Artificial emotional deep Q learning for real-time smart voltage control of cyber-physical social power systems. (15th June 2023)

• Record Type: Journal Article
• Authors: Yin, Linfei; He, Xiaoyu
• Is Part Of: Energy. Volume 273(2023)
• Publication Date: 2023-06-15
• Subjects: Cyber-physical social power systems -- Artificial emotion -- Q learning -- Deep learning -- Real-time smart voltage control; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2023.127232
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: The volatility of renewable energy leads to numerous voltage changes in a short period, thus affecting the quality of the power supply. A real-time smart voltage control framework of cyber-physical social power systems is proposed to replace the traditional multi-timescale "reactive power optimization and voltage regulation" framework. This work combines artificial emotion, deep learning, and Q learning as an artificial emotional deep Q-learning algorithm. The proposed framework and algorithm are simulated in the developed parallel smart voltage control platform. The proposed algorithm has a strong-robust capability of real-time online updating, learning, and decision-making. Furthermore, four virtual systems simulating the actual smart voltage control systems are built. Compared to the conventional proportional-integral-derivative, the voltage deviation of the proposed algorithm is reduced by 68.95%. Among six uniform time-scale algorithms, the proposed algorithm has the highest control performance and the lowest control error. Besides, the parameters of the proposed algorithm are continuously optimized to enhance the control performance through the continuous interaction of the parallel systems. The numerical results verify the effectiveness and feasibility of the real-time smart voltage control framework of cyber-physical social power systems. Highlights: Multi-timescale "reactive power optimization and voltage regulation" is replaced. Smart voltage control framework of cyber-physical social power systems is proposed. An artificial emotional deep Q-learning algorithm is proposed. Four virtual systems simulating the actual smart voltage control systems are built. Parameters of the method are continuously optimized to enhance control performances.