In an earlier post BlueLightning suggested measuring the fully charged battery capacity by comparing the change in kWh with the change in SoC while charging. I have done something similar using measurements of amps to the battery and the SoC collected from my CZero now and then over the last 4 years. This covers mileage from 6000 km to almost 80000 km. Up to 50,000km the data was collected with Canion at about 1 second intervals. After 50,000km the data was collected with OBDZero at about 4 second intervals. I first divided the data into measurement sets without gaps. This resulted in about 400 measurement sets most of which were quite short. I then used this formula to compute Ah for each timestep:
Ah = (A0 +A1)*dH/2
Where A0 is the amps measured in the previous timestep and A1 is the amps measurement in the present timestep. dH is the timestep (either 1 sec. or 4 sec.) in hours.
For each continuous measurement set I added up the Ah values and the changes in SoC. I then computed the battery capacity using this formula:
CapAh = 100*Sum(Ah)/Sum(dSoC)
Where Sum(Ah) is the accumulated amp-hours to the battery, dSoC is a 0.5% change in SoC and Sum(dSoC) is the total change in SoC in each continuous measurement set.
Most of the original measurement sets were too short to give meaningful results. By setting a minimum of 20% change in SoC for a useful set, 100 sets remained. The capacities computed from these sets are shown as points in the graph above.
The loss of capacity over time is easy to see but there are ups and down along the way. The most obvious of these are the capacity measurements between 40 and 41 Ah at a bit more than 40,000 km. As noted in the graph these measurements were record shortly after the car was serviced at 40,000 km. I believe that as part of the service the car’s battery capacity estimate was reset to 41 Ah or higher. This agrees with bobakka’s observation that the battery capacity seemed to be too large when their car was serviced. This also shows that in reality I’m computing the car’s estimate of the capacity not the true battery capacity. The less marked ups and downs in the capacity may be due to temperature changes and of course errors in the measurements.
The red and green lines in the graph are two models fitted to the data. The green line is:
CapAh = 50.4 - 1.14*ln(km)
And the red line is:
CapAh =40.11 - 0.000036*km
The green natural log line fits the data a bit better than the red linear model. However neither fit is good and I need more data in other to be certain of which model gives the best description of aging. There are other reasons to believe that a natural log model best describes battery aging so with a bit of luck I will have a useable battery for some years to come.