Category: Physics

Summary: Finding how unevenly fast support can be distributed across islands before an otherwise adequately supported microgrid loses stability.

Average support levels are not the whole story in islanded microgrids, because strongly uneven allocation can create weak spots even when the mean looks adequate. This experiment asks how much dispersion in fast-support allocation a modular islanded grid can tolerate before its small-signal stability fails.

The model constructs dense swing Jacobians with fixed mean support and then uses GPU iterative deepening to bisect the support-dispersion threshold. Stability is determined from the leading eigenvalues of a non-symmetric system matrix, turning the question into a direct threshold estimate rather than a broad parameter survey.

That emphasis isolates heterogeneity as the control variable. The value is in distinguishing whether instability comes from insufficient total support or from distributing that support too unevenly across the islands.

Method: GPU dense non-symmetric eigensolves with iterative deepening and bisection on support-allocation dispersion at fixed mean support.

What is measured: Critical dispersion threshold, leading-eigenvalue stability boundary, localization statistics, and bracket width.