Search papers, labs, and topics across Lattice.
This paper introduces a novel heteroscedastic, asymmetric, group-conditional split-conformal framework that utilizes a bootstrap-diverse XGBoost ensemble for generating climate-invariant prediction intervals in solar and wind forecasting. The proposed method addresses the limitations of existing probabilistic approaches by providing reliable uncertainty quantification without the need for site-specific recalibration, thus enabling effective integration of renewable energy sources across diverse climates. Evaluated across four distinct climatological sites and multiple forecasting horizons, the framework achieves near-nominal coverage and reduces the Interval Score by up to 35% compared to traditional methods, demonstrating its robustness and adaptability.
A single model can reliably forecast solar and wind generation across diverse climates without the need for recalibration, achieving up to 35% better uncertainty quantification than existing methods.
Reliable uncertainty quantification is essential for integrating solar and wind generation into modern power systems, where operators must weigh risk rather than act on point forecasts alone. Existing probabilistic methods, however, often either lack finite-sample validity or require per-site recalibration, so a single model rarely transfers across the diverse climates of a dispersed generation fleet. This paper proposes a heteroscedastic, asymmetric, group-conditional split-conformal framework built on a bootstrap-diverse XGBoost ensemble, producing prediction intervals that adapt in width to local difficulty while retaining distribution-free coverage guarantees. A single fixed specification, with no per-site or per-horizon tuning, is evaluated across four climatologically distinct sites spanning both hemispheres, at horizons of 1 to 12 hours, for both solar irradiance and wind speed. The framework holds near-nominal coverage on both targets and reduces the Interval Score by up to 35% relative to competitive baselines, with the calibration and sharpness of its intervals shown to be properties of the method rather than of site-specific tuning.
