Rated miles vs Wh/mi

[Boatguy]

Finally from this I think it is hard to pin down a car's Wh/RM from measurements.

Actually your results are quite close to mine. You used 1 RM for each 275Wh consumed, very close to the 273 that I observed and well within the rounding error.

I also went back and looked at some data I had collected for a trip to/from Yosemite. That round trip showed one RM consumed for every 275Wh outbound and 273Wh on the return for an average of 274.

I'm getting comfortable with the idea that Tesla clicks off an RM for about every 274Wh consumed, apparently on both the S70D and S90D.
While actual Wh/mi consumption per odometer mile is going to vary based on terrain, weather, driving style, etc., since my projected miles are based on 290Wh/RM, my true Projected miles should be about 5% lower (i.e., 275/290 = .95) and a car like the S70D which assumes 300Wh/mi actually has about 8% less Projected miles (275/300 = 92%) than is shown on the Energy graph.

This also implies that with a 100% charge of 289RM my usable battery pack is only 79.5kWh. Of course this is predicated on Tesla accurately reporting the total energy consumed and miles driven. In any case, Tesla's numbers do not cross foot. They really should clean this up in 8.0 and more accurately report Projected miles.

This is pretty easy to measure, perhaps a few others can report their Wh/RM.

 

[hacer]

...
I'm getting comfortable with the idea that Tesla clicks off an RM for about every 274Wh consumed, apparently on both the S70D and S90D.
While actual Wh/mi consumption per odometer mile is going to vary based on terrain, weather, driving style, etc., since my projected miles are based on 290Wh/RM, my true Projected miles should be about 5% lower (i.e., 275/290 = .95) and a car like the S70D which assumes 300Wh/mi actually has about 8% less Projected miles (275/300 = 92%) than is shown on the Energy graph.
...
.

I don't think it is constant. I think the SOC algorithm is complicated, and the car uses different Wh/RM values depending on how/what state the SOC algorithm is in. I once had a 16 mile trip fresh off of an 80% charge where I achieved actual 247 Wh/mi and the RM went down 17 miles! I don't think it could do that if there were a fixed Wh/RM constant. I hypothesize that when the battery is in the region of SOC where there is very little voltage change with SOC change that the algorithm uses coulomb counting (integration of current) times voltage (also accounting for voltage drop from internal battery resistance) with a fudge factor to determine SOC. The fudge factor probably depends on several things e.g. temperature, present SOC, history, etc. I think that fudge factor is trying to be "conservative" (while discharging) meaning under-estimate the SOC so that they don't lull you into over-taxing your usages and get yourself into trouble. When the SOC starts to go into the zones where the voltage is changing appreciably with SOC, the algorithm is better able to get to a true handle on SOC and doesn't try to be conservative. The effect is to see non-linear RM vs kWh used behavior.

A generic lithium-ion battery discharge curve looks like

Where the large voltage change with current (1C, 3C, 5C are different current draws) primarily are caused by cell internal resistance. In the middle part of the SOC curve, the voltage barely changes so it's hard to know where you truly are. Once you get into the high or low parts of the curve you can be much more certain about the SOC. Since Tesla keeps protective reserves at both the high and low ends, they have even less of the voltage-dependent regions to work with.

At least this can explain why for a whole bunch of trips, my SOC seemed to decrease much more quickly than expected because of the narrow change in SOC near the flat part of the curve caused the SOC estimator to be "conservative" (but always still with signficant range available so maybe not critically important to be super accurate) but once I charged to a high SOC, or discharge to a lowish SOC, the expected energy vs SOC was more in line with expectations.

 

[Boatguy]

I once had a 16 mile trip fresh off of an 80% charge where I achieved actual 247 Wh/mi and the RM went down 17 miles! I don't think it could do that if there were a fixed Wh/RM constant.

Yes, can't argue with that. That's similar to the mystery of how I could lose 5RM while the car was parked for 5hrs.

At the same time, your trip data, and my data for three trips, covering close to 1,000 miles, all produce a Wh/RM of about 273-274. Let's see if more empirical data shows a lot of variation, or converges on a number.

What we do know is that neither the Rated or Projected miles does a good job of forecasting how far the car can actually be driven. And to confound everything, the Trip Planner operates in SOC, so there is no way to cross check it with the other data presented. Tesla either has a really horrendous crew designing the UX around power and miles, or is intentionally confusing the issue by presenting a stream of numbers which can't be cross checked with each other.

