Networks and coverage

The API covers 15 electricity networks across 8 states and territories: New South Wales, Victoria, Queensland, Western Australia, South Australia, Tasmania, Australian Capital Territory and Northern Territory. Each network publishes its outages differently — JSON, a map layer, a web page — and each leaves different things out. The table shows, for each one, how often its outages carry each field.

Shares are the proportion of each network’s recorded outages that carry the field, from GET /v1/networks as at 20 Oct 2026. Ask the endpoint for today’s figures.

A share of none means the network does not publish that field at all, and the API returns null for it on every one of that network's outages. That is the network's gap, not a fault in the data, and it is why GET/v1/networks reports these shares live: read them before comparing networks.

What each network does and does not publish

Victoria

• AusNet Services publishes points, not areas — an approximate location for each outage, labelled as such everywhere. It publishes no cause. Its feed also lists cancelled and forward-dated planned work for months, so most of what it publishes at any moment is not a live outage; status=active is what is off. It is the only Victorian network that routinely publishes planned work ahead of time.
• CitiPower / Powercor is one feed for two networks; network.brand says which. It lists current outages, so planned work appears once it is under way rather than ahead of time.
• Jemena publishes areas and named streets, and keeps completed outages in its feed for a short while. Restoration estimates are on a minority of its outages.
• United Energy publishes no start time and no status. Every outage it lists is current, so a listed outage is active and is restored when it leaves the feed; starts_at is null, and time filters use when we first saw it. Customer counts are per outage, with no split by locality.

New South Wales

• Ausgrid publishes points and town names without postcodes, and lists planned work up to a week ahead.
• Endeavour Energy publishes areas for planned work. Its faults are published as tiny circles, eleven to twenty-two metres across, which are recorded as points. It lists planned work ahead.
• Essential Energy serves regional New South Wales and southern Queensland; each outage carries the state it is in. It publishes areas and no postcodes, lists only current outages and publishes no status — so a cancelled Essential outage cannot be told apart from a restored one.

Queensland

• Energex and Ergon Energy Network publish areas for faults and points for planned work, with town names but no postcodes and no split of customers by town. Ergon publishes no cancelled status. Both feeds sit behind a bot filter that occasionally refuses a poll; the odd failed poll is normal for them and costs nothing but a few minutes' freshness.

South Australia

• SA Power Networks publishes areas, town names with postcodes, and planned work about a fortnight ahead. It publishes no cancelled status and no restoration time, so restorations are observed. A planned job that has finished stays scheduled for a few hours, until the day rolls over, rather than being guessed restored from its window.

Tasmania

• TasNetworks publishes areas. It publishes no start time field, so the time the network created the record stands in. Its restoration estimate is usually "to be confirmed", which is null. A brand-new fault can appear briefly under a placeholder reference before the network assigns its real one; that shows as one short outage restored and a new one beginning.

Western Australia

• Western Power publishes areas, the only per-suburb split of customers that sums exactly, and planned work ahead. It publishes no cause in its feed, and no cancelled status or restoration time, so restorations are observed.
• Horizon Power publishes points whose positions are rounded to about one to eleven kilometres — never treat one as an address — and no cause.

Northern Territory

• Power and Water publishes a web page, not a data feed. It carries no customer counts, no status and no year on its dates; the year is worked out from the weekday, and the status from the clock. Locations are the centres of the suburbs it names, so they are points. The same page lists water outages, which are filtered out of the electricity data. Planned work is closed at its scheduled end, so a job that overran is recorded as finishing on time.

Australian Capital Territory

• Evoenergy publishes the richest record of any network: areas, customer counts, a real status, and separate scheduled, actual and expected times, so its restorations are reported. It publishes no postcodes, and a cause only for planned work.

Feed health

GET/v1/networks reports each feed's status right now:

While a feed is stale or failing, the API keeps returning that network's last recorded outages. A quiet network with a failing feed is not a network with no outages.

A degraded feed's data is current, but it has gaps. We close an outage at the first successful poll after it leaves the feed, so while polls are failing a restoration can be recorded late, and an outage that starts and ends between two successful polls may not be recorded at all.