MACHINE LEARNING BASED OPTIMIZATION OF EV VEHICLE SOC IN BATTERIES EMPOWERING SELF GOVERNANCE AND ENERGY MANAGEMENT

Abstract

This study investigates how local self-government institutions can effectively support Electric Vehicle (EV) adoption by integrating advanced governance mechanisms with Machine Learning–based State of Charge (SOC) estimation for Lithium Cobalt Oxide () batteries. A comprehensive dataset was collected from 14 municipal corporations and 22 urban local bodies (ULBs) across South India, comprising 3,200 EV trip logs, 1,150 smart-charging station entries, and 760 battery diagnostic profiles. To enhance SOC prediction accuracy within real municipal operating environments, a Hybrid LSTM–CNN–RNN SOC Estimation Framework (LCR-SOCEF) was developed. The LSTM component models temporal variations in battery behaviour, CNN extracts feature patterns from dynamic charging–discharging cycles, while RNN captures sequential governance-linked operational dependencies, such as traffic density, public-transport routing, and charging-infrastructure accessibility. Results demonstrate that ULBs with stronger institutional coordination, higher infrastructure readiness, and proactive EV governance policies achieved SOC accuracy improvements up to 5.3% for batteries. These findings highlight how ML outputs can serve as decision-support tools for municipal mobility planning, enabling better scheduling of public EV fleets, optimized charging networks, and improved citizen transport services. The study aligns directly offering a governance-centred, technology-integrated framework to strengthen sustainable, decentralized urban mobility transitions.