| 引用本文: | 郑逸凡,周赣,傅萌,王子恒,冯燕钧.基于图形处理器加速的电网全网拓扑分析算法[J].电力自动化设备,2022,42(5): |
| ZHENG Yifan,ZHOU Gan,FU Meng,WANG Ziheng,FENG Yanjun.Global power network topology analysis algorithm based on GPU acceleration[J].Electric Power Automation Equipment,2022,42(5): |
|
| 摘要: |
| 随着电网规模的扩大以及电力系统互联水平的提高，电网分析计算的计算量显著上升。图形处理器(GPU)等并行计算设备能够提高大电网分析的计算效率，但电网分析计算中的串行电网拓扑分析已经成为影响并行算法计算效率的重要因素，亟待得到加速。作为电网拓扑分析加速问题的难点，全网拓扑分析可采用并行算法进行加速。为此，提出了一种基于GPU加速的电网全网拓扑分析算法：将电力系统网架模型抽象为利于并行解耦调用的压缩数组形式；以前驱数组法为基本思想，设计了优化并行的厂站拓扑分析方法，以从电网特性中挖掘厂站间及节点间的并行性；以此为核心，提出了基于GPU加速的电网全网拓扑分析算法。算例的测试结果表明，相比传统串行算法、多核中央处理器并行算法，所提基于GPU加速的电网全网拓扑分析算法运用于规模较大的电网时能分别取得7.242倍、2.148倍的加速比，显著加快了大电网全网拓扑分析的速度，进而提高了各类电网分析计算的整体效率，具有潜在的工程应用价值。 |
| 关键词: 大电网 厂站拓扑分析 并行计算 图论 图形处理器 |
| DOI：10.16081/j.epae.202202027 |
| 分类号:TM744 |
| 基金项目:国家自然科学基金资助项目(51877038) |
|
| Global power network topology analysis algorithm based on GPU acceleration |
|
ZHENG Yifan1, ZHOU Gan1, FU Meng1, WANG Ziheng2, FENG Yanjun1
|
|
1.School of Electrical Engineering, Southeast University, Nanjing 210096, China;2.China Electric Power Research Institute, Nanjing 210003, China
|
| Abstract: |
| With the expansion of power grid scale and the improvement of interconnection level of power system, the calculation amount of power grid analysis increases significantly. GPU(Graphics Processing Unit) and other parallel computing devices can improve the calculation efficiency of large power grid analysis, but the serial power network topology analysis has become an important factor affecting the calculation efficiency of parallel algorithms, which needs to be accelerated. As the difficulty of power grid topology analysis acceleration, parallel algorithm can be used to accelerate the global network topology analysis. Therefore, a global power network topology analysis algorithm based on GPU acceleration is proposed. The power system grid model is abstracted into a compressed array form, which is convenient for parallel decoupling and invocation. Taking frontier array method as the basic idea, an optimized parallel substation topology analysis method is designed to explore the parallelism among substations and nodes from power grid characteristics. Based on this, a global power network topology analysis algorithm based on GPU acceleration is proposed. The test results of examples show that compared with the traditional serial algorithm and multicore central processing unit parallel algorithm, the proposed global power network topology analysis algorithm based on GPU acceleration can achieve 7.242 times and 2.148 times acceleration ratio respectively when applied to large-scale power grid. The proposed algorithm significantly speeds up the global topology analysis of large power grid and improves the overall calculation efficiency of all kinds of power grid analysis, which has potential engineering application value. |
| Key words: large power grid substation topology analysis parallel computing graph theory GPU |
