基于改进非对称三重训练的风电并网系统暂态稳定自适应评估  

Adaptive Assessment on Transient Stability of Wind Power Grid-connected System Based on Improved Asymmetric Tri-training

作　　者：孙坚[1,2] 张安祥 SUN Jian;ZHANG Anxiang(College of Electrical Engineering and New Energy,China Three Gorges University,Yichang 443000,China;Hubei Engineering Research Center for Smart Energy Technology(China Three Gorges University),Yichang 443002,China)

机构地区：[1]三峡大学电气与新能源学院,宜昌443000 [2]智慧能源技术湖北省工程研究中心(三峡大学),宜昌443002

出　　处：《电力系统及其自动化学报》2024年第8期150-158,共9页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(52077120);三峡大学科学基金资助项目(KJ20A016)。

摘　　要：为进一步提高暂态稳定评估模型的特征提取能力和模型在系统运行工况改变后的适应性,构建具有注意力机制的双路径卷积网络,以判别暂态稳定情况,得到更好的暂态稳定评估效果。当拓扑结构和运行方式变化过大时,通过时域仿真生成大量无标签样本,以双路径卷积网络作为三重训练基分类器;通过融合非对称三重训练和主动查询策略自适应调整基分类器参数,逐步提取源域与目标域之间的公共特征,减少标签样本的需求。最后,算例分析验证了所提方法的有效性。To further improve the feature extraction capability of a transient stability assessment(TSA)model and its adaptability after changes in the system operating conditions,a dual-path convolutional network with an attention mechanism is constructed to distinguish the transient stability and obtain better TSA results.When the topology structure and operation mode change too much,a large number of unlabeled samples will be generated through time-domain simulations.The dual-path convolutional network is used as a tri-training base classifier,and the parameters of the base classifier are adaptively adjusted by integrating asymmetric tri-training and active query strategies.The common features between the source and target domains are gradually extracted,thereby reducing the demand for labeled samples.Finally,the effectiveness of the proposed method was verified through the analysis of examples.

关 键 词：暂态稳定评估 非对称三重训练 主动学习 电力系统 

分 类 号：TM712[电气工程—电力系统及自动化]