Weather based battery charge compensation

Hi,

I have my battery set to charge to 80% overnight on my octopus go tariff. If the weather the next day is good then my battery will hit 100% charge from the solar panels and then start exporting to the grid.

Would it be possible to add a setting to the battery charge amount for weather compensation, such that if its to be a cloudy day tomorrow, I can have my battery charge an extra 5-15% overnight and if its going to be sunny it would charge ₁₀% less overnight. The dashboard as a weather card in it so the information looks to be available.

Thanks

Paul

I believe this is in development, sure I read somewhere they want to build this. You might need to link your Solcast account to get the benefits of this if you haven't already.

Yes there's a "coming soon" setting in the system mode area of the app for "solar forecast automation" which I will suspect will do just what PaulC posted.

A few of us have written our own software to do this. Mine runs on a Raspberry Pi and pulls the Solcast forecast. Seems pretty reliable for the most part (my overnight charge is down at about 30% at present). www.github.com/salewis38/palm

and another similar to @stevelewis can be found here https://github.com/lockedj/solarbatteryoptimize

Yorkie71 For it to be effective, it'll need more than a tick box to enable it. The Solcast forecasts are a best guess (you can actually download the P10, P50 and P90 forecasts, a weighted average seems to work best). Then there's how consumption is forecast and how much margin to include to be on the safe side.

stevelewis I agree. It would be a good opportunity to factor in some machine learning based on historical consumption data. After all, GE do have all that do hand. If I think about how I adjust the charging, I'll glance at the forecast and adjust the battery level if needed, but I'll take into account my previous estimations, (normally not accurate but I'm learning) and also factor in that it will be a high drain day; Wednesday is normally wash day for example. The feature needs to replicate that process. As you say, solar forecast alone, while better the nothing, could be better.

Yorkie71 interesting to see the parallel thread on FB at present. I've been running my own code to do just this for a couple of months with some success. Making it work generically across multiple scenarios, such as different tariffs, array/battery size ratios, usage habits, etc would be the obvious next stage.

stevelewis Yes I saw your thread on FB. Maybe you should license your code to GE! 🤣 Or share some ideas. Some solar AI automation would be really helpful.

Yorkie71 both stevelewis and my code is opensource, so free for others to play with - not sure there is a business there but its been fun building it and if others can use it.... I've been running mine now for over a year and been very happy with it. It is dependant on the forecast - I use openweathermap api and look at cloud cover rather than solcast. There have been a few days where its been way out but for the vast majority has been good. I keep a watch on the solcast forecast and suspect it will be no better and may be worse based on eye balling its forecast for the next day - but no empirical evidence.

locked @Yorkie71 Good fun and I may save £50 a year but I won't be retiring on my code either 😄

New version of mine this evening on github is skinnied down to just doing the SoC calcs. Curious about how well it works for others or whether I've developed a point solution... location, ratio of PV/battery and daily usage patterns. Then there are the permutations on export, peak and off-peak rates.

Not sure which is better out of Solcast and openweathermap, a weighted average of the 10% and 50% Solcast forecasts at midnight seems to cover most changeable weather conditions for the next 12-18 hours.

I'd be intrigued to know how successful these solutions are proving to be. I was thinking about it a few months ago but then shelved it because it dawned on me that the marginal costs of over-charging were so much lower than the marginal costs of under-charging.

This was in my specific circumstances, perhaps: overnight I had 5p/kWh and an export rate of 4p/kWh. Therefore overcharging by a kWh would only waste a penny (maybe about 1.2p including system losses?). If the automation under-charged overnight, however, I'd be losing about 13p/kWh.

I concluded it would be better to just set the overnight charge %ge on the high side rather than take the risk.

As of today my rather nice Octopus Go rate has expired and I'm on 7.5p overnight and 40p during the day, still with 4p export, so I'm reconsidering. While the numbers on the overcharge side have increased, that 40p daytime rate means the undercharge risk has increased rather more. However, I'm willing to give it a go if you guys are finding it works successfully. Presumably there is - or it would be easy to add - a bit of margin to habitually overcharge by?

Thanks.

very happy with my home automation that optimises how much to top the battery up overnight at cheap rate. It takes into account a number of factors including weather forecast (cloud cover for each hour, others used solar forecast from solcast), expected home energy usage and decides how much to top up and does so on the conservative side to minimise any expensive import. It has worked really well for the last year, full details including code can be found here :- https://github.com/lockedj/solarbatteryoptimize The only time import at expensive rate is when the load demand is more than solar + battery provide (limited by the inverter)

locked

garethhann Very happy here too. Solcast provides 3 estimates with 10/50/90% probabilities and I take a weighted average of the 10% and 50% estimates to be on the safe side. I also have a cobfigurable seasonal bottom limit. I've had one day where the weather forecast changed wildly overnight, otherwise it's been working solidly and getting close to 100% SoC most days.