 

[hacer]

...
What we do know is that neither the Rated or Projected miles does a good job of forecasting how far the car can actually be driven. And to confound everything, the Trip Planner operates in SOC, so there is no way to cross check it with the other data presented. Tesla either has a really horrendous crew designing the UX around power and miles, or is intentionally confusing the issue by presenting a stream of numbers which can't be cross checked with each other.

I think when SOC gets fairly low, it gets more accurate. It's hard to be accurate when SOC is high because who knows what driving you will do?

I agree with you that projected miles isn't good because it seems to be using a messed up power rating. Rated miles was never promised to be actual miles so it's harder to fault. I think the underlying problem here is in the SOC estimator and I'm a bit forgiving about that (now) because it is a tremendously difficult estimation problem in the long run. For all of my data that I've recorded, including the last trip, there has been a direct correspondence between SOC and RM. It has been RM = 240 * SOC; it's been this way (within 1% uncertainty in SOC) whether I beat rated miles at 270 Wh/mi or get less than rated miles at 247 Wh/mi. So the underlying issue with RM is SOC estimation. I initially wanted it to be simple, or at least when the SOC is in the middle range I wanted delta-SOC to equal kWh consumed as a percent of available battery, especially for mild temperature conditions that we've had here since I got my car.

Of course with large temperature changes, delta SOC will not equal kWh consumed. Another factor is that the internal resistance of the battery is SOC dependent, current density dependent as well as temperature dependent and I'm starting to think it plays a bigger roll in SOC estimation than I previously thought.

Performance cars doing a take off can have over 100V dropped across the internal resistance of the battery! Assuming this happens at the max battery current of 1,500 Amperes that's 67 milliOhms of resistance. The power loss during takeoff in that resistance is 150 kW (22% of total power delivered!), certainly not negligible. Your Tesla can easily measure the power and energy delivered by the battery but figuring out its internal loss is a lot harder. In more normal driving you might only have 40 milliOhms and be drawing 200 Amps, for a 1.6 kW loss (total guesses on these numbers) sometimes and other times have -50A and 90 W loss. But during normal driving you don't have the luxury of zeroing the current to see the resting battery voltage and even if you did there are long time-constant diffusion kinetics going on inside the cell that alter the potential immediately after current is stopped. so you won't get a decent internal impedance estimate. It is uncommon that the instantaneous power consumption equals the average consumption so you need to be accurate over a wide fluctuation. Bottom line is there is a lot of complexity and it's hard to be accurate when the battery voltage is at a nominal value.

All that said, Tesla really should alter the estimated range calculation to better align with the car's actual nominal energy per mile, whatever that number is.

 

[hiroshiy]

I think when SOC gets fairly low, it gets more accurate. It's hard to be accurate when SOC is high because who knows what driving you will do?

I agree with you that projected miles isn't good because it seems to be using a messed up power rating. Rated miles was never promised to be actual miles so it's harder to fault. I think the underlying problem here is in the SOC estimator and I'm a bit forgiving about that (now) because it is a tremendously difficult estimation problem in the long run. For all of my data that I've recorded, including the last trip, there has been a direct correspondence between SOC and RM. It has been RM = 240 * SOC; it's been this way (within 1% uncertainty in SOC) whether I beat rated miles at 270 Wh/mi or get less than rated miles at 247 Wh/mi. So the underlying issue with RM is SOC estimation. I initially wanted it to be simple, or at least when the SOC is in the middle range I wanted delta-SOC to equal kWh consumed as a percent of available battery, especially for mild temperature conditions that we've had here since I got my car.

Of course with large temperature changes, delta SOC will not equal kWh consumed. Another factor is that the internal resistance of the battery is SOC dependent, current density dependent as well as temperature dependent and I'm starting to think it plays a bigger roll in SOC estimation than I previously thought.

Performance cars doing a take off can have over 100V dropped across the internal resistance of the battery! Assuming this happens at the max battery current of 1,500 Amperes that's 67 milliOhms of resistance. The power loss during takeoff in that resistance is 150 kW (22% of total power delivered!), certainly not negligible. Your Tesla can easily measure the power and energy delivered by the battery but figuring out its internal loss is a lot harder. In more normal driving you might only have 40 milliOhms and be drawing 200 Amps, for a 1.6 kW loss (total guesses on these numbers) sometimes and other times have -50A and 90 W loss. But during normal driving you don't have the luxury of zeroing the current to see the resting battery voltage and even if you did there are long time-constant diffusion kinetics going on inside the cell that alter the potential immediately after current is stopped. so you won't get a decent internal impedance estimate. It is uncommon that the instantaneous power consumption equals the average consumption so you need to be accurate over a wide fluctuation. Bottom line is there is a lot of complexity and it's hard to be accurate when the battery voltage is at a nominal value.

