A good review of lithium-ion battery cost reductions

“Re-examining rates of lithium-ion battery technology improvement and cost decline” – by Ziegler & Trancik

Open source article : https://pubs.rsc.org/en/content/articlelanding/2021/ee/d0ee02681f

It is clear that in the past few decades, lithium-ion batteries have improved in terms of various metrics – energy density, specific energy, power density, etc. while costs have also reduced significantly.

This publication by Micah Ziegler and Jessika Trancik from MIT explores the correlation of the decline in price with improved technology (Moore’s Law), cumulative production (Wright’s Law), annual production (Goddard’s Law) and R&D activity as measured via patent filings.

What is the main takeaway ?

“Learning rate” – which correlates the decline in prices with increasing production, is defined here as the price decline for battery cells upon doubling of the cumulative market size.

Note that a good correlation was found with cumulative, and not annual sales.

For price scaled by energy capacity, the analysis shows a learning rate of ~ 20 – 24%, with a higher rate for the cylindrical cells. That is, the price of battery cells per kWh of capacity has decreased by 24% (for cylindrical cells) with every doubling of the cumulative production.

We all know battery costs are reducing – what’s new ?

Perhaps the methodology here is as valuable as the takeaways. The authors have gone beyond energy capacity and considered other metrics to evaluate the progress with batteries.

(1) There are various sources of battery price data available in the literature (the authors mention having considered > 1,700 sources). A simple extrapolation of the individual sources leads to significant differences in the future costs. The authors show that battery costs could have reached $75/kWh anywhere from 2009 to 2027. This can partly explain the various projections made for electric vehicles to reach cost parity with internal combustion engines.

(2) Battery capacity is only one metric, and the authors then consider other performance attributes such as energy density (kWh/l), specific energy (kWh/kg) and power density (kW/l). When including these other measures, the learning rates are even higher, going up to 31% for cylindrical cells.

So can we use this to extrapolate to the future?

As the authors mention in the paper :

Even Wright noted in his seminal study of airplane costs that “time saving” was a difficult-to-value metric required to compare travel in a plane to that in a car.

Extrapolating is going to be difficult given so many variables at play and the difficulty to include all metrics of value to the end application.

One of the main limitations of such a study is not including what we don’t know yet – namely new battery chemistries and how quickly they will become mainstream. Also as OEMs race to accelerate electrification to meet upcoming CO2 targets, at risk is a medium term (next decade ?) shortage of raw materials, driving up prices. But the methodology here is solid and can be used again to make a revised assessment.

Check out the monthly newsletter covering the latest on sustainable transportation technologies and regulations. Sign up if you like such content.

Like it ? Share it !

Tagged electrification