A Hybrid 1D-CNN-LSTM Model for Grid Demand Prediction and Adaptive EV Charging Management

Electric Vehicle (EV) charging requires grid demand management along with accurate charging behavior predictions to optimize energy distribution and prevent grid congestion. This study proposes a Deep Learning (DL) based system that integrates State of Charge (SoC) monitoring in real-time to enhance grid demand forecasting and enable adaptive charging decisions. Proposed system incorporates Convolutional Neural Network (1D-CNN) for extracting localized patterns with Long Short-Term Memory (LSTM) to obtain sequential interdependencies. The proposed system dynamically adjusts charging modes (fast/slow) based on real-time grid demand predictions to balance energy loads effectively. The model was trained and evaluated using historical EV charging dataset and achieved 93.9% predictive accuracy. Results show that incorporating SoC as a key input feature significantly enhances the system's decision-making capability and charging efficiency. This research highlights the adaptive charging strategies and provides a scalable solution for real-world EV charging management.