An optimal scheduling method for power data centre arithmetic resources based on edge computing

Traditional power scientific computing acceleration is mostly carried out in a single server or a simple server cluster, which is limited by computing resources and acceleration performance. Therefore, in this paper, the optimal scheduling of arithmetic resources from complex power scientific computing tasks to data centres is investigated. Firstly, through in-depth analysis of several important power science computing solution models, we extract their common acceleration kernel, i.e., the decomposition-coordination computing paradigm originated from the node conductivity matrix, in order to construct a new model of multi-task computation-communication equilibrium directed acyclic graph corresponding to it. Following the analysis of this task flow directed acyclic graph and the characteristics of the data centre network architecture, the proposed minimum cut binary tree algorithm is used to prioritize the mapping of the virtual machines where the key tasks are located to the physical machines, taking into account the computational time complexity of the tasks at each node in the task flow directed acyclic graph and the weight of the data interactions between the tasks, in order to minimize the overall solution time of the scientific computation. After the research on CloudSim data centre simulation platform, we conducted a series of hybrid running experiments on scientific computing for IEEE 118 to case 13659_pegase very large-scale power systems. The experimental results show that our algorithms perform well in both lightly or heavily loaded data centre scenarios, and effectively improve the computational acceleration performance by increasing the physical machine resource utilization and reducing the data centre network throughput. These findings provide important theoretical guidance and technical support for solving the data centre real-time solution of complex power scientific computing for large-scale power grids.