Quantum Safe Federated Reinforcement Learning for Intelligent Energy Efficiency Optimization in IoT Enabled Smart Grids

A smart grid is as an electricity system that manages digital and refined mechanism to monitor and handle the transfer of electricity from various sources, optimizing efficiency and reliability while reducing costs and environmental impacts. The rapid growth of smart grids raises total energy demand and renewable energy combination for smart, adaptive, and energy-efficient resource distribution strategies. Traditional energy management methods, often fail to handle real-time grid dynamics, leading to suboptimal energy distribution, high operational costs, and significant energy consumption. But main challenges in smart grids are optimizing energy efficiency and controlling electricity generation, transmission, and distribution. This paper introduces an AI-powered approach to optimize energy in smart grids using Generalized Linear Regressive Quantum-Safe Federated Reinforcement Learning (GLRQS-FRL). The main aim of GLRQS-FRL model is to perform the energy efficiency optimization in IoT-enabled smart grids with minimal computation and communication overhead. To begin with, GLRQS-FRL model collects the smart grid data from the dataset. After the acquisition phase, data preprocessing is carried out to transform the raw dataset into cleaned format based on missing data handling and outlier removal. Followed by, two phase linear regression is employed to determine the most relevant features and remove the others. Finally, the Quantum Federated Reinforcement Learning performs the optimal energy usage prediction by employing Azadkia-Chatterjee correlation coefficient with the selected relevant features with higher accuracy. In addition, Quantum differential privacy model is employed to further protect sensitive data. Finally, accurate energy efficiency optimization results are predicted with minimal error. Experimental consideration of proposed GLRQS-FRL model is conducted using various evaluation metrics such as accuracy, RMSE, NMSE, R ² score, computation overhead and communication overhead. The quantitatively analyzed results reveal that the proposed GLRQS-FRL model attains higher accuracy in smart grid optimization with minimal overhead as well as lesser error compared to traditional deep learning methods.