Energy Storage Systems (ESSs) Part I: Comparative Technology Model & Analysis for the Levelized Cost of Storage (LCOS)

In the end of 2016, Lazard published the second release of their Levelized Cost of Storage (LCOS) analysis, comparing cost trends in energy storage systems across application and technology. We had questions that we wanted to explore further. Given the various use cases for battery storage technology: what technologies were applicable to which situations? And why?

Lazard’s analysis provides various use cases by altering inputs to fit different use scenarios. We’re giving you the power to explore your own.

Trivium created our own LCOS model to explore battery costs further. Now, we’re releasing it for public use, complete with the analytical bells and whistles. Below, we’ll show you how to fly it so that you can adjust it to your own analysis.

[emaillocker id=”3778″]


Without storage capacity, the inability of solar PV or wind resources to provide dispatchable power will prevent renewables from providing a complete alternative to their fossil fuel counterparts. Fortunately, as technology advances, the costs of storage technology are beginning to near levels of economic viability.

From an economic perspective, storage technologies are compared by two key indicators: the annualized life cycle cost per kW, and the levelized cost of storage (LCOS). The annualized life cycle costs provide an “all in” estimate of the entirety of the costs across the project’s lifetime, normalized to an annual figure. The levelized cost of storage is the cost grid operators would need to charge in order to cover all economic conditions and costs associated with the project.

Running the Analyses

This excel model contains three different analytical tools for evaluating energy storage technologies:

  • LCOS and Annualized Life Cycle Cost Comparative Analysis: Excel models are generally geared to run single cases (technologies) through a model at a time. This tool runs each technology through the model and delivers the cost outputs, enabling the user to compare technologies.
  • Technology-specific Sensitivity Analysis: Sensitivity analyses are used to answer “what if” questions. For example, this tool allows the user to select a specific storage technology, and ask what the LCOS output would be if the discount rate was higher or lower across a range of possible rates. Sensitivities are explored across charging costs, discount rates and round-trip efficiencies.
  • Gravity Analysis & Tornado Chart: While the Sensitivity Analysis looks at changes to a specific input across a range of values, a gravity analysis examines single changes to many of the model variables in order to determine which variables have the greatest impact on the resulting cost outputs. A tornado chart is a visual representation of the level of “gravity” a certain variable may have on an output.

Analyzing the LCOS and Life Cycle Costs across Technology

The first analysis we will explore is the comparative cost table (as seen below).

What It Is: The technology cost table compares the cost outputs associated with each energy storage technology included on the Data sheet. Trivium has provided sample data for nine different technologies, which can be edited and updated as needed by the user.

How It Works: This table runs each technology found on the Data sheet through the model calculations (found on the Model sheet), and delivers the cost outputs.

How to Use It: All inputs for this table are found on the Data sheet, which holds assumptions for each storage technology, and the hypothetical project. After making any differences or adjustments to the assumptions on the Data sheet, press the “Run Table” button to update the comparisons.

Exploring the Storage Technology Comparison

To conduct this analysis, a few project-level assumptions need to be made. These will be held constant as the technology-specific variables are run through the model. In the simulation above, we assume the following as constants:

  • 20 year project life (independent of technology life, which may require replacements)
  • 10% discount rate
  • $0.04 per kWh for the storage charging cost
  • 5% electricity cost escalator

On some level, these project-level inputs are arbitrary – costs and rates are can reflect any number of market scenarios worldwide, and should be updated in accordance with the scenario being explored. Please feel free to change inputs on the data sheet as you see fit.

As we will find in the sensitivities section (below), some technologies might be better suited to environments where charging costs are higher (such as the Island use case in Lazard’s analysis), while others might be better suited to handle changes in required discount rates.

In the table outputs, three key metrics are explored: total capital expenditures, the annualized technology life-cycle cost per kW (and its component parts), and the levelized cost of storage in $/MWh. Our focus will be on the latter two as they provide a basis for comparability between technologies. The annualized life cycle costs are adjusted for time value of money, and compares the cost of a single unit of stored energy capacity inclusive of all expected life time costs (such as Replacements, Operations & Maintenance, etc).

The levelized cost of storage figure provides the price per MWh that the storage system would need to sell at in order to achieve financial viability. We chose to show this figure in MWh rather than kWh so that we could more easily visualize the relationship between the LCOS figure, and the life cycle costs for the respective technology being examined. These costs can be explored further in the graphs below.

