摘要
针对现阶段机器学习在风电并网系统暂态电压稳定评估的快速性、准确性方面存在的不足,提出了一种基于grcForest模型的风电并网系统暂态电压稳定评估方法。首先针对输入特征数目随着级联森林层数的增加可能出现的梯度增长或梯度减少的问题,采用残差网络对其进行优化,保证了层数增加后的模型依旧能保持最初的学习能力;其次分析风电并网系统暂态电压的关键影响因素,结合暂态故障构建输入特征;再通过评估模型离线训练,完成模型的参数设置和性能优化;最后把构建完成的输入特征应用于grcForest风电并网系统暂态电压稳定评估模型,结合数据对模型进行评估验证。IEEE10机39节点系统的仿真分析验证了该方法的快速性和准确性。
Aiming at the shortcomings of machine learning in the rapidity and accuracy of transient voltage stability assessment of wind power grid-connected system at the present stage,the paper proposes a transient voltage stability assessment method for the system based on grcForest model.Firstly,for the problem that the number of input features occurs gradient increases or decreases with the increase of cascaded forest layers,a residual network is used to optimize it,ensuring that the model with increased layers can still maintain initial learning ability.Then the key factors affecting the transient voltage of the system are analyzed,and the input features are constructed combined with transient fault.After that,through the offline training of an evaluation model,the parameter setting and performance optimization of the model are completed.Finally,the constructed input features are applied to the grcForest model-based transient voltage stability evaluation model for the system,and the model is evaluated and verified with the data.The rapidity and accuracy of the method are verified by the simulation analysis of IEEE 10-machine 39-bus system.
作者
陈康
王泽
郭永吉
CHEN Kang;WANG Ze;GUO Yongji(College of Electrical and Information Engineering,Lanzhou University of Technology,Lanzhou 730050,China)
出处
《智慧电力》
北大核心
2023年第1期31-37,共7页
Smart Power
基金
国家自然科学基金资助项目(51867015)。
作者简介
陈康(1996),男,河北保定人,硕士研究生,主要研究方向为新能源并网系统电压稳定性。
