[Paper Review] Energy Spatio-Temporal Pattern Prediction for Electric Vehicle Networks

Qinglong Wang|arXiv (Cornell University)|Feb 14, 2018

Electric Vehicles and Infrastructure33 references3 citations

TL;DR

This paper proposes a spatio-temporal prediction framework for estimating aggregated electricity energy stored in electric vehicles (EVs) across city-scale regions using spatial and temporal features derived from real Beijing taxi trajectories. It combines a neural network-based spatial predictor with a linear-chain CRF-based temporal predictor, achieving a normalized mean square error (NMSE) below 0.1 across all regions, demonstrating high accuracy for smart grid integration and EV charging management.

ABSTRACT

Information about the spatio-temporal pattern of electricity energy carried by EVs, instead of EVs themselves, is crucial for EVs to establish more effective and intelligent interactions with the smart grid. In this paper, we propose a framework for predicting the amount of the electricity energy stored by a large number of EVs aggregated within different city-scale regions, based on spatio-temporal pattern of the electricity energy. The spatial pattern is modeled via using a neural network based spatial predictor, while the temporal pattern is captured via using a linear-chain conditional random field (CRF) based temporal predictor. Two predictors are fed with spatial and temporal features respectively, which are extracted based on real trajectories data recorded in Beijing. Furthermore, we combine both predictors to build the spatio-temporal predictor, by using an optimal combination coefficient which minimizes the normalized mean square error (NMSE) of the predictions. The prediction performance is evaluated based on extensive experiments covering both spatial and temporal predictions, and the improvement achieved by the combined spatio-temporal predictor. The experiment results show that the NMSE of the spatio-temporal predictor is maintained below 0.1 for all investigate regions of Beijing. We further visualize the prediction and discuss the potential benefits can be brought to smart grid scheduling and EV charging by utilizing the proposed framework.

Motivation & Objective

• To model and predict the spatio-temporal patterns of electricity energy carried by large-scale EVs in urban regions.
• To enable smarter interaction between EVs and the smart grid by forecasting available energy for scheduling and V2G services.
• To improve energy management efficiency for EV charging service providers through accurate, region-specific energy predictions.
• To evaluate the performance of combined spatial and temporal predictors against individual models using real-world trajectory data.

Proposed method

• Extract spatial features from EV trajectories using a neural network to model regional energy distribution patterns.
• Model temporal energy changes using a linear-chain conditional random field (CRF) to capture time-series dynamics of state-of-charge (SOC).
• Combine spatial and temporal predictors via an optimal combination coefficient that minimizes normalized mean square error (NMSE).
• Train and evaluate the spatio-temporal predictor on real GPS and SOC data from Beijing taxis, covering 16 city regions.
• Use a weighted fusion strategy to balance spatial and temporal predictions, reducing error through complementary strengths.
• Visualize real-time aggregated energy predictions across Beijing’s 4th Ring Road to demonstrate practical utility.

Experimental results

Research questions

• RQ1 How accurately can the spatio-temporal pattern of aggregated EV energy be predicted across city-scale regions?
• RQ2 What is the contribution of spatial versus temporal features in predicting future EV energy availability?
• RQ3 Can a combined spatio-temporal predictor outperform individual spatial or temporal models in prediction accuracy?
• RQ4 How does prediction performance degrade over longer time horizons (e.g., 2–3 hours ahead)?
• RQ5 What are the practical implications of accurate energy pattern prediction for smart grid scheduling and EV charging services?

Key findings

• The spatio-temporal predictor maintains an NMSE below 0.1 across all 16 investigated regions in Beijing, indicating high prediction accuracy.
• The combined predictor outperforms both the individual spatial and temporal predictors, demonstrating complementary error reduction.
• For 3-hour-ahead predictions, the NMSE of the spatio-temporal predictor reaches a maximum of 0.09, showing robust long-term forecasting capability.
• The temporal predictor alone shows slight performance improvement over longer intervals, indicating its stability in time-series forecasting.
• The spatial predictor degrades more significantly with longer prediction horizons due to fading spatial correlation over time.
• Visualization results confirm that central urban regions (within 3rd Ring Road) have higher aggregated energy, aligning with real traffic and energy usage patterns.