-
-
Notifications
You must be signed in to change notification settings - Fork 0
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Aerosol Data And Analog Ensembles #2
Comments
This paper has a simple MLP improve RMSE from 38.24 to 37.78 when incorporating CAMS forecast, and down to 32.75 when also including AERONET and an angstrom(?) coefficient to help correct the CAMS forecast. Not entirely sure what the angstrom coefficient is, but still an improvement, although fairly small from just CAMS. These forecasts were done over the Middle East, where there is a lot of dust, and fairly clear skies in general. |
This is the AERONET that is mentioned https://aeronet.gsfc.nasa.gov/new_web/download_all_v3_aod.html Updated weekly it seems? Not sure if useful for our forecasting |
This project actually seems quite relevant to what we are doing, they built an end-to-end forecasting system for operators in the US. Analog Ensemble gave an improvement of ~17% in RMSE for their power forecasts for 0-72 hours. |
@dantravers @JackKelly Still going through and finding things, but some interesting results in the literature so far. Aerosols seem to improve forecasts, and the Sun4Cast project linked above had a lot of moving parts, but interesting results, seems to suggest Analog ensembles improved deterministic and probabilistic forecasts. |
This one suggests when including aerosol index, the MAPE decreased by halfish for both sunny and cloudy days |
This section of the Solar Energy Forecasting book is relevant on aerosols... (if you want to buy your own copy then pls expense it to ocf!) And this paper is relevant https://pubs.rsc.org/en/content/articlelanding/2018/ee/c8ee01100a |
Great. Let’s pick it up in the next ML meeting. Thanks for the info, ***@***.***> @***@***.***!
From: Jack Kelly ***@***.***>
Sent: Tuesday, April 18, 2023 6:12 PM
To: openclimatefix/pv-pseudo-experiments ***@***.***>
Cc: dantravers ***@***.***>; Mention ***@***.***>
Subject: Re: [openclimatefix/pv-pseudo-experiments] Aerosol Data And Analog Ensembles (Issue #2)
This section of the Solar Energy Forecasting book is relevant on aerosols...
<https://user-images.githubusercontent.com/460756/232853117-31417edb-2673-4700-85b5-4f06a5e3a115.jpg>
<https://user-images.githubusercontent.com/460756/232853118-70ad1085-3da9-4488-9b1b-6b8d6bff84fe.jpg>
<https://user-images.githubusercontent.com/460756/232853119-53fe9462-ff80-4ad9-b459-9107266bacbe.jpg>
<https://user-images.githubusercontent.com/460756/232853120-133ce277-e94b-4617-8bf2-726bd29d36f3.jpg>
—
Reply to this email directly, view it on GitHub <#2 (comment)> , or unsubscribe <https://github.com/notifications/unsubscribe-auth/AJGGYOZMHXOTJK4ST5OHLUDXB3DOBANCNFSM6AAAAAAXCXMVQ4> .
You are receiving this because you were mentioned.Message ID: ***@***.***>
|
We want to determine if it is worth including aerosol data in our experiments and forecasts. This issue is just recording some of the literature and ideas about how to use it.
Arxiv/Blog/Paper Link
https://www.researchgate.net/profile/Miguel-Prosper/publication/360978841_Development_of_a_solar_energy_forecasting_system_for_two_real_solar_plants_based_on_WRF_Solar_with_aerosol_input_and_a_solar_plant_model/links/6297124fc660ab61f85695e7/Development-of-a-solar-energy-forecasting-system-for-two-real-solar-plants-based-on-WRF-Solar-with-aerosol-input-and-a-solar-plant-model.pdf
Detailed Description
This paper uses ECMWF CAMS aerosol forecast over India and southern Spain to determine the effects of using it with GFS forecast data. It has a ablation study on how useful the aerosols were, and found a ~10.9% MAE improvement vs just GFS solar information, and had more realistic split between diffuse and direct solar radiation on days with high aerosol forecasts, as well as a 9% NMAE improvement on forecasting temperature. Both underestimated the amount of clouds (which hopefully including satellite cloud imagery would help fix.)
Context
This paper seems like a fairly large improvement including the aerosol forecast.
The text was updated successfully, but these errors were encountered: