Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Car’s energy consumption (lack of) accuracy

This site may earn commission on affiliate links.
The car's dashboard does NOT show the TOTAL consumption of the energy that entered the battery!
This is a bold statement, but also a very verifiable one.

For 1 month, after each daily drive,
  1. I took note of the kWh used, reported in the car's dashboard (Model S 85D)
  2. I charged every night to exactly 70%, to top of that day's consumption,
  3. After the charge, I took note of the kWh added to the battery (you can get that number from the car’s dashboard after charge completion, or from the car’s API).
You would expect that the kWh in 1 (used) and 3 (re-added to the battery) would be the same, right? THEY ARE NOT!

For every single daily drive, I needed to add more kWh to the battery (refilling it to 70%), than what the car said were the kWh consumed in the previous day.

To be perfectly clear: I’m not talking charging efficiency. Only the “in” and “out” of the battery. You can read more about that in this other thread.

The plotted data shows a very interesting pattern:

drive_eff-png.315569

  • If you drive gently (~160 Wh/km), for each 50 kWh the car says you used, you can actually re-add around 52 kWh to the battery (~96% usage accuracy).
  • If you drive like a maniac (~280 Wh/km), for each 50 kWh the car says you used, you can actually re-add around 59 kWh to the battery (~84% usage accuracy).

I don’t know the definitive explanation for this, but:
  • It’s crystal clear that you can add more energy to the battery of the Model S, than the energy the Model S says is leaving the battery… and that can’t be right!
  • Because the energy entering the battery is coherent with the expected charging efficiency (see this thread), I can only conclude that the Model S is not reporting the TOTAL energy leaving the battery
  • Looking at the pattern of the graph above, my speculation / suspicion, is that the Model S is NOT measuring the battery’s discharge (in)efficiency:
    • If you discharge a lithium (or any other) battery at a high C-rate, you won’t be able to achieve its rated energy capacity. For example, you’ll have energy being wasted to heat;

I’m creating this topic:
  • To hear if you can replicate the experience (charge to X%, drive, take note of the kWh used during that drive, charge to the same X%, take not of the kWh added to the battery and see if it matches)
  • Hear your other possible explanations for this.
The car is doing a lousy job at showing its owners the actual energy cost to drive it. You need to add around 30% to the kWh x electricity rate calculation, to account for charging efficiency (see this other thread) and for the car’s lack of accuracy measuring energy consumption.

That being said, I love my Model S. I really think it’s the best car in the world and wouldn’t change it for anything (besides a P100D!)
 

Attachments

  • drive_eff.png
    drive_eff.png
    36 KB · Views: 1,881
Simply put, you can indeed put more energy into a battery than it will give back. Matter of fact, you will never hit 100%.

Why do you think that Tesla cools the batteries when charging? Because they get hot. Why do they get hot? Because energy is not used and converted to heat. And the faster you charge the batteries, the more heat they will create.

Take a any rechargeable battery, connect to a charger, measure the charge current, it will never reach 0.
 
Simply put, you can indeed put more energy into a battery than it will give back. Matter of fact, you will never hit 100%.

Why do you think that Tesla cools the batteries when charging? Because they get hot. Why do they get hot? Because energy is not used and converted to heat. And the faster you charge the batteries, the more heat they will create.

Take a any rechargeable battery, connect to a charger, measure the charge current, it will never reach 0.

1) As you should imagine, someone that's uses a meter, reads the car API, reads car data from the CAN bus, and opens a topic in TMC does know that there are inefficiencies, does know why Tesla cools the battery, and does know (and wrote on the 1st post) that a higher C-rate increases inefficiency.
2) I'm not talking about charging (that's the other thread).
3) Nothing that you point out justifies Tesla not taking discharge inefficiency into consideration in the calculation of the car's consumption.
 
Of course the battery itself doesn't have perfect efficiency, so it is impossible for energy out = energy in.

But it's also widely known that the consumption meter doesn't show vampire drain (although the remaining range does account for it).

