Ecars stats - a work in progress

Black_Knight

innrain wrote: »

That is pretty neat. Now you just need a mobile app where the user select the charger they want to go to and get advised if the charger is likely to be free @ arrival time. Get notified if the charger is getting used while driving towards it with an revised likelihood of being free @ arrival time. Considering our network will remain behind the EV take out for a very long time the app would be useful enough. Find a way to monetize this and you're flying with Bezoos to the moon

So I heard you say... add "buy me a coffee" to the page. Gotcha. The expense to be fair isn't much. €20 a month to host the VM everything is running on. The personal time I put in is probably the biggest expense but at least I get some enjoyment from it when it goes well.

Never really thought about mobile, though it's a very valid point. Who uses a laptop/desktop anymore!? That'll be ways away though. Front end (obviously) is not my forte. I'll put it on the list.

Predicting when a unit would be available is a bit more complicated. You'd want to be getting average charge durations at different times of the day and applying that to the current status of the unit. That's machine learning territory. Won't be going that direction. I really just want somewhere that folks can digest their ability to use a charging site at different days and times. Baby steps

listermint

You need to connect with a data analysis grad probably somewhere like NCI for example. There's potential product in here with the mobile first being the approach. The machine learning part is alot less cumbersome than it sounds a decent project orientated grad could be quite interested in this.

Black_Knight

listermint wrote: »

You need to connect with a data analysis grad probably somewhere like NCI for example. There's potential product in here with the mobile first being the approach. The machine learning part is alot less cumbersome than it sounds a decent project orientated grad could be quite interested in this.

Probably right, just right now for me it's not worth the time and expense to do anything like that. I'll always look to add more to it, but right now it's about getting something basic out there.

Ideally i'd move all this to run on some cloud services, ideally as many free ones as possible, but it was a battle between getting something out soon vs ramping up on new tech.

Black_Knight

You can toggle between socket types now.

Calahonda52

Black_Knight wrote: »

You can toggle between socket types now.

Great, it might be good to include the link each time..:
easier than scrolling up to find it

Black_Knight

Calahonda52 wrote: »

Great, it might be good to include the link each time..:
easier than scrolling up to find it

You mean you've not bookmarked it? :pac:

I've asked mods to add http://test.ecars-stats.com/ to the original post on this thread.

Oh, and i've made the dropdown more searchable.

Up next is probably:
A site summary for the week/month - #of units, #of each socket, % of time not dead
Map site downtime this week/month
Handle removed units (backend work)
Handle slow chargers (fingers crossed my backend work/data gathering is correct)

And of course, what fallout comes from the esb map update tomorrow.

Black_Knight

A crucial milestone. :pac::pac::pac:

https://www.buymeacoffee.com/ecarsstats

Calahonda52

Thanks again
There is no data for Barlow Nissan , a dealer charger, just wonder why its there, Randles in TRALEE, another dealer is not listed which is what I expect

Black_Knight

Calahonda52 wrote: »

Thanks again
There is no data for Barlow Nissan , a dealer charger, just wonder why its there, Randles in TRALEE, another dealer is not listed which is what I expect

Hey. So I can probably handle that. Thanks for pointing out out.
The list of sites is, currently, just every site which ESB categories as FAST or ULTRA_FAST, regardless of whether there's usage data behind it. The lookup is just a bit more efficient this way. I'll probably leave them, but I need to handle scenarios where there's no data better:
Sites without a certain socket/plug
3rd party sites with no data


http://test.ecars-stats.com/


Edit: randles in Tralee works fine for me.

Calahonda52

Black_Knight wrote: »

Hey. So I can probably handle that. Thanks for pointing out out.
The list of sites is, currently, just every site which ESB categories as FAST or ULTRA_FAST, regardless of whether there's usage data behind it. The lookup is just a bit more efficient this way. I'll probably leave them, but I need to handle scenarios where there's no data better:
Sites without a certain socket/plug
3rd party sites with no data


http://test.ecars-stats.com/


Edit: randles in Tralee works fine for me.

Thanks again, at the risk of being a PITA, its Randles Killarney you have listed, which is an ESB rig on the garage site, whereas Randles Tralee is inside the garage compound and not a public CP.

