I'm testing out timed discharges as I'm on Octopus Flux. Ideally I would have liked to do an export this evening, but it didn't happen because the solar forecast was too low. WonderWatt calculated it as 8.1kwh for tomorrow, which is very low. Solcast is forecasting something in the region of 33 kwh, which is closer to what I would expect.

@"godfreym29"#p12 Hi, it is an estimate that relies on the cloud cover factor, among a few other things, you can see in the FAQ: https://community.wonderwatt.com/d/5-wonder-watt-faq It does look like cloud cover is predicted to be 88% for your location. That basically has a big impact. How has it been on previous days? Our aim also is to introduce the personal correction factor which you can configure if the estimate is always over or under.

Same question from me in that I had to go back yesterday to manual GivEnergy timed export for flux as wonderwall had solar forecast at 3.2 whereas solacast had it above 10. I know that weather prediction is changeable so it might be the different in the range of solar predicted in the following day, but in truth in 80% of the days, I generated greater than even higher solacast prediction so even their prediction can be low.

@"Dai"#p15 Thanks Dai - so a correction factor/multiplier of 1.4 or such might work well for you?

@"admin"#p17 it might so long as it was allowable to be big enough, even solar install is different in location and efficiency of the panels, but it’s a balance of too many customisations available to user (increasing the complexity) and setting the base algorithms the best they can be. The figure used in my prediction yesterday which caused the automation not to fire was 3.2kw for the day, I passed that by 10am this morning and I am south west pointing roof which therefore would normally peak at around 2pm. I would try it again tonight but being a special day being the last really high export rate price for flux, aim to manually dump everything tonight so haven’t set the trigger to fire. Will try again the start of next week but maybe it would be good to see somewhere in the app in the short term what the prediction is for the following day and allowing the user to make a judgement in advance rather than relying on looking at the logs to see the failure reason after the event which in many occasions be too late?

@"admin"#p14 Hi, so far today I've generated just over 2 5kWh and may get another 2 or 3 before it stops for the day. Quite different from the 8 that WW forecast. I did a manual Export last night and managed to refill the battery today and still exported 13 kWh

@"godfreym29"#p19 Yep, it's a rough estimate, and largely based on cloud coverage. You could of course always set the threshold lower. Great suggestion in terms of seeing a 'current prediction' for the next day. Though we probably need to keep the number of real-time weather lookups limited for those purposes to a set number of times per day, as there is a cost involved. We'll put it on the ever-growing to-do list! Thanks!

I've set my threshold to 1kWh for the forecast (bit extreme I know!) but has been working well for me for the last few days.

Like @"MartynK"#5 set an extreme low solar threshold yesterday to ensure it fired and it worked fine. @"admin"#1 mentioned the additional cost to them of providing a solar forecast in advance each day to a user and we are beta testing phase so it’s a balance between cost and speed for any software changes Maybe a quicker alteration for smart export is via the UI allowing the user to select to export based on an existing SoC level at the time of peak period OR the SoC & solar forecast. Every time the smart export fires, the solar forecast is provided in the activity log, then over a period of time user will build up understand over a period of time what the “right” solar forecast is for them and from developer cost perspective you don’t incur the extra cost of multiple external calls at this time. At the moment, if others are like me because it’s new we are at this point of time learning the best way to use the software and maybe after a couple of weeks of activity logs we would get to know the right solar forecast thresholds for our site and usage patterns.

@"Dai"#p25 Yeah, by setting the threshold as low as 1kWh - you effectively achieve exactly that, it basically only takes into account the SoC, as the kWh estimate is always above 1. Like @"MartynK"#5 's solution really.

@"admin"#p30 yep, realised similar as soon as I hit the send key (it was early, maybe too early, on a Sunday morning😉), will run it at solar prediction of 1 for the week and a high battery percentage report back as my battery this time of the year can hold 2 days of home consumption so I have a fallback

How is the solar forecast calculated?

