Contextual retrieval-augmented tree ensemble Boosting for Multi-Horizon Solar PV Power Forecasting

Saddam Hussain

doi:10.1016/j.egyai.2026.100855

Contextual retrieval-augmented tree ensemble Boosting for Multi-Horizon Solar PV Power Forecasting

2026-07-25

0

OA

AI

S

Saddam Hussain *

A

Amjad Ali

S

Sikandar Abdul Qadir

F

Farrukh Baig

DOI：10.1016/j.egyai.2026.100855

Original

Save

View PDF

Abstract

En 中文

• CRATEBoost: unified AI framework integrating retrieval-augmented gradient boosting with probabilistic forecasting. • Regime-aware analogue retrieval embedded directly into natural-gradient optimization of joint mean–variance learner. • Outperforms 8 SOTA baselines by 4–14% nMAE and 1.5–15% CRPS across multi-horizon forecasts. • Operationally deployable for high-PV-penetration grids: unit commitment, reserve sizing, ramp management. • Advances AI methodology for renewable energy forecasting through unified probabilistic design.

Keywords:

Photovoltaic power forecasting

probabilistic forecasting

retrieval-augmented gradient boosting

natural gradient boosting

UMAP

Ensemble Model Output Statistics

Contextual retrieval-augmented tree ensemble Boosting for Multi-Horizon Solar PV Power Forecasting

Abstract

AI Summary

Journal

Organization