All that said, Tesla really should alter the estimated range calculation to better align with the car's actual nominal energy per mile, whatever that number is.

I feel that RM calculations are off in the beginning of driving, so driving just 16 miles will have much higher RM consumption. Don't know why, but that's my gut feeling, as my commute is very short (5 miles, zero traffic jam).

So is that like this?
Classic cars 300Wh/m in the US, 320Wh/m in EU and JP
AP cars
S70D S90D 275Wh/m
P90D 290Wh/m

 

[Boatguy]

I feel that RM calculations are off in the beginning of driving, so driving just 16 miles will have much higher RM consumption. Don't know why, but that's my gut feeling, as my commute is very short (5 miles, zero traffic jam).

So is that like this?
Classic cars 300Wh/m in the US, 320Wh/m in EU and JP
AP cars
S70D S90D 275Wh/m
P90D 290Wh/m

The S90D definitely computes Projected Miles at 290Wh/mi. I continue to think that it's actually subtracting RM at a rate of 275Wh/RM. I'm driving about 90mi tomorrow, returning Monday so I'll track the consumption and RM and report back.

 

[hacer]

I feel that RM calculations are off in the beginning of driving, so driving just 16 miles will have much higher RM consumption. Don't know why, but that's my gut feeling, as my commute is very short (5 miles, zero traffic jam).

So is that like this?
...

Well my hypothesis of "conservative" SOC estimation would behave that way IF the starting SOC is in a middle range (maybe 40% to 85%). If don't charge for a few days and let your SOC go down to say 30% then do your commute, you might see a different answer.

 

[hacer]

So just to muddy the waters, this weekend I was playing with the REST API to observe my car charging for the purpose of eventually determining the most efficient current setting for charging. I have only one sample so far, but the API provides a couple of values with interesting names (these are Tesla's names): "charge_energy_added" and "charge_miles_added_rated" which if you plot one vs. the other you get a straight line with slope of 295 kWh/rated mile.

So I was taking a sample every 3 seconds and integrating the product of "charger_actual_current" and "charger_voltage" yielded 4.62 kWh power delivered from the wall which agrees very well with my meter reading 4.6 kWh. The "charge_energy_added" ended at 4.33 kWh. I hadn't driven much (14.6 miles since last charge, 4kWh used since last charge).

So I conclude that there was 330 Wh of vampire drain. The battery current during charging was pretty stable averaging 30.7 Amps. Even with 0.066 Ohms of battery resistance, the power loss in the battery would only be 62.8 W compared to 10,745 W being delivered to the battery, so almost all of the energy delivered to the battery should be usable energy at such a low charge current. This was all with the charger set to 48A although most of the time it was drawing 47.

So yet a new kWh/mi number built into my car: 295.

 

[Boatguy]

So just to muddy the waters, this weekend I was playing with the REST API to observe my car charging for the purpose of eventually determining the most efficient current setting for charging. I have only one sample so far, but the API provides a couple of values with interesting names (these are Tesla's names): "charge_energy_added" and "charge_miles_added_rated" which if you plot one vs. the other you get a straight line with slope of 295 kWh/rated mile.

So I was taking a sample every 3 seconds and integrating the product of "charger_actual_current" and "charger_voltage" yielded 4.62 kWh power delivered from the wall which agrees very well with my meter reading 4.6 kWh. The "charge_energy_added" ended at 4.33 kWh. I hadn't driven much (14.6 miles since last charge, 4kWh used since last charge).

So I conclude that there was 330 Wh of vampire drain. The battery current during charging was pretty stable averaging 30.7 Amps. Even with 0.066 Ohms of battery resistance, the power loss in the battery would only be 62.8 W compared to 10,745 W being delivered to the battery, so almost all of the energy delivered to the battery should be usable energy at such a low charge current. This was all with the charger set to 48A although most of the time it was drawing 47.

So yet a new kWh/mi number built into my car: 295.

I don't quite understand your point about vampire drain being 330Wh. Presumably this charging took about 40min during which vampire load should be less than 30Wh. If we assumed it was about 30Wh, then you drew 4.6kWh from the wall and put 4.36 into the car for 95% efficiency which is the same thing that I computed for my car a few weeks ago.