Its a really interesting resource showing demand, sites like Blanch SC is astounding usage

ps Hutton and Meade ( nissan garage) looks like another blank as does IKEA Belfast (Brexit so no X border )

Black_Knight

Calahonda52 wrote: »

Thanks again, at the risk of being a PITA, its Randles Killarney you have listed, which is an ESB rig on the garage site, whereas Randles Tralee is inside the garage compound and not a public CP.

Its a really interesting resource showing demand, sites like Blanch SC is astounding usage

ps Hutton and Meade ( nissan garage) looks like another blank as does IKEA Belfast (Brexit so no X border )

Ah sorry. You're saying you expect Randles in Tralee to not be listed. I follow you now.

Yeah, ESB removed a bunch of Nissan garages from the app last year, but not all of them. Not sure why they're there in the first place.

Fix incoming....

Black_Knight

Black_Knight wrote: »

Fix incoming....

Fixed.

Now to get back to my real job. :eek:

innrain

Calahonda52 wrote: »

Thanks again, at the risk of being a PITA, its Randles Killarney you have listed, which is an ESB rig on the garage site, whereas Randles Tralee is inside the garage compound and not a public CP.

Its a really interesting resource showing demand, sites like Blanch SC is astounding usage

ps Hutton and Meade ( nissan garage) looks like another blank as does IKEA Belfast (Brexit so no X border )

H&M are not Nissan anymore. They are Hyundai. And the charger is disconnected. The 3rd party chargers from ecars is a legacy list which was not updated in at least 3 years. Lidl Drogheda is an example.

The one in Killarney is even badly located on the app sending people to a B&B. Fortunately Google Maps knows better.:)

Black_Knight

Copy pasta'd some css templating to make it less horrible looking.

http://test.ecars-stats.com/

Changes:
- The UI clearly
- Handled stations with no data for a specified plug.
- Backend work done to get unavailability of a site over the last 4 weeks, though unavailability over the last 1 week might be a bit more useful. I'll have a think on this.

Enjoy. Any feedback is always welcome.

innrain

If only businesses from around Stillorgan Hotel would capitalize on this data. This charger is in use at any hour of the day and any day of the week (almost). There are a lot of apartments in the area.
So above 50% is most likely in use and around 25% you got good chance of finding it free? Which is the highest number? I found Lucan Saturday morning @10:00 and evening @ 16:00 87.5% Can you do the color scale unique for all the chargers? That would help in comparing chargers. Right now you assign the darkest red to the peak value of the charger. What if instead you assign it for 100%. The best is to find the peak value of all chargers but that would require more code.

Black_Knight

innrain wrote: »

If only businesses from around Stillorgan Hotel would capitalize on this data. This charger is in use at any hour of the day and any day of the week (almost). There are a lot of apartments in the area.
So above 50% is most likely in use and around 25% you got good chance of finding it free? Which is the highest number? I found Lucan Saturday morning @10:00 and evening @ 16:00 87.5% Can you do the color scale unique for all the chargers? That would help in comparing chargers. Right now you assign the darkest red to the peak value of the charger. What if instead you assign it for 100%. The best is to find the peak value of all chargers but that would require more code.

It's on the long list. I've noticed that, had a very brief go at fixing it, but not got further. Tonight was spent enjoying a few beers with a friend (first time in a long time) so I'll try get to that soon.

Black_Knight

innrain wrote: »

Can you do the color scale unique for all the chargers? That would help in comparing chargers. Right now you assign the darkest red to the peak value of the charger. What if instead you assign it for 100%. The best is to find the peak value of all chargers but that would require more code.

