(Scientific) study on the calculated vs computer fuel economy discrepancy

Cody, don't sweat the time of day temperature differences. As long as the station's tanks are underground the fuel temperature doesn't vary enough to matter during 24 hours.

 

28000 miles on 2010 since 12/18/09.
Average mpg error of 65 tanks is 5.0%.
Most often filled to first click, but have shifted to second click.
Lifetime mpg (calculated) is 50.1 mpg.
Temperature range is -18F to 111F, with no perceived difference in mid error as a function of temperature, though I have not tested that. I color code the tanks by season, and it just looks about the same all the way through.

 

Very interesting read, thank you for sharing.

So I did see a few formulas but is there an average estimate we can all agree on... aka will the mpg reader on the Prius be off by 6% most of the time? Or do temperatures play to much of an impact to give an average?

 

bobwilson4web gets 6%, and his would be far more accurate for vehicle(s?) tested than my 5%, but it is right in that 5-6% range. Exception: The first 2010 deliveries seemed to show a 10% error.

 

These are my averages from 86 tank fillups over 2 years and 34k miles.

Gallons = 8.5
MPG Calc = 47.7
MPG Disp = 50.7
Error = 3.0
Error % = 6.3
MPH = 37
Temp = 63

The temps vary from high teens to 100+F. I did not see a noticeable correlation between temps and % error. It seemed to have more to do with different pump cutoffs, rounding off dollar costs, driving patterns between fillups, etc.

Over the 86 fillups all of the variations would appear to have been averaged out.

 

Interesting, but I think (sorry conspiracy theory in me) a lot of the discrepencies are the inaccuracy of the gas pumps. I got "ripped" at a station back 7 years ago when I drove 80 plus miles round tip to work each day. I had a taurus and got some funky fuel. filled up my 17 gallon tank and when it hit 18 i stopped and went in to complain. guy wouldn't do anything so I called weights and measures and they tested it and said It was off and the station refunded my entire purchase... and very apologetic. Odd thing was I got the best gas mileage from that fill up I ever got on that car. Car seemed to run fine ...

After that experience, I used to take a small 1 gallon gas can and fill it up to make sure it was in the ball park before filling up my tank (I know .. Weird). I would not do this with stations that I regularly filled up at (though I tried to use same pumps as much as possible...)

I know... I'm strange...


Nice thing is with the prius, not having that problem with no 17 gallon tanks... (-:

 

This same thing happened to me a few years back. I put 13 gallons into a 12 gallon tank. I never went back to that station again....

 

I just got an 2011 Prius II and I've kept the numbers (except for temp) on my other cars, so I'll just start keeping that info and pass it along when there enough data points.

 

Extremely interesting read.

I will go through it one more time and apply a similar analysis on my data (I have recorded each single fuel up since I bought the car more than 2 years ago). I have 63 fuel ups spanning 3 different countries, on very different weathers, up to the first click, or up to the brink and fuelled in very different gas stations (no brand - independent, or seven sisters brands).

My current average error is 6,36%.
FYI: my brother's Yaris is around 4%....
I will provide you the data as well, in TXT format as you requested, if you wish, so that you can run your analysis on my data as well as see if you reach the same conclusions.
The temperature for me is the average over the month of the average temperature. Since I fuelled a few times while on the alps in summer, I might have an average temperature of 25C but I fuelled up at 10C...

 

I have tried to plot my data in the same way as you, but I am getting weird results - I think this has to do with the fact that I applied as "temperature" values, the overall value of where I live and not where I fuelled up. And when I drive from Germany to Italy, e.g., there can be substantial temperature differences if I fuel up in Italy, Austria or Germany. And most often I would fuel up in Austria on the way back (colder) and use the rest of the fuel in Germany (warmer).
I think this falsifies the data, with respect to your analysis.
I will see if I can associate more accurate values to my tanks, but I cannot yet find a weather website that provides historical daily data for a specific region/city. I can only have an average monthly value...
I will send you in any case the raw data in TXT format to see if maybe I am making some mistake of sorts.

 

Try Weather Forecast & Reports - Long Range & Local | Wunderground : Weather Underground and go to "trip planner" .

You can enter the city, country of interest and a date span and it will give you historical averages for the time period you selected.

It's one of my favorite trip planning tools.

 

Or simply this...
History : Weather Underground

xs650 - thanks!

 

I should have thought of that one. It's actually better for what you want to do since it gives history for specific dates of a given year while my method give average history for a range of days as far back as their records go.

 

I've PMd my first two years of data to you in english and metric. Used your method for the average temp. Boston NWS data is pretty close to what we get here.

 

You show great diligence in collecting and presenting data and it hopefully will encourage others to keep records in a similar manner.

As far as the "vehicle fuel volume sensor" is concerned, I would assume that fuel injector delivery is what is monitored since this measure is essentially instantaneous, and I would assume that the measure is volumetric, i.e., proportional to injector open interval. So, if the fuel is colder, then for a given volume, a slightly greater mass is injected. If the filling station meters were exactly volumetric also, then all would cancel out. But all the filling station pumps that I see in use today are digital and perhaps they are required to do some temperature compensation in their readings so that you get what amounts to a litre at some STP. This would account for the slightly larger "winter" discrepancy.

There are other possible factors here. If the tank is plastic, then its dimensions (and volume) will change with temperature, and being plastic it may actually deform to a slightly greater volume when its full.

 

I always wondered why they just don't figure out your FE using the ECM data. They know the mileage that the vehicle travels, they know the flow rate of the injectors, and they use a CAN databus system that is capable of sharing information between modules easily. By knowing the flow rate, injector pulsewidth, and mileage, they can very accurately give you fuel usage. Even if they didn't want to use actual injector pulsewidth, they can use the MAF signal and calculate fuel usage from the MAF reading, also very accurately.

 

This *is* how it is done. They do calculate from the injector data and miles driven.

Tom