基于Elastic Net的暂态稳定裕度在线评估  

Elastic Net Based Online Assessment of Power System Transient Stability Margin

作　　者：宓登凯 王彤[1] 相禹维 杜文娟[1] MI Dengkai;WANG Tong;XIANG Yuwei;DU Wenjuan(State Key Laboratory of Alternate Electric Power Systems With New Energy Resources(North China Electric Power University),Changping District,Beijing 102206,China)

机构地区：[1]新能源电力系统国家重点实验室(华北电力大学)

出　　处：《电网技术》2020年第1期19-26,共8页Power System Technology

基　　金：国家自然科学基金资助项目(51637005);中央高校基本科研专项资金(2018MS006);国家电网公司科技项目(系统保护动作行为全过程智能分析、控制决策与评价关键技术研究(SGTYHT/17-JS-199))~~

摘　　要：随着电力系统规模的扩大和新能源技术的广泛应用,电力系统暂态稳定特性更加复杂,在线暂态稳定裕度评估面临严峻挑战。提出一种基于Elastic Net的电力系统暂态稳定裕度在线评估的新方法。该方法无需计算形成输入特征,直接面向量测数据,将系统稳态运行节点电压幅值与相角作为样本特征,利用Elastic Net算法完成特征筛选并构建量测数据与故障极限切除时间之间的映射关系,进一步将原始特征映射至高维空间以提高预测模型的准确性。最终,实现根据系统稳态运行信息对暂态稳定裕度的快速预测。对接入风电场的新英格兰39节点系统和IEEE 118节点系统进行算例分析,结果表明,该方法通过少量数据训练即可有效筛除无关特征并具有较高的预测精度。With expansion of power system and continuous proliferation of renewable energy generation,the transient stability characteristics of power system become more complicated,and online transient stability margin assessment faces severe challenges.This paper proposes a new method based on Elastic Net for online evaluation of power system transient stability margin.Instead of calculated features,the measurement data before fault were employed as inputs.The system steady-state bus voltage amplitude and phase angle are taken as sample features.Then the Elastic Net algorithm is used to complete the feature selection and construct a mapping relationship between measurement data and critical clear time.The original features are further mapped to a high dimensional space to improve accuracy of the prediction model.Finally,fast prediction of the transient stability margin based on the steady state operating information of the system is achieved.Case study on New England 39-bus system connected to wind farm and IEEE 118-bus system shows that the proposed approach can effectively select features and has high prediction accuracy through a small amount of data training.

关 键 词：暂态稳定裕度 在线评估 Elastic Net 特征筛选 

分 类 号：TM721[电气工程—电力系统及自动化]