“LCOE is useful as a starting point, but is widely abused. It is used to mislead a lot of people about the relative value and cost of different energy systems. It’s like saying a car costs a lot more than a bicycle, so we should all buy bicycles. But they are providing different services. Now we are seeing solar produced at times when it has almost no value. A low LCOE doesn’t solve that problem.”
“The prevailing approach that relies on comparisons of the ‘levelized cost’ per MWh supplied by different generating technologies, or any other measure of total life-cycle production costs per MWh supplied, is seriously flawed. It is flawed because it effectively treats all MWhs supplied as a homogeneous product governed by the law of one price.”
1. Comparison is Intuitive
It’s intuitive. We all do it. On the supermarket shelf, across a choice of brands, we weigh quality against cost and then select the product for us which we feel is the best value.
Since 2009, labels have featured unit pricing — per kg or per litre — which levelises costs for each product and makes choices even clearer.
When different sources of electrical energy are levelised — per megawatt hour (MWh) — to try to compare their costs, it initially appears just as straightforward.
(Note: value range for Nuclear is based on the single data point of Georgia Power's Vogtle First-of-a-kind AP1000 project)
Lazard’s LCOE+ report is a major annual contribution to the discourse on energy, and was recently referenced by the Sydney Morning Herald, The Guardian, The Conversation, the ABC, the IEEFA, and analysis commissioned by the Clean Energy Council.
The levelised costs of energy are calculated by Lazard with reference to assessment of real world projects, with all input values provided in the appendices. In principle, anyone can plug these into the LCOE formula and replicate Lazard’s results.
Courtesy of the Australian Energy Regulator. A web calculator is available from the US NREL.
And this is a recognised advantage of LCOE:
“The levelized cost of energy (LCOE) calculator provides a simple way to calculate a metric that encompasses capital costs, operations and maintenance (O&M), performance, and fuel costs…” (NREL).
As a straightforward calculation, LCOE requires no fitting or modelling, or simulation of grids. The Electric Power Research Institute expands on this:
“What makes the LCOE metric useful is that it combines investment and operating costs and plant performance into a single metric that can be compared with projects that have different lifetimes and cost structures. Also contributing to LCOE’s widespread use is its ease to calculate, requiring only a few key inputs to determine anticipated costs averaged over a project’s lifetime output.”
The Australian Energy Regulator justified its recent employment of LCOE in the same way:
“LCOE is a common estimation technique used for comparing the lifetime costs of different generation technologies. …we chose LCOE for its simplicity, which makes it accessible, transparent and comparable.”
By considering each key input value by itself, however, something counter-intuitive becomes apparent: Lazard’s chart (above) includes energy sources with significantly different levels of output. Obviously, a geothermal plant will generate power very differently to onshore wind, which will generate power very differently to a gas-fired peaking plant, and so on. Despite its usefulness in some respects, an LCOE figure can’t provide information about an energy source’s value in supplying its market.
Lazard does caution that:
“Except as illustratively sensitized herein, this analysis does not consider the intermittent nature of selected renewables energy technologies or the related grid impacts of incremental renewable energy deployment.”
For the avoidance of any confusion, this quote is explicitly about the dissimilar values of solar and wind energy to conventional, dispatchable sources within grids. So, why does Lazard directly compare them?
To be entirely fair, this direct comparison has been widely adopted despite all warnings of flaws and abuse. For example, the OECD NEA’s Projected Cost of Generating Electricity includes a similar chart, based on a different dataset. The US Energy Information Administration provides the LCOE estimates for different technologies in a set of tables, but in this case clearly delineates dispatchable technologies from resource-constrained technologies, noting:
“Because load must be continuously balanced, generating units with the capability to vary output to follow demand (dispatchable technologies) generally have more value to a system than less flexible units that use intermittent resources to operate (resource-constrained technologies). We list the LCOE values for dispatchable and resource-constrained technologies separately because they require a careful comparison. We include the solar PV hybrid LCOE under resource-constrained technologies because, much like hydroelectric generators, solar PV hybrid generators are energy-constrained and so are more limited in dispatch capability than generators with essentially continuous fuel supply.
[…] The duty cycle for intermittent resources is not operator controlled, but rather, it depends on the weather, which does not necessarily correspond to operator-dispatched duty cycles. As a result, LCOE values for wind and solar technologies are not directly comparable with the LCOE values for other technologies that may have a similar average annual capacity factor, and we show them separately as resource-constrained technologies.”
“Solar PV hybrid”, specified as “coupled with a four-hour battery storage system”, i.e. “firmed renewable energy”, [see footnote] features a correspondingly higher LCOE , yet is still considered as resource-constrained technology by the EIA. The EIA notes:
“making direct comparisons of LCOE [...] across technologies [is] problematic and misleading as a method to assess the economic competitiveness of various generation alternatives.”
So, we can’t say we weren’t warned: it’s misleading to directly compare the LCOEs of dispatchable technologies with resource-constrained technologies in assessments of economic competitiveness.
Figure from Lazard’s Levelized Cost of Energy+ (2024)
Furthermore, Lazard has recently started assessing firmed renewable energy costs. The examples of large-scale solar plus 4 hour battery storage are among the highest estimates, and it is further explained that, “Total LCOE, including firming cost, does not represent the cost of building a 24/7 firm resource on a single project site”. Appropriately, “24/7 firm” technologies are not included on the chart: four hours of storage isn’t sufficient to be realistically dispatchable across 24 hour operation:
“...LCOE was originally developed to compare costs of dispatchable baseload nuclear and coal plants with the same capacity factors (similar generation attributes)...”
… Rather like how we don’t compare the price per kilo for some bananas versus a frozen pizza when we’re shopping for the week.
2. “Extending” LCOE
The CSIRO’s Low Emissions Technology Roadmap was published in 2017, overwhelmingly emphasising solar and wind energy in all pathways. Potential issues caused by solar/wind were indifferently addressed with a collection of “enablers” including storage, transmission & distribution, and peaking gas.
The next year, the CSIRO assessed various approaches to extending LCOE so that the “balancing costs” associated with achieving reliable and stable electricity supplies from solar and wind technologies could be included. As the report detailed, LCOE was still considered essential:
“...for the much larger community of non-modeller electricity industry stakeholders who want to understand why electricity models give the results that they do, or why investors are making certain technology choices, an LCOE measure is considered a very useful short cut.”
… Nevertheless, throughout the study the CSIRO stressed that the transparency and accessibility of the methodology was crucial:
“Without LCOE it is difficult to compare technologies without using sophisticated models and the results of those electricity market models become more of a “black box”, to be taken on trust, to non-practitioner audiences.”
In GenCost 2019-20, energy storage capacities coupled with solar and wind accounted for “balancing costs” (to some extent) in a similar fashion to Lazard and the EIA analysis. In GenCost 2020-21 and onwards, the “hybrid dispatch and balancing solution portfolio optimisation model” was implemented for shares of “Wind & solar PV combined” from 50% to 90% “with integration costs”.
Sources:
Two unintended consequences have arisen from this method of “extending” LCOE.
Firstly, a careful analysis of the GenCost report prior to 2024 has previously revealed that the bulk of integration costs — the anticipated investments in new storage, gas-fired peaking, and transmission line capacity — were treated as sunk costs, therefore excluded from the optimisation model, and therefore not added to “Wind & solar PV combined — Variable with integration costs” LCOE ranges. This is reflected by further explanation provided in the most recent GenCost 2023-24 report:
“To calculate the integration cost of variable renewables, we therefore start by allowing them free access to any existing flexible capacity (that has not retired). Next, we need to add the cost of any extra capacity the project needs to deliver reliable electricity.”
The confident, nation-wide reporting on the current superior costs of “firmed” renewable energy, as concluded by the CSIRO, doesn’t account for this.
Secondly, the lead author has stated in an online session that the CSIRO doesn’t intend to make the optimisation modelling public:
“In terms of all the detail around the modelling of the electricity system in 2030 and 2023 with these different renewable shares, we've been reluctant to dump a whole lot of that modelling detail out there because we're trying to support AEMO and the ISP process. What we don't want to create is a competing set of modelling that sits next to the ISP with a whole lot of different, slightly different outcomes.”
GenCost includes tables of key input data for LCOE calculations in its appendices, which is simple and transparent. In contrast, the optimisation model for integration costs, necessary to derive lower, economically superior LCOE values for “renewables backed by storage and transmission” is a black box.
This is the situation we now face: input data for stand-alone electricity generation technologies from GenCost (or a similar source) can be entered into an LCOE formula such as the AER's, and give a $/MWh result, yet the critical inputs for “firmed renewables” are inaccessible and the methods for deriving them are beyond the maths knowledge level of most interested Australians.
3. The way forward?
The line between the usefulness of LCOE, and its flaws and abuse, isn’t easy to draw or agree upon.
The EIA’s guidance regarding the division of dispatchable and resource-constrained technologies may be appropriately conservative, as it doesn’t attempt to transform solar or wind into dispatchable sources of electricity.
Alternatively, LCOE's inclusion in GenCost is distinctly less justified than the rest of the report’s cost modelling, and it isn’t utilised by AEMO’s Integrated System Plan. So, why even have it?
The CSIRO’s own most recently stated primary intentions could be the basis of a move to retire LCOE from the GenCost project entirely, and avoid future issues with narrow usefulness, inaccessible modelling assumptions, and misleading technology comparisons:
“The purpose of GenCost is to provide key input data, primarily capital costs, to the electricity modelling community so that they can investigate complex questions about the electricity sector up to the year 2055. LCOE data can only answer a narrow range of questions. It is provided for the purpose of giving stakeholders who may not have access to modelling resources an indication of the relative cost of different technologies on a common basis.
… If the LCOE does not answer a stakeholder’s question then they may need to commission their own modelling study. Making data available that can be used in modelling studies is the primary goal of GenCost.”
Footnote: This is an LCOE estimate for “firmed” renewable energy, but doesn’t include costs for additional transmission and other infrastructure.
Yorumlar