Path informed adaptive trend analyzer using Hilbert Huang transform for electric vehicle driving range prediction

Aljafari, Belqasem; Sudhakar Babu, Thanikanti; Selvarajan, Shitharth · 2026 · Crossref

DOI: 10.1038/s41598-026-37995-w

archive: archived pipeline: cataloged verified

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

Summary

This paper addresses the critical challenge of accurately predicting the driving range of electric vehicles (EVs), a key factor in mitigating "range anxiety" and optimizing energy management. Existing methods, including physics-based models and standard machine learning or deep learning approaches, often fail to capture the nonlinear, dynamic, and non-stationary characteristics of real-world driving data. These limitations result in poor generalization across different vehicle classes, battery conditions, and environmental scenarios. To overcome these issues, the authors propose a novel hybrid deep learning architecture designed to integrate high-level feature extraction, adaptive sequence modeling, and attention weight optimization. The proposed framework consists of three primary components. First, the Hilbert-Huang Auto Recurrence Transform (HART) extracts rich temporal and recurrent features from raw telemetry data. Second, the Path-Informed Adaptive Inverted Trend Hub Range Analyser (PAITHRA) serves as the predictive model, combining a Pathformer with adaptive feature routing to handle complex driving sequences. Third, the Chaotic Algae Sparrow Hyper-tuner (CASH) acts as a dynamic optimizer for attention weights, ensuring the model focuses on the most relevant input features. The model was evaluated using two extensive datasets: the EV Energy Consumption Dataset and the Full Electric Vehicle Dataset 2024, which encompass diverse driving styles, battery states, and weather conditions. The experimental results demonstrate that the proposed model significantly outperforms conventional deep learning models. It achieved a Prediction Error of 2.5, a Mean Absolute Error (MAE) of 0.8 kW, and a Root Mean Square Error (RMSE) of 1.3 kW. The model attained a Validation Accuracy of 99% and an R² Score of 0.991. Furthermore, the framework is computationally efficient, requiring only 35 minutes for total convergence across all epochs and an inference time of just 8 milliseconds per sample. These metrics indicate superior accuracy and robustness compared to existing state-of-the-art methods. The significance of this work lies in its ability to provide precise, context-specific range predictions that adapt to real-world uncertainties. By effectively handling heterogeneous inputs such as onboard telemetry, GPS traces, and environmental factors, the model supports better route planning, reduces unnecessary charging cycles, and extends battery life. This advancement contributes to increased user confidence in EVs, facilitates efficient fleet management, and promotes the broader adoption of sustainable transportation by minimizing energy waste and lifecycle emissions.

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.

• telematics crash prediction