Logo do repositório
 
Publicação

End-to-end decomposition-aware neural architecture search with sparse expert transformers for thermal energy forecasting

2026Artigo de revisão Acesso aberto
Ver registo simples
datacite.subject.fosEngenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática
dc.contributor.authorSeman, Laio Oriel
dc.contributor.authorStefenon, Stefano Frizzo
dc.contributor.authorMatos-Carvalho, João Pedro
dc.contributor.authorNied, Ademir
dc.contributor.authorYow, Kin-Choong
dc.contributor.authorStefenon, Stefano Frizzo
dc.contributor.editorIEEE
dc.date.accessioned2026-09-23T15:04:19Z
dc.date.available2026-09-23T15:04:19Z
dc.date.issued2026
dc.description.abstractThis study presents a unified framework for thermal energy forecasting that integrates signal decomposition, Neural Architecture Search (NAS), and sparse transformer modeling. The method employs a hyperparameter-optimized hierarchical Variational Mode Decomposition (VMD) to enhance signal representation through adaptive noise suppression and mode separation. A differentiable preprocessing module, referred to as HeadComposer, performs joint search over normalization, decomposition, and convolutional transformations within a bilevel optimization procedure. The forecasting backbone consists of a transformer architecture incorporating gated residual connections for adaptive token selection and Mixture-of-Experts (MoE) feedforward layers for conditional computation. Experiments are conducted on real-world thermal power generation data comprising approximately 10,000 test samples, using an input look-back window of 96 time steps and a multi-step prediction horizon of 24 steps ahead. Results demonstrate consistent improvements over state-of-the-art baselines, achieving a mean absolute percentage error of 5.94%, corresponding to reductions of approximately 12-15% relative to competing methods. The proposed design demonstrates the effectiveness of combining decomposition-based preprocessing (VMD and HeadComposer) with NAS-driven architecture adaptation and sparse expert routing (based on MoE) for modeling non-stationary thermal generation processes.eng
dc.identifier.citationSeman, L. O., Stefenon, S. F., Matos-Carvalho, J. P., Nied, A., & Yow, K.-C. (2026). End-to-End Decomposition-Aware Neural Architecture Search With Sparse Expert Transformers for Thermal Energy Forecasting. IEEE Access, 14, 67765–67787. https://doi.org/10.1109/ACCESS.2026.3689436
dc.identifier.doi10.1109/ACCESS.2026.3689436
dc.identifier.eissn2169-3536
dc.identifier.urihttp://hdl.handle.net/10400.21/23183
dc.language.isoeng
dc.peerreviewedyes
dc.publisherIEEE
dc.relation.hasversionhttps://ieeexplore.ieee.org/document/11501672
dc.rights.urihttp://creativecommons.org/licenses/by/4.0/
dc.subjectMixture-of-experts
dc.subjectNeural architecture search
dc.subjectVariational mode decomposition
dc.titleEnd-to-end decomposition-aware neural architecture search with sparse expert transformers for thermal energy forecastingeng
dc.typereview article
dspace.entity.typePublication
oaire.citation.endPage67787
oaire.citation.startPage67765
oaire.citation.titleIEEE Access
oaire.citation.volume14
oaire.versionhttp://purl.org/coar/version/c_970fb48d4fbd8a85
person.affiliation.nameInstituto Politécnico de Lisboa, Instituto Superior de Engenharia de Lisboa
person.familyNameStefenon
person.givenNameStefano Frizzo
person.identifier.orcidhttps://orcid.org/0000-0002-3723-616X
relation.isAuthorOfPublicationb0383f14-261a-42cf-a223-7c571c84a241
relation.isAuthorOfPublication.latestForDiscoveryb0383f14-261a-42cf-a223-7c571c84a241

Ficheiros

Principais
A mostrar 1 - 1 de 1
Miniatura indisponível
Nome:
End-to-End_Decomposition-Aware_Neural_Architecture_Search_With_Sparse_Expert_Transformers_for_Thermal_Energy_Forecasting.pdf
Tamanho:
3 MB
Formato:
Adobe Portable Document Format
Ver/Abrir
Licença
A mostrar 1 - 1 de 1
Miniatura indisponível
Nome:
license.txt
Tamanho:
4.03 KB
Formato:
Item-specific license agreed upon to submission
Descrição:
Ver/Abrir

Coleções

ISEL - Eng. Elect. Tel. Comp. - Artigos