摘要
随着新能源大规模接入电网,传统调度模式难以应对系统高随机性与复杂性,电网侧储能系统的优化调度成为提升电网灵活性与可靠性的关键。本研究提出一种基于改进深度强化学习的电网侧储能调峰控制策略:通过融合可再生能源出力、负荷需求及储能设备参数构建多源数据输入层,设计兼顾短期调峰效益与长期全生命周期成本的奖励函数,使智能体通过与微网环境交互学习最优调度策略。基于园区级微网测试系统的案例表明,该策略较传统调度方法,全生命周期成本降低11.9%~34.6%,电池寿命延长22.55%~37.36%,同时新能源综合消纳率提升至92.3%,微网峰谷差降幅达36.36%。该策略为现代电网中电网侧储能系统的动态智能管理提供数据驱动方案,助力提升电网运行效率与新能源消纳能力。
The large-scale integration of renewable energy into the power grid presents challenges for traditional scheduling methods,which struggle to manage the system's high randomness and complexity.Optimizing the scheduling of grid-side energy storage systems is crucial for enhancing the power grid's flexibility and reliability.Accordingly,this study proposes a peak-shaving control strategy for grid-side energy storage using an improved deep reinforcement learning algorithm.The strategy features a multi-source data input layer that integrates renewable energy output,load demand,and energy storage equipment parameters.A reward function is designed to balance short-term peak-shaving benefits with long-term full-life cycle costs,allowing the agent to learn optimal scheduling strategies through interaction with the microgrid environment.Case studies conducted on a park-level microgrid test system demonstrate that this strategy,compared with traditional scheduling methods,reduces full-life cycle costs by 11.9%to 34.6%,extends battery life by 22.55%to 37.36%,increases the comprehensive renewable energy consumption rate to 92.3%,and decreases the peak-valley difference of the microgrid by 36.36%.This strategy offers a data-driven approach to the dynamic and intelligent management of grid-side energy storage systems in modern power grids,thereby improving operational efficiency and enhancing the ability to absorb renewable energy.
作者
杨瑞锋
韩昱
YANG Ruifeng;HAN Yu(Xinzhou Power Supply Company,State Grid Shanxi Electric Power Company,Xinzhou O34000,Shanxi,China)
出处
《储能科学与技术》
北大核心
2026年第1期166-176,共11页
Energy Storage Science and Technology
基金
国网山西省电力公司科技项目资助(5205H0230001)。
关键词
改进深度强化学习
电网侧储能
奖励函数
优化调度
全生命周期
improved deep reinforcement learning
grid-side energy storage
reward function
optimized scheduling
full life cycle
