矢量化动态最优潮流计算的步长控制内点法实现

Step-controlled Primal-dual Interior Point Method Implementation for Vectorial Dynamic Optimal Power Flow Calculation

实现动态最优潮流(dynamic optimal power flow,DOPF)的矢量化计算。通过将同类型、同时段的优化变量集中排列,建立动态最优潮流的矢量化模型,并采用步长控制内点法进行求解。各时段的梯度矩阵和海森矩阵具有与导纳矩阵相关的稀疏特性,在计算过程中保持不变。通过设计稀疏矩阵结构和内存分配策略提高Karush-Kuhn-Tucker(KKT)系统的形成速度。分析爬坡约束和购电量合同约束对求解KKT系统的影响,对比测试多个优化排序算法,指出近似最小度(approximate minimum degree,AMD)和列近似最小度(column approximate minimum degree,COLAMD)算法求解该模型KKT系统具有很高的效率。对节点数从14到1040共5个测试系统12～96时段的DOPF模型进行仿真计算,验证所提算法的正确性和高效性。基于步长控制内点法的矢量化方法提高了DOPF程序的计算速度和收敛性。

A vectorial implementation of dynamic optima/ power flow （DOPF） was presented. The vectorization of DOPF was established by arranging the control variables and state variables according to the variable types and time intervals. A step-controlled primal-dual interior point framework with upper and lower inequality constrains was used to solve this DOPF model. The gradient and Hessian matrices of each time interval had relative non-zeros position with the admittance matrix, which was constant during iterations. Hence a sparse data structure and memory allocation strategy was utilized to accelerate the construction of Karush-Kuhn-Tucker （KKT） system. The effect of ramping rates and generation contract constrains on solving KKT system was analyzed. Through computation statistics, it is confirmed that approximate minimum degree （AMD） reordering algorithm is most efficient with only ramping rate constrains, and column approximate minimum degree （COLAMD） reordering algorithm is most efficient with both ramping rate and generation contract constrains. Numerical simulations on test systems ranging in size from t4 to 1 040 buses over 12-96 time intervals validate the correctness and efficiency of the proposed method. Vectorization technique with step-controlled primal-dual interior point method improves the calculation speed and convergence performance of DOPE