Yes, vampire drain does account for some of the difference but, as you can see from the correlation between discharge rate and the error of the car's reported consumption, the main reason must be in the discharge efficiency.

And, as a matter of fact, the vampire drain SHOULD ALSO be included in the car's consumption. For example, any ICE car that has consumption during the time its switched off, will use more gas to power the alternator to recharge the 12v battery, as soon as it's turned on again. That increased gas consumption will show up on the car's average consumption.... unlike what happens on the Model S, where that "parasitic" consumption isn't accounted in any way. And that isn't the transparent way to go.
 
  • Like
Reactions: Dave EV
Yes, vampire drain does account for some of the difference but, as you can see from the correlation between discharge rate and the error of the car's reported consumption, the main reason must be in the discharge efficiency.

And, as a matter of fact, the vampire drain SHOULD ALSO be included in the car's consumption. For example, any ICE car that has consumption during the time its switched off, will use more gas to power the alternator to recharge the 12v battery, as soon as it's turned on again. That increased gas consumption will show up on the car's average consumption.... unlike what happens on the Model S, where that "parasitic" consumption isn't accounted in any way. And that isn't the transparent way to go.
I think most people agree that it would be nice if the energy meter would include the vampire drain, but the fact that it doesn't isn't motivated by any intentional non-transparency. I believe that the reason it doesn't is that the MCU is sleeping during that time and there is no CAN bus traffic during that consumption (in order to minimize the consumption). The battery ECU is still capturing it internally, so it should be possible for the software to forward the "missing" data to the MCU. Maybe Tesla will do that someday but it doesn't seem to be high on their priority list. They still have quite a lot of bugs in the audio system for example that I'm sure even more people care more about.
 
  • Like
  • Helpful
Reactions: nwdiver and Rocky_H
I don't want to march into the parade. I am also curious about these things.

Just want to say that afaik these kw readings are just approximations and not measured directly, but calculated. And given the accuracy achieved and the intent for the owner (know how much range is left, know how much to charge, ...) they are surprizingly accurate.

Much more so than many other cars (EVs or other). The range projection is really reliable.
 
I’ve been keeping a log similar to yours. I charge to a certain limit and then after one, two, or three days of driving, I recharge to that limit and compare what the car said it used to the amount of energy it took to reach the limit. The actual energy usage is typically 25% to 30% more than the stated amount after two or three days. It has been as close as 5% on trips when I recharge the same day. On the other end of the extreme, it’s been off as much as 80% when the car was parked at the airport for a week. The additional usage appears to be “vampire” losses; it would be good if Tesla addressed this. I’m driving a 2012 Model S, but the good news is the Model 3 my wife drives does a much better job. The vampire losses with it have been almost negligible from what I can tell, but I haven’t done a similar study.
 
Another aspect that I think does not get reported, but should not enter into your observations at this time of the year, is battery heating. So I think both battery heating and battery cooling go unreported and can account for some or much of the difference you see. I also wonder how much of a difference the temperature of the battery effects the charging efficiency even outside of the window of extra heating/cooling required.

My vampire drain is only 1-2% per day so I doubt that would add up to 20-30% unless you took 10 days or more to do your comparison.
 
  • Like
Reactions: nwdiver
I too was surprised by the numbers. Prior to collecting the data, I would have said my vampire losses were 2-3% per day. For fun I decided to keep track of the numbers, and that’s what I’m seeing. I’d be curious what others see.

I started collecting the data around March, so I don’t know what the battery heating would do to the numbers. I suspect it would make them worse, even with Range Mode on. I recently turned off “Always Connected” to see if that makes a difference. I do have a habit of checking on the car a few times a day, so I’ll need to stop that.:(
 
  • Like
Reactions: RuiP
ICE cars do not show the total energy put into the gas tank or the amount of fuel used when idling, etc.
Yes there are inefficiencies in charging/discharging the battery. Also, AC, heating, parasitic losses take their toll on range.
So yes, you put more energy into the battery than you get out moving the car. Why are you surprised?
 
