Deep learning models show varying uncertainty handling in load forecasting

By PulseAugur Editorial · [1 sources] · 2026-07-14 13:02

Researchers have developed a new framework for probabilistic load forecasting in smart buildings, focusing on how to best incorporate uncertainty from reconstructed input features. The study compares a post-hoc residual-quantile scheme with an integrated in-model quantile-learning scheme using three deep learning backbones: recurrent, hybrid recurrent, and Temporal Fusion Transformer (TFT) models. Results indicate that the optimal uncertainty placement is dependent on the model architecture, with the TFT model showing the most reliability when using integrated quantile learning, achieving narrower prediction intervals and better calibration. AI

IMPACT This research could improve the accuracy and reliability of energy load forecasting in smart buildings by better handling input uncertainty.

RANK_REASON The cluster contains a research paper detailing a new framework and comparative analysis of deep learning models for load forecasting. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Deep learning models show varying uncertainty handling in load forecasting

COVERAGE [1]

arXiv cs.LG TIER_1 English(EN) · Heung Seok Jeon · 2026-07-14 13:02

Learning-based Probabilistic Load Forecasting with Post-hoc and In-model Uncertainty

Smart-building load forecasters are often trained offline on dense, multivariate, high-frequency data, but deployment may provide only hourly, feature-limited inputs. Missing features must then be reconstructed, and their errors can propagate through the model. If this input unce…

COVERAGE [1]

Learning-based Probabilistic Load Forecasting with Post-hoc and In-model Uncertainty

RELATED ENTITIES

RELATED TOPICS