchronise 35 minutes later.
What intrigued me was the question,” How did the frequency control get to be so bad?”. I recalled a discussion with an industry colleague, a control technician from a thermal power station. He had said something questioning the control requirements in the FCAS specification and how it could be interpreted in two different ways. He disagreed with what was being required, as in his opinion “it would not control frequency”. As with most things electrical, measurements reveal everything, so we started looking into our PQ measurements. We could see extended periods where the frequency was just not constant (at least not at 50 Hz!).
I am fast arriving at the conclusion that the FCAS market has caused a significant number of unintended consequences which I am sure the designers of the FCAS market either did not contemplate or dismissed as unlikely to occur. Technological advances in control capability, the interconnectivity of control systems and the ability to alter the control mode of a generating unit in real time, these were unlikely to have been considered an issue in 1999. They are real now. The ability to switch and disable governors, or become frequency insensitive (turbine follow mode) between one dispatch interval to another is real now. The advent of switching controllers allows for “frequency services” to be dissected between frequency bands, and the service provided looks exactly like the market offer. So the design of a frequency service which started in a spreadsheet to identify the energy increase from governor action that the market would pay for, has now been reverse engineered into the controls. The services in the market, their structure, size and shape are exactly followed by the controls. The power system consequences of this type of control were not examined, as it was not thought it would be applied to the governors.
Prior to the NEM, all units had mandatory governor performance, tight deadbands and droop control. Each state operated its own area of control with AGC and interconnector flows controlled through tie-line bias. In a long skinny power system with weak interconnection, this is important. On introduction of the NEM, it was decided that the eastern seaboard could be controlled using one area of control, and the interconnector flow between states would be adequately controlled through the 5 minute market dispatch. So at the time the FCAS market was introduced, generators had mandatory governor control and the market was treated as an “enablement market”, meaning if you were enabled you were paid, but everyone provided the control.
For many years the FCAS market has worked, because control systems are not upgraded all at once. After about thirteen years of the FCAS market, control upgrades appear to have taken place in accordance with the market design. Once the upgrades became prevalent, the frequency control deterioration is evident. But here’s the problem; it is still meeting the Frequency Operating Standards (FOS). At the time the market was introduced, the Reliability Panel widened the FOS and added a percentage of time measure.
So now the market is asking, “What’s the problem?”
There are individuals, both regulatory and market, thinking that the oscillatory behaviour on the 10th Feb was caused by the loads! (Doesn’t Generation + Interconnection MW = Load? So it can be concluded that the frequency disturbance is caused by the load!) There is now a serious debate in the market over whether there is a problem if the frequency moves around within the normal operating band of 300 mHz. Figures 2 and 3 are derived from the 4 second data used to calculate “causer pays” factors in the FCAS Market.
As a system control engineer, I contemplate the loss of being able to study, predict and trust the response of the system when frequency controls are enabled and disabled on units across the country every 5 minutes. I can think of several engineers who are turning over in their grave. For light relief, I refer you to the sketch from Clarke and Dawe on the Energy Market: https://www.youtube.com/watch?v=ELaBzj7cn14 .
Seriously, the loss of primary control on the synchronous fleet is significant and there is talk of fast frequency services being provided by batteries, wind farms or demand side response!. Recent rule changes require network owners to ensure sufficient “inertia” and “system strength” be retained. So politicians are arguing that renewables are the cause of the loss of “system stability”. I know that on the highway of electricity, we are asking the Mini Coopers to correct for the loss of control on the Mac trucks. I say this because the control action on the large synchronous units is like having a driver that can’t reach the pedals and who only tries to steer once every five minutes. Given the length of the grid and the weak interconnections, perhaps it is time to start asking every generator to simply do their fair share of control and get back to a control philosophy that makes sense.
(with thanks to Ryan Jennings for the production of the charts.)