SPECTRA: State-Space Exogenous Context and Temporal-Frequency Resolution Architecture for Probabilistic Energy Forecasting

2026-07-22

AI Summary

Existing probabilistic forecasting methods for energy systems struggle to jointly model multi-scale structures, exogenous variables, and uncertainty. This work proposes a novel state-space-based architecture that, for the first time, adopts deterministic–stochastic separation as a core design principle. The approach adaptively decomposes the target series into trend-cycle and high-frequency residual components, each fused with aligned exogenous contextual information, and then refines them jointly through a multi-resolution spectral-temporal state space model. Furthermore, it estimates calibrated quantile bounds using complementary representations. Evaluated across 18 experimental settings encompassing load, electricity price, photovoltaic, and wind power forecasting, the method achieves the best continuous ranked probability score (CRPS) in 14 cases, yielding an average CRPS reduction of 5.74% and a 7.27% improvement in upper-tail quantile risk.

Abstract

Modern power systems increasingly require probabilistic forecasts amid interacting uncertainties from renewable intermittency, flexible demand, market volatility, and weather-dependent generation. However, existing methods often treat multi-scale decomposition, exogenous-variable alignment, and probabilistic output as separate steps, obscuring how predictable structures and uncertainty-bearing fluctuations jointly shape the forecast distribution. This paper proposes a state-space exogenous-context and temporal-frequency resolution architecture for general probabilistic energy forecasting. Its central premise is that trend-periodic components primarily determine the baseline trajectory, whereas high-frequency residuals and external perturbations govern the spread and asymmetry of forecast uncertainty. Accordingly, the architecture adaptively separates deterministic and residual streams, aligns exogenous context with both, refines the deterministic backbone through multi-resolution spectral-temporal state-space modeling, and estimates ordered quantile boundaries from their complementary representations. Experiments on load, price, solar, and wind forecasting achieve the best continuous ranked probability score in 14 of 18 settings, reducing average CRPS by 5.74\% and upper-tail quantile risk by 7.27\% over the strongest baselines. These results support deterministic-stochastic separation as an effective design principle for general probabilistic energy forecasting.

Problem

• probabilistic energy forecasting
• multi-scale decomposition
• exogenous variables
• forecast uncertainty
• renewable intermittency

Innovation

• state-space modeling
• probabilistic forecasting
• multi-resolution spectral-temporal
• exogenous context alignment
• deterministic-stochastic separation