广东电力2026，Vol.39Issue(8)：26-35,10.DOI:10.3969/j.issn.1007-290X.2026.08.003
基于深度强化学习的海上风电并网系统备用容量优化配置方法
Reserve Capacity Optimization Configuration of Offshore Wind Power Grid-connected Systems Based on Deep Reinforcement Learning
摘要
Abstract
In response to the challenge of balancing economy and stability in traditional reserve capacity optimization methods in the context of large-scale offshore wind power integration,an improved deep reinforcement learning approach is proposed.Based on the deep deterministic policy gradient(DDPG)algorithm,a reward function incorporating transient stability constraints is designed,and a prioritized experience replay mechanism is introduced to optimize the efficiency of training sample utilization.A curriculum learning strategy is employed to implement progressive training to ensure model robustness.Simulation tests are conducted on a modified IEEE 39-bus system integrated with a 2 000 MW offshore wind farm cluster under typical daily and extreme fault scenarios.The results show that,under typical daily scenarios,the proposed method reduces the wind curtailment rate from 12.3%in traditional deterministic optimization to 4.2%,and increases the reserve adequacy rate to 97.5%.In extreme fault scenarios,the minimum frequency drops to 49.7 Hz,the minimum voltage is maintained at 0.88 p.u.,and the recovery time is shortened by more than 60%compared with traditional methods.Ablation experiments validate the effectiveness of the stability reward function and the prioritized experience replay mechanism,which enhance performance by 14.1%and improve training efficiency by 25%,respectively.This method effectively addresses the reserve capacity optimization problem under high offshore wind power penetration and provides a new technological approach for the safe and stable operation of power systems.关键词
海上风电/备用容量优化/深度强化学习/暂态稳定约束/课程学习/深度确定性策略梯度Key words
offshore wind power/reserve capacity optimization/deep reinforcement learning/transient stability constraint/curriculum learning/deep deterministic policy gradient(DDPG)分类
信息技术与安全科学引用本文复制引用
卢洵,刘新苗,金楚..基于深度强化学习的海上风电并网系统备用容量优化配置方法[J].广东电力,2026,39(8):26-35,10.基金项目
中国南方电网有限责任公司科技项目(Q202301020) （Q202301020）
