基于MGS-LGBM算法的电力系统暂态稳定评估

多粒度级联轻型梯度提升机(MGS-LGBM)具有超参数设置简单、模型泛化能力强、分类准确率高、训练评估快等特点。为提高电力系统暂态稳定评估的准确性和快速性,将MGS-LGBM引入电力系统暂态评估中。首先通过时域仿真提取原始数据,构造能够反映系统稳定情况的23维特征量,输入MGS-LGBM模型中,稳定结果作为输出量,利用模型中的多粒度扫描和级联结构对样本特征和结果进行高效并行训练。通过新英格兰10机39节点系统仿真验证MGS-LGBM算法,通过与其它机器学习算法比较,算法在提高暂态评估准确率的同时兼顾快速性,且在含有无关特征和训练集较少的情况下仍能保持较好的评估性能。

Power System Transient Stability Assessment Based on MGS-LGBM

In order to improve the accuracy of power system transient stability assessment,this paper introduces multi-grained cascade light gradient boosting machine(MGS-LGBM),characterized by simple hyper-parameter setting,strong model generalization ability,high classification accuracy and fas training evaluation,into power system transient assessment.Firstly,the original data is extracted by time domain simulation,and the 23-dimensional feature quantity that can reflect the stability of system is constructed as the input of MGS-LGBM model,the stable result is used as the output,and the multi-granularity scan and the cascaded stage are used to train the features and results in parallel efficiently.Through the simulation of New England 10 machine 39 bus system,MGS-LGBM algorithm has higher transient evaluation accuracy than other machine learning algorithms,and can still maintain a good evaluation with irrelevant features and a small number of training sets.