With regard to Wh/RM, I took a trip yesterday and returned today, about 80miles in each direction. I started out with 100% charge and got home with about 33%. In both directions I consumed one RM for each 273Wh. On the three round trips that I've tracked, a total of about 400mi, actual Wh/mi ranged from 286 - 302, but all consistently consumed one RM per 273Wh.

I also lost 9RM overnight. The car sat in the sun for about 7hrs with the temperature in the low 80's. This was similar to the vampire loss of 5RM I saw in 5hrs two weeks ago, again with the car sitting in the sun: HVAC obviously off in both cases. This is much higher than the 2-3RM/day I usually see parked in my garage. Another owner had a similar experience related in this thread Close Call...

 

[msnow]

I don't quite understand your point about vampire drain being 330Wh. Presumably this charging took about 40min during which vampire load should be less than 30Wh. If we assumed it was about 30Wh, then you drew 4.6kWh from the wall and put 4.36 into the car for 95% efficiency which is the same thing that I computed for my car a few weeks ago.

With regard to Wh/RM, I took a trip yesterday and returned today, about 80miles in each direction. I started out with 100% charge and got home with about 33%. In both directions I consumed one RM for each 273Wh. On the three round trips that I've tracked, a total of about 400mi, actual Wh/mi ranged from 286 - 302, but all consistently consumed one RM per 273Wh.

I also lost 9RM overnight. The car sat in the sun for about 7hrs with the temperature in the low 80's. This was similar to the vampire loss of 5RM I saw in 5hrs two weeks ago, again with the car sitting in the sun: HVAC obviously off in both cases. This is much higher than the 2-3RM/day I usually see parked in my garage. Another owner had a similar experience related in this thread Close Call...

Thanks for the link, I just read most of the posts but not all. I'd offer this: while parked in my garage on warm days I've heard my climate control go on several times which I *assumed* was on to cool the battery. I'm not saying that would cause all the vampire you are seeing but it could be part of it.

 

[hacer]

I don't quite understand your point about vampire drain being 330Wh. ...

I started the day with 80% charge, when I plugged in to charge the kWh since last charge was 4 kWh. It took 4.33 kWh to bring it back to the original 80%. That is 330 Wh more "charge_energy_added" than the trip meter said I drew from it, so I conclude that that was the entire day's vampire drain which didn't get counted in the trip. I am assuming that any power used during charging isn't counted as "charge_energy_added".

 

[Boatguy]

I started the day with 80% charge, when I plugged in to charge the kWh since last charge was 4 kWh. It took 4.33 kWh to bring it back to the original 80%. That is 330 Wh more "charge_energy_added" than the trip meter said I drew from it, so I conclude that that was the entire day's vampire drain which didn't get counted in the trip. I am assuming that any power used during charging isn't counted as "charge_energy_added".

Ah, ok, that makes sense.

 

[Boatguy]

I started the day with 80% charge, when I plugged in to charge the kWh since last charge was 4 kWh. It took 4.33 kWh to bring it back to the original 80%. That is 330 Wh more "charge_energy_added" than the trip meter said I drew from it, so I conclude that that was the entire day's vampire drain which didn't get counted in the trip. I am assuming that any power used during charging isn't counted as "charge_energy_added".

If you have occasion to be parked outside for at least 5-6hrs on a typical SoCal summer day, I'd be curious to know the loss while you were parked. I also suspect the HVAC, but it's a pretty big hit as the OP in the other thread learned.

 

[msnow]

If you have occasion to be parked outside for at least 5-6hrs on a typical SoCal summer day, I'd be curious to know the loss while you were parked. I also suspect the HVAC, but it's a pretty big hit as the OP in the other thread learned.

I parked at the beach here in SoCal for 6 hours on a 75 degree day. Maybe lost a mile or two.

 

[hacer]

If you have occasion to be parked outside for at least 5-6hrs on a typical SoCal summer day, I'd be curious to know the loss while you were parked. I also suspect the HVAC, but it's a pretty big hit as the OP in the other thread learned.

Well here in Maryland we don't get many typical SoCal summer days (a pity really), although we have had a few nice days recently.

I usually walk to work (leaving my Tesla parked in the garage at home) but sometimes I drive to work and then the car sits in the parking lot for many hours. I'll try to look at this the next time I drive in.