One Year of Battery Capacity Data

Since acquiring my 2015 Spark EV last summer, I've been collecting a variety of data from the car. My motivation has been to identify any battery capacity loss over time. Since my roundtrip commute is about 60 highway miles, which typically uses between 80 and 90% of the battery, losing 5-10% of the battery capacity could preclude me from using the car for my commute.

Among the data I collected, I recorded the usage percentages and energy used. These are displayed on the "energy details" page on the center infotainment display. With some simple math, this data should be a surrogate for the battery's capacity. For example, if during a trip I have used 82% of the battery and it indicates that 14.7 kWh were used, that implies the available battery capacity is 17.9 kWh (14.7/0.82). The one important caveat is that this math only works from a fully charged state. If the battery is only partially recharged, the car does not reset this data and the numbers are misleading.

Over the last year, I have taken 125 data points at the end of my commute. This was collected somewhat sporadically in time, so I binned the data into months as shown in the chart below. Each month has the individual and average usable capacity estimates. The chart also includes the number of samples for each month.



Two things immediately stand out:

1) The individual capacity estimates vary considerably, from 16.4 to 18.5 kWh. I have not been able to identify the cause of the variability within this range. It does not appear to correlate to ambient temperature, average speed, number of passengers, etc. I might be able to say that fall is better than spring, but again I don't know why that would be since the actual driving conditions are very similar then, especially when compared to the extremes of the summer and winter.

2) Within the variability of the data, there has been no apparent loss in battery capacity over the 14 months and the 14,500 miles I have driven the car.

An additional piece of information that I can glean from this data is the usable capacity of the '15 and '16 batteries. The average battery capacity appears to be about 17.7 kWh.

Zoomit said:

Since acquiring my 2015 Spark EV last summer, I've been collecting a variety of data from the car. My motivation has been to identify any battery capacity loss over time. Since my roundtrip commute is about 60 highway miles, which typically uses between 80 and 90% of the battery, losing 5-10% of the battery capacity could preclude me from using the car for my commute.

Among the data I collected, I recorded the usage percentages and energy used. These are displayed on the "energy details" page on the center infotainment display. With some simple math, this data should be a surrogate for the battery's capacity. For example, if during a trip I have used 82% of the battery and it indicates that 14.7 kWh were used, that implies the available battery capacity is 17.9 kWh (14.7/0.82). The one important caveat is that this math only works from a fully charged state. If the battery is only partially recharged, the car does not reset this data and the numbers are misleading.

Over the last year, I have taken 125 data points at the end of my commute. This was collected somewhat sporadically in time, so I binned the data into months as shown in the chart below. Each month has the individual and average usable capacity estimates. The chart also includes the number of samples for each month.

Two things immediately stand out:

1) The individual capacity estimates vary considerably, from 16.4 to 18.5 kWh. I have not been able to identify the cause of the variability within this range. It does not appear to correlate to ambient temperature, average speed, number of passengers, etc. I might be able to say that fall is better than spring, but again I don't know why that would be since the actual driving conditions are very similar then, especially when compared to the extremes of the summer and winter.

2) Within the variability of the data, there has been no apparent loss in battery capacity over the 14 months and the 14,500 miles I have driven the car.

An additional piece of information that I can glean from this data is the usable capacity of the '15 and '16 batteries. The average battery capacity appears to be about 17.7 kWh.

Click to expand...

I calculate my battery capacity for my 2015 Spark EV 2LT the same way and over the last 5 months I averaged 17.8 kWh for 41 charging sessions. I believe regeneration can have an effect on the capacity calculation but I do not know a good way to measure it. My wife usually drives the 2015 and always drives in D.

I also have a 2014 Spark EV 2LT and it is averaging 18.8 kWh. I drive the 2014 100% in L so I am getting much more regeneration contribution. One thing I have noticed with the 2014 is my full-charge range is always between 102 to 112 miles.

Regen could certainly affect the capacity estimate, but I've noticed that during prolonged downhill regen, both the percentages and the energy used numbers do go down. This would lead me to think the estimate is still close.

Zoomit said:

Regen could certainly affect the capacity estimate, but I've noticed that during prolonged downhill regen, both the percentages and the energy used numbers do go down. This would lead me to think the estimate is still close.

Click to expand...

I agree.

