Renhai Feng, Khan Wajid, Tariq Afshan, Abbas Muhammad, Yousaf Muhammad Zain, Aziz Abdul, Bajaj Mohit, Abdullah Mustafa, Tuka Milkias
School of Electrical and Information Engineering, Tianjin University, Tianjin, 300072, China.
Center for Renewable Energy and Microgrids, Huanjiang Laboratory, Zhejiang University, Zhuji, 311816, Zhejiang, China.
Sci Rep. 2025 Apr 3;15(1):11499. doi: 10.1038/s41598-025-94419-x.
As power systems around the world shift to incorporate more renewable energy sources, particularly wind power, maintaining grid stability becomes increasingly challenging due to the inherent variability of these sources. This paper introduces a novel bi-level robust optimization framework that enhances the capabilities of adaptive Under-Frequency Load Shedding (AUFLS) in managing the uncertainties brought by high penetration of wind energy and dynamic participation of electric vehicles (EVs). Central to this framework is an innovative adaptive non-parametric Kernel Density Estimation (AAKDE) technique, which sharpens the accuracy of wind power fluctuation predictions. This method enables more precise and efficient control of load-shedding events, which is crucial for preventing frequency drops that can lead to grid instability. This research proposes a strategic shedding queue mechanism that systematically prioritizes the discharge of EVs based on their real-time state-of-charge and charging behavior. This prioritization minimizes user discomfort and taps into the potential of EVs as flexible energy resources, thus providing substantial support to grid operations. To enhance the responsiveness of our AUFLS approach, we integrate a reinforcement learning model that adjusts in real time to grid conditions, optimizing decision-making for frequency stabilization. Our extensive MATLAB/SIMULINK simulations on an upgraded IEEE 39 bus test system demonstrate a significant reduction in load shedding requirements. Compared to traditional AUFLS methods, our approach cuts load shedding by over 50%, effectively maintains system frequency within safe operational limits, and shows superior performance in scenarios of high renewable variability and EV integration. This research highlights the potential of adaptive non-parametric methods in transforming AUFLS strategies, paving the way for smarter, more resilient power systems equipped to handle the complexities of modern energy landscapes.
随着世界各地的电力系统转向纳入更多可再生能源，特别是风能，由于这些能源固有的波动性，维持电网稳定性变得越来越具有挑战性。本文介绍了一种新颖的双层鲁棒优化框架，该框架增强了自适应低频减载（AUFLS）在管理风能高渗透率和电动汽车（EV）动态参与所带来的不确定性方面的能力。该框架的核心是一种创新的自适应非参数核密度估计（AAKDE）技术，它提高了风电波动预测的准确性。这种方法能够更精确、高效地控制减载事件，这对于防止可能导致电网不稳定的频率下降至关重要。本研究提出了一种策略性减载队列机制，该机制根据电动汽车的实时荷电状态和充电行为系统地对其放电进行优先级排序。这种优先级排序将用户不适感降至最低，并挖掘了电动汽车作为灵活能源资源的潜力，从而为电网运行提供了有力支持。为了提高我们的AUFLS方法的响应能力，我们集成了一个强化学习模型，该模型可根据电网状况实时调整，优化频率稳定的决策。我们在升级后的IEEE 39节点测试系统上进行的广泛MATLAB/SIMULINK仿真表明，减载需求显著降低。与传统的AUFLS方法相比，我们的方法将减载量降低了50%以上，有效地将系统频率维持在安全运行范围内，并且在高可再生能源波动性和电动汽车集成的场景中表现出卓越的性能。本研究突出了自适应非参数方法在转变AUFLS策略方面的潜力，为配备应对现代能源格局复杂性的更智能、更具弹性的电力系统铺平了道路。
