Capacity factor is often discussed when evaluating and comparing the efficiency and performance of solar farms. However, looking just at this metric can be misleading as it ignores many underlying technical and commercial factors, as solar farm specifications almost never align for a simple apples-to-apples comparison.

As many of us in the industry will be meeting at the Clean Energy Council’s Large-Scale Solar Forum in Brisbane on Wednesday, I thought I would publish this article as a timely reminder of the many important development and operational factors that can affect the performance of a solar farm.

The analysis presented in this article draws upon a range of data, primarily from our recently released GSD2022, and explores just a handful of reasons why comparing solar farms on capacity factor alone is overly simplistic and can be misleading. In doing so I run through a list of metrics that help paint a more complete picture of how to view generator performance.

Capacity Factor: What it measures, and what it misses

Let’s begin by having a look at what capacity factor is, before we delve into what it isn’t.

For a given length of time, a generator’s capacity factor is simply its electricity generated divided by the theoretical maximum that it could have produced if it ran at full capacity. It’s a metric that is especially discussed when comparing solar farms as they all operate on relatively similar and predictable schedules.

Below, I’ve made a simple chart to compare solar units in the NEM based solely on capacity factor over one calendar year. For convenience, I’ve filtered out all units that have a maximum capacity of around 30MW or less and have excluded any units that connected to the NEM after 1^st January 2022.

It’s common to see rankings of units from ‘best’ to ‘worst’ based on capacity factor. This simple example ranks all solar units in the NEM by their capacity factor over 2022.

Source: GSD2022 Data Extract

The chart above makes for a simple and quick-to-understand comparison but glosses over a lot of important details.

This is because capacity factor really only captures two variables:

1. Generation – More specifically, final generation output. This means it skips over all of the differing factors that impact each individual unit’s output such as location, network constraints, bidding behaviour, etc.
2. Capacity

The measure completely ignores other equally important aspects of asset performance such as revenue (e.g. spot revenue, LGC revenue, etc.), costs (operating costs, construction costs, market costs, etc.) or even the underlying value of the electricity produced (e.g. electrical losses, etc.).

Case study: A tale of two solar farms

To help demonstrate my point, I thought it was best to present a case study to illustrate just some of the complexities that are involved in understanding solar farm performance. In this analysis I’m comparing the performance of two solar units in South Australia throughout last calendar year:

1. Bungala One – Part of the larger Bungala Solar Farm, operating since 2019 and located near the town of Port Augusta, SA with a maximum capacity of 110MW.
2. Tailem Bend – A solar farm East of Adelaide that has been operating since 2018 with a maximum capacity of 95MW. Stage 2 of the larger solar farm is currently under development.

I’ve chosen these two units as they are located in the same region (hence are subject to the same spot price), have a relatively similar maximum capacity, and were constructed around the same time – but have quite different capacity factors.

There was a 6.87 percentage-point absolute difference (or 36.4% relative difference) in capacity factor between Bungala One and Tailem Bend in 2022 — with Bungala One near the ‘top’ of the rankings and Tailem Bend near the ‘bottom’. In this article, we look beyond capacity factor to explore how each asset actually performed throughout the year.

Source: GSD2022 Data Extract

Technical factors

Many decisions that impact performance get made during the design and construction phase of development, and the following reasons demonstrate why not all solar farms are created equal.

Factor 1: Latitude

The first factor that affects solar generation is the raw energy potential of the solar farm’s location. And to a large extent, this comes down to the latitude of its location.

As the sun’s trajectory relative to the earth follows a cosine curve, broadly speaking, it is the latitudes between the Tropics of Cancer and Capricorn that receive the most overhead sunlight. As a very rough rule of thumb in the NEM, the further south from the Tropic of Capricorn (which runs through the Queensland town of Rockhampton) that a site is located, the less solar irradiance it should receive. One small caveat is that coastal areas have higher levels of moisture content which means that they are more likely to have more frequent cloud cover.

Bungala One receives significantly more raw energy than Tailem Bend, in part, due to its more northerly latitude, but that’s not necessarily the most important factor to determine performance.

