Summary

  • Ulrich Speidel’s APNIC account separates three jobs often sold as one: generation by panels, duration from batteries, and uninterrupted passage through a grid transition.
  • His grid-connected, islandable design still needs a UPS for equipment that cannot lose power when the system disconnects from or returns to the grid.

A solar brochure likes a photograph of panels in sunlight. A network operator needs to know what happens in the seconds after the grid drops. In his 4 November APNIC article, Ulrich Speidel makes the unglamorous point: an islandable photovoltaic system and an uninterrupted power supply perform different jobs. Buying one does not silently include the other.

Speidel, a University of Auckland computer-science lecturer, begins before the purchase order. He asks for a list of everything the site actually powers: servers, switches, routers, radios, UPS equipment, cooling, lighting and any loads hidden in the same circuit. Peak demand defines what the system must deliver at once; average demand helps describe what it must sustain. A rooftop panel’s peak wattage is a laboratory rating, not a promise about a shaded, warm or cloudy afternoon. A quote built around the panel number while ignoring the site’s load is already answering the wrong question.

He then distinguishes a standalone solar island from a grid-connected installation that can detach during an outage. In the latter, panels, batteries and grid can all supply power; the load, batteries and grid can all receive it. That flexibility can put surplus generation to work while the grid is healthy. It also creates a handoff. Speidel explains that the system synchronizes with the grid’s AC frequency and, when it islands or reconnects, has a short inverter shutdown. The UPS must carry equipment that cannot tolerate that interruption.

The operational stakes are plain. A battery can hold enough energy for hours and still fail to cover a brief break in delivery. A router that restarts can interrupt a service before anyone asks whether the batteries were full. Speidel has published a design argument, not a measured failure at a named network site. The missing record is an instrumented transition test showing inverter interruption, UPS ride-through and actual load continuity. Until a buyer has that record, the strongest claim on the invoice is capacity, not continuity.

His warning against treating solar as an idle backup also has an incentive behind it: equipment that never works while grid power is available earns little toward its cost. Tariff and export terms will decide the economics at a particular site. The common decision is narrower and harder: inventory the load, model the generation and storage, then prove the handoff that keeps the network alive.