Build a better GOM

[TonyWilliams]

I want simple Rated Range displayed with the existing GOM.

It would simply use GIDs measured at as follows:



Full charge = (GID - 6) / (281 - 6)) = 1 * 21.3 = 21.3kWh

[where the actual GID is 281]



Half = (GID - 6 ) / ( 281 - 6)) = 0.5 * 21.3 = 10.65kWh

[where the actual GID is 138]



Turtle = (GID - 6 ) / ( 281 - 6)) = 0

[where the actual GID is 6]



So, the GOM would display:

Range = (4 miles per kWh) * 21.3 kWh = 85.2 miles,
When the car is at or very close to Turtle (it's not indexed on Gids), then the meter will show 0 miles.

A method to change the "4 miles per kWh to something else would be ok.

 

[palmermd]

I want simple Rated Range displayed with the existing GOM.
So, the GOM would display:

Range = (4 miles per kWh) * 21.3 kWh = 85.2 miles,
When the car is at or very close to Turtle (it's not indexed on Gids), then the meter will show 0 miles.

This is by far the best way to do it. I've been amazed that Nissan has not thrown away the algorithm in favor of this much simpler method.

Or do like Tesla and provide a button to change between the algorithm and the simple calculation.

 

[kovalb]

Rated range is used by Tesla in the Model S. It just means they use a fixed energy economy (miles/kWh) to calculate it. I believe it is based on the EPA rating, hence the name rated range. The Model S has very little variation in energy economy between city and highway driving. It also has a long range. Combined these facts make rated range doable for them. (EPA MPGe 88 city / 90 hwy - 13MY 85kWhr model)

The Nissan LEAF has much higher energy economy in low speed city driving than in high speed freeway driving. Therefore it is necessary to know how the car is being driven in order to estimate range with any accuracy. This is also true of other EVs I know of besides Tesla's. (EPA MPGe 129 city / 102 hwy - 13MY)

Modifications were made to the 13MY LEAF to address the range estimate. The initial accuracy improved noticeably while variation only slightly.

Please consider the scenario where you have been driving around town at lower speeds until the SOC is getting low. You then hop on the freeway and travel at high speed. Do you really want the range to indicate that you can make it to your destination when you cannot? Unfortunately due to the significant increase of energy economy at low speeds there is no simple answer for this. It can also be an issue when going down a hill and then back up.

This is not to say that offering a range estimate based on fixed energy economy is a bad idea; so long as it is in addition to an algorithm that tries to show the actual condition it can be very useful. Please note that the Chevy Spark has this feature and yet there are still references by Spark owners to their "guess-o-meter".

 

[TickTock]

I'm trying to come up with a better algorithm that will appeal to more people. This forum tends to be a bit more educated than many other forums so while we may be all comfortable with Tony's range chart and Stoaty's driving tips, my objective is to come up with something that eliminates the need to do math (to put it bluntly). For this, I think we need three numbers: Best case, Worst case, and Current. Best case and Worst case will vary person to person so the car needs to keep track of your round trip efficiency. This way you can, at a glance, know if you need to be crawling home or if you can afford the luxury of the freeway. For current, I presently us the rolling average efficiency of the last 10 minutes of driving. This way you know if you are on a track to beat the best case if you are pushing the range limits or if you are exceeding the worse case if you are pushing the speed limits. However, I am finding it difficult to code the round trip efficiency without access to gps data. The LeafSpy, I think, could do it taking advantage of the phone's GPS, but hardware meters are limited to what's available on the canbuses. The best thought I have come up with so far is to just go with a 24 hour efficiency period. Default interval will start/stop at 2am but users on the night shift can program a different time if needed.

 

[TickTock]

OK. I am going to try checking and resetting the efficiency counters the first time the car is turned on after 3am and having been charged (any amount). This way, even if you are out late, it will continue to collect the round trip data and only close out the day when you plug in. It is not fool-proof. If you are out past 3am and plug into a public charger, it will think you are back home which could lead to an erratic datapoint if you happen to live at the top (or bottom) of a mountain.

 

[GregH]

Anyone have ideas on how to detect when a Leaf is at its home charger?

Perhaps if you set the GPS to make a noise when approaching your home, there might be a message on the AV-CAN at the time the noise is produced..?

 

[TickTock]

Unfortunately, I can only monitor two canbusses. GPS may very well be there. Who know what other riches are to be found there...

 

[TickTock]

OK. I've found two simple ways to improve the GOM.

