aiqbench analysis
This study evaluates the trade-offs between forecast accuracy and computational cost when building heterogeneous ensembles for photovoltaic power prediction. The research demonstrates that static fusion methods can outperform boosting in specific datasets but highlights significant redundancy among model components, suggesting a need for stricter component-wise evaluation to optimize inference speed. These findings are relevant to AI practitioners optimizing machine learning pipelines for renewable energy applications where efficiency is critical.
Ensemble Complexity in Photovoltaic Forecasting
arXiv:2609.15049v1 Announce Type: cross Abstract: An ensemble can improve photovoltaic forecasts while adding components that contribute little or increase computation. We assess these effects through matched comparisons and ablations of a fixed heterogeneous predictor bank. Hourly experiments use GEFCom2014 and three additional public datasets, with chronological partitions and three seeds. Under retrospective ERA5 assistance, static fusion reduces scaled mean absolute error against matched boosting by 1.11%, 4.41%, and 1.63% on PVDAQ, OPSD, and Ausgrid; only OPSD remains supported after multiple-comparison correction. Weather gating offers no consistent incremental benefit. Exploratory member removals show group-level dependence alongside individual redundancy. A separate, previously inspected fifteen-minute case replaces one neural member with a tree predictor: normalized error falls by 1.72%, but measured inference is slower. These findings support component-wise evaluation with explicit limits on weather availability and test-set reuse.
Abstract:An ensemble can improve photovoltaic forecasts while adding components that contribute little or increase computation. We assess these effects through matched comparisons and ablations of a fixed heterogeneous predictor bank. Hourly experiments use GEFCom2014 and three additional public datasets, with chronological partitions and three seeds. Under retrospective ERA5 assistance, static fusion reduces scaled mean absolute error against matched boosting by 1.11%, 4.41%, and 1.63% on PVDAQ, OPSD, and Ausgrid; only OPSD remains supported after multiple-comparison correction. Weather gating offers no consistent incremental benefit. Exploratory member removals show group-level dependence alongside individual redundancy. A separate, previously inspected fifteen-minute case replaces one neural member with a tree predictor: normalized error falls by 1.72%, but measured inference is slower. These findings support component-wise evaluation with explicit limits on weather availability and test-set reuse.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2609.15049 [cs.LG]
(or arXiv:2609.15049v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.15049 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 227 words
Related coverage
ORBIT-2 based Weather and Climate Downscaling and Downscaled Global Forecasts on AMD Instinct
AMD · Jun 8, 2026
Horizon-specific Expert Fusion for Photovoltaic Power Forecasting
arXiv · Sep 16, 2026
Global Attention with Linear Complexity for Exascale Generative Data Assimilation in Earth System Prediction
arXiv · Sep 2, 2026
LD-EnFF: Latent-Dynamics Ensemble Flow Filtering for Data Assimilation with Sparse Observations
arXiv · Oct 6, 2026