Source: 2020 The World Bank, Global Solar Atlas 2.0, Solargis

For our two solar units in question, we can see from the map above that Bungala One’s more northerly latitude contributes to its location having a higher raw solar irradiance.

Factor 2: Capacity

In the NEM there are two measures of capacity that analysts can use when calculating capacity factor – ‘Maximum Capacity’ or ‘Registered Capacity’. Paul McArdle has written an explainer distinguishing the two measures, and when it is appropriate to use each. For solar farms, registered capacity usually denotes the total capacity of the unit’s panels, and maximum capacity denotes the size of its inverter. Therefore, it would be more correct to use maximum capacity when analysing solar farms as output is limited by the inverter (hence in our earlier charts we have stuck to maximum capacity in our calculations).

The maximum and registered capacities of Bungala One and Tailem Bend. We’ve seen other sources use either measure of capacity used in their calculation of capacity factor, leading to differing results.

Source: GSD2022

In the NEM there is an increasing trend for solar farms to oversize the capacity of their solar panels relative to their inverter. Having a higher Max-to-Registered-Capacity ratio results in a more pronounced capped shape of output where output is generally higher in the early morning and late afternoon periods than would be for a solar farm with a simple 1:1 ratio.

From using the tables above, my math here tells me that Bungala One has a Max-to-Registered-Capacity ratio of 1:1.23 while Tailem Bend’s is 1:1.14. This means that if all other things were equal, Bungala’s output would be slightly higher during the shoulder periods of daylight.

Factor 3: Angle of incidence

Solar farm designs can employ variations of fixed, single or even dual-axis tracking.

The majority of solar farms in the NEM use single-axis tracking on their panels, with some of the newer solar farms such as Columboola using bifacial single-axis tracking. Typically older solar farms and those located within cyclone-prone areas (such as at Sun Metals) were constructed without tracking, and thus are fixed. There are currently no solar farms big enough for NEM registration that use dual-axis tracking, but The University of Queensland has built a <1MW dual-axis solar array on their Gatton Campus for research purposes. Wine connoisseurs might also be aware that two dual-axis solar trackers were installed at the Jacob’s Creek Winery in SA in 2011.

The charts below have been taken from our GSD2022 and show both the range of output for each day over the year (min-to-max in green); and the maximum daily availability (in black) for both Bungala One and Tailem Bend.

Tailem Bend’s output is less consistent, but Bungala One’s availability has a noticeable dip in winter

Source: GSD2022

There are clearly some differences visible between the two units:

1. Bungala One has a very pronounced dip in availability in winter, while Tailem Bend does not.
2. Tailem Bend has many days where maximum output does not rise close to maximum capacity (i.e. white space visible in the chart) whereas Bungala One reaches close to maximum output most days.
3. Tailem Bend also saw more days where availability was also limited well below maximum capacity.

My current hypothesis for why Bungala One shows a drop in output during winter, while Tailem Bend does not, is that it comes down to panel orientation and use of tracking. Bungala One employs single-axis tracking where its solar panels face East each morning, and track West throughout the day. Meanwhile, Tailem Bend’s panels are northward facing and are fixed (at approximately 19° according to ARENA’s benchmarking PV performance report). Perhaps one of our more learned readers can comment below to help me understand if this hypothesis is correct?

During summer months, the sun’s trajectory in the sky is higher so panels that track East-to-West are significantly more effective at capturing direct sunlight. In winter there are naturally less hours of sunlight, but the suns trajectory is also lower and further north (relative to South Australia) – this means that panels that track East-to-West receive less direct sunlight during the middle-of-the-day peak in winter than they do in summer.

The sun’s trajectory over both solar farms in summer vs winter. The pair of trajectories does not differ substantially between the two units despite a separation of roughly 3 degrees in latitude.

Source: Sun Surveyor

My two graphics below attempt to summarise this hypothesis to show the panel orientation and tracking for both units and the resulting impact on their summer vs winter output.

Bungala One’s East-to-West single-axis tracking causes its generation profile to receive a ‘winter shrink’ compared to summer.

Source: Google Earth and NEMreview

By contrast, it would appear Tailem Bend’s northward-facing fixed panels means that its generation profile gets a ‘winter pinch’ compared to summer.

Source: Google Earth and NEMreview

Ultimately the choice of fixed vs single-axis tracking comes down to economics and suitability. As we can see in the charts above, Bungala One’s single-axis tracking means that it captures a higher volume of energy in total than a fixed set-up such as at Tailem Bend.

Operational factors

Once built it is then only operational decisions that can positively (or negatively) impact a unit’s performance and subsequent returns. And motivations behind these operational decisions are sometimes difficult to ascertain as commercial arrangements are generally confidential by nature.

Factor 4: Negative prices

When discussing performance, specifically financial performance, it’s of great importance to analyse how effective a solar farm is at generating when electricity is most valuable, and inversely, how effective a solar farm is at bidding themselves out of the market when prices are negative.

Nick Bartels has already explained, and used, the RRP Energy Price Harvest metric from the GSD2022 to demonstrate how effective an individual unit is at capturing high or low spot prices.

Throughout 2022, Tailem Bend generated significantly less than what it had available during negative prices compared to Bungala One.

Source: GSD2022 Data Extract

In the charts above, both unit’s capacity factors (dark purple) and availability factors (light purple) are binned into price bands of the South Australian Regional Reference Price.

It is worth noting that the existence of the LGC scheme (to be discussed later) augments the revenue of renewable generators, and can contribute to their profitability. Depending on the specifics of the PPA structure the interaction of Spot Revenue and LGC revenue might affect the way in which the solar farm operates. For instance, it might still be profitable for a solar farm exposed to spot prices to operate at spot prices slightly negative if the revenue earned on LGCs helps to offset the costs paid to generate at negative spot prices.  For this reason, we see a number of solar farms and wind farms bidding at what amounts to ‘negative LGC’ prices.

The bars in the highlighted red area of the charts show negative prices, and we can observe:

1. That both Solar Farms have tried to avoid generating at negative prices in 2022 (i.e. we see light purple in the bars in the negative price buckets);
2. But that Tailem Bend has been more active in trying to avoid negative spot prices.

Bringing this back to the focus on capacity factor, we surmise that this difference in behaviour (i.e. more volume sacrificed at Tailem Bend to avoid negative prices):

1. Will have contributed to a lower capacity factor at Tailem Bend
2. But might have contributed to a higher aggregate spot revenue earned by the station over the year.  But remember again that we don’t have visibility of either solar farm’s PPA to understand the degree to which this is a motivator.

Factor 5: Network constraints

Network constraints can have a significant impact on the generation figure used when calculating capacity factor, and in order to understand this, some analysis must be done to examine how much an individual unit was constrained, and for what reasons.

It is important to understand that a unit can be ‘constrained down’ (i.e. curtailed), but can also be ‘constrained up’ in any given dispatch interval. The constraint equations that are processed by NEMDE dictate these outcomes.

The ‘top five’ constraint equations that affected each unit throughout the year, and the amount of hours the unit spent on the Left Hand Side (LHS) of each corresponding constraint when it bound.

Source: GSD2022

For those familiar with the NEM dispatch process, the screenshot above shows the ‘top five’ constraint equations (by hours bound) and the amount of time that each unit appeared on the LHS of the equation when each of those constraints were bound. We can calculate from these tables that Bungala One appeared on the LHS of bound constraints around 167% more of the time than Tailem Bend.

For those who aren’t so familiar with the inner workings of the dispatch process, myself and the team here at Global-Roam have done our best to try to translate the AEMO’s constraint data into curtailment figures for each of these two units. Below I have used a two-dimensional scatter plot to visualise the incidences of curtailment at each solar farm. To assist in understanding these charts, I have created this graphic to provide a quick explainer on how to interpret them.

The red dots highlight the extent to which Bungala One was adversely affected by network constraints throughout 2022.

Source: AEMO MMS

In contrast, Tailem Bend was more active when attempting to avoid negative prices.

Source: AEMO MMS

The charts illustrate the slight differences in bidding strategies employed by each operator. It would appear that Tailem Bend was more active when attempting to avoid an expected negative price by bidding out. Bungala One’s location on the network, and its bidding behavior, resulted in it being ‘constrained off’ far more often.

Closer inspection of the data shows me that almost all incidences where Tailem Bend was ‘constrained on’ (i.e. the orange dots) occurred during the November 2022 SA islanding event – which was coincidentally caused by a transmission tower failure near the town of Tailem Bend. As Allan O’Neil noted at the time, ‘energy’ prices during the islanding event remained stable whilst FCAS prices skyrocketed to the Administered Price Cap. Such a mismatch of ‘energy’ and FCAS prices could provide an incentive for some units (of any fuel type) to bid out of the market to avoid the risk of a high Contingency Raise FCAS bill.

Factor 6: FCAS costs

All generators must pay their share of FCAS costs.

In the GSD2022 we included a table for each unit to estimate (using our own calculation) of what FCAS costs they will have incurred through each month of the year.

The month of November – which contained the SA islanding event – has been highlighted in the table below to draw your attention to how severe the costs were for each solar unit, and indeed all SA generators, during the islanding.

* Asterisk in the table denotes that the unit changed FCAS Portfolio during the month. See the GSD2022 for more detail.

Note: AEMO recovers regulation costs by FCAS Portfolio, and hence not always on a unit-by-unit basis. In our GSD2022, where a unit was part of a larger portfolio, we have attempted to calculate that unit’s share of FCAS costs.

Bungala One’s FCAS bill was almost $1M higher than Tailem Bend’s over 2022, owing mainly to higher FCAS (Regulation and Contingency) costs incurred during the SA islanding event in November.

Source: GSD2022 Data Extract

In the table above I’ve shown FCAS costs together, but readers need to be clear that sometimes FCAS costs are best considered separately.

There are two bundles of FCAS costs in the NEM relevant to generators:

Regulation Costs (Raise and Lower)

Regulation FCAS costs are allocated to both generators and loads through the well-known — and often misunderstood — Causer Pays framework. As Harley MacKenzie has explained in detail in this article, this mechanism is complex, and importantly, there is a timing mismatch between when a unit’s Causer Pays Factor (CPF) is calculated and when the corresponding costs are finally allocated. That mismatch can influence the incentive for a unit to respond to frequency deviations.

As Jonathon Dyson has discussed in the context of “The Rise of the Machines”, this timing issue has contributed to an emerging trend: some Semi-Scheduled units have adopted self-forecasting systems with the explicit aim of achieving a low (or even zero) CPF — thereby paying very low or no Regulation FCAS costs. Jonathon also notes an important behavioural by-product. A unit seeking a minimal CPF may reduce its output at times of day when system frequency is typically low. While this strategy can reduce Regulation FCAS charges, it can also reduce the unit’s overall capacity factor. Quantifying the size of this effect is challenging, however, because the key variable for CPF calculation is availability, not just dispatch-interval output, and availability itself may be artificially lowered under certain strategies.

This behaviour appears consistent with the 2022 outcomes shown earlier:

• • GSD2022 estimates that Bungala One paid almost $900,000 in Regulation FCAS charges for the year;
  • Tailem Bend, by contrast, paid $0.

A reasonable hypothesis is that Tailem Bend made use of self-forecasting through 2022 while Bungala One did not. However, to be clear:

• • This analysis has not examined the specific extent to which either solar farm used self-forecasting;
  • Nor has it assessed the specific impact on capacity factor.

Accordingly, the above observations should be treated as indicative only.

Contingency Raise Costs

Contingency Raise FCAS is treated entirely separately from Causer Pays. Generators simply pay their proportional share of Contingency Raise costs whenever they are generating — there is no CPF mechanism involved at all.

Following the South Australia islanding event in November 2022, generators in SA had a clear incentive to reduce their energy-market output (and, for those capable, increase their Contingency Raise enablement) in order to avoid unfavourable exposure to high Contingency Raise prices relative to energy prices. Naturally, such behaviour reduces annual and monthly capacity factors.

The scatter plots referenced earlier suggest that this may have been one of the motivations for the units examined. At the same time, some generators were also “constrained on” during certain intervals, complicating the picture.

In November 2022, Tailem Bend paid less in Contingency Raise FCAS than Bungala One — a pattern consistent with what was observed in other months of the year as well.

A third bundle of FCAS costs — Contingency Lower Costs — exists, but is allocated to loads, not generators. As such, this category is not relevant to the comparison presented here.

Other considerations

The reasons listed above are just some of the drivers for ‘how’ capacity factor can be misleading, but to answer the overarching ‘why’ listed in this article’s title, we must consider known and unknown motivations and incentives for generator behavior.

LGCs

An additional revenue stream for solar farms comes from Large Generation Certificates that can generated, banked and sold – but the exact returns from this stream are much harder to calculate.

Whilst some brokers (such as the good people at Green Energy Trading) publish a daily spot price for the value of an LGC, generators are not forced to sell each certificate at the time of generation and exact specifics of transactions or stored amounts is not publicly available.

It is important however to note that LGC revenue can skew market behavior, and therefore generation performance.

PPA and/or portfolio arrangements

The high majority of solar farms are built and sold on long-term PPAs, contracted at a fixed price. The terms of a PPA are typically confidential by nature, which limits our ability to understand why generators behave in certain ways.

Tailem Bend was constructed throughout 2018 by Equis Energy, and a 22-year PPA was signed with Snowy Hydro. While the specifics of that contract are not public, RenewEconomy reported in 2019 that “Tailem Bend has an off-take agreement with Snowy Hydro that broadly requires it to switch off when wholesale electricity prices go into negative territory”.

Bungala Solar Farm meanwhile was constructed by Reach Solar starting in 2017, and signed a ’long-term’ PPA with Origin Energy. The specifics of that contract are also not public but Origin has disclosed that it would acquire the associated LGCs generated by the solar farm as part of the PPA.

Origin Energy reported that it’s bundled PPA prices were around the $70-$80 per MWh mark when it signed a long-term PPA with Bungala One.

Source: Origin Energy

A half-year results presentation from Origin Energy in February 2017 shows that the average PPA contract price at the time that the Bungala PPA was signed was around $70-$80 per MWh. It’s worth noting that the counter-party, Enel (the operator of the solar farm), wrote-down a large portion of that contract value nearly three years later, after it encountered significant delays to have the solar farm fully connected.

Final verdict: Was capacity factor a good measure of performance?

Using our GSD2022 data extract I can calculate that, net of FCAS costs, Bungala One generated approximately $16.03M of spot revenue in the market in 2022, while Tailem Bend generated approximately $13.37M.

If we divide that number by the maximum capacity at each solar farm – Bungala One generated $145,727 per MW of installed capacity, while Tailem Bend generated $140,705 per MW of installed capacity. We can therefore see that normalised spot revenue results by this measure were not all that different despite what our basic, original, interpretation of capacity factor might have led you to expect.

Spot revenue, however, is not the be-all and end-all when it comes to defining an asset’s value to the market or assessing its performance. As I’ve stated, commercial arrangements play a larger role in the actual financial implications for each asset owner, and the operators of each solar farm may be motivated by other unknown incentives.

Key takeaways

The NEM is a very complex place, which means that casual observations are increasingly at risk of being overly simplistic.

The analysis presented above is only a start at trying to look ‘under the hood’ of how two solar farms are performing but I hope it has provided the following key takeaways:

• Operational decision-making can be just as important as decisions made during development. As we have seen, generator output and financial returns can be just as impacted by bidding strategy and behaviour, as they can be by design choices.
• Commercial arrangements and incentives largely dictate generator behaviour. But unfortunately, we are not always aware of these motivations as the details of these arrangement are typically confidential.
• Capacity Factor might be easy to calculate and interpret, but there is almost always more to the story. And while we can’t always take the time to assess and reassess generators on a comprehensive case-by-case basis, we should at least take caution when trying to compare or even rank them.