[boltage]

What I have noticed when the battery is nearly full, the state-of-charge percentage shown in MyChevrolet falls more slowly than expected based on kWh used since last charge shown on the center screen. For example, if the center screen says 3 kWh used since last charge (to 100% FULL), MyChevrolet says that it is at 96% (4% used, implying a 75 kWh battery, obviously not the case) rather than the expected 95% (5% used).

On the other hand, when the battery is somewhere near 50%, the percentage seems to drop a lot faster per kWh. A similar calculation of kWh used divided by fraction used gives an implied battery size of only about 53-54 kWh. But then at lower levels like 25%, the calculation gives an implied battery size of 55-56 kWh.

Anyone else notice this?

 

[wesley]

This has been empirically confirmed. Displayed SoC does have somewhat nonlinear relationship with the raw SoC.

Let's take a look at the second graph I posted in a thread roughly half a year ago.

Detailed battery monitoring results up to 24,099 km

My Bolt EV reached the 1 year milestone a couple of days ago, so I gathered up all the data I've been collecting over all these times and made some graphs and analyses. First of all, the car has been driven for 24,099.4km (14,974.7 miles) for its first 365 days. It's not quite as extensive as...

www.chevybolt.org

If you charge to 100% full (displayed SoC) on an L2 charger, it will charge to slightly more than 96% raw SoC. However, displayed SoC will not drop below 100% until raw SoC goes below 95%. This is the "blip" that you see on the top right side of the plot. Therefore you have an extra 1% usable, which is the difference you noticed.

You will also notice that the plot's slope becomes steeper as you go near 0% SoC. Compare the slope between 0-25% SoC and 75-100% SoC. That's why it feels like the battery drains faster with less charge.

 

[boltage]

That extra 1% explains how that happens if charged to FULL.

However, if I charge to 88% (hilltop reserve), there is still a nonlinearity, where the percentage per kWh is largest in the middle of the range, compared to at the top or bottom.

 

[wesley]

That extra 1% explains how that happens if charged to FULL.

However, if I charge to 88% (hilltop reserve), there is still a nonlinearity, where the percentage per kWh is largest in the middle of the range, compared to at the top or bottom.

In my case it was happening at the bottom, as mentioned in my earlier post. The slope seems to be most gradual (least percentage per kWh) around 75-100%, moderate at 25-75%, and steepest around 0-25%. This occurs independent of the maximum charge level settings used. If you have an OBD-II dongle, you could collect the data and plot a graph like the one I did to see if it indeed has steeper slope at the middle in your case.

 

[Vertiformed]

This has been empirically confirmed. Displayed SoC does have somewhat nonlinear relationship with the raw SoC.

You will also notice that the plot's slope becomes steeper as you go near 0% SoC. Compare the slope between 0-25% SoC and 75-100% SoC. That's why it feels like the battery drains faster with less charge.

FWIW, my curve fit to the data is

• hvsoc = 4 + 1.01659 dsoc - 0.0009659 dsoc^2

This argues that, if your battery has a 57 kWh usable capacity,

• When you're at 75% displayed, you've actually only used 23% (13.1 kWh) not 25% (14.25 kWh), with 77% (43.9 kWh) remaining.
• When you're at 50% displayed, you've actually only used 47.3% (27 kWh) not 50% (28.5 kWh), with 52.6% (30 kWh) remaining
• When you're at 25% displayed, you've actually only used 73% (41.6 kWh), you really have 27% remaining (15.37 kWh) not 25% (14.25 kWh).

In other words, it steals about 1 kWh from you at the top end and gives it to you at the bottom end. But all that assumes raw SoC is the more accurate measure, which is at best unproven.

 

[wesley]

FWIW, my curve fit to the data is

• hvsoc = 4 + 1.01659 dsoc - 0.0009659 dsoc^2

This argues that, if your battery has a 57 kWh usable capacity,

• When you're at 75% displayed, you've actually only used 23% (13.1 kWh) not 25% (14.25 kWh), with 77% (43.9 kWh) remaining.
• When you're at 50% displayed, you've actually only used 47.3% (27 kWh) not 50% (28.5 kWh), with 52.6% (30 kWh) remaining
• When you're at 25% displayed, you've actually only used 73% (41.6 kWh), you really have 27% remaining (15.37 kWh) not 25% (14.25 kWh).

In other words, it steals about 1 kWh from you at the top end and gives it to you at the bottom end. But all that assumes raw SoC is the more accurate measure, which is at best unproven.

Neat. My quadratic fit at the moment stands at soc_r = 5.01 + 1.01soc_d - 0.00102soc_d^2. First and second degree terms are nearly the same, but the constant apparently differs by 1.

 

[Vertiformed]

Neat. My quadratic fit at the moment stands at soc_r = 5.01 + 1.01soc_d - 0.00102soc_d^2. First and second degree terms are nearly the same, but the constant apparently differs by 1.

Mine is hard-wired to be 4% raw SoC at 0% displayed and 96% raw SoC at 100% displayed.

On the one hand, you've gone lower than me, on the other, my lowest point is 13.33% displayed and 17.59% raw SoC, which is a better fit with mine (17.38% raw) than with yours (18.29% raw).

 

[Vertiformed]

@wesley, since I found it for another post, I should note that my endpoints are stolen from this data source, which says

• -3.2% uSoC = 1% hvSoC (and car won't go below 1% hvSoC)
• 0% uSoC = 4% hvSoC
• 6.52% uSoC = 10% hvSoC
• 10.87% uSoC = 14% hvSoC
• 100% uSoC = 96% hvSoC