2024基于机理与人工智能混合驱动的新型电力系统智能分析与调控策略研究报告

Industry Overview - The report focuses on the development of intelligent analysis and control strategies for new power systems driven by a combination of mechanism and artificial intelligence [1] - The core objective is to achieve "carbon peak and carbon neutrality" by building a new power system characterized by high proportions of renewable energy integration, power electronics, and energy storage devices [4] - Key challenges include the dynamic, random, and uncertain nature of grid operations, which require advanced online safety assessment and intelligent scheduling control methods [4] Key Research Areas - Multi-temporal and spatial dimension power prediction technology has been developed using integrated machine learning models, applied in various scenarios such as residential, industrial, and commercial loads, as well as renewable energy generation and charging stations [2] - Intelligent decision-making technologies based on deep reinforcement learning have been developed to address control challenges in power system planning and scheduling, including optimization of power equipment configuration and control of reactive voltage, active power, network loss, and topology [2] - Digital twin modeling and parameter intelligent identification technologies have been established for complex power equipment, including traditional generators, DC systems, wind power, and composite load models [2] Applications and Case Studies - A data-driven, AI-based grid brain technology framework has been applied in Jiangsu Power Grid, addressing issues such as voltage violations, power loss reduction, and power flow constraints [14] - The intelligent control system deployed in Jiangsu Power Grid achieved a 99.41% effective instruction rate, with an average network loss reduction of approximately 3.6412% [16] - Reinforcement learning-based methods have been used for automatic parameter calibration of generator and excitation models, achieving good performance after 400 training iterations [20] Advanced Technologies - Deep reinforcement learning has been applied to real-time AC optimal power flow, deriving fast solutions for secure and economic grid operations [22] - AI-based autonomous topology control has been developed to maximize time-series available transfer capabilities, considering uncertainties in grid operations [37] - A comprehensive planning, early warning, and control platform integrating AI, HPC, and big data technologies has been established for multi-objective power system simulation, verification, and control [41] Distributed Resource Management - Virtual power plants (VPPs) have been developed to aggregate distributed renewable generation, energy storage, and load subsystems, providing capacity and ancillary services to improve grid economy and reliability [25] - VPPs enable intelligent adjustment of various load devices, addressing challenges related to the randomness and volatility of new energy integration [25] - Reinforcement learning-based net load volatility control has been applied in active distribution power networks, effectively reducing peak-valley differences and overall fluctuations [32]