Windlab has conducted a simulation of a 96% renewable national electricity market (NEM). The goal of the study was to show that very high renewable penetration levels can be achieved by expanding wind and solar generation, which is firmed by existing hydro and readily achievable levels of storage.
It differs from other 100% renewable studies as it is based primarily on actual wind, solar and demand data from AEMO. Other studies have relied on simulated data.
Wind, solar and demand data was obtained from AEMO at a 30-minute resolution for a three-year period: Nov 2016 to Nov 2019. The wind and solar data were scaled up so that on average they provided a little more energy than demand.
The wind and utility solar data in some states was sparse, particularly in the early years of this simulation. To improve geographic diversity, simulated wind data in QLD, central TAS and west and northern NSW was added to the actual wind generation.
Simulated utility solar data was used to supplement actual utility solar data in all states except NSW. The rooftop solar data had good representation in all states over the study period, so no supplementary simulated data was required.
The wind, solar and demand data was scaled in the following way:
– NEM Demand, no rescaling, same as Nov 2016-2019;
– NEM wind increase from ~7 GW to 38 GW, generating 62% of NEM demand;
– Utility PV increase from 2.7 GW to 16 GW, generating 18% of NEM demand;
– Rooftop PV increase from ~9.5 GW to 35 GW, generating 23% of NEM demand.
Figure 1 illustrates how the rescaling looks on a NEM-wide basis. A different rescaling parameter was used in each state, with Figure 2 showing the resultant capacities and penetration levels.
- To help match supply and demand on a 30-minute basis for most days over the three-year period, it was found that 24GW/81GWh of short-term storage was required.
For long-term storage it was assumed the 2GW/350GWh Snowy2 was fully constructed, along with the transmission to enable the power to be supplied to either NSW or VIC. Within each state, it was assumed all generation could be transmitted without curtailment to demand in that state. For interconnectors, the following were assumed:
– Energy Connect interconnector from SA to NSW (800 MW);
– Marinus stage 1 (750 MW VIC-TAS);
– QNI Medium Upgrade (885/760 MW);
– Interconnectors upgraded with Stage 1 recommendations from ISP (moderate increase in NSW-QLD and VIC-SA transfer limits).
Figures 4-6 show generation traces for some fortnights from the three years of simulations. In addition to the wind, solar, hydro and storage generation, we have shown traces marked “other”. This is demand that couldn’t be matched with supply from the other sources.
In the short to medium term, it is expected that this demand could be met by gas generation or reduced through demand management. In the longer term, it expected that this demand could be met by supply from biofuels, hydrogen fuelled generators, or additional long-term storage such as Tasmania’s Battery of the Nation projects. Over the three years, 4% of supply came from “other”.
Figure 7 shows one year of simulations illustrating daily average power levels. It can be seen from this plot that existing hydro is able to fill much of the gaps in days with poor wind or solar generation. This is supplemented by generation from Snowy 2.0. However, on some days there remained a shortfall where generation was required from “other” in order to match supply with demand.
Figure 8 shows the wind and solar generation over the entire three years of simulation, again using daily averaging. The figure makes it clear that most of the shortfall in supply occurs in winter. The figure also shows storage levels in Snowy 2.0, showing that the storage is typically depleted in each autumn or winter, but then is filled again each spring.
When aiming to achieve very high renewable penetration, a key goal is finding which technologies in which states perform the best on those days that have a large shortfall in wind and solar supply. These days are typically cloudy winter days that have little wind across much of the NEM.
Figure 9 shows the normalised performance of wind and solar in each state during the 10% of days which had the largest generation deficits. Each technology is normalised by its average daily output, so if a technology has a normalised generation value of 100%, it means that on those days it is generating the same amount as its average.
Not surprisingly, in the worst 10% of days, all of the technologies in all of the states were underperforming with normalised generation <100%. The best performer was wind in QLD, which had a normalised generation level of 87%, closely followed by utility solar in QLD with 85%.
This illustrates that QLD is a particularly important state if we want to progress to very high renewable penetration. Its wind is still performing relatively well when it is calm across much of the NEM. Its solar also continues to perform relatively well in winter.
To add support to the importance of QLD wind, a correlation analysis has been done between wind and rooftop PV data for the last 7 months. The data was limited to just 7 months so that it could be done using solely generation data from AEMO, without any reliance on simulated data.
There are only two wind farms with publicly available data in QLD, Mt Emerald in the far north and Coopers Gap wind farm in the south of the state. Coopers Gap only started generating above 5 MW in mid July 2019, so the data was restricted from then onwards.
Table 1 shows the results of the correlation analysis at 30-minute resolution and also at daily resolution. Looking at the 30-minute table, it is apparent that solar is highly correlated with solar in all other states. It is also apparent that solar in all states is negatively correlated with wind in all states. This is because wind generation is biased to night-time generation.
Another key result from the 30-minute analysis is that Mt Emerald in far north QLD is negatively correlated with wind in every other state and has essentially zero correlation with Coopers Gap wind farm in southern QLD.
Looking at the daily correlation matrix, it is again apparent that solar is highly correlated with solar in all other states. Solar is negatively correlated with wind, suggesting that cloudy days tend to be more windy than average.
Furthermore, we can once again see that Mt Emerald is the standout performer in terms of being negatively correlated with wind or solar in every other state. It is clear that wind in far north QLD is an important component of a mostly renewable NEM. It has a tendency to outperform on days when wind and solar is lacking in the rest of the NEM.
To summarise, this study has indicated that a very high penetration rate of renewables on the NEM is possible with readily achievable levels of storage and interconnector upgrades. It has also shown that QLD wind is of particular importance due to its lack of correlation with wind or solar in all other states.