State of Charge: a boring topic? [The Battery Cycle #3]

Four years ago, with insights from Claudius Jehle, CEO of volytica diagnostics GmbH, we launched “The Battery Cycle” series of articles, aiming to shedding light on the complexities of Li-Ion batteries.
The series provided a huge amount of valuable insights for anyone involved in electric mobility.
Now, four years later, we return to those same topics—still thanks to the collaboration with Claudius Jehle—enriched by real-world data and lessons learned.
We welcome your questions, feedback and contributions at info@sustainable-bus.com.

In our last article, we explained both parallels and differences of the various subspecies of Li-Ion batteries such as NMC, LFP and LTO. The underlying basic principle is always the same, which we explained using the – to some extent ironical, but ultimately very helpful – analogy of parachuters for ions, congesting at the exit of their vessel (aka anode) during discharge, their adventurous and tedious travel through electrolyte and separator, their crowding during landing and finally slow migration into and settling in the opposite electrode (aka cathode).

Today we will stay figurative to explain the seemingly boring topic State of Charge (SOC) and give some interesting insights and take-aways. In fact, it is surprisingly difficult for the onboard electronics (which we will refer to as BMS, battery management system) to determine the SOC even if these systems have signals like cell voltages, currents and temperatures directly at hand.

State of Charge: how full is it?

Unfortunately, our microscopic (i.e. on atom/ion level) parachuter analogy becomes clumsy when explaining a macroscopic effect like SOC – so think of a glass with a certain content (trillions of parachuters). The maximum possible content is the capacity. The SOC is simply content per capacity, i.e. 150ml in a 300ml glass results in 50% SOC. Ha, we’re done, great.

Of course it doesn’t stop here, unfortunately. The only signals the BMS can really measure, in the very sense of the word, are the cell voltages and the current flowing through the cells (okay, and temperature, but that’s not needed for now). Yes, it doesn’t “measure” SOC, it’s estimating it!

Why measurement is hard

In the analogy of a glass, the cell voltage is synonymous for the current filling height. If you can accurately measure this, and you have a translation table to convert height into content, you should be good. That’s exactly how a BMS determines SOC…

…once in a while, at least, because most BMS cannot do this continuously: there is something that I call “foam”¹. The liquid in the glass is not water, it’s foamy beer. When drafting, gulping or shaking (i.e. when using the battery, or charging it), a layer of foam builds up, making it difficult even for the trained eye to measure filling height to the fraction of millimetres. This phenomenon and the characteristics of this “foamy voltage” (mostly called “overpotentials”) are highly dependent on chemistry, immediate past current flow,

1^ Disclaimer: There is no real foam, again just an analogy. For the geekier readers: The congestion of ions at the gates to the electrodes, i.e. them slowing down before entering the anode or cathode, and their then ensuing slow and tedious migration into that solid material, is causing the macroscopic effect of the settling voltages (foam) after current was flowing. It’s called the overpotential voltages that build up when current is flowing and slowly reduce after this has stopped. Exactly like drafting a beer.

The foam/overpotentials slowly decay, just like on a beer. The BMS frequently needs phases of idle and rest of typically 30-60 minutes, to be able to reset the SOC. In the busy times in-between, the only way the BMS can somehow keep track of SOC is to count how much beer (=current) went in and out and adding/subtracting it from the last known reset value. This is called “Coulomb Counting”, and the longer the time from last reset, the farther off this tally taking will be¹.

Determining SOC is not an easy task

Guess what, we’re still not done. So not only needs the BMS frequent and prolonged phases of rest to reset, and in the meantime roughly tracks SOC by tally taking, the shape of the glass, the mentioned ‘translation table’ is yet another obstacle.

In the left of the illustration, a cylindrical glass is shown. Easy game: 2x as much filling height means 2x the SOC, absolutely linear relationship. Sadly, that’s not the case for Li ion batteries: as alluded already in the last article, depending on chemistry, this (and other properties) of cells varies significantly. The “glass shapes” are highly irregular and manifold across chemistries.

