Probabilistic end-to-end irradiance forecasting through pre-trained deep learning models using all-sky-images
<p>This work proposes a novel approach for probabilistic end-to-end all-sky imager-based nowcasting with horizons of up to 30 min using an ImageNet pre-trained deep neural network. The method involves a two-stage approach. First, a backbone model is trained to estimate the irradiance from all-...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2024-01-01
|
| Series: | Advances in Science and Research |
| Online Access: | https://asr.copernicus.org/articles/20/129/2024/asr-20-129-2024.pdf |
| _version_ | 1827392533523595264 |
|---|---|
| author | S. Chaaraoui S. Houben S. Meilinger |
| author_facet | S. Chaaraoui S. Houben S. Meilinger |
| author_sort | S. Chaaraoui |
| collection | DOAJ |
| description | <p>This work proposes a novel approach for probabilistic end-to-end all-sky imager-based nowcasting with horizons of up to 30 min using an ImageNet pre-trained deep neural network. The method involves a two-stage approach. First, a backbone model is trained to estimate the irradiance from all-sky imager (ASI) images. The model is then extended and retrained on image and parameter sequences for forecasting. An open access data set is used for training and evaluation. We investigated the impact of simultaneously considering global horizontal (GHI), direct normal (DNI), and diffuse horizontal irradiance (DHI) on training time and forecast performance as well as the effect of adding parameters describing the irradiance variability proposed in the literature. The backbone model estimates current GHI with an RMSE and MAE of 58.06 and 29.33 W m<span class="inline-formula"><sup>−2</sup></span>, respectively. When extended for forecasting, the model achieves an overall positive skill score reaching 18.6 % compared to a smart persistence forecast. Minor modifications to the deterministic backbone and forecasting models enables the architecture to output an asymmetrical probability distribution and reduces training time while leading to similar errors for the backbone models. Investigating the impact of variability parameters shows that they reduce training time but have no significant impact on the GHI forecasting performance for both deterministic and probabilistic forecasting while simultaneously forecasting GHI, DNI, and DHI reduces the forecast performance.</p> |
| first_indexed | 2024-03-08T17:33:17Z |
| format | Article |
| id | doaj.art-05abbf9a467d4e93a46eaff6326467a8 |
| institution | Directory Open Access Journal |
| issn | 1992-0628 1992-0636 |
| language | English |
| last_indexed | 2024-03-08T17:33:17Z |
| publishDate | 2024-01-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | Advances in Science and Research |
| spelling | doaj.art-05abbf9a467d4e93a46eaff6326467a82024-01-02T13:25:13ZengCopernicus PublicationsAdvances in Science and Research1992-06281992-06362024-01-012012915810.5194/asr-20-129-2024Probabilistic end-to-end irradiance forecasting through pre-trained deep learning models using all-sky-imagesS. Chaaraoui0S. Houben1S. Meilinger2International Centre for Sustainable Development (IZNE), University of Applied Sciences Bonn-Rhein-Sieg, 53757 Sankt Augustin, GermanyInstitute for Artificial Intelligence and Autonomous Systems (A2S), University of Applied Sciences Bonn-Rhein-Sieg, 53757 Sankt Augustin, GermanyInternational Centre for Sustainable Development (IZNE), University of Applied Sciences Bonn-Rhein-Sieg, 53757 Sankt Augustin, Germany<p>This work proposes a novel approach for probabilistic end-to-end all-sky imager-based nowcasting with horizons of up to 30 min using an ImageNet pre-trained deep neural network. The method involves a two-stage approach. First, a backbone model is trained to estimate the irradiance from all-sky imager (ASI) images. The model is then extended and retrained on image and parameter sequences for forecasting. An open access data set is used for training and evaluation. We investigated the impact of simultaneously considering global horizontal (GHI), direct normal (DNI), and diffuse horizontal irradiance (DHI) on training time and forecast performance as well as the effect of adding parameters describing the irradiance variability proposed in the literature. The backbone model estimates current GHI with an RMSE and MAE of 58.06 and 29.33 W m<span class="inline-formula"><sup>−2</sup></span>, respectively. When extended for forecasting, the model achieves an overall positive skill score reaching 18.6 % compared to a smart persistence forecast. Minor modifications to the deterministic backbone and forecasting models enables the architecture to output an asymmetrical probability distribution and reduces training time while leading to similar errors for the backbone models. Investigating the impact of variability parameters shows that they reduce training time but have no significant impact on the GHI forecasting performance for both deterministic and probabilistic forecasting while simultaneously forecasting GHI, DNI, and DHI reduces the forecast performance.</p>https://asr.copernicus.org/articles/20/129/2024/asr-20-129-2024.pdf |
| spellingShingle | S. Chaaraoui S. Houben S. Meilinger Probabilistic end-to-end irradiance forecasting through pre-trained deep learning models using all-sky-images Advances in Science and Research |
| title | Probabilistic end-to-end irradiance forecasting through pre-trained deep learning models using all-sky-images |
| title_full | Probabilistic end-to-end irradiance forecasting through pre-trained deep learning models using all-sky-images |
| title_fullStr | Probabilistic end-to-end irradiance forecasting through pre-trained deep learning models using all-sky-images |
| title_full_unstemmed | Probabilistic end-to-end irradiance forecasting through pre-trained deep learning models using all-sky-images |
| title_short | Probabilistic end-to-end irradiance forecasting through pre-trained deep learning models using all-sky-images |
| title_sort | probabilistic end to end irradiance forecasting through pre trained deep learning models using all sky images |
| url | https://asr.copernicus.org/articles/20/129/2024/asr-20-129-2024.pdf |
| work_keys_str_mv | AT schaaraoui probabilisticendtoendirradianceforecastingthroughpretraineddeeplearningmodelsusingallskyimages AT shouben probabilisticendtoendirradianceforecastingthroughpretraineddeeplearningmodelsusingallskyimages AT smeilinger probabilisticendtoendirradianceforecastingthroughpretraineddeeplearningmodelsusingallskyimages |