1) As mentioned elsewhere, use only round-trip efficiency history for predicting future range. Since the current implementation uses last trip, this can result in gross over-estimating or underestimating if your last trip involved a large net elevation change. Doesn't even have to be large. A 200 foot delta in your commute can result in 1kWh difference in your consumption going and returning
2) Not all gids are created equal. I know it has been mentioned anecdotally that you seem to get more miles per gid at low charge states by several members, but I only recently started actually plotting the Wh/g and the data supports this observation very decisively. On my car (presently at 68% SOH) the top 30% of capacity gets only ~60Wh/gid while the lower 40% that I use gets between 120 and 80 Wh/gid. The average is always between 78 and 80, as expected. There is about 20% on the very bottom I haven't characterized yet (I usually don't go much below LBW). I have checked this both ways (starting with 100% charge at home and starting with 100% charge at work). Anyway, this also results in poor range estimation because you end up under-estimating the remaining charge at low charge states.

I have already added #1 above which improved things but continued to underestimate my remaining DTE when returning home. This perplexed me because it really should be *over-estimating* my return trip since I live uphill from work (~200ft so it should take ~0.9kWh more to go home than to go to work which the neither the Leaf nor the CANary has any knowledge of). I am hoping addressing #2 will start to get it to make sense. Instead of just using a gids*.08 formula, I have implemented a look-up table to keep a running average if the true available kWh at each gid (with an weighted update at the end of each round trip for the entries observed during that trip). We'll see in a week or so if that helps.

 

[DaveinOlyWA]

OK. I've found two simple ways to improve the GOM.

1) As mentioned elsewhere, use only round-trip efficiency history for predicting future range. Since the current implementation uses last trip, this can result in gross over-estimating or underestimating if your last trip involved a large net elevation change. Doesn't even have to be large. A 200 foot delta in your commute can result in 1kWh difference in your consumption going and returning
2) Not all gids are created equal. I know it has been mentioned anecdotally that you seem to get more miles per gid at low charge states by several members, but I only recently started actually plotting the Wh/g and the data supports this observation very decisively. On my car (presently at 68% SOH) the top 30% of capacity gets only ~60Wh/gid while the lower 40% that I use gets between 120 and 80 Wh/gid. The average is always between 78 and 80, as expected. There is about 20% on the very bottom I haven't characterized yet (I usually don't go much below LBW). I have checked this both ways (starting with 100% charge at home and starting with 100% charge at work). Anyway, this also results in poor range estimation because you end up under-estimating the remaining charge at low charge states.

I have already added #1 above which improved things but continued to underestimate my remaining DTE when returning home. This perplexed me because it really should be *over-estimating* my return trip since I live uphill from work (~200ft so it should take ~0.9kWh more to go home than to go to work which the neither the Leaf nor the CANary has any knowledge of). I am hoping addressing #2 will start to get it to make sense. Instead of just using a gids*.08 formula, I have implemented a look-up table to keep a running average if the true available kWh at each gid (with an weighted update at the end of each round trip for the entries observed during that trip). We'll see in a week or so if that helps.

i have also noticed that not all GIDs are created equally. at first i thought the LEAF needed a warmup cycle since I noticed that the first 10-15 GIDs did not go nearly as far as any other ones did every morning but I almost always leave with a full charge but later noticed that only occasionally is the phenomena (about 30-35% of the time) repeated at lower SOCs even with the car sitting 6+ hours.

 

[TickTock]

That is you present SOH & max gids? Good to hear from another driver since I could not be sure if the phenomenon I am monitoring wasn't unique to batteries near EOL (I am presently at 66% SOH). Your experience sounds very similar except I have not ever seen 60 Wh/gid at the lower charge states - only when at or near full. Pretty soon, I'll be getting a new battery (probably going to stretch this one through summer to get the most out of it) and I'll, then, be able to tell how a new battery behaves.

OK. I've found two simple ways to improve the GOM.

1) As mentioned elsewhere, use only round-trip efficiency history for predicting future range. Since the current implementation uses last trip, this can result in gross over-estimating or underestimating if your last trip involved a large net elevation change. Doesn't even have to be large. A 200 foot delta in your commute can result in 1kWh difference in your consumption going and returning
2) Not all gids are created equal. I know it has been mentioned anecdotally that you seem to get more miles per gid at low charge states by several members, but I only recently started actually plotting the Wh/g and the data supports this observation very decisively. On my car (presently at 68% SOH) the top 30% of capacity gets only ~60Wh/gid while the lower 40% that I use gets between 120 and 80 Wh/gid. The average is always between 78 and 80, as expected. There is about 20% on the very bottom I haven't characterized yet (I usually don't go much below LBW). I have checked this both ways (starting with 100% charge at home and starting with 100% charge at work). Anyway, this also results in poor range estimation because you end up under-estimating the remaining charge at low charge states.