1) As you should imagine, someone that's uses a meter, reads the car API, reads car data from the CAN bus, and opens a topic in TMC does know that there are inefficiencies, does know why Tesla cools the battery, and does know (and wrote on the 1st post) that a higher C-rate increases inefficiency.
Yes but I'd also assume that you know that there is no way to completely measure it, and even incomplete measurement would require a calorimeter for the battery which would be an astonishingly stupid thing to add to the car. Further I'd assume that you know that batteries are incredibly complex, besides the heat losses there are irreversible chemical changes that take place within the cells, that the internal resistance of the cell changes (nonlinearly) over age, temperature, state of charge, rate of discharge, and is different between charge and discharge all in ways that are difficult to model. I'd also assume that you know it is hard to measure the internal resistance for a variety of reasons, not the least of which is that not all of the cell potential drop while delivering current is due to resistance. It's also not feasible to measure the internal resistance while providing demand-based output power. But I guess you don't know these things, so I've just listed (some of) them for you.
3) Nothing that you point out justifies Tesla not taking discharge inefficiency into consideration in the calculation of the car's consumption.
I'd say that what I just pointed out justifies it. The battery SOC estimator, which is internal to the battery ECU is, by most accounts, the best in business at estimating actual remaining energy available but it also struggles through the mid-range region where battery operates most of the time and the voltage hardly changes with state of charge; as such, when it starts to see significant voltage changes, it can rapidly update its estimate of SOC. How would that be relayed to the energy consumption meter? My guess is you would complain about why is there some sudden giant load on the battery even though "nothing changed in how I was driving".

It seems that the energy meter is simple and just measures the time-integrated power flow at the battery terminals and only when the MCU is on. Perhaps they should throw in a nominal resistance estimate to slightly improve the results but it would still be imperfect. I don't fault them for not doing that.
 
ICE cars do not show the total energy put into the gas tank or the amount of fuel used when idling, etc.
That is not correct.
If you fill your tank with 50 liters and drive for 500 KM until the gas runs out, every single ICE car will report a 10l/100Km average. ICE cars obviously also report the fuel used when idling when calculating the average fuel consumption.

Yes there are inefficiencies in charging/discharging the battery.
For sure, nothing new here.

So yes, you put more energy into the battery than you get out moving the car. Why are you surprised?
What's surprising, is that parasitic consumption and discharge inefficiency are omitted from the energy consumption figure. They could and should be included. Otherwise, Tesla show rename that indicator from "consumption" to "a big part of your car's consumption"

I'm not even talking about accounting for wall socket consumption (accounting for charge inefficiency). Even though I think they also could and should take take into the calculation of the car's consumption.
 
Yes but I'd also assume that you know that there is no way to completely measure it, and even incomplete measurement would require a calorimeter for the battery which would be an astonishingly stupid thing to add to the car. Further I'd assume that you know that batteries are incredibly complex, besides the heat losses there are irreversible chemical changes that take place within the cells, that the internal resistance of the cell changes (nonlinearly) over age, temperature, state of charge, rate of discharge, and is different between charge and discharge all in ways that are difficult to model. I'd also assume that you know it is hard to measure the internal resistance for a variety of reasons, not the least of which is that not all of the cell potential drop while delivering current is due to resistance. It's also not feasible to measure the internal resistance while providing demand-based output power. But I guess you don't know these things, so I've just listed (some of) them for you.

I'd say that what I just pointed out justifies it. The battery SOC estimator, which is internal to the battery ECU is, by most accounts, the best in business at estimating actual remaining energy available but it also struggles through the mid-range region where battery operates most of the time and the voltage hardly changes with state of charge; as such, when it starts to see significant voltage changes, it can rapidly update its estimate of SOC. How would that be relayed to the energy consumption meter? My guess is you would complain about why is there some sudden giant load on the battery even though "nothing changed in how I was driving".

It seems that the energy meter is simple and just measures the time-integrated power flow at the battery terminals and only when the MCU is on. Perhaps they should throw in a nominal resistance estimate to slightly improve the results but it would still be imperfect. I don't fault them for not doing that.