Acejames87

Is some sort of factor being added mine is underestimating by about 1.8x. Only been running a couple of days but would like to actually use forecast if it would become a bit more accurate

JMA

Acejames87 @admin

I been reading all the posts related to the forecast and I also feel there's something off in the calculation of the prediction. Mine is also 1.8 to even 2.0 times off from the prediction on WW. I've two arrays, a 4.2kWp (SE facing, 35 deg) and a 2.1kWp (NW facing 35 deg) array. They produced on Oct 13 a total of 17.42 kW. At this time of year when applying a strong correction to the NW facing one, an equivalent single array would be 5.04 total kWp if they were all SE facing. So basically with a 5.04 kWp 35 deg slope, SE facing equivalent I produced 17.42 kW. If I put that "equivalent" in the WW setttings then WW's prediction is less than half of what I really get.
@admin : If you can share your calcs/approach I'm happy to provide you with real life data to allow calibration.

admin

Hi all - we're actively looking into this. Forecast during summer was on the whole fairly accurate.

The calc does take into account sun's position too, during winter time, but it looks like it's currently underestimating by a good margin.

Will keep you posted.

admin

@SilverArt @Dai @JMA @Acejames87 @MartynK

A fix for the recent solar forecast underestimation will be rolled out shortly today.

See: https://community.wonderwatt.com/d/33-solar-calculation-fix-to-be-rolled-out-16-october

felmer

My Solar (5.9kWh)/Battery (11.5kWh) has been in since end of Nov and just looking to move to a flux or agile octopus tariff.
Been taking a look at Wonder Watt and its solar prediction for charging overnight, but its way above what I'm generating.
Roof pitch 35 deg, correct lat/long set, 40 deg west of due south (so set for south-west). 100% charge rate and target. Located in Surrey.
Examples:
4th Jan - WW predicted 5.2kWh, actual 1.2kWh
5th Jan - WW predicted 4.2kWh, actual 3kWh
6th Jan - WW predicted 5.4kWh, actual 0.7kWh
Current generating hours are ₀₈:30-14:30. At the mo high cloud and generating 0.70kWh up to 1.9kWh when the sun breaks through. It's been cloudy every day for last few weeks, but on 22nd Dec when there was last lots of sunshine I generated 8.5kWh, so think my system is working Ok (?).
In which case why WW is over-estimating by so much, and best way to adjust for it to make sure battery gets appropriately charged at night?

admin

felmer Yeah, it's a fair question, and we've been tweaking the forecast a few times, especially when going into winter.

However, it's been pretty good for spring/summer, so we're also quite conscious to not mess with it willy-nilly, and make it OK for winter, but potentially screw it up for spring and summer sun.

We'll run some unit tests though and see if we can adjust the penalty for sun angle during winter months some more.

Cheers!

SilverArt

It really does vary dependant on the time of year and. IMHO at the moment it is probably worth ignoring the forecasts and just to as you are doing, charge to 100%.

In the next couple of months that may change due to the weather so just keep your eye on it and possibly adjust the kWh setting in WW for the panels to see if you can tweak that to get forecast a little closer to reality but it really does vary day to day.

admin

But generally, in those low forecasts like 5.2 prediction, actual 1.2 etc. let's not forget it's weather based, mainly cloud coverage.

Your decision based cut-offs shouldn't be on granular levels like that IMO. It should be like 'greater than 10kwH expected' do X. Not, if over 4.7kWh yield do X.

Main point, we know some of our long winter months in the northern hemisphere suck. I'd be happy with 5kWh each day in December.

SilverArt

admin Your decision based cut-offs shouldn't be on granular levels like that IMO. It should be like 'greater than 10kwH expected' do X. Not, if over 4.7kWh yield do X.

Agreed, I was using low <7kWh, medium = 7 - 15kWh and high > 15kWh to select which historic usage data to use for my daily forecasting (which I needed before you got the Giv. charge rate sorted.). It was accurate enough to enable me to manually manage charge rate effectively, in fact the 'high' setting, served no useful purpose. It would certainly be a little dependant on actual installation and usage patterns.

psiloc

Hi there,

Love the application, it's exactly what I've been looking for.

