新能源高占比电力系统电磁暂态并行仿真的优化分网方法

Optimal grid-division method of electromagnetic transient parallel simulation for high proportion of renewable energy power systems

为提升含大规模多类型电力电子设备的大型电力系统电磁暂态仿真效率,提出基于层次聚类预分网、预分网下的聚合分网优化问题数学建模和基于改进非支配排序遗传算法(NSGA-Ⅱ)优化求解的并行仿真优化分网框架及方法。首先基于层次聚类网络约减对大型系统进行化简,降低原始系统的复杂程度。然后基于典型仿真耗时实测对元件相对计算量进行统一表征,将预分网后的聚合分网问题转化为多目标优化问题。并通过改进NSGA-Ⅱ得到优化分网的Pareto解集,再基于加权欧氏距离最小值法获得最优分网方案。最后通过对比同一算例系统下不分网、基于Metis工具的分网以及所提方法的分网结果,验证了所提方法的有效性。

In order to improve the efficiency of electromagnetic transient simulation of large-scale power system with large-scale and multi-type power electronic equipment,a parallel simulation optimal grid-division framework and method based on hierarchical clustering grid pre-division,mathematical modeling of aggregation grid-division optimization problem under pre-division and optimization solution based on improved nondominated sorting genetic algorithm-Ⅱ(NSGA-Ⅱ)are proposed.Firstly,the large-scale system is simplified based on hierarchical clustering grid-reduction to reduce the complexity of the original system.Then,based on the time-consuming actual measurement of typical simulation,the relative calculation of components is characterized in a unified way,and the aggregation grid division problem after grid pre-division is transformed into a multi-objective optimization problem.The Pareto solution set of optimal grid-division is obtained by improving NSGA-Ⅱ,and then the optimal grid-division method is obtained based on the minimum weighted Euclidean distance.Finally,the effectiveness of the proposed method is verified by comparing the results of the same example system under the situation of no grid-division,grid-division by Metis tool and griddivision by the proposed method.