Building on the momentum from part 1 of our series where we examined the wide-ranging behind-the-meter (BTM) and in-front-of-the-meter (IFTM) uses for stationary storage, we're now shifting focus to investigate a crucial techno-economic factor in stationary storage: the Levelized Cost of Storage (LCOS), as well as it’s “little brother” the Annuitized Capacity Cost (ACC). Brought to you by our stationary storage expert Gaël Mourouga!
This article is primarily influenced by the research paper Projecting the Future Levelized Cost of Electricity Storage Technologies by Schmidt, Melchior, Hawkes, and Staffell. Schmidt and Staffel have also created the Energy Storage Analysis Toolkit website, allowing you to experiment with various parameters and visualise the results.
I contacted Oliver Schmidt, who is now visiting lecturer at Imperial College London and consultant in the field of renewable energy and storage, to have a chat on his article and online tool. He was nice enough to send me a pre-print of his upcoming open-access book “Monetizing Energy Storage - A toolkit to assess future cost and value” which I warmly recommend if you’re interested in the topic of stationary storage.
For our readers well-versed in concepts like spot prices, round-trip efficiency, discount rates, and O&M costs, feel free to skip directly to section 5 of this article “Visualisation of LCOS”. Otherwise, I've got you covered! The first section will provide a short introduction to the intricacies of LCOS calculations.
Overall, I’d highly encourage you to read the article and, once finished, head over to the website to explore the graphs yourself, as I did while writing this piece.
1. What is LCOS?
The Levelized Cost Of Storage (LCOS) is an indicator that answers the question “What is the average cost of discharging one MWh of energy for my system?”. It can be calculated simply by adding all costs (from installation to end-of-life) and dividing by the total amount of MWh discharged during the system lifetime.
In a nutshell, it can be seen as the minimum value at which electricity should be sold over the system's lifetime to reach a positive return on investment.
In that sense, it is very similar to the Levelised Cost Of Electricity (LCOE) for generation systems such as wind and solar.
2. What is ACC?
The Annuitized Capacity Cost (ACC) looks very much like the LCOS but answers a slightly different question, which is “What is the average cost of providing one MW every year for my system?”. Similarly to the LCOS, it can be calculated by adding all costs and dividing by the lifetime power (in kW-years) of the system.
In a nutshell, it is similar to the LCOS but for power.
3. Which one should I use?
In both cases, the cost calculation is the same (see next section). The only thing that differs is the quantity by which you divide the costs, namely lifetime energy or lifetime power capacity.
If you are operating in a market where you place bids on energy (where prices are expressed in $/MWh) and you are confident about the yearly energy output of your system, then you should use LCOS.
If you are operating in a market where you place bids on power (where prices are expressed in $/MW) and you are confident about the yearly power capacity output of your system, then you should use ACC.
In the following, I will only manipulate LCOS in order to limit the number of equations and graphs but feel free to play with ACC as well on the Energy Storage Analysis Toolkit.
4. How do I calculate LCOS?
In scientific notation, the calculation of LCOS would look something like this:
Where n refers to a given year in the operation of the system, N is the system lifetime, and r is the discount rate.
I will define each component that enters this calculation in more detail in the following subsections. I will also illustrate the costs with examples from lithium-ion, pumped hydro and flow battery technologies.
Discount rate
The Discount Rate r accounts for a certain rate of interest to convert future costs and cash flows into their Net Present Value (NPV).
For example, assuming a discount rate r=10%, a cost C=$1000 during the 4th year of operation would actually amount to $683 in Net Present Value.
Investment cost
The investment cost consists of costs related to the installation of the system, scaled to both energy and power. For example, a renewable energy provider may want to co-locate with a battery system and ask an LFP battery maker to provide it with a 1MW/4MWh BESS.
The investment cost is therefore given by:
This represents the total amount of cash required to install the system, summed over the construction time of the system (relatively low in the case of a lithium-ion battery, but potentially much higher in the case of a pumped hydro system).
These costs are separated into power and energy, as a 1MW/2MWh and a 1MW/4MWh BESS may not cost the same. This is especially true for pumped hydro or flow batteries, where energy can be added completely independently from power.