Just wanted to echo the last few reports of the solar being over estimated. Last night the solar forecast was 4.8kWh, but I'm only on track to achieve about 1kWh. It did, however, correctly decide to charge the battery to 100% which is ultimately all that matters.

Just throwing it out there, but what are the chances of the forecast algorithm using actual generation data from the inverter vs. what was forecast to determine a bias over time for my account and apply it to the forecast?

admin

psiloc It's quite interesting actually, today in sunny Wales(!!!), WW forecasted 18.9kWh and we're currently on 17.3kWh and the day isn't over yet...

It's tricky really as it is based on weather forecast/cloud cover. Historic or actual generation would likely make little difference as the weather changes compared to the forecast just differ too much most of the time.

psiloc

admin No worries! Thanks for the reply.

You'll have heard this lots of times already I'm sure but if you could implement dynamic schedules that use the Octopus API for Agile and Intelligent you really would have a killer app on your hands here.

admin

psiloc Thanks for the kind words, much appreciated.

We're already testing Agile Octopus dynamic 30 min charge slots out, see the thread here:

https://community.wonderwatt.com/d/63-agile-octopus-integration/6

Not sure on Intelligent tariffs (not actually on Octopus ourselves) - are you talking about the dynamic cheap slots Octopus send out now and then? We would need to understand whether there's an API for that, I suppose there would be, tho might need Octopus API token.

psiloc

admin Yes exactly.

Similar to Agile, your users would want their battery to know about the smart charging slot to either charge the home battery or at least, prevent it from discharging into the EV. I suspect it's probably slightly simpler to implement than Agile to be honest because you won't have to worry about the half hourly unit price, just the fact it's off-peak and when.

I don't know if it's useful to you but this home assistant plugin claims to be able to get this data from the Octopus API, if there might be some code you could pinch:

https://github.com/megakid/ha_octopus_intelligent

admin

psiloc That's great, thanks for the pointer!

psiloc

Not expecting a reply, but just wanted to report that my solar prediction for today is 5.6kWh and there is not a cat in hells chance I will generate anything close to that :) we're forecast almost total cloud cover all day.

I know you're very busy (and working for free) so don't think I'm trying to put pressure on, but I just wanted to say that if you have access to my account/location/setup then feel free to use it for analysis.

admin

psiloc Hi - your predicted cloud cover for today was 72%, which (amongst other things like your PV setup) resulted in that specific kWh prediction. Weather forecasting is unfortunately not always reliable. 😔

psiloc

Can I ask what the solar forecast would have been for me on the 9th? If you can go backwards? That was our last sunny ish day and I generated 8.7kWh

admin

psiloc If you can find out what your cloud coverage percentage was that day, we can calculate it for you.

At 50% cloud coverage, it would have predicted: 7.6kWh
At 45% cloud coverage: 8.0kWh
At 40% cloud coverage: 8.4kWh
At 30% cloud coverage: 9.1kWh

So if you had a fairly sunny day, that seems a fairly reasonable estimation.

psiloc

Thanks! About 20% cloud cover average on that day from the look of some random website I just found. So yeah pretty close! A little on the high side again though so I'm starting to think I just lower my solar capacity setting to correct it.

At this point I'm just asking stuff because I'm enjoying thinking about it to be honest, so don't mind me if you don't want to, but have you considered putting a manual forecast correction slider somewhere in the settings page? Like in my case now rather than fudging my other settings I'd just correct it to -1.5kWh or something.

This could also be how people who know they get shading on their panels could deal with that.

admin

psiloc It's all good! :-)

Yeah, we have a correction factor in the backend for each user, but currently not surfacing it.

The truth is, if you were to tweak it from day-to-day based on the previous day, it would be rather counter productive, as the main point is for the automation to do the heavy lifting and for the user to be pretty much hands-off as much as possible.

« Previous Page Next Page »