Currently running a test on both side by side.
On the left - Forecast.Solar
On the right - Solcast
Both graphs are for yesterday, today, and tomorrow, both are sized to the same scale, both use the same system settings (although forecast.solar is using an applied horizon and I have to apply my own shading adjustment to Solcast). Both data sets come from my own API calls, forecast solar is run every hour (with hourly data points), Solcast I run (two planes) once for back estimates and then four times during the day for forecasts (with half-hourly data points).
It is high summer here in the UK, and for the past couple of weeks I am seeing the actual generated each day at about 90% of the forecast for Solcast.
I have been running with the forecast.solar integration for over two years, and have analysed the forecast to actual match for the past 12 months.
If anyone is interested, this is my summary analysis, for the year 21 June 2023 to 21 June 2024.
Comparing actual generated energy with the daily forecast (forecast value captured before sunrise as an average during 01:00 to 05:00)
Now plotting the ratio of actual/forecast, with ‘tolerance’ bands of 75% and 125%. As you can see, forecast over-estimates during the winter (I get about half of the forecast) and under-estimates during the summer (I get 1.5 to 2 times the forecast)
And finally, to answer the question “how much more actual do I get each day above the forecast?”
‘0’ is forecast is spot on. In practical terms, to be able to realistically use a solar forecast for home energy control, I was hoping for something within +/- 5 kWh consistently across the year, certainly in the summer when my peak production is close to 25 kWh on a good day, and perhaps +/- 2 kWh in the winter when my peak production is only 3-4 kWh per day.
A fascinating subject, hours and hours of work over years, and I am still no closer to getting a forecast figure I feel confident in using!
Edit.
I really should add that I have two planes, facing East-North-East and West-South-West, and on a single storey garage. Winter horizon shading is an issue, so I accept a reduction of 1 to 2 kWh during the morning between mid October and mid February on a typical 5-8 kWh day due to the low sun path, although I also get some compensation from light reflected back from the upper-storey house wall. All in all, difficult to set up and compensate for.