| 引用本文: | 王齐辉,孙国强,陈胜,卫志农,臧海祥.基于深度学习的最优线路开断分布鲁棒优化调度方法[J].电力自动化设备,2025,45(5):218-224. |
| WANG Qihui,SUN Guoqiang,CHEN Sheng,WEI Zhinong,ZANG Haixiang.Distributionally robust optimal dispatching method of optimal transmission switching based on deep learning[J].Electric Power Automation Equipment,2025,45(5):218-224. |
|
| 摘要: |
| 随着新能源渗透率的提高，源荷时空分布不均匀问题愈发突出，且新能源固有的不确定性影响电网安全稳定运行。提出考虑光伏不确定性的最优线路开断分布鲁棒优化模型。为提高模型求解效率，提出一种基于深度学习的最优线路开断分布鲁棒优化调度方法。采用列与约束生成算法求解得到大量样本；以负荷、光伏出力及不同场景下的光伏初始概率为输入，以最优线路开断方案为输出，构建神经网络；针对离线训练后的神经网络，以新的负荷、光伏出力和不同场景光伏初始概率为输入，得到对应的最优线路开断方案；将得到的最优线路开断方案代入物理模型进行求解以提高物理模型的求解效率。改进的IEEE 118节点算例验证了所提分布鲁棒模型的有效性以及数据驱动方法的高效性。 |
| 关键词: 输电线路开断 分布鲁棒优化调度 列与约束生成算法 神经网络 |
| DOI：10.16081/j.epae.202412027 |
| 分类号:TM73 |
| 基金项目:国家自然科学基金面上基金资助项目(52077060) |
|
| Distributionally robust optimal dispatching method of optimal transmission switching based on deep learning |
|
WANG Qihui, SUN Guoqiang, CHEN Sheng, WEI Zhinong, ZANG Haixiang
|
|
School of Electrical and Power Engineering, Hohai University, Nanjing 211100, China
|
| Abstract: |
| Along with the increasing penetration of renewable energy source, the issue of uneven spatial and temporal distribution of supply and demand becomes more pronounced, and the inherent uncertainty of renewable energy source affects the safe and stable operation of power grid. A distributionally robust optimal model of optimal transmission switching is proposed considering photovoltaic uncertainty. In order to improve the model solving efficiency, a distributionally robust optimal dispatching method of optimal transmission switching is proposed based on the deep learning. The column and constraint generation algorithm is adopted to obtain a large number of samples. Taking the load, photovoltaic output, and photovoltaic initial probabilities under different scenarios as the input, and the optimal transmission switching scheme as the output, a neural network is constructed. Aiming at the offline trained neural network, the new load, photovoltaic output, and photovoltaic initial probabilities under different scenarios are taken as the input, and the corresponding optimal transmission switching scheme is obtained, which is substituted into the physical model for solving to improve the solving efficiency of physical model. The effectiveness of the distributionally robust model and the efficiency of the proposed data-driven method are verified by the modified IEEE 118-bus system example. |
| Key words: transmission line switching distributionally robust optimal dispatching column and constraint generation algorithm neural network |
