您的位置：首页 > 论文页面
基于在线自适应直推式支持向量机的电力系统 暂态稳定评估
发表时间：2019-10-31 浏览量：1227 下载量：178
| 全部作者： | 陈鑫磊，王韶 |
| 作者单位： | 重庆大学输配电装备及系统安全与新技术国家重点实验室 |
| 摘 要： | 现有基于机器学习的暂态稳定评估模型无法在线更新，对实际系统的适应能力差且参数优化计算代价大。针对以上不足，本文结合直推式学习和增量学习的思想，提出一种基于在线自适应直推式支持向量机（online adaptive transductive support vector machine，OATSVM）的电力系统暂态稳定评估方法。通过广域测量装置实时监测系统状态，当故障发生时，根据相量测量单元提供的故障前后实测信息，构建一组能反映暂态稳定特性的28维系统特征作为评估模型的输入量。利用在线直推式支持向量机（transductive support vector machine，TSVM）对新增样本进行增量学习，在评估稳定结果的同时实现模型的在线更新。为加快参数优化的计算速度，在参数摄动下对模型进行自适应优化。最后对IEEE-68节点算例系统进行仿真与分析，从预测精度和计算时间两方面验证所提方法的准确性和有效性。 |
| 关 键 词： | 电气工程；电力系统；暂态稳定评估；直推式支持向量机；在线学习；广域测量系统 |
| Title： | Transient stability assessment of power system based on online adaptive transductive support vector machine |
| Author： | CHEN Xinlei, WANG Shao |
| Organization： | State Key Laboratory of Power Transmission Equipment & System Security and New Technology, Chongqing University |
| Abstract： | The existing machine learning-based transient stability assessment model can not be updated online and have poor adaptability to actual systems, and parameter optimization calculations are costly. In view of the above deficiencies, a novel transient stability assessment method for power system based on online adaptive transductive support vector machine (OATSVM) is proposed in this paper, by combining the idea of transductive learning and incremental learning. System status is monitored through wide area measurement system in real time. When the fault occurs, a set of 28-dimensional system features which can reflect the transient stability characteristics is constructed as the input of the evaluation model, according to the measured information before and after the fault provided by the phasor measurement unit. Newly added samples are incrementally learned by online transductive support vector machine (TSVM). Evaluating stable results and model updating are executed simultaneously. In order to speed up the parameter optimization, the model is adaptively optimized under parameter perturbation. Finally, using the simulation and analysis of IEEE-68 bus example system, the accuracy and validity of the proposed method are verified from two aspects of the prediction accuracy and calculation time. |
| Key words： | electrical engineering; power system; transient stability assessment; transductive support vector machine; online learning; wide area measurement system |
| 发表期数： | 2019年10月第5期 |
| 引用格式： | 陈鑫磊，王韶. 基于在线自适应直推式支持向量机的电力系统 暂态稳定评估[J]. 中国科技论文在线精品论文，2019，12（5）：735-746. |
7
评论数 0请您登录
暂无评论
下载全文
多维论文