文章摘要
文云峰,赵荣臻,肖友强,等.基于多层极限学习机的电力系统频率安全评估方法[J].电力系统自动化,2019,43(1):133-140. DOI: 10.7500/AEPS20180629012.
WEN Yunfeng,ZHAO Rongzhen,XIAO Youqiang, et al.Frequency Safety Assessment of Power System Based on Multi-layer Extreme Learning Machine[J].Automation of Electric Power Systems,2019,43(1):133-140. DOI: 10.7500/AEPS20180629012.
基于多层极限学习机的电力系统频率安全评估方法
Frequency Safety Assessment of Power System Based on Multi-layer Extreme Learning Machine
DOI:10.7500/AEPS20180629012
关键词: 频率安全  极限学习机  低惯性系统  机器学习  人工智能  大数据
KeyWords: frequency safety  extreme learning machine(ELM)  low inertia system  machine learning  artificial intelligence(AI)  big data
上网日期:2018-11-09
基金项目:国家自然科学基金资助项目(51707017);重庆市基础科学与前沿技术研究项目(cstc2017jcyjAX0422);中央高校基本科研业务费专项资金资助项目
作者单位E-mail
文云峰 湖南大学电气与信息工程学院, 湖南省长沙市 410000 yunfeng.8681@163.com 
赵荣臻 重庆大学电气工程学院, 重庆市 400044  
肖友强 云南电网规划建设研究中心, 云南省昆明市 650011  
刘祯斌 中国建筑第七工程局有限公司, 深圳市 518116  
摘要:
      可再生能源发电的随机性、间歇性和低惯性特征导致含可再生能源电力系统的频率安全问题凸显。利用时域仿真进行频率安全评估存在计算量大、耗时长等缺陷,难以满足多重复杂不确定因素“组合数爆炸”下的频率安全快速评估需求。为了实现频率安全的快速分析与预测,提出一种基于多层极限学习机(ML-ELM)的频率安全在线评估方法。该方法通过深层架构建立输入与输出之间的非线性映射关系,并在自下而上的逐层无监督训练过程中,引入自动编码器算法和正则化系数,逐层优化输入层与隐含层之间的权重矩阵,以使ML-ELM有效表征复杂函数、提高预测精度和泛化能力。在IEEE RTS-79系统上开展算例测试,将测试结果与时域仿真和浅层神经网络方法所得结果进行比较,验证了所提方法的准确性和泛化能力。
Abstract:
      The random, intermittent and weak inertia characteristics of renewable energy generation have led to a prominent problem in the frequency safety of high-rate renewable energy power systems. The use of time-domain simulation for frequency safety assessment has the disadvantages of large amount of calculation and long time. It is difficult to meet the rapid assessment requirement of frequency safety under the “combined explosion” of multiple complex uncertainties. In order to realize online analysis and prediction of frequency safety, a method based on multi-layer extreme learning machine(ML-ELM)is applied. The non-linear mapping relationship between the input layer and the hidden layer is built by the deep structure theory and in the layer-wise unsupervised training, automatic encoder algorithms and regularization coefficients are introduced to optimize the weight matrix between the input layer and the hidden layer, so that the ML-ELM can effectively represent complex functions and improve predictive accuracy and generalization ability. Case studies of the IEEE RTS-79 system demonstrate the rapidity, high accuracy and well generalization ability of the proposed method.
查看全文(Free!)   查看附录   查看/发表评论  下载PDF阅读器