How much of this was home vs DCFC? I can only calculate % using DCFC since I never charge to full at home (or anywhere else). DCFC used to be about mid 16 to low 17 with rare 18, but now it's mostly low 16 to mid 17. If that's to be believed, I've lost about 1/2 kWh in 1.5 years with 2015.

None of this was using DCFC (it's never been used on the car), but the charging source shouldn't change the accuracy of the numbers. The battery just needs to be completely recharged and the car will reset the numbers.

High use of DCFC could increase battery degradation, apparently, but I don't know the particular risks with either of the Spark EV batteries (MY14 & 15/16).

SparkevBlogspot said:

since I never charge to full at home (or anywhere else).
//cut //
I've lost about 1/2 kWh in 1.5 years with 2015.

Click to expand...

Why do you never charge to full?

Doesn't the battery load balance only when fully charged? If you never recharge to full, could it be that battery is reporting a lower capacity because it can't do that load balancing?

Some people do not charge to full because full charges and deep discharges may have an impact on battery life (particularly if they sit at full charge).

I have not been tracking my battery usage diligently on my 15 Spark. However, I have noticed readings that roughly correlate to what others were seeing (~18.4-18 kWh when new, ~17.6 recently). Car is about 10 months on the road (a little less, actually) and about 11k miles (a little more, actually). The other day I took a trip and measured 17.0 and was a little worried.

So, I thought about it, did a couple tests, took some measurements, and just took two long trips where I managed 18.0 calculations again.

Mainly what I did was to return to driving the way I did when I first got the car. One of the things I have not seen anybody list is some form of discharge rate. Granted, without adding some stuff on we don't really have that info, but we do have a proxy we can use. The m/kW gauge gives us something similar (though you would also need to include drive time to get a real discharge rate). When I first got the car I was playing the energy saving game and getting mid-high sixes for m/kW (got 7.2 or more on the 18.4 kWh measurement). Then, after a few months, I started driving more naturally (slower than my previous car, but close to normal traffic speed). I was getting mid-high fives for m/kW when my measurements were ~17.6 kWh. I got about 4.8 m/kW on the trip I got the reading of 17.0 kWh.

Discharge rate and capacity is not quite as dramatic as with lead batteries, but several studies have shown some correlation. Add to that the fact that EVs do not have a constant discharge rate plays into this even further. Battery capacity is advertised under specific conditions (including discharge rates). Changing those conditions alters capacity. One of the speculations that I read for why lithium batteries are not as effected by discharge rate is the higher heat factor that lithium batteries experience at higher discharge rates. In one study a 50 Ah battery discarged at 5 A could be 30 C cooler than the same battery discharged at 50 A. Capacity was minimally affected. The heat helping to retain the capacity. I didn't want to spend the $40 bucks to see the rest of the study, but if true it could also be that the BTMS could decrease capacity if cooling the batteries too far for optimal performance. In that study, they were focused on lead acid, so I am not sure how many samples they took of the lithium batteries. Also, the abstract said discharge rate was negligible for lithium, but 1 or 2% looks negligible compared to the effect discharge rate has on lead acid batteries.

Part of the reason it is hard to see a pattern in temps is that it is one of several variables that has an effect on battery capacity. Trying to find a specific pattern in a specific variable without first accounting for the other variables is very difficult.

A few things I have noticed about either the car or how people are calculating capacity.

1) Using meters on chargers (or the house) to determine battery capacity. This may be a way to help determine loss, but it is a horrible way to determine capacity. Recharging, in and of itself, is not 100% efficient. Thus the kWh listed on the meter is what was pulled, but not what went in. Further, unless you only take readings from a single source. Unless you have personally calibrated them, there is no way to see how accurate the meters are. If two machines are off, one +1% and the other -1% you could have calculations off by .4 kWh with rounding. The car's meter may not be 100% accurate, but it would consistently have the same error.

2) Haphazard readings can potentially have a dramatic effect on how results look. If we assume 18 kWh usable capacity, then an error of just 5% comes pretty close to 1 kWh of capacity. It could be very easy to introduce a 5% error is not careful. While taking lots of random samples should account for this, it is still easily possible to show readings that indicate a trend that really does not exist. Two ways this error can be introduced (and stacked).

Not taking the reading at the exact same percentage each time. The percentage is calculated on the fly using certain variables, probably an assumed capacity, and discharge since last full charge. The problem is that actual capacity is in flux given outside variables that almost certainly are not accounted for (temp when off, how long sitting off while not charging, etc...), thus the calculation is likely flawed from the get go. Any error in the calc is magnified the further from 100% discharge you take the reading. If you just say, take the reading anywhere after 40% you could be introducing an easy percentage or two of error right there. To get the most accurate info it should always be taken at the same percentage of discharge.