I have already added #1 above which improved things but continued to underestimate my remaining DTE when returning home. This perplexed me because it really should be *over-estimating* my return trip since I live uphill from work (~200ft so it should take ~0.9kWh more to go home than to go to work which the neither the Leaf nor the CANary has any knowledge of). I am hoping addressing #2 will start to get it to make sense. Instead of just using a gids*.08 formula, I have implemented a look-up table to keep a running average if the true available kWh at each gid (with an weighted update at the end of each round trip for the entries observed during that trip). We'll see in a week or so if that helps.

i have also noticed that not all GIDs are created equally. at first i thought the LEAF needed a warmup cycle since I noticed that the first 10-15 GIDs did not go nearly as far as any other ones did every morning but I almost always leave with a full charge but later noticed that only occasionally is the phenomena (about 30-35% of the time) repeated at lower SOCs even with the car sitting 6+ hours.

 

[nosuchthing]

Spark's GOM is what Leaf's should have been. It's absolutely fabulous. Nissan needs to reverse-engineer its algorithm and put it in 2015.

 

[GRA]

Spark's GOM is what Leaf's should have been. It's absolutely fabulous. Nissan needs to reverse-engineer its algorithm and put it in 2015.

And it also shows max, min and current range, just as TickTock is talking about.

 

[TickTock]

This is going to be harder than I thought. For the past few weeks I have been seeing wh/gid plots similar to this:

the noise varies run to run but the general overall behavior remained consistent. However, this morning, I forgot to plug in and had to leave home with only a 90% charge. I got the following profile:


Much more consistent values and near 80 for the upper gids. I wonder if the last 5% of charging is what is screwing up the gid estimator. Anyway I'll see if I can repeat this behavior. I may have to have multiple lookup tables depending in the initial charge state or something goofy like that.

OK. I've found two simple ways to improve the GOM.

1) As mentioned elsewhere, use only round-trip efficiency history for predicting future range. Since the current implementation uses last trip, this can result in gross over-estimating or underestimating if your last trip involved a large net elevation change. Doesn't even have to be large. A 200 foot delta in your commute can result in 1kWh difference in your consumption going and returning
2) Not all gids are created equal. I know it has been mentioned anecdotally that you seem to get more miles per gid at low charge states by several members, but I only recently started actually plotting the Wh/g and the data supports this observation very decisively. On my car (presently at 68% SOH) the top 30% of capacity gets only ~60Wh/gid while the lower 40% that I use gets between 120 and 80 Wh/gid. The average is always between 78 and 80, as expected. There is about 20% on the very bottom I haven't characterized yet (I usually don't go much below LBW). I have checked this both ways (starting with 100% charge at home and starting with 100% charge at work). Anyway, this also results in poor range estimation because you end up under-estimating the remaining charge at low charge states.

I have already added #1 above which improved things but continued to underestimate my remaining DTE when returning home. This perplexed me because it really should be *over-estimating* my return trip since I live uphill from work (~200ft so it should take ~0.9kWh more to go home than to go to work which the neither the Leaf nor the CANary has any knowledge of). I am hoping addressing #2 will start to get it to make sense. Instead of just using a gids*.08 formula, I have implemented a look-up table to keep a running average if the true available kWh at each gid (with an weighted update at the end of each round trip for the entries observed during that trip). We'll see in a week or so if that helps.

 

[TomT]

Yep, it is the best GOM I have seen in any EV to date!

Spark's GOM is what Leaf's should have been. It's absolutely fabulous. Nissan needs to reverse-engineer its algorithm and put it in 2015.

 

[TonyWilliams]

Yep, it is the best GOM I have seen in any EV to date!

Spark's GOM is what Leaf's should have been. It's absolutely fabulous. Nissan needs to reverse-engineer its algorithm and put it in 2015.

Ditto.

http://youtu.be/TPsgnI4vRoQ" onclick="window.open(this.href);return false;

 

[TickTock]

Here's what the full round trip looks like. This is actually filtered (I average over +/-4 gids to see the overall avarage Wh/gid becasue otherwise it is too noisy to tell much).

I reach work with about 100 gids so that really variable section is my return trip after the car has sat half full for 10 hours. Although the Wh/g reached 80 by the time I get to work, the net average over the top half when I reach work is always around 70 Wh/gid (there are a lot at 65 Wh/gid whenI first start out). By the time I get back home, the net average over the entire used range is closer to 80 (77 here).

Here's another graph from the next day. General shape is similar with minor deviations. Merely coincidental that the average is also 77.0, though.

 

[garygid]

I set the proximity alarm to bark when about 50 feet from my house,
but I have not looked in the CAR or EV logs for the bark event.

However, your 3 AM to 3 AM "trip" boundry should work well,
I would expect.