Very low forecasts for "tomorrow"

[yebaws]

Over the last week or so, I've noticed that the "tomorrow" forecasts have become unrealistically low.

For example, yesterday, the "tomorrow" forecast was 0.00. The actual production today was very low, but still 0.3Kwh, and the solcast "today" forecast was 1.3 Kwh. This is more in line with what I'd expect as we have a lot of hills around us, so winter production is always a fair bit lower than the forecast. For tomorrow, the "tomorrow" forecast is showing as 0.01Kwh.

Looking back at the sensor histories over the last week, all of the "tomorrow" forecasts have been under 0 (ie 0 point something), whereas all of the "today" forecasts have been between 1 & 2Kwh. It's not an integration issue as my solcast dashboard is showing the same figures. I haven't changed any settings in Solcast recently.

Interested to know if anyone else has had this? Maybe it's just a case of Solcast becoming a bit inaccurate in the "tomorrow" forecasts during the short winter days when solar radiation is very low, although I can't say I noticed this happening in December.

I poll Solcast at midnight, as that's when my automation needs to make a decision about battery charging. It's all fairly irrelevant at this time of the year as it's going to be a 100% charge every night anyway, but I just find it odd that I've never noticed these super low figures in previous winters.

[gcoan]

If the solcast website is giving the same as you are getting in the integration, then sounds more like a question for Solcast. Double check your configuration is correct on Solcast.

Just comparing my solar forecast tomorrow to the actual solar today for this month, mine is giving a pretty good correlation of tomorrow forecast to actual solar the following day (apart from a couple of days this week where the panels were covered in frozen snow and it took 2 days to thaw out!)

[BJReplay]

@yebaws thanks for starting a discussion rather than raising an issue.
@gcoan thanks as always for responding with facts from your side of the world.

@yebaws, unless you've turned on automated dampening, all the integration does is report what solcast is forecasting - no more, no less.

If you've turned on Automated dampening, then as noted in the readme Any rooftop site misconfiguration can have a significant impact on reported forecast, but that should be corrected in the rooftop site configuration. It is highly recommended to prove that the configuration is correct, and that forecasts are reasonably accurate on good generation days before attempting to configure automated dampening.

But as you and Geoff have already pointed out, it's not good generation season now.

At midnight or thereabouts, Solcast may have seen a lot of cloud 6-18 hours away, some of which has cleared the next day. That's hardly surprising - it's only a forecast. 300Wh is within a reasonable margin of error.

[gcoan]

My sequence was:

1. set forecast_history_max_days to 3650

in other words don't expire history. Why not, computing power is cheap

2. got repair notice about this setting having being deprecated

3. changed forecast_history_max_days to history_max_days, still set to 3650

4. Repair notice cleared, but two messages in the log so not sure what has been set or not:

2026-01-11 20:36:51.615 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option set history_max_days: 3650
2026-01-11 20:36:51.615 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option default set history_max_days: 730
2026-01-11 20:36:51.615 DEBUG (MainThread) [custom_components.solcast_solar.util] Removing advanced deprecation issue

[autoSteve]

Well buger me... What is set?

I am tipping the NON (edit) default, but... Yeah, ... what?

Good catch. That is a bug right there. Replicated.

The tell is the duplication again as I noted earlier. First go gets it right, second one buggers it up.

2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option proposed automated_dampening_model_days: 14
2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option proposed history_max_days: 3650
2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option set automated_dampening_adaptive_model_minimum_history_days: 7
2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option set automated_dampening_generation_history_load_days: 14
2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option set automated_dampening_adaptive_model_configuration: True
2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option set estimated_actuals_log_ape_percentiles: [10, 50]
2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option set estimated_actuals_log_mape_breakdown: True
2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option set history_max_days: 3650
2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option set reload_on_advanced_change: True
2026-01-14 00:44:09.552 DEBUG (MainThread) [custom_components.solcast_solar.solcastapi] Advanced option default set history_max_days: 730
2026-01-14 00:44:09.553 DEBUG (MainThread) [custom_components.solcast_solar.util] Removing advanced deprecation issue

[gcoan]

Well buger me... What is set?

I am tipping the NON (edit) default, but... Yeah, ... what?

that was what I wondered, what is set?

Good catch. That is a bug right there. Replicated.

Happy to contribute to your future fixes pipeline, can't have you with nothing to do 😭

[autoSteve]

For the record, I do believe that I have fixed this bug, @gcoan. Target v4.5.1. It was a slippery one.

[gcoan]

well done, have to admit I had forgotten about it !

Cheers

[yebaws]

Auto-dampening modelling of base factors pays no attention to the forecast, because that can be very wrong. It pays attention to past estimated actuals from Solcast, which tend to be way less wrong.

Ahh...that makes a lot of sense

A Home Assistant integration for a distance measurement sensor perched above one panel might work, or might not. Something cobbled together using an ESP32?

Sounds like one of those things that we do for the challenge or trying to get it to work rather than any great effect it will have improving our lives - I'm all for it!