To get images into your head, LFP looks like a broad, bellied, bulbous wine decanter and NMC has rather an elongated candy-shape (narrow, fairly straight, narrow). This lookup-table to translate height (voltage) to content is determined once by the manufacturer in a lab and programmed into the BMS; it is called the “open-circuit voltage characteristics”, or also OCV curve.

One battery, one number?

The main challenge, as one can easily imagine, comes during operation: 5mm of height in the belly-part can be a good 20 or 30% of total liquid content! In fact, LFP is so broad that between ca. 20 to 90% the voltage virtually doesn’t change. The slightest measurement error can result in extreme SOC errors. And then there is also this foam!

Imagine the BMS of an LFP battery joyfully finds itself one of those deadly needed idle periods and prepares for a reset – it waits until foam has settled (hoping that the asset remains at rest) but then realizes that it came to rest in the bellied part! The voltage measurement could equally translate to 20% or 40% SOC – an ordeal!

It gets worse: “A battery is 1,000 batteries pretending to be one”

I will wrap things up in a moment, but for the full truth, you need to bear with me for another moment. What you have learnt so far is that, particularly for LFP, accurate SOC determination is sometimes sheer luck. The often heard “…but Mr. Jehle, the BMS has full access to everything in microsecond resolution” is fair but doesn’t change physics.

So far, we spoke specifically about how the BMS determines cell-SOC. But you are interested in the battery system SOC. … the (Master-)BMS has the thankless task to come up with one single representative SOC for a flock of hundreds or thousands of individual cell values. And yes, they differ. Not in theory, but in practise.

For reasons we will dedicate at least one article to, this “disbalance” (also called imbalance or asymmetry) causes a challenge. The BMS cannot do the obvious and take an average of all cell SOC, because when discharging, the cells below average will be empty first and demand a stop at >0% SOC, and vice-versa when charging. So oftentimes there is sophisticated logic at play when the BMS performs this task, and believe me, sometimes making things worse.

So what?

Wait, what does this mean in practice? We will dig a bit deeper in the following articles about the most detrimental effects, namely sudden standstill (“sudden depletion events”) and inhomogeneous degradation, but one thing is universal: I estimate the average SOC error to be somewhere between 5-10%, reaching up to 10-20% in e.g. low-cost systems, high-duty (few breaks) applications and especially LFP/LTO systems. This is significant – a battery that’s off by 10% is, basically, 10% oversized!

Inaccurate SOC causes batteries to be oversized, operations to be erratic, assets not used to their true potential, money being wasted and frustration rising.

Hands-on tip #3

Action items and best practices for an accurate SOC:

• High-quality products (e.g. hi-accuracy voltage and current sensors)
• Periodic and sufficiently long breaks (for resetting and, consequently, balancing)
• Particularly for LFP/LTO applications: these resetting and balancing breaks should be at >80% or <20% SOC, to be outside the “belly region”, making it tremendously easier for the BMS to function, e.g. after charging.
• Make use of centralized battery monitoring with advanced SOC recalculation and warning functionalities, which will help you identify “off” assets, schedule balancing and maintenance and single out faulty assets.

But: high standstill SOC is one of the most detrimental degradation facilitators. Depending on the cell chemistry, there is a sweet spot somewhere between 80-100%. Ask your supplier or, rather, an independent expert such as diagnostics providers or academic labs for directions.

Know your OCV curve – demand transparency.

Always request a high-resolution OCV curve for the cells used in your battery system, ideally <5% SOC steps, in both directions. Need help defining this for your tender? Contact volytica .

1^ There are two fundamental reasons why Coulomb Counting is inaccurate: Firstly, because every current sensor has tiny offsets and biases; if you keep adding them up, they accumulate and will cause the assessment to “drift away”. I bet when you drink a large glass of beer and had to guess the exact amount of remaining beer just by counting your sips, you’d fail. At least I would. And secondly, the BMS needs the exact maximum capacity (glass size), which shrinks as the battery ages (we will cover than in the article about SOH), which is an even more challenging task, and BMS-SOH is notoriously wrong.