基于轻量级梯度提升机和生成对抗网络的含风电电力系统频率稳定评估  

Frequency Stability Evaluation of Power System Containing Wind Power Based on Light Gradient Boosting Machine and Generative Adversarial Network

作　　者：赵冬梅[1] 郑亚锐 谢家康 郭育村 ZHAO Dongmei;ZHENG Yarui;XIE Jiakang;GUO Yucun(School of Electrical and Electronic Engineering,North China Electric Power University,Changping District,Beijing 102206,China)

机构地区：[1]华北电力大学电气与电子工程学院,北京市昌平区102206

出　　处：《电网技术》2022年第8期3181-3190,共10页Power System Technology

基　　金：国家重点研发计划项目(2017YFB0902600)。

摘　　要：针对目前电力系统频率稳定评估研究未考虑新能源和系统拓扑变化的问题,提出一种考虑风速特征的基于轻量级梯度提升机(light gradient boosting machine,lightGBM)和生成对抗网络(generative adversarial network,GAN)的含风电电力系统频率稳定评估方法。首先分析风电对频率稳定的影响,其次采用lightGBM对频率变化率,暂态频率极值和准稳态频率3个指标建立预测模型,引入注意力机制对输入特征排序降维,通过预测指标综合判断系统频率稳定性。系统拓扑发生改变时,采用GAN产生大量相似样本对模型进行更新。在含风电新英格兰10机39节点系统和含风电IEEE118节点系统上的仿真结果表明,所提方法比传统机器学习方法精度更高,速度更快,泛化性能更好。且考虑风速特征后不同算法的模型精度均大大提高。Aiming at the problem that the new energy sources and system topology changes are not considered in the current power system frequency stability assessment researches, a method for frequency stability assessment of the wind power system based on the light gradient boosting machine(light GBM) and the generative adversarial network(GAN) is proposed with the consideration of the characteristics of wind speed. Firstly, the influence of wind power on the frequency stability is analyzed. Secondly, the light GBM is used to establish a prediction model for the three indicators of the frequency change rate, the transient frequency extreme value and the quasi-steady-state frequency. An attention mechanism is introduced to rank the input features and reduce their dimension Finally, the prediction indicators are combined to determine the system frequency stability. When the system topology changes, the GAN is used to generate a large number of similar samples to update the model. The simulation results on the 39-node system with wind power IEEE10 machines and the IEEE118-node system with wind power show that the proposed method has higher accuracy, faster speed and better generalization performance than the traditional machine learning methods. And after considering the wind speed characteristics, the model accuracy with different algorithms is greatly improved.

关 键 词：风电 电力系统 频率稳定 轻量级梯度提升机 生成对抗网络 

分 类 号：TM614[电气工程—电力系统及自动化]