Interrelationships between urban travel demand and electricity consumption: a deep learning approach

Movahedi, Ali; Parsa, Amir Bahador; Rozhkov, Anton; Lee, Dongwoo; Mohammadian, Abolfazl Kouros; Derrible, Sybil · 2023 · Crossref

DOI: 10.1038/s41598-023-33133-y

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the interrelationships between urban travel demand and electricity consumption, positing that these infrastructure systems are interdependent because individuals are either in buildings consuming electricity or traveling on roads. The research aims to enhance the sustainability and resilience of infrastructure networks by quantifying these links. Specifically, the authors examine correlation, temporal dynamics, spatial relationships, and the predictive capability of traffic data for electricity load forecasting in Chicago. The methodology employs Long Short Term Memory (LSTM) deep learning networks to model electricity consumption patterns based on traffic volume. The study utilizes two merged datasets from November 2017: aggregated residential electricity usage data at 30-minute intervals for 28 zip codes along major Chicago expressways, and traffic volume data from 211 loop detectors on the same network. The analysis proceeds in four stages: calculating Pearson correlation coefficients at loop detector, zip code, and citywide levels; testing various time windows (up to 12 hours) to determine optimal temporal inputs for prediction; evaluating spatial relationships by varying the distance between traffic detectors and zip code centroids; and training over 250 LSTM models to predict electricity consumption using traffic data from the same and nearby zip codes. The results reveal complex, non-linear interrelationships. Correlation analysis shows that while citywide Pearson coefficients are low (0.14–0.16), local correlations vary significantly, with higher values observed in northern and central Chicago compared to the south, likely due to expressway boundary effects. Temporal analysis indicates that an 8-hour time window of traffic data provides the most robust prediction of electricity consumption, outperforming shorter windows and suggesting that traffic patterns capture lifestyle elements like workdays rather than just immediate rush-hour dynamics. Spatial analysis demonstrates that model performance generally decreases as the distance between traffic detectors and the target zip code increases, confirming that local traffic data is a stronger predictor than distant data. The LSTM models successfully identified significant predictive power in traffic data for electricity load forecasting across various zip codes. The significance of this work lies in demonstrating that traffic volume is a viable proxy for estimating electricity demand, offering a novel approach to short-term load forecasting. By leveraging existing transportation data, utilities and planners can improve grid stability and operational efficiency. The findings underscore the importance of analyzing infrastructure systems as interconnected entities rather than isolated networks, providing a framework for future research on multi-infrastructure resilience and sustainability in urban environments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• exposure measurement