The first graph compares technologies across their annualized life cycle costs, with an additional line exhibiting the levelized storage costs in $/MWh. The second graph takes a closer look at the breakdown in lifecycle costs by examining the component life cycle costs as a percentage of the total. This graph becomes immensely helpful once we begin exploring sensitivity analyses.

Sensitivity Analysis in Storage Technology Costs (Technology-specific)

Section 3 of the Analysis sheet examines energy storage costs at a technology-specific level by exploring cost outputs across a range of sensitivities.

What It Is: Sensitivity analyses allow the user to explore how model outputs – in this case, the annualized life cycle costs and LCOS – change when key input parameters are changed.

How It Works: These tables are created using a background program in VBA. The program replaces the input applied in the model (from the Data sheet), with a range (or ranges) of sensitivity inputs defined above or next to the sensitivity table. The outputs calculated from running the sensitivity variable through the model are written to the table, and then the original model input is restored.

How to Use It: Choose a storage technology that you would like to explore from the dropdown box, and optionally, adjust the sensitivity inputs (in blue only). Press the “Run Sensitivity Tables” button to run the analysis. The heatmap can be unapplied by clicking on the check box.

Exploring the Energy Storage Cost Sensitivities

Lazard’s LCOS publication breaks down a number of use cases for each technology by altering model parameters. For example, an Island microgrid might cost significantly more than a mainland microgrid even if the technology is the same because – all else equal – the cost to charge the storage system (in $/kWh) is significantly higher. While we didn’t divide our analysis by such use cases, we created a foundation for you to conduct your own exploration of energy storage economics through sensitivity analysis. The key sensitivities we’ve explored in this model are those relating to the project discount rate, the cost of charging the technology, and in a separate table, the “round trip” efficiency (kWh in/kWh out) of the battery. By analyzing these sensitivities, we can examine the resulting costs across numerous circumstances. While we have focused our LCOS exploration on the charging cost and discount sensitivities, there are many more variables that bare exploration – this is only a high-level overview.

Layered on top of each sensitivity table, we included an optional heat map. As visual aids, these heat maps end up being quite useful when exploring how different technologies respond to different sensitivities. For example, select “Compressed Air Energy Storage” (CAES – shown above) from the list-box and run the sensitivity table using the button. Apart from looking quite like a Senegalese flag, you’ll notice that – like other technologies with high O&M costs relative to CapEx – the LCOS output is significantly more sensitive to the cost of electricity than it is to the discount rate. This will certainly be the case for CAES because – as we can see from our charts above – it has the highest variable O&M costs across all technologies listed due to additional natural gas requirements to utilize the technology.

Now we’ll examine a technology with comparatively lower O&M to CapEx costs such as Sodium-Sulfur (above). Do you see the difference in the heatmap? We can see that for this technology, the discount rate (applied over the project life) has a significantly higher impact on the levelized cost calculations than the cost of electricity. This makes sense: the electricity price (as an input into O&M costs) is escalated at 2.5% while the discount rate remains at 8.5%. The greater the CapEx intensity relative to the O&M costs, the greater a role the discount rate will play in the cost of the technology over time. If we wanted to explore these variables further for a specific technology, we can head over to the data sheet and alter the inputs to our liking, or create additional sensitivity tables. This, however, can get extensive, and tends to be a bit of an academic exercise unless a specific situation is being analyzed so that the sensitivities can be better chosen.

Gravity Table Analysis & Tornado Diagram (Technology-specific)

In the previous section, we explored three sensitivities that might be applicable to multiple technologies. This section takes a deeper dive into analyzing which parameters of the energy storage system might be the most “sensitive” to adjustments, enabling the user to identify key inputs to explore.

What It Is: A gravity table ranks the model’s most impactful inputs, and the tornado diagram provides a visual representation of these impacts.

How It Works: At the core, a gravity table is a table of sensitivities. While the sensitivity tables above explored a range of changes to specific inputs, a gravity table explores single changes across a range of inputs in order to determine which input creates the most significant changes in the cost outputs.

