An Empirical Benchmark of Deep Time-Series Models for Smart Meter Energy Forecasting

Behnaz Kavoosighafi, Maria Eidenskog, Wiktoria Glad, Katerina Vrotsou

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 13 functions out of this paper's own repositories and ran 13 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

Repositories linked to this paper

• behnazkavoosi/Energy-Time-Series-Library

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

We benchmark nine deep learning forecasting models on two high-resolution real-world smart meter datasets. • We show that forecasting accuracy improves with longer historical context only up to a saturation point, while accuracy consistently declines as the prediction horizon increases. • We demonstrate that lightweight architectures achieve competitive accuracy at lower computational cost, and that architectural differences become significant mainly at longer horizons and on more heterogeneous data. • We show, through a subgroup analysis, that Transformer-based models appear to offer an advantage on the most under-represented population segments, but this pattern is not statistically robust and does not hold consistently across datasets.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2608.18675")
get_code_for_paper("2608.18675")
have("2608.18675")

Connect an agent — have() is free.