基于CPU-GPU异构的电力系统静态电压稳定域边界并行计算方法

CPU-GPU Heterogeneous Computing for Static Voltage Stability Region Boundary in Bulk Power Systems

为提升区域互联电力系统静态电压稳定域边界(SVSRB)的构建效率,该文以直接法为基础,提出一种基于CPU-GPU异构的静态电压稳定域边界并行计算方法。该方法首先依据SVSRB拓扑特性,基于边界追踪算法实现直接法求解鞍结分岔(SNB)点时初值的高效选取,克服直接法对初值敏感这一瓶颈;然后结合CPU-GPU异构平台,将直接法求解SNB点计算量较大、计算耗时占比高的修正量求解部分由GPU完成,其他逻辑性强但计算量较低的部分由CPU完成,以实现SNB点的并行求解,降低直接法计算量大、计算复杂度高的不足,从而提升SVSRB的搜索效率;最后以WECC3机9节点测试系统,波兰电网2737节点和3120节点测试系统,欧洲电网7092节点、9241节点、11624节点和13659节点测试系统算例对该文所提方法进行了分析与验证,结果表明所提并行计算方法可实现电力系统静态电压稳定域边界的快速、准确搜索。

To enhance the computational efficiency of the static voltage stability region boundary(SVSRB)in bulk power systems,a CPU-GPU heterogeneous computing technique is developed in this paper to parallel explore the SVSRB by using the direct method.According to the topological characteristics of SVSRB,a boundary tracking based direct method is proposed to explore the saddle node bifurcation(SNB)on the SVSRB by using the adjacent explored SNB as the initial point which significantly copes with the initial values choice of direct method.Then,the CPU-GPU heterogeneous computing is employed to accelerate the computational efficiency of direction method in which the mismatches of power flow are parallel solved in the GPU and the rest is solved in CPU.This reduces the computational time cost of direction method in SVSRB tracking.The WECC 9-bus,Polish 2737-bus,Polish 3120-bus,European 7092-bus、European 9241-bus、European 11624-bus and 13659-bus test systems are used to evaluate the performance of the proposed method.The results confirm that the proposed CPU-GPU heterogeneous computing based direct method performs great computational accuracy and efficiency for exploring the SVSRB in the bulk power grids.