Why a data-quality consultation is the interesting one
Most energy consultations ask what should be built. This one asks what should be known, which on an isolated grid is the harder question. The SWIS has no neighbour to lean on. When rooftop solar swings the load there is no interconnector to absorb the error, so the accuracy of the forecast is not an administrative matter. It is the operating margin.
80.8%
The highest share of the implied underlying-demand proxy met by rooftop solar at its daily peak, across the retained AEMO WEM operating record for the SWIS.
The retained operating record makes the stake concrete. Across 1,019 retained AEMO WEM operating days, rooftop photovoltaic output at its daily peak has averaged 52.4 per cent of the implied underlying-demand proxy and reached 80.8 per cent at its maximum. On the steepest day in that record, operational demand climbed 2,412 MW from the daily minimum to the evening peak as that solar output fell away.
| The operating stake | Reading | Basis |
|---|---|---|
| Rooftop PV at daily peak, average | 52.4% | Share of implied underlying demand |
| Rooftop PV at daily peak, maximum | 80.8% | Share of implied underlying demand |
| Largest retained evening ramp | 2,412 MW | Daily minimum to evening peak |
| Retained operating days | 1,019 | 2 Oct 2023 to 17 Jul 2026 