International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 09 | Sep 2025
p-ISSN: 2395-0072
www.irjet.net
Real Time AI-Based Bidirectional Energy Communication Between Electric Vehicles and Smart Grid System 1Prof. Shubham V. Patil, 2Prof. Adesh V. Patil 1Electrical Enginering, Sanjeevan Group of Institutions-Polytechnic, Panhala, Maharashtra, India 2Computer Science and Engineering,Dr. D. Y. Patil Technical Campus Talsande, Polytechnic Talsande,
Maharashtra, India ---------------------------------------------------------------------***--------------------------------------------------------------------Abstract—Growing electric-vehicle (EV) adoption Since 2020, research has explored AI-enhanced smartpressures distribution networks that were never grid integration. Deep reinforcement learning (DRL) designed for millions of mobile batteries. Recent dispatchers demonstrated 22% frequency-deviation literature (e.g., Fayiz et al., 2023[1]; Jiang et al., reduction during V2G services[1]; budget-aware 2023[2]) shows that deep- learning controllers can incentive RL improved user adherence by 31%[2]. orchestrate grid-to-vehicle (G2V) and vehicle-to-grid Transfer-learning CRNNs lowered mean-absolute-error (V2G) exchanges, yet field-ready, low- latency in station-load forecasting by 4%–8% with limited data frameworks remain sparse. This paper proposes a (Zhou et al., 2025[4]). On the battery side, Hybrid-AI hybrid convolutional-recurrent neural network BMSs combining neural and symbolic modules (CRNN) with transfer-learning support that resides on extended cycle-life by 15%– 20% (Sudhapriya & an edge controller inside the charging station. The Jaisiva, 2025[5]). Despite progress, gaps persist: architecture cooperates with a Hybrid-AI BatteryFew works fuse real-time AI predictions with lowManagement System (HAI-BMS) to predict state-oflatency edge execution and standardised V2G/G2V charge (SoC), optimise power bids, and enforce cyberprotocols. Cyber-physical resilience remains undersecure ISO 15118/OCPP messaging. MATLAB/Simulink addressed, even though coordinated cyber-attacks on simulations— parameterised with data from recent chargers were detected within 5 s using LSTM-DRL studies—show up to 27% peak-load shaving and 18% digital twins (Shi et al., 2023[6]). faster charging compared with rule-based baselines. Conceptual hardware results drawn from validated Most studies use single-site simulations, limiting testbeds confirm <200 ms end-to-end latency, meeting transferability. IEC 61851-23 requirements. This paper closes these gaps by proposing an edgeKeywords—EV, Battery Management, AI, Smart Grid resident CRNN-Transfer-Learning controller that collaborates with an HAI-BMS and communicates via 1. Introduction ISO 15118/OCPP to the smart grid. Rising EV penetration and renewable targets are The remainder proceed as follows. Section 2 details the converging to create unprecedented stress on architecture and control logic. Section 3 presents electricity networks. Global EV stock surpassed 40 MATLAB/Simulink simulation results. Section 4 million in 2024, adding ≈ 500 TWh annual demand[3]. outlines hardware validation using literature testbeds. Conventional unidirectional charging exacerbates Section 5 concludes and sketches future work. evening peaks; conversely, coordinated bidirectional exchange can transform parked EVs into distributed storage that cushions renewables’ volatility (Fayiz et al., 2023[1]). © 2024, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 55