Replication: October 2024
Andrew Grogan's spreadsheet, shared in his LinkedIn post, uses data from Open Electricity. To check it independently, I rebuilt the same month from AEMO's raw data for every generating unit, using the NEMOSIS Python package, and ran his model on both versions.
The AEMO version of the data (data/nem_2024-10.js) has not been uploaded yet, so the comparison cannot be shown.
1. Electricity by source over the month
GWh over the month. Green means within 2%, amber within 10%, red more than 10%. Charging and pumping are shown as negative numbers because they take electricity from the grid.
2. Demand every half hour
3. The same scenario on both datasets
Andrew's model, run once on his data and once on the AEMO data, with the same settings. Change the settings to see whether the two still agree.
Coal is switched off and gas is kept at Andrew's settings in both runs. Existing batteries keep their real behaviour from each dataset.
Why the two versions differ
- Batteries. AEMO's unit data records much more battery charging and discharging than the extract in the spreadsheet. Many batteries were being moved onto new AEMO registration rules in 2024, and some may be missing from the older extract. The AEMO figures are consistent with real battery losses, with about 78% of charged energy returned. They also line up with AEMO's own reporting: its Quarterly Energy Dynamics report puts average battery output at 90 MW for October to December 2024. The AEMO rebuild gives about 81 MW for this month, while the spreadsheet extract gives about 14 MW.
- A new wind farm. Port Augusta's wind farm was ramping up in October 2024. It appears in AEMO's unit data and accounts for most of the small extra wind.
- Gas labels. Total gas matches closely, but AEMO and Open Electricity label a few gas units differently, so some output moves between combined cycle and open cycle gas.
- Bioenergy. Sugar mill and landfill gas plants are grouped together in the AEMO version.
- Small generators. Some very small generators are not in AEMO's unit data, and Open Electricity estimates some of these.
Method
- AEMO tables:
DISPATCH_UNIT_SCADA(output of every unit every 5 minutes) andROOFTOP_PV_ACTUAL(rooftop solar every 30 minutes), downloaded with NEMOSIS. - Each unit is matched to a fuel type using AEMO's registration list, with a short manual list for units that have closed or changed registration.
- 5 minute values are averaged into half hours. Demand is total generation minus battery charging and pumped hydro pumping, the same as in Andrew's sheet.