A distribution utility we worked with had the problem every DISCOM now shares: rooftop solar was arriving faster than engineering could study it. Each application — can this array connect here? — meant pulling feeder data, building a model, running a load-flow, checking thermal limits, and imagining the worst-case outage. The work was sound. It just didn’t scale, and the queue grew whether or not anyone was working it.

The goal wasn’t to remove engineers from the loop. It was to stop making them redo the same solve by hand for every routine application, so their judgement landed on the cases that actually needed it.

The setup

We loaded the utility’s network through the CIM adapter and stood up a digital twin per feeder using the pandapower bridge. Current grid state came from the connectors already in place — GIS for geometry and topology, SCADA and AMI for measurements, DERMS for the existing solar and battery inventory. No new data platform; the twin read what the utility already had.

The screen

Each pending application became an intervention study. The twin injected the proposed DER and ran the three checks that decide feasibility:

  • an AC power-flow solve for voltage-band violations at every affected node;
  • an IEEE 738 thermal check for conductor overload under realistic weather and load;
  • an N-1 contingency pass so approvals held even with a neighbouring element out.

Clean applications came back feasible with the supporting numbers. Marginal ones came back with a precise violation — which bus, which conductor, under which contingency — and a scoped upgrade instead of a flat “no.”

The engineers didn’t review fewer studies carefully. They stopped reviewing the ones that were never in doubt.

The outcome

Routine, clearly-feasible applications cleared automatically, with a governed record and an audit trail. Only the genuinely constrained feeders reached a human, already scoped. The interconnection queue stopped being a backlog measured in months and became a throughput the utility could plan around — and every decision, feasible or not, left a replayable trail for the regulator.

Hosting capacity didn’t get easier; it got automated. The same physics, the same checks, the same standards an engineer would apply — just run on every application, the moment it arrives, instead of whenever the queue allowed.

Key takeaways

  • The twin read the utility’s existing CIM topology and live connector data — no new platform.
  • Each application ran voltage, IEEE 738 thermal and N-1 checks automatically.
  • Feasible cases cleared with evidence; constrained cases arrived pre-scoped to engineers.
  • The queue shifted from a months-long backlog to a governed, auditable throughput.