Protection Channel Competition and Network Feedback in Correlated Converter Disconnections

After a fault on a weak network is cleared, several grid-following converters sharing a common bus can disconnect in quick succession. Which units disconnect, and through which protection, depends on competing AC and DC overvoltage functions, on the memory of measurement filters and timers, and on the voltage change caused by the first disconnection; a voltage-peak check or a prescribed outage list cannot capture these interactions. This paper proposes a model-specific diagnostic framework that carries converter states, filtered measurements, timer ages, and connection states from one protection event to the next and explains the outcomes by three conditions: the voltage change caused by a disconnection, the completion or reset of a pending timer, and the DC energy remaining after blocking. All combinations of bounded AC setting changes are re-simulated in closed loop on a three-converter positive sequence benchmark. Raising AC thresholds never increased the number of disconnected units, which is a structural property of this benchmark rather than a physical guarantee. However, in 830 of 23,328 single-setting comparisons, it shifted the disconnection to another unit and in 40 from the AC to the DC protection of the same unit; separate setting changes did not have additive effects. A fixed-topology relay evaluation identified different disconnected units in 613 of 5832 runs, which demonstrates a logical necessity check within the same equations rather than evidence that the full model is physically more accurate. Under the averaged model, the relief given by the first disconnection depends strongly on the assumed current interruption law. The findings lack EMT or hardware validation and yield neither protection settings nor outage probabilities.