Consumption Vs. Temperature (Graphs)

[Doogals]

Thought it would be interesting to compare my fuel consumption data points with the average outside air temperature. I assumed the curvature would look similar to each other in terms of peaks and valleys, but with consumption going up as the temperature went down.

However I was surprised to see how irregular the temperature graph was, compared to how smooth the consumption graph is. Maybe I'm missing something here, but any thoughts on this?

[GasSavers_BEEF]

well, a few things.

They actually correlate fairly well (inversely anyway). if you were to rescale the first graph to a range of 8-13, it would probaby match up better.

there is another factor in there too depending on location and that is summer blend vs winter blend. I ran into this during one of my experiments with a WAI. right in the middle of the experiment, I gained like 4-5mpg on average. That really messed things up.

you also have the factor of driving tendancies and route. if you drove the same place all the time and never drove differently, then this wouldn't be a factor. We always go different places and react differently on a daily basis depending on traffic, road conditions, and weather. Within the past year or so I have done quite a few performance upgrades to one of my cars and the mileage has dropped due to that. I know it is my heavey foot doing it but it is still a factor.

if you had the raw (excel data) I could probably make them look almost identical just rescaling them. there will be some deviation and I can't really account for all those deviations but they are probably more similar than you think.

[Doogals]

Good suggestion, starting to look closer. Also easier to read with the markers hidden.

[GasSavers_BEEF]

if you put the l/100km to negative (cheesy way to invert the numbers) you can see the way they line up.

if you make a second excel spread sheet (I assume you are using excel) you can just copy the first and manually insert a negative in front of all those within that column. you can also use a pivot chart to plot them both on the same graph. you could even put one or the other on a secondary axis.

[GasSavers_BEEF]

Ok, I eyeballed it (and added the negatives) and put together this cheesy pivot chart. For whatever reason, it started in January and ended in December. Hopefully this looks right and it didn't do the curve fit so it is a bit chopy.

[Doogals]

First off thanks for all your help beef, I combined both data series into one line chart with two Y axes, and didn't have to change the L/100 km to negative values as there was an option for 'values in reverse order'.

By adjusting the maximum/minimum values for each axis I can fine tune how the two curves line up with each other, but is this an accurate way to compare the data? Feels like I'm 'cheating' the results?

[GasSavers_BEEF]

I understand what you mean about the whole "cheating" thing. Nothing is ever 100% conclusive. All experiment have a level of uncertainty. You still have other variables that you can't really account for or change (at best, you can minimize them). Things like weather conditions (rain, wind, etc.), driving style (our styles always vary some), and route (other than the everyday drive to work).

I think the data is accurate for what you want. You are looking for a correlation. A correlation doesn't have to be 100%. You are also comparing things that are measured in different units so it is hard not to manipulate the data to some extent. you are also just manipulating scale and not the data itself.

You are looking for change (not necessarily percent of change). the fact that the change seems to match up means you have correlation even if the amount of change is very small for one vs the other. they do seem to inversely track each other. Also, you are doing this for your own edification and as such, there is nobody to fool or mislead.

Just my thoughts.

[Doogals]

My driving pattern seems to be pretty consistent, here is a comparison of my previous vehicle to my Patriot for the same period of time (but different years).

2007 Pontiac G6 Coupe (3.9L, 6MT) May to December 2011
2012 Patriot (2.4L, 5MT) May to December 2012

[GasSavers_BEEF]

I more meant (myself as an example), sometimes I go out to eat for lunch and sometimes not, I go to the grocery store only so often (just 5 minutes from my home), sometimes I need to run small errands. Shorter runs kill MPG especially when the car starts cold vs already warmed up.

also, with those different routes come different driving conditions as far as traffic patterns. I can drive on country roads going 50mph and get 4mpg better though in the city I am lucky to see 20 on some days during the lunch hour. I drive very consistent as well but you have to go with the flow of traffic in a given situation.

the hope is that over a month's time, it will average out.

I also wonder, what country are you from? I ask as you are looking at consumption as L/100km and not MPG.

[alvaro84]

I use l/100km too, but in this case I'd probably switch to mpg - being reciprocal to l/100km it would be more useful when I find the correlation between temperature and fuel consumption. It would probably give a bit different shape to the graph, and I think it would be more accurate than just negate the l/100km one. Km/l would suffice too, of course.

My mileage log at cleanmpg and fuelly returns a similar shape anyway, FE peaks often being somewhat late behind temperature peaks. Seasonal changes are clearly visible.