Quantum-inspired distributed policy-value optimization learning with advanced environmental forecasting for real-time generation control in novel power systems

With the increasing weight of wind and photovoltaic power (WPP) in grids, the uncertainty of WPP has an increasing impact on power systems. The access to WPP raises the difficulty of controlling the grid frequency. The existing automatic generation controllers have difficulty in reducing the impact of a high proportion of WPP integration on grid frequency, difficulty in autonomously adjusting power balance in different areas of distributed novel power systems, and slow system operation speed. To address instability, weak adaptability, and slow processing rate caused by large-scale WPP access to novel power systems, this study proposes the quantuminspired distributed policy-value optimization learning methods for power generation control in distributed systems and adopts the advanced environmental forecasting method into the multi-temporal distance forecasting-distributed Grover policy-value optimization (MF-DGPVO) methods. In this study, the MF-DGPVO methods combine quantum thought with the training process of value optimization and policy optimization to reduce data operations and improve network operation speed. In this study, the MF-DGPVO methods are simulated in the two- and four-area novel power grids. Compared to proportional-integral-derivative, Q learning (QL), state-action-reward-state-action (SARSA), sliding mode control, and fuzzy logic control, the MF-DGPVO methods are more than 30% smaller in frequency deviations and more than 50% smaller in control errors. Besides, the proposed MF-DGPVO approaches obtain greater flexibility than QL and SARSA.