Measurement-Oriented Coherent Interval Estimation for Power Load Uncertainty Using Unified Paired-Quantile Multitask Learning
IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 1-14
DOI: 10.1109/tim.2026.3684681 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Heng-Yi Su und 1 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Gefördert von: National Science and Technology Council.
Abstract (Originalsprache)
Reliable and coherent characterization of uncertainty in power load measurements is essential for modern energy systems operating under high-resolution metering environments. In practice, the nonstationary and heterogeneous nature of load measurements complicates the construction of well-calibrated and probabilistically coherent prediction intervals (PIs) across multiple coverage levels. To address this challenge, this article formulates coherent interval-based load uncertainty estimation (CI-LUE) as a measurement-oriented multicoverage quantile estimation problem under a unified paired-quantile multitask learning (UP-MTL) paradigm. Building upon this formulation, a probabilistic adaptive multigate mixture-of-experts (PA-M2oEs) framework is developed to dynamically allocate expert capacity across quantile levels based on task-dependent estimation difficulty, thereby enhancing probabilistic coherence and adaptivity under measurement uncertainty. Within the PA-M2oE architecture, a lightweight modified temporal convolutional network (LM-TCN) is incorporated to improve temporal representation learning under nonstationary measurement conditions. The proposed framework is optimized using a composite objective that jointly enforces quantile monotonicity and expert diversity, yielding ordered and well-calibrated PIs without reliance on distributional assumptions or postprocessing. Experiments on real-world power load measurement datasets demonstrate consistent improvements in interval reliability, sharpness (SH), and probabilistic coherence across multiple coverage levels, validating the effectiveness of the proposed CI-LUE framework.
Förderung
- National Science and Technology Council
Zitieren
APA
Su, H., & Lin, T. (2026). Measurement-Oriented Coherent Interval Estimation for Power Load Uncertainty Using Unified Paired-Quantile Multitask Learning. IEEE Transactions on Instrumentation and Measurement, 75, 1-14. https://doi.org/10.1109/tim.2026.3684681
Vancouver
Su H, Lin T. Measurement-Oriented Coherent Interval Estimation for Power Load Uncertainty Using Unified Paired-Quantile Multitask Learning. IEEE Transactions on Instrumentation and Measurement. 2026;75:1-14. doi:10.1109/tim.2026.3684681
BibTeX
@article{Su2026,
author = {Su, Heng-Yi and Lin, Tzu-Chieh},
title = {Measurement-Oriented Coherent Interval Estimation for Power Load Uncertainty Using Unified Paired-Quantile Multitask Learning},
journal = {IEEE Transactions on Instrumentation and Measurement},
year = {2026},
volume = {75},
pages = {1--14},
doi = {10.1109/tim.2026.3684681}
}
EUVIMED