Security Analysis and the SAR

gridoxide::security_analysis (--features cgmes) takes a CGMES model and a decoded CO dataset, solves the base case and each contingency, holds each terminal's current against its CurrentLimits, and writes what it finds as a dataset of the SecurityAnalysisResult (SAR) profile.

import gridoxide

analysis = gridoxide.security_analysis(profiles, co_files)      # sharing="capability"
print(analysis.solved, analysis.stalled, analysis.violations()[:1])
xml = analysis.to_sar(publisher=..., is_version_of=..., rights_holder=...,
                      generated_at=..., title=..., description=...)
gridox cgmes-security profiles/*.xml --contingencies co/*.xml --sar out.xml \
    --publisher <iri> --version-of <iri> --rights-holder <name> --generated-at <time>

The building blocks are public: read_contingencies, current_limits, terminal_flows, monitor, machines, capability_weights. solve solves one state as the analysis solves each: it returns the buses, each component's outcome, each TopologicalNode's bus, the terminal flows, and what the options did (the buses reactive limits took off voltage control, the shift sharing the imbalance gave each bus, the machines it took past their limits).

What it reports

A SAR holds the violations, a violating terminal's per limit exceeded, plus the base case of each terminal a contingency violates. A contingency's own equipment is not monitored: it carries nothing. PowerFlowResult.value is in percent, as the profile defines it. A contingency that does not converge has no result, as the profile has no place for one; stalled lists them.

A TATL allows more than the PATL for a while, so one at or below its PATL is not a limit, and is left out. Otherwise terminals are held against every limit they have.

Where the buses come from

SecurityOptions::topology (topology= in Python) says where the network's buses come from: the TP's TopologicalNodes (BusBranchFromTp, the default), or the switches as the SSH sets them, every closed one merged (NodeBreakerFromEq, "switches"). From the switches a switch opened or closed in the data, as a remedial action does, makes a different network; the TP, written for the switches as they were, does not see it. Everything else is the same: taps, regulation, terminal flows, limits. On ReliCapGrid's IGMs whose TP agrees with their switches (Belgovia, Britheim, Galia, Svedala) the two give the same voltages to within 1e-6 p.u.; Espheim's TP leaves equipment off that its switches connect, and from the switches one component holds two angle references.

Curative actions

run_with_curative takes, by contingency, the remedial actions to apply after it: each a Curative, its RemedialAction's mRID and the changes it makes (an object's Class.attr set to a value). The contingency is solved, then solved again with all its actions applied together; the violations of that state carry the actions in Violation::applied, and the SAR writes a RemedialActionApplied for each action on each such result. Which actions go with a contingency, and their values, the caller decides from the RA data. A switch an action sets is seen only from the switches (topology).

Who takes the imbalance

What a state schedules seldom equals what it needs, and the difference has to land somewhere. That choice can move the answer more than anything else here, and no CGMES attribute settles it:

sharingWhoNote
capability (default)the machines in service, by GeneratingUnit.maxOperatingP
referencethe angle reference the model declaresloads what is near it
equalevery generator bus alikeignores machine size

GeneratingUnit.normalPF would be the model's own statement, but in ReliCapGrid it is scaled per IGM, so it cannot share across a merged model. Generator limits do not stop a share; past_limits lists where one goes past them. None of these is a dispatch: reserves and control areas are not modelled.

Limits

  • Reactive limits are off by default (enforce_q_limits): a model whose voltage targets they cannot all hold stalls with them, in the base case and so everywhere.
  • Only current limits are monitored, not voltage or power limits.
  • Remedial actions are not applied.
  • A contingentStatus other than outOfService is an error.

See the 2026 SV-IOP for how this does on ReliCapGrid.