CAJ | 학술논문

随着超大规模区域互联电网的发展，智能电子设备和相量测量单元广泛应用，如何实现对所产生的PB级大数据的高速处理成为完成实时(超实时)计算的关键。云计算作为一种新型的互联网计算模式，为实现电力系统大数据分析和复杂电网高效并行计算提供了可能。针对电力系统基本计算单元对角加边模型(block bordered diagonal form，BBDF)和分解协调并行算法，提出一种低能耗数据中心的优化映射和并行计算方法。依据任务间计算耦合性，将分解协调并行算法进行拆分，并提出依据任务计算复杂度的任务到虚拟机偏好绑定放置方法。随后建立以虚拟机的CPU利用率、内存利用率为约束条件，以节能为目标的Bin-Packing模型，求解 BBDF 分解协调并行计算到数据中心映射的最优配置。通过 CloudSim 平台对 IEEE 118节点电网模型和含有538节点和1133节点的大规模电网进行仿真计算。结果表明，应用虚拟机技术的数据中心计算在时间和系统能耗方面都优于传统单机多线程并行计算。IEEE 118节点算例计算时间降低42.32%，随着系统规模增大，1133节点实际电网计算时间降低75.8%。
수착초대규모구역호련전망적발전，지능전자설비화상량측량단원엄범응용，여하실현대소산생적PB급대수거적고속처리성위완성실시(초실시)계산적관건。운계산작위일충신형적호련망계산모식，위실현전력계통대수거분석화복잡전망고효병행계산제공료가능。침대전력계통기본계산단원대각가변모형(block bordered diagonal form，BBDF)화분해협조병행산법，제출일충저능모수거중심적우화영사화병행계산방법。의거임무간계산우합성，장분해협조병행산법진행탁분，병제출의거임무계산복잡도적임무도허의궤편호방정방치방법。수후건립이허의궤적CPU이용솔、내존이용솔위약속조건，이절능위목표적Bin-Packing모형，구해 BBDF 분해협조병행계산도수거중심영사적최우배치。통과 CloudSim 평태대 IEEE 118절점전망모형화함유538절점화1133절점적대규모전망진행방진계산。결과표명，응용허의궤기술적수거중심계산재시간화계통능모방면도우우전통단궤다선정병행계산。IEEE 118절점산례계산시간강저42.32%，수착계통규모증대，1133절점실제전망계산시간강저75.8%。
With the advent of large-scale regional interconnected power grids as well as the utilization of the phasor measurement unit (PMU) and intelligent electronic devices (IEDs) in electrical power systems becoming more common, the analysis of petabyte sized data has become a primary focus in research communities. Cloud computing, which is a form of data storage on the Internet, increases the possibility to implement a tool for analyzing large datasets in power systems and parallel computing in complex grids. This paper proposed a new optimized method for the mapping of block bordered diagonal form (BBDF) and decomposition-coordination algorithms for cloud computing data centers. Based on the computational complexity of coupling between tasks, the decomposition-coordination algorithm had been split to perform different tasks, and to judge the amount of calculation. A binding placement algorithm also was presented as a method to map the tasks into virtual machines (VM). A new energy-efficient Bin-Packing model was built for the final mapping step, which is the process of transferring the data from the VMs to the data centers. This will be performed while ensuring that the constraints of CPU and memory utilization rate are in check. IEEE 118 node grid model, as well as two large-scale power systems, which utilizes 538 and 1133 nodes systems, were calculated through the CloudSim platform. The results suggest that data centers using the virtual machine technology are more effective than the use of traditional parallel computing methods in terms of time and system energy consumption. In the IEEE 118 nodes system, the total <br> processing time has reduced to 43.2%, and with the system size increasing, the 1133 nodes system experienced a 75.8%decrease.