Done.
- Darkest colour is now 100%, rather than busiest hour.
- Unit unavailability is now shown (for the last 7 days only).
- Found a "bug" though. The times are all UTC, so i'll have to think about fixing that properly, or hackily. :pac:

http://test.ecars-stats.com/

the_amazing_raisin

Do you have stats for the AC units hidden anywhere or is it just DC only? I'd like to see the usage of Plunkett train station in Waterford as I suspect only one plug is ever in use since one of the spaces is almost inaccessible (to my parking abilities at least)

Black_Knight

the_amazing_raisin wrote: »

Do you have stats for the AC units hidden anywhere or is it just DC only? I'd like to see the usage of Plunkett train station in Waterford as I suspect only one plug is ever in use since one of the spaces is almost inaccessible (to my parking abilities at least)

I'll open it up sure. Main reason I "hid" them was because the dropdown was massive, and not searchable. Obviously you'll have to select the untethered connector. Bad UX, but I can fix that in my own time.

Disclaimer - SCP are handled a little different. I'll have to double check my logic for them but my gut tells me they should be ok.

You should be able to search all chargers now.

You'll see some 12.5% occupancy on plunkett train station. That's basically of the last 4 weeks, on one of the Sundays, 1 socket was in use for the full hour. If both sockets were in use it should say 25%. If both sockets were in use every Sunday at those same hours it would say 100%. I'll double check the logic later to be sure.

Black_Knight

Getting close to something like going live.

Site is now on https://ecars-stats.com/ and i've reverted it back to just showing FAST chargers.

http://test.ecars-stats.com/ will be a bit more test like from now on. It has the slow chargers on it too, but I want to get a bit more confidence in the logic behind their data before I go releasing it.

innrain

There is a issue with park pointe. There are 2 in the drop down list and one displays Mayfield.
Frankfield has the highest number 91.7 that I found. So no chance of getting a charge Sunday afternoon there. It is good the airport ones are nearby-ish.
What color scheme are you using? Is it from this list?
https://plotly.com/python/builtin-colorscales/

Black_Knight

innrain wrote: »

There is a issue with park pointe. There are 2 in the drop down list and one displays Mayfield.
Frankfield has the highest number 91.7 that I found. So no chance of getting a charge Sunday afternoon there. It is good the airport ones are nearby-ish.
What color scheme are you using? Is it from this list?
https://plotly.com/python/builtin-colorscales/

Cheers, i'll have a look at that.

Colour scheme is probably something I could adjust. I put zero thought into it. I set it up, it looked decent, I left it.

Black_Knight

Looks like ESB ecars updated the site ID for park Pointe, and reused that ID for mayfield. Annoying... that's not how IDs work Ecars devs! I'd say it happens rarely, and I think my stats should be ok to handle it - stats are per unit, rolled up to give a usage stat for the site. Only bother is when esb relocate a unit, but that data would filter out of the stats I display over time.

I'll try fix up park pointe tonight, and any others (and futureproof for when this happens again).

Black_Knight

Fixed. Should be futureproofed too.

Found another bug though. Because I keep all the data, if ESB replace a unit, or change it's ID, I end up calculating like there's 2 units there, hence why park pointe has a lot of 50% unavailability. That should be 100%. Should be an easy enough fix... tonight probably...

Fixed.
https://ecars-stats.com/

rotoq

Excellent work. Do you have access to some ecars API or are you scraping their site?

Black_Knight

https://boards.ie/discussion/comment/117615502#Comment_117615502

Yes and no. Their site calls APIs to return the data. I'm just using those same APIs.

Black_Knight

New "features" added:

• Slow chargers added
• "Unavailability" heatmap is "realtime" up to the last hour. So if you're wondering "has that unit been working well today" you can now see.

Work in progress:

• Displaying "new units" added in the last week

Black_Knight

https://boards.ie/discussion/comment/117617197#Comment_117617197

Of course as soon as I say it I notice a bug. Should be fixed now, but i'm not going to retrofit the fix to the existing data. It'll fix itself over time now.

(The bug was what some units, mostly SCP, were not getting their data for both sockets stored correctly and thus a 100% offline unit would show as 50% offline).

Black_Knight

Feels like more than a month since I added something to the site, but anyways. I've added a page for new units added in the last 30 days. It's crudely done for now.

https://ecars-stats.com/new_units.html

30 days might be a bit long, so open to feedback. I'll see about adding the date added, or at least list them in order of latest to oldest install.

garo

https://www.boards.ie/discussion/comment/117770320#Comment_117770320

A but hard to read. Maybe make it into a table? And date added would be useful.