There's a very simple way to do it:
  1. The car's BMS calculates, in real time, the amount of energy left in the battery. Meaning, in every single moment, the car has information of the kWh left in the pack. That’s were the SoC % indicator comes from.
  2. The car also measures the miles / KMs covered.
  3. Therefore, the indicator should be a very simple software solution:
(kWh_remaining t-x – kWh_remaining t) / (Odometer t – Odometer t-x) ,

Where x is the moving average intended to calculate consumption (from instantaneous to lifetime)​

For sure, like every single battery SoC calculator on earth, the energy capacity of the battery is estimating kWh based on voltage readings of the pack and some very complex algorithms. However, like you point out, it’s a very, very accurate one: when was the last time you anyone noticed their Model S SoC % indicator jump around numbers??


Would this indicator give a 100,0% accurate indication of the car’s energy consumption? Maybe it would have an <1% margin of error.

Would that give an IMMENSELY MORE accurate indication of the car’s real energy consumption, in comparison to the current method that totally ignores discharge inefficiency and parasitic consumption, underestimating real consumption for 5% to +30%? It definitely would!
 
Example of Efficiency difference between 120V and SuperCharging.

View attachment 315699 View attachment 315700

Darmie, in this topic, we are only covering consumption accuracy, not charging efficiency!
For the first, it's a more well know fact that you only get ~85%-95% efficiency when charging.
What's less known - and what's being discussed here - is that the car doesn't account all the actually energy used, in its energy consumption calculation.

A simple example:
- You pull 50 kWh from the wall and only 45 kWh end up in your battery. Charging efficiency => what you are pointing out.
- You then drive to empty, and the car you tell you that you only used 40 kWh of energy. It omits that 5 kWh were lost (i) in the discharge process and (ii) when you have your car parked, and it's communicating with "motherbase"

So, a lot of owners, who are showed the 40 kWh consumption, have no idea the car actually used 45 kWh, nor that it took 50 kWh from the wall socket to actually charge.
 
but the good news is the Model 3 my wife drives does a much better job. The vampire losses with it have been almost negligible from what I can tell, but I haven’t done a similar study.
My Model 3 is similar to the Model S numbers reported here - I typically charge a couple times a week when remaining range gets down to 100 miles and recharge to 70%.

Energy from the wall to charge is typically 25-35% more than what the dash indicates which is quite a bit of error. I would expect about 10% difference in the car was counting accurately due to charging inefficiency.
 
I should have clarified this a bit:
  • I have all the energy saving features turned on
  • I NEVER check my car with the mobile app
  • I also turn off the cabin control when I park the car in the garage.
  • I never use the mobile app to precondition the car before use

Also for the sake of clarity regarding the conclusions of the 1st post, I do exactly the same things to minimize vampire drain. Additionally, almost all my data points are from daily driving & charging.
These 2 aspects minimize the impact that vampire drain could have on the car’s consumption error.

Exemplifying this with a negative example:
- Drive 50 km over the course of 2 weeks, using 15% of the battery
- After those 2 weeks, top of the battery to the initial 70%.
- Simultaneously, at a 1%/day vampire drain, over those 2 weeks, you’d have lost 14% to vampire drain.
- In this example, you’d be getting a 50% consumption “error” just because of vampire drain (14% on top of a reported usage of 15%)!
However, repeat the same above example (use 15% on a 50km drive => then top of to 70%), but this time just in 1 day, and the vampire drain weight drops from 50% to 1% / 15% = 6,6%.

Trying to eliminate the vampire drain as a sole explanation for the consumption error I'm getting => in my dataset, I choose only daily drives of >80km, in order to dilute the impact of the vampire drain:
- 9 drives, with an average of 170km driven in a single day;
- Average consumption reported by the car: 37,9 kWh (223 wh/km reported)
- Average energy that entered to top of the battery: 41,5 kWh
- Consumption error in relation to the useful full pack: (41,5 - 37,9) / 72,5 = 5%
- These 5% are way higher than my daily vampire drain (which is less than 1%), especially for days when you’re driving for more hours.