基于IRAM的大型电力系统特征值在线并行计算策略  

Parallel computing strategy of online eigenvalue calculation based on IRAM for large-scale power system

作　　者：王永昆[1] 鲍颜红[1,2] 方勇杰[2,3] 郑伟[1] 徐伟[1] 

机构地区：[1]国网电力科学研究院/南京南瑞集团公司,江苏南京210003 [2]华北电力大学电气与电子工程学院,北京102206 [3]中国电力科学研究院,北京100192

出　　处：《电力自动化设备》2014年第5期144-148,共5页Electric Power Automation Equipment

基　　金：国家电网公司大电网重大专项资助项目课题(SGCC-MPLG001-003-2012)~~

摘　　要：针对离线小干扰稳定分析无法满足电力系统时变的情况,提出了一种基于隐式重启动Arnoldi算法(IRAM)的特征值在线计算策略。该策略根据前一断面小干扰分析结果,预估当前断面特征值分布情况,建立了位移点选择的多目标优化数学模型,并采用线性加权和法求解该模型。按照预估计算量大小逐个分配计算任务,相比现有的任务分配方式,可最大限度地减少CPU闲置时间。对南方电网在线数据进行了分析计算,计算结果表明,所提策略可对小干扰稳定关注的区域内的特征值进行全面搜索,相对于平均任务分配策略和现有的不平均任务分配策略,其耗时更短。Since the offline small signal stability analysis is not suitable for the time-varying power system, a strategy of online eigenvalues calculation based on IRAM （Implicitly Restarted Arnoldi Method） is proposed,which predicts the eigenvalue distribution of present time fraction according to the small signal analysis of previous time fraction to build a muhi-target optimized model with searching point selection. The linearly weighted sum method is applied to solve the model. Calculation tasks are assigned sequentially according to their estimated calculation loads,which minimizes the idle time of CPUs. Online data of China Southern Power Grid are analyzed and calculated,which shows that,the eigenvalues in area concerning the small signal stability are entirely searched and its time consumption is shorter than that of either even or the existing uneven task assignment strategy.

关 键 词：电力系统 小干扰稳定 线性加权和 在线计算 并行计算 稳定 特征值 

分 类 号：TM712[电气工程—电力系统及自动化]