Logo image
Back
A Hybrid GNN and DRL Model for Optimizing Energy Utilization in Solar-Powered IoT Smart Buildings and EV Charging Stations
Conference proceeding

A Hybrid GNN and DRL Model for Optimizing Energy Utilization in Solar-Powered IoT Smart Buildings and EV Charging Stations

Geeta Khatri, Murali Karri, Bharanidharan R, Ravi Kant, P. Nagasekhara Reddy and Anvesh Perada
2025 International Conference on Information, Implementation, and Innovation in Technology (I2ITCON), pp 1-7
04 Jul 2025
DOI: https://doi.org/10.1109/I2ITCON65200.2025.11210633

Abstract

Adaptation models Adaptive Control Deep reinforcement learning DRL Policy Learning Electric vehicle charging Energy Utilization EV Charging GCN Graph Neural Network Graph neural networks IoT Load Balancing Long short term memory Real-time systems Smart buildings Solar Energy Optimization TimeSeries Forecasting Weather forecasting Energy Efficiency Solar Energy
The rapid growth of smart cities creates a significant difficulty in the effective management of solar energy within IoT-enabled buildings and electric vehicle (EV) charging infrastructures, due to their dynamic, non-linear, and geographically distributed nature. This study presents an innovative hybrid model that combines Graph Neural Networks (GNN) with Deep Reinforcement Learning (DRL) to enhance energy efficiency and electric vehicle load distribution in solarpowered smart settings. Graph Neural Networks (GNNs) include spatial dependencies in energy distribution, whereas Deep Reinforcement Learning (DRL) adaptively acquires optimal energy allocation algorithms over time. The system was trained and analysed utilising real-world-inspired datasets from Kaggle, resembling solar power generation, energy consumption, and electric vehicle demands. In comparison to baseline models like LSTM, DQN, and XGBoost, the proposed model demonstrated enhanced performance, attaining a Mean Squared Error (MSE) of 𝟎. 𝟎 𝟏 𝟐 , a Root Mean Squared Error (RMSE) of 𝟎. 𝟏 𝟎 𝟗 , and an electric vehicle charging success rate of 97.2 %. Energy efficiency attained 94.6 %, illustrating the model's ability to execute intelligent, adaptive judgements in fluctuating settings. These findings highlight the model's capacity to revolutionise smart grid operations and improve sustainability. The hybrid GNN-DRL framework facilitates advanced energy management systems, with following efforts focused on real-time implementation and integration with predicted meteorological data for enhanced efficiency.

Metrics

1 Record Views

Details

Logo image