Second, not taking readings at the same event. By this, I mean taking the reading at an observable event. If we always take readings at 50%, we can still cause a significant error. The fact that the percent used is a whole number (where we don't know where it shifts, is it 49.5 or 50 or 50.9?) means there is a little bit of play in our calculations. In my car, I have observed that for a given percent of usage, the actual used can vary by .3 kWh. At 50% it might read 8.9 when it switches from 49% and be at 9.1 by the time it switches to 51%. So unless I take the reading when either the percent changes or the next used value change after a percent change, then I am inducing a fairly significant error.

With enough random samples, this should even out. However, in theory depending on how reading are taken, I could show a greater loss than exists or hide the severity of a loss. Without strict consistency in how I take readings, calculations are about as good as the GOM.

3) Regen. Generally, this doesn't seem to have an effect because the numbers often reduce with significant regen. However, how does that calculation happen? If it looks solely at the energy sent to the battery then lots of regen would actually show a lower capacity (because charging is not 100% efficient). If they account for charging effeciency, is it a constant in the equation they use? If so, that calc is flawed too (charging is less efficient the higher the input power).

I was in Grass Valley this summer. I charged fully at the CHP, everything reset to 0 and headed out. After coming down the hill and getting to the SMUD station, my car showed 50% and 12.1 kWh used. That would give a reading of 24.2 kWh capacity. We all know that isn't correct. At some point, the constant regen totally screwed up the percentage and used calculations. What I don't know is if it was just the result of constant regen (only trip so far with that kind of regen, even coming down from Sonora was not even close), or was it the accumulation of multiple artifact errors in the calculation (the latter meaning regen always throws off the calc, just not enough to notice most of the time).

4) Multiple usage categories may effect total usage percentage (by up to 1.5%). If driving/acc shows 30%, climate shows 5%, and conditioning shows 1%, is that 36% or 37.5%? Does each category round? Is each category calculated separately compared to total assumed capacity? If they are, then 30 could be 30.49, 5 could be 5.49 and 1 could be 1.49 and total 37.47 instead of the displayed 36%. We can certainly make assumptions based on what would be the best and smartest thing to do, but do we know they took the best route?

5) Trying to force battery capacity data into a line graph. These graphs basically represent two variables. Capacity is affected by at least four variables. By removing variables without first making them constants (impossible in a real world driving scenario), it can be easy to see false patterns. Granted, I realize it might be the best we have handy, but ultimately still pretty flawed.

Yes, good points and useful examples. I recognize them but apparently didn't feel motivated to elucidate them like you have. "Ultimately still pretty flawed" I can live with that. I'm admittedly a numbers guy, like you I suspect, and recognize the limits to what can be reasonably extracted from driving the Spark EV.

In a way, I'm thankful for not having the access that LEAF owners do to their car's performance data. Otherwise we'd all be armed with a lot more trivialities to argue over!

Welcome to the forum

Thanks. I am more of a numbers guy too.

And I realize that while I talked about the effect of things, I didn't actually do the math. But, this is one of those times where the theory and the real world applications are vastly different. There are too many variables whose effects change based on other variables to give a consistent specific answer. The different equations might all be within a few percentage of each other, but at small starting values small changes can have seemingly more drastic effects.

The reality is some of the numbers needed are battery chemistry specific (and may not be published). And even then, the answer is still technically, "it depends." If you discharge at this rate at this temperature your capacity is X, but if you discharge at that rate at that temperature then the capacity is Y. It's not the same as simply pouring in X gallons of gas.

If I had to boil it down (because many people may not want to read that much), for the most accurate calculations readings should be taken consistently (same percent and at same event, i.e. Right when the percentage changes) for each reading. Temperature and m/kWh should also be recorded. And even then, without some heavy math, only readings where all four values are the same (percent, event, temp, discharge - only event having to be exact, the others within small ranges) might truly be comparable.

Probably one or two other things I'm missing too. The electric bike forums spend way more time trying to understand battery calculations than the EV forums seem to. But then, they have more control since a lot of that stuff is still fairly DIY custom.

