最优潮流的原对偶内点法矢量化实现

Vectorization Primal-dual Interior Point Implementation of Optimal Power Flow

为提高计算速度,采用矢量化技术实现最优潮流计算。通过将同类型的优化变量集中排列,建立最优潮流模型的矢量化表达形式。采用原对偶内点算法求解该模型,建立梯度矩阵及海森矩阵线性组合的矢量化计算公式。求解修正方程时,对系数矩阵进行近似处理,对修正方程系数矩阵采用LDLT算法进行分解。采用近似最小度(AMD)算法对系数矩阵进行排序,减少分解所产生的注入元。基于C/C++开发电力系统矢量运算支持库,设计动态稀疏存储策略进一步提升最优潮流程序的计算速度。对多个测试系统进行仿真计算表明:矢量化可简化最优潮流的程序逻辑并提高程序运行速度。

Vectorization technology is used to improve the efficiency of optimal power flow （OPF） implementation. The vectorization expression of OPF is established by arranging the control variables and state variables according to the variable type. In solving the OPF model with primal-dual interior point method,the gradient matrix and the vectorization expression of Hessian matrices linear combination are setup. The coefficient matrix of correction equation is dealt approximately; hence correction equation can be solved by LDLr decomposi- tion. Approximate minimum degree （AMD） reordering algorithm is used to reduce the fill-in elements in LDLr decomposition. The power system vectorization calculation library is developed based on C/C＋ ＋. The dynamic sparse storage strategy in this library is utilized to improve calculation speed. Numerical simulations on test systems confirm that vectorization simplifies the OPF implementation and saves execution time.