How to Use It: In order to use the gravity table, select a technology, and choose which “deltas” (the degree of change to be applied the input) that you’d like to analyze, and run the table. Apart from the base inputs applied on in the Data sheet, the gravity table requires the most technology-specific judgment in terms of which “deltas” – the figures inputted should ideally be realistic, and relatable in size to one another. This is not always an easy task.

Exploring the Gravity Sensitivity Tests

We will explore the gravity table analysis by first selecting the Nickel-Cadmium storage technology.

The “Current Value” row shows what the current assumptions are (from the Data sheet) for the technology under analysis. The row below, “Delta”, allows us to define the changes we might explore in the inputs. This is where subjective judgment – and perhaps, research – comes in.

The first sensitivity is for the technology’s Round Trip Efficiency. While this input was explored in the previous section’s sensitivity tables, it was not done so in relation to other inputs. As the current value is 90%, we might explore the impact of a 10% loss or gain in this efficiency – perhaps as technology improves, we might expect increases in this efficiency. By making our delta value 10%, the analysis will apply value 80%, 100% respectively. When we run the table, we see that out LCOS range with a 10% delta is $852 under an 80% assumption, and $696 under a 100% efficiency assumption. This process repeats for the rest of the variables being examined.

From the modeling perspective, one variable warrants special discussion: the number of battery replacements during the life of the project. This variable differs because it is a calculation (row 54 in the Data sheet) rather than the driving input: the operational life of the technology. We used the number of replacements in place of the technological life to spare the user from figuring out how many years might be required to incur an additional replacement. One thing, however, is clear: the number of replacements in a project plays a dramatic impact on the costs. We can see this relative to our other changes through a Tornado Diagram, which is a visual representation of the Gravity table above.

A key takeaway from the chart above is that for Nickel-Cadmium, the number of replacements required over a 20 year project plays a massive impact on the cost outputs.

References, Data, and Other Notes on “Flying” the Model

The spreadsheet model is divided across three sheets: Analysis (which we’ve covered above), Data, and Model.

The Data sheet is organized between Global inputs, and Technology-specific inputs for energy storage system (ESS) technologies. The Global inputs define project-oriented parameters that remain constant across technology type. The Technology-specific Inputs are organized by ESS technology, but only one technology can be “applied” in the model (except on the Analysis page). To choose a technology, the operator can select the chosen technology to evaluate in the dropdown box contained in the Dashboard section of the page.

The Model sheet houses the calculations required in order to determine the two key model outputs: the total annualized life cycle costs (and the respective components), and the levelized cost of storage. After exploring numerous calculations, we decided on choosing a model and data from a book by Francisco Díaz-González, Andreas Sumper, and Oriol Gomis-Bellmunt, which can be found here: Energy Storage in Power Systems. Data has been converted from EUR to USD from exchange rates provided at time of publishing.

Methodology Notes and Model Limitations

One of the key parameters we decided to exclude from this model is battery degradation. In almost any energy project, the degradation of a project unit over time is an important factor in the project’s economics. But with energy storage technologies, the degradation of a storage unit is tends to be highly dynamic, and largely dependent on usage patterns. For example, a lithium ion battery that runs through 10000 cycles at only a 50% depth of discharge will degrade along a significantly different curve than a a scenario that looked at 2000 cycles at a 90% depth of discharge. Exploring these differences across the variety of technologies presented is beyond the scope of this project.

Another set of parameters left unexamined (at least, at an explicit level) by this model are the technology cost reductions that might be assumed for replacements. In our data, a Lithium Ion battery has 7 years of operational life before the battery storage unit (independent from power capacity and BOP costs). Over a 20 year hypothetical project life, the battery will (hypothetically) be replaced twice at 7 year intervals. Given the level of research going into lithium ion technology, steep cost reductions might be expected over the next decade or two, and the replacement costs may be significantly lower than present price quotations. While this cost reduction curve is not explicitly modeled, it can be implied through lower Replacement Cost per kWh assumptions.

Wrapping Up

This report – the first of a series – has explored some of the technicalities of evaluating energy storage projects, and has provided readers with a basic tool to evaluate the feasibility requirements of such projects. As costs for storage capacity decline, energy storage technologies will undoubtedly play a much larger role on projects. Going forward, future Trivium reports (and open sourced models) will explore the financial implications of energy storage. We plan to delve into battery implications for mini grids, and to explore policy options for financing energy storage in emerging markets.