Replacement cost
In my example of renewable co-location, the project owner probably wants the BESS to have a 25-year warranty of operation, to match solar and/or wind installations’ projected lifetime. Most LFP cells have a lifetime of 10-15 years, and the cells would probably have to be replaced halfway through the project, incurring replacement costs.
This equation is very similar to the previous one, except the discount rate extends later in the future (construction time + number of replacements * replacement time). Also, the replacement nominal cost for power and energy might be lower than the investment cost, depending on which elements of the storage system have to be replaced.
O&M cost
In the same example of renewable co-location, it would be important to keep the LFP cells within their Key Performance Indicators (KPIs) such as round-trip efficiency or rated power, which implies some Operation and Maintenance (O&M) cost throughout the system lifetime.
In this equation, maintenance costs add up over the years, and the right side of the numerator includes capacity fading mechanisms.
In a lithium-ion battery, capacity will decrease over time due to SEI formation layer and loss of active material. In a pumped hydro system, the capacity loss may include water evaporation. In a flow battery system, crossover from one reservoir to the other will cause capacity fading, although this decrease is generally reversible.
Maintenance operations for lithium-ion batteries include capacity and efficiency testing, thermal scans, as well as checks on the thermal management system like cleaning of the filters. In a pumped hydro system, they will include managing cavitation damage on the steel turbine, or replacement of the circuit breakers. In a flow battery system, they may include electrolyte rebalancing to recover faded capacity, or tank/tubing maintenance due to corrosion.
End-of-Life cost
At the end of the project, the storage system should be recycled or decommissioned, according to the system size and local regulations.
Lithium-ion EOL cost will include dismantling, shredding and recycling through a mixture of pyrometallurgical or hydrometallurgical techniques to recover elements such as Nickel, Cobalt and Lithium. Pumped hydro EOL cost will include decommissioning of the plant, possibly including assessment of natural land transformation. Flow battery EOL costs include battery disassembly and recycling of the separate components (generally regarded as easier than Lithium-ion battery components to recycle).
Annual charged electricity
In the most simple case, the annual charged electricity would amount to the nominal energy of the system, multiplied by the expected number of cycles over a year.
In a more accurate estimation, however, we can also account for the Depth-of-Discharge (for a Lithium-ion battery, cycling between 20% and 80% of the maximum capacity is typically expected to prolong the lifetime compared to the full range), the self-discharge rate (a full battery would empty itself in a couple of months if left idle), as well as degradation mechanisms (cycle degradation through SEI layer formation and time degradation such as corrosion).
This yields the following formula:
The round-trip efficiency is typically high for Lithium-ion systems (c. 90% in a lab), where most losses are through internal resistance and reaction overpotentials. Accounting for the Battery Management Systems (BMS), the round-trip efficiency tends to decrease c. 80%. Pumped hydro is also quite efficient, most losses being mechanical in nature due to pumping freshwater up and down. Flow batteries, on the other hand, have a lower round-trip efficiency (c. 60%) since they need to both pump a liquid and drive an electrochemical reaction, while also requiring a BMS.
Charging cost
Charging cost accounts for the cost of charging the system over its lifetime
The electricity price is usually assumed constant, equal to $50/MWh for In-Front-Of-The-Meter applications and $100/MWh for Behind-The-Meter applications.
Total discharged electricity
This is one of the most important parameters in the LCOS equation: it is the quantity of electricity that we need to sell at a minimum of $[LCOS] for the storage system to be worth it.
From the formula, we can see that round-trip efficiency and self-discharge rate will have a direct impact on the LCOS, as they directly affect the revenue stream of the battery.
5. Visualisation of LCOS
Using the Energy Storage Analysis Toolkit website we can now enter values for all the parameters introduced in the previous section, and visualise how each cost participates in the projected LCOS. These are the default values for a Lithium-ion system in an “Arbitrage” application:
And this is the resulting LCOS estimation, with a visualisation of how the different costs contribute to LCOS:
This result is on the higher end of the calculation from another reference in LCOS calculations, Lazard’s 2021 LCOS report, which would estimate an LCOS between $131 and $232 per MWh for a comparable case.
The website also provides uncertainty analysis, where you can enter ranges of likely variations for the different costs, and it will plot the impact on the LCOS calculation in the form of a box plot
To obtain this figure, I assumed that investment costs for Li-ion would be likely to go down by around 20% and that electricity prices would be likely to go up by around 70%. The investment cost decrease ultimately seems to dominate the LCOS, although there is quite a bit of variance that can be introduced by highly variable electricity prices (which is likely to be the case in a renewable-dominated future).
6. Comparison of different technologies
The website allows us to estimate the best technologies for a different number of cycles and discharge times, based on LCOS estimations:
Here, I used the default values for each technology, but you can modify e.g. the investment cost or round-trip efficiency to better reflect what you think the technology is capable of.
I was a little bit surprised to see that flow batteries were not competitive against lithium-ion at higher discharge durations, but if you decrease the investment cost (or shift it to O&M cost, as the electrolyte leasing model would allow) you will see that they become more competitive.
7. LCOS evolution over time
The authors also provide a “Lifetime Cost Projection” calculator, which allows us to choose positive or negative variations in either technology characteristics (investment cost, round-trip efficiency, etc…) or application characteristics (electricity prices, annual cycles) and visualise the impact on LCOS as a function of time.
First, I wanted to see what would be the projected LCOS if we assumed that Lithium-ion investment, replacement costs and end-of-life costs would decrease by 8% per annum, while round-trip efficiency and lifetime would go up by 2% per annum. I kept the confidence interval for all parameters around 3%. Note that you need the parameters from the Section “Visualization of LCOS” to obtain the same results as me.
And this is the resulting decrease of LCOS over time, with a confidence interval:
Not bad, we could reach the milestone of $100/MWh around 2034, give or take 4 years! This is in line with most projections, barring some major materials constraints in the next decade.
But then, I wanted to model what would happen if electricity prices increased and, more importantly, became more variable.
As you can see, I only assumed an average increase of electricity prices of 3% per annum (compared to -8% of cost reduction), and a relative increase of uncertainty of 5% per annum for future electricity prices (which, if we add factors like climate change, renewable penetration on the grid and fossil fuel price variability, doesn’t seem like such an unlikely value).
The result is the following evolution of LCOS:
As we can see, this increase in electricity prices could potentially prevent the system from ever reaching the landmark $100/MWh. Electricity prices are therefore pretty impactful on LCOS estimation and evolution, and ultimately on estimating product-market fit for different storage technologies in different applications.
The calculations shown here in terms of electricity price estimation were pretty simple, so we could benefit from more accurate revenue forecasting, in complement to LCOS estimations, for evaluating the potential of storage projects.
Conclusion
The LCOS is a very practical indicator, for many reasons: it manages to capture differences between technologies by accounting for their most important characteristics, while also being fast to compute, and it provides very convenient visualisations for different systems over time.
One of the limitations of LCOS calculations, however, is that it only takes a fixed value for the electricity costs, and outputs a minimum revenue. In other words, it assumes you always buy electricity at a certain price and calculates the price you need to sell the electricity at to make a positive return on investment.
There are some cases where it works really well:
If you want to sign a Power Purchase Agreement (PPA) with a Transmission System Operator (TSO), typically in renewable colocation applications, where you want to fix a buying/selling price for a certain amount of time.
In Behind-the-Meter applications where you get fixed peak and off-peak prices, as well as potential feed-in tariffs, and want to estimate the profitability of a storage system.
In the case of trading on electricity markets, however, storage systems can participate in different markets (ancillary services, wholesale markets), where prices can fluctuate pretty wildly based on daily, monthly and seasonal variations. More importantly, electricity trading works as an auction-based system, where you need to submit bids in advance before the prices are known.
Therefore, while you can definitely assume an average buying price for electricity and calculate your system’s LCOS, simulating actual trading on the electricity markets has the potential to better reflect expected revenues over the system's lifetime.
What’s next?
Congratulations on making it this far! In my next article, I will show how Reinforcement Learning algorithms are well-suited for this task. If this is of interest to you, stay tuned!
🌞 Thanks for reading!
📧 For tips, feedback, or inquiries - reach out
🌐 Follow us on Twitter, LinkedIn, and our website
😇 Learn more about angel investing through Intercalation Collective
My big issue with this approach is that the number of cycles per year has an outside influence on the result, and is a system property, not a technology specific property. It makes comparisons very hard as you have to basically go back to the root assumptions (which you may not be given) and normalize for an appropriate charge cycle.