Heterogeneous Weighted Graph Partitioning and Decoupling Optimization Strategy for Electromagnetic Transient Parallel Simulation of AC/DC Distribution Networks

• • The empirical computational cost model, built on floating-point operation counts of matrix LU decomposition, achieves a simulation time fitting goodness-of-fit R² > 0.97 across three large-scale AC/DC distribution network test cases (IEEE 34-node, IEEE 123-node, European LV). This high predictive accuracy enables reliable pre-simulation estimation of parallel overhead, eliminating trial-and-error resource allocation in engineering deployment. • • Under the optimal partition configuration determined by the proposed screening strategy, the parallel speedup ratio reaches 12.11–15.77, substantially exceeding conventional natural partitioning schemes. This performance level directly translates to reduced simulation wall-clock time for transient stability assessment, allowing grid operators to evaluate high-penetration power electronic scenarios within operational planning windows. • • The heterogeneous weighted graph model maps matrix dimensions of diverse device models (converters, photovoltaics, energy storage) to node weights, addressing the device heterogeneity that causes load imbalance in traditional degree-based partitioning. This graph-theoretic representation captures computational complexity differences that prior methods ignored, preventing partition bottlenecks that degrade parallel efficiency. • • The multi-objective partitioning scheme simultaneously minimizes tie-line variable count and balances partition computational overhead, a dual consideration absent in transmission-grid-oriented partitioning algorithms. For AC/DC distribution networks with short lines and tight electrical coupling, this joint optimization reduces both per-step communication overhead and serial bottleneck time in the boundary coordination equation solve.