中国电机工程学报
中國電機工程學報
중국전궤공정학보
ZHONGGUO DIANJI GONGCHENG XUEBAO
2015年
17期
4318-4325
,共8页
田鹏飞%刘崇茹%贠飞龙%洪国巍
田鵬飛%劉崇茹%贠飛龍%洪國巍
전붕비%류숭여%원비룡%홍국외
隐式重启动Arnoldi算法%关键特征值%动态扩维%搜索圆
隱式重啟動Arnoldi算法%關鍵特徵值%動態擴維%搜索圓
은식중계동Arnoldi산법%관건특정치%동태확유%수색원
implicitly restarted Arnoldi method%critical eigenvalues%dynamically increasing the dimension%searching circle
文章对带位移逆变换的隐式重启动Arnoldi(implicitly restarted Arnoldi,IRA)算法进行了改进,提出了动态增加Krylov 子空间维数求取指定圆内特征值的方法。论文从隐式重启动机理出发,在锁定已收敛特征值的基础上,动态增加特征值个数和子空间维数,扩大搜索圆的半径,实现指定大小的搜索圆内部所有特征值的有效求解。进而将电力系统关键特征值所在区域按一定规则分割,分割得到的所有小区域利用搜索圆覆盖求解。该方法不需人工干预,并行实现后效率更高,且搜索机制规避了漏解的现象。最后,状态空间为570阶和5272阶的电力系统的关键特征值计算结果表明,所提方法不仅高效,且可靠实用。
文章對帶位移逆變換的隱式重啟動Arnoldi(implicitly restarted Arnoldi,IRA)算法進行瞭改進,提齣瞭動態增加Krylov 子空間維數求取指定圓內特徵值的方法。論文從隱式重啟動機理齣髮,在鎖定已收斂特徵值的基礎上,動態增加特徵值箇數和子空間維數,擴大搜索圓的半徑,實現指定大小的搜索圓內部所有特徵值的有效求解。進而將電力繫統關鍵特徵值所在區域按一定規則分割,分割得到的所有小區域利用搜索圓覆蓋求解。該方法不需人工榦預,併行實現後效率更高,且搜索機製規避瞭漏解的現象。最後,狀態空間為570階和5272階的電力繫統的關鍵特徵值計算結果錶明,所提方法不僅高效,且可靠實用。
문장대대위이역변환적은식중계동Arnoldi(implicitly restarted Arnoldi,IRA)산법진행료개진,제출료동태증가Krylov 자공간유수구취지정원내특정치적방법。논문종은식중계동궤리출발,재쇄정이수렴특정치적기출상,동태증가특정치개수화자공간유수,확대수색원적반경,실현지정대소적수색원내부소유특정치적유효구해。진이장전력계통관건특정치소재구역안일정규칙분할,분할득도적소유소구역이용수색원복개구해。해방법불수인공간예,병행실현후효솔경고,차수색궤제규피료루해적현상。최후,상태공간위570계화5272계적전력계통적관건특정치계산결과표명,소제방법불부고효,차가고실용。
On the basis of implicitly restarted Arnoldi (IRA) method, a method, in which the dimension of Krylov subspace is dynamically increased to compute eigenvalues in a specified circle, is proposed. In the implementation process of the proposed method, firstly, the radius of searching circles is dynamically expanded through increasing the number of eigenvalues and the dimension of Krylov subspace, according to the mechanism of implicitly restarting. Secondly, the region where the target eigenvalues located is divided into small computing units, which are covered by specified searching circles. The proposed method has higher efficiency after parallel implementation and avoids eigenvalues missing by searching mechanism compared with the existing methods. Furthermore, no manual intervention is needed in the proposed method. Two systems with 570 and 5 272 state variables are tested in this paper, and the results indicate that the proposed method is efficient, reliable and practical.