I would think cell balancing to some degree occurs on every charge cycle. Of course, I don't have hard evidence for this, but working on some DIY battery projects, I activate cell balance on all charge cycles from simple "lossy resistor" to max efficiency DC-DC type. There's really no reason to wait until the battery is full (by that, I assume one of the cells is "full").

I don't charge to full, because I live on top of a long hill. I add 1 or 2 bars (probably 1.5 bars) by the time I get down the hill, so I use regen rather than friction brakes. On DCFC, charge taper beyond about 85% incurs more cost than charging at home not to mention wasted time, so I only go to about 80%. Added side benefit as 67goat mentioned is better battery life, though with the amount of DCFC I use, I doubt it's much.

Even with all that DCFC and running almost to empty many times, I've only "lost" 1/2 kWh, and even that's dubious. That suggests DCFC has little impact on battery life.

What I would be curious about is battery capacity loss with it constantly charged to 100%, disconnected from charger, then let it bake in 100F+ heat. This would be like Leaf in AZ where people just keep it plugged in, but due to lack of active battery cooling on Leaf, the battery just bakes in heat, thus leading to quick degradation. Due to liquid heat capacity, SparkEV might not be as bad as Leaf, but 100% + heat is probably bad combination. Active cooling on SparkEV would save it from such abuse if it's plugged in, though.

I'd don't have that kind of history, but I have not exactly babied my batteries.

I charge to full every day at work. However, they only stay fully charged for a couple hours while plugged in. I live in an apartment, so I rarely charge at home (maybe six times since last October, and only 12 amps @ 120 for 12 hours or less).

On hot days that occur on the weekend, my batteries have no protection.

I use DCFC in the neighborhood of 2-10 times a month. I only need it for long trips (going to the airport, going to the Bay Area, going south to see my family, and I go to see my friend 2-4 times a month - 120 miles round trip). I think 3 DCFC events in a single day is my record.

Lately I have been driving 65-70 on the freeway (except long trips where I travel 55-62).

I have drained the battery to 2-4 miles probably about ten times.

I am approaching 12k in just under 10 months.

Other than weather, temperature, and changes in driving style, I have not noticed a decrease in effective range. I just made a 103 mile trip on a single charge with almost 20 miles left on GOM, which is actually the best single charge I have ever had. I had one other 103 mile trip, but only had 6 miles left on GOM.

If you're keeping it plugged in, and EVSE doesn't disconnect, TMS should keep the battery happy at any temperature. But if you're going to let it bake in the sun and not plugged in or no power from EVSE (ie, public chargers), better to let it bake at mid level state of charge rather than at > 80%.

Speaking of baking, I let mine bake often, but they're always at 50% or less. When you're out and there's no chargers (which is most of the time), it has to bake.

I get that it would be better to have it at 50% or have it plugged in, but at home plugged in is not an option. And 50% charge is sketchy in case I need to go somewhere in an emergency. It takes me 50+% to get to the SMUD DCFC (though having the new station in Galt is a big help). It could easily take me 50% to get to the Salida or Tracy DCFC stations under the right road conditions (Salida is usually about 35%, but a hot, windy day with fast moving traffic could push that past 50%).

Having an EV as my only vehicle, while living in an apartment, and not living on the coast is not the easiest. I make it work, but it will be much better after I buy a house and can plug in at home when needed.

Zoomit said:

Since acquiring my 2015 Spark EV last summer, I've been collecting a variety of data from the car. My motivation has been to identify any battery capacity loss over time. Since my roundtrip commute is about 60 highway miles, which typically uses between 80 and 90% of the battery, losing 5-10% of the battery capacity could preclude me from using the car for my commute.

Among the data I collected, I recorded the usage percentages and energy used. These are displayed on the "energy details" page on the center infotainment display. With some simple math, this data should be a surrogate for the battery's capacity. For example, if during a trip I have used 82% of the battery and it indicates that 14.7 kWh were used, that implies the available battery capacity is 17.9 kWh (14.7/0.82). The one important caveat is that this math only works from a fully charged state. If the battery is only partially recharged, the car does not reset this data and the numbers are misleading.

Over the last year, I have taken 125 data points at the end of my commute. This was collected somewhat sporadically in time, so I binned the data into months as shown in the chart below. Each month has the individual and average usable capacity estimates. The chart also includes the number of samples for each month.

Two things immediately stand out:

1) The individual capacity estimates vary considerably, from 16.4 to 18.5 kWh. I have not been able to identify the cause of the variability within this range. It does not appear to correlate to ambient temperature, average speed, number of passengers, etc. I might be able to say that fall is better than spring, but again I don't know why that would be since the actual driving conditions are very similar then, especially when compared to the extremes of the summer and winter.

2) Within the variability of the data, there has been no apparent loss in battery capacity over the 14 months and the 14,500 miles I have driven the car.

An additional piece of information that I can glean from this data is the usable capacity of the '15 and '16 batteries. The average battery capacity appears to be about 17.7 kWh.

Click to expand...

i

This is cool! If you used excel for this, would be willing to share the file so I could start tracking my own data?

It will be interesting to see what happens in Oct, because that is the time of year I have normally gotten my best readings, but I just made a round trip from the charging station in Salida to Jamestown (technically it was Sonora, but really only in post office terms). 98 miles round trip. I had 8% left when I got back to the charging station (8-10-15 on the GOM). Calculated out, capacity would be right at 17kwh. I currently have about 21K miles on the car.

Temp in the valley when I left was 83. Temp was in the 60s in the valley when I returned. I didn't have any ESVEs with me, so I drove up mostly about 50 mph and down at about 45 mph. I was very concerned about the range. I haven't made that trip in quite some time, so I was extra cautious due to range. I came back late in the evening, so there wasn't much traffic to piss off with my slow driving.

67goat said:

It will be interesting to see what happens in Oct, because that is the time of year I have normally gotten my best readings, but I just made a round trip from the charging station in Salida to Jamestown (technically it was Sonora, but really only in post office terms). 98 miles round trip. I had 8% left when I got back to the charging station (8-10-15 on the GOM). Calculated out, capacity would be right at 17kwh. I currently have about 21K miles on the car.

Temp in the valley when I left was 83. Temp was in the 60s in the valley when I returned. I didn't have any ESVEs with me, so I drove up mostly about 50 mph and down at about 45 mph. I was very concerned about the range. I haven't made that trip in quite some time, so I was extra cautious due to range. I came back late in the evening, so there wasn't much traffic to piss off with my slow driving.

Click to expand...

I use the same EVgo station in Salida. My average battery capacity for the last 12 DCFC charging sessions, from 3/10/17 to 4/12/17 [17,700 miles now on the ODO] is 17.56 kWh. [2015 Spark EV 2LT]

Personally, I think the amount of regen put back into the battery between charging cycles is what is causing the calculations to vary so much. To get a more accurate reading, I think it would be necessary to fully charge the vehicle to reset the energy info screen to zero, drive in D on a windless day, over a flat road for 60 miles using a cruise control setting of 55 mph. Then use the energy info screen data to make the battery capacity calculation. I plan to try this to see if I am right.

I never look at the input from the charging station. Just the used from the car panel. Regen may affect it through rounding errors, but usually regen will decrease the kWh used and the percentage, so it should not affect the total too much.

The only rime that wasn't true was when I came down from Grass Valley. It was almost completely regen down the hill. It definitely screwed up the calc that time.

67goat said:

I never look at the input from the charging station. Just the used from the car panel. Regen may affect it through rounding errors, but usually regen will decrease the kWh used and the percentage, so it should not affect the total too much.

The only rime that wasn't true was when I came down from Grass Valley. It was almost completely regen down the hill. It definitely screwed up the calc that time.

Click to expand...

I have noticed when I add charge but not to the point of resetting the energy info screen, the percent used drops but the number of kWh used does not change. This screws up the calculation too.

67goat said:

I never look at the input from the charging station. ...
... I came down from Grass Valley. It was almost completely regen down the hill....

Click to expand...

I agree, with charging losses and TMS operating to either cool or heat the pack, as required, I don't see the point of monitoring what goes in. It's not all for battery charging.
What goes out is what's important.

For instance: Everyday I get the Onstar txt saying 'charging is compete'. Then 10-20 mins later I'll get the Chargepoint txt saying 'using very little power and may be complete'.
I know if I don't use the CP card at unplug I'll get a txt 'car unplugged' within 10 seconds so it is not network delay.
When I look at a daily usage report from the my CP acct I'll see the 10-20 mins of ~600 watt usage after it drops from the 3.3kW usage, when I get the Onstar txt.
I assume that is TMS or Battery Management continuing to use power for 'housekeeping'.

I love the Grass Valley area!!! :mrgreen: There's gold in them there hills. Seriously, backpacking gold prospecting / sifting opportunities. I met some friends that have the gear to do it.