An Integrated Intelligent Framework for Electric Vehicle Energy Management and Charging Infrastructure Using HEDQN and IoT‐Based Adaptive Sparse Autoencoder With Multilevel Attention Mechanism

K. P., Ranjusha; B., Amutha; Balasundaram, Rebecca; Venkatesan, Ramalingam; Ponce-Silva, Mario

An Integrated Intelligent Framework for Electric Vehicle Energy Management and Charging Infrastructure Using HEDQN and IoT‐Based Adaptive Sparse Autoencoder With Multilevel Attention Mechanism

K. P., Ranjusha ORCID: https://orcid.org/0009-0000-5636-8195, B., Amutha ORCID: https://orcid.org/0000-0001-9875-4852, Balasundaram, Rebecca, Venkatesan, Ramalingam ORCID: https://orcid.org/0000-0003-2193-6644 and Ponce-Silva, Mario (2026) An Integrated Intelligent Framework for Electric Vehicle Energy Management and Charging Infrastructure Using HEDQN and IoT‐Based Adaptive Sparse Autoencoder With Multilevel Attention Mechanism. International Transactions on Electrical Energy Systems, 2026 (1).

[thumbnail of International Transactions on Electrical Energy Systems - 2026 - K. P. - An Integrated Intelligent Framework for Electric.pdf]

Preview

Text
International Transactions on Electrical Energy Systems - 2026 - K. P. - An Integrated Intelligent Framework for Electric.pdf - Published Version
Available under License Creative Commons Attribution.
| Preview

Official URL: https://doi.org/10.1155/etep%2F3828884

Abstract

In this research paper, a combined intelligent framework of two complementary systems for electric vehicle (EV) infrastructure is proposed. The first system focuses upon a hierarchically enhanced deep queue networking (HEDQN) model that decides on EV charging stations, routes vehicles and schedules power from the grid optimally, using hierarchical reinforcement learning in a cloud‐connected IoT network. The second system is an IoT‐based energy management system (EMS) for hybrid electric vehicles (HEVs) based on adaptive sparse autoencoder with multilevel attention network (ASA‐MANet) and a revised random variable‐based innovative American Zebra optimization algorithm (RRVI‐AZOA) for energy management. Intelligent energy distribution, accurate state‐of‐charge (SoC) prediction and dynamic power sharing between the internal combustion engine (ICE) and electric motor are achieved with real‐time data from vehicle subsystems with an external infrastructure. Experimental results display that the integrated framework is able to predict SoC with 96.4% accuracy, reduces the wait time of EV charging by 61%, increase charging station utilization from 78% to 92%, decrease peak grid‐load variation from 34% to 14%, extend battery life by 22% and enable energy feedback from EVs to the grid with 68% of the load shifting during the off‐peak period and saving ₹1200 per annum per EV. The experimental results demonstrate the framework as a scalable and sustainable solution for next‐generation electric mobility.