A Reinforcement-Learning-Based Secure Demand Response Scheme for Smart Grid System
Aparna Kumari, Sudeep Tanwar
Abstract
Smart grid (SG) systems necessitate secure demand response management (DRM) schemes for real-time decisions making to increase the effectiveness and stability of SG systems along with data security. Motivated from the aforementioned discussion, in this article, we propose Q-SDRM, a secure DRM scheme for home energy management (HEM) using reinforcement learning (RL) and ethereum blockchain (EBC) to facilitate energy consumption reduction and decrease energy costs. In cooperation with RL, <inline-formula> <tex-math notation="LaTeX">$Q$ </tex-math></inline-formula>-learning is adopted to make optimal price decisions using Markov decision process (MDP) to reduce energy consumption, which benefits both consumers and utility providers. Then, Q-SDRM uses ethereum smart-contract (ESC) to deal with data security issues and incorporate with off-chain storage interplanetary file system (IPFS) that handles data storage costs issue. Experimental results reveal the effectiveness of the proposed Q-SDRM scheme, which significantly reduces energy consumption and energy cost. The proposed scheme also provides secure access to energy data in real time compared with state-of-the-art approaches regarding different evaluation metrics, such as scalability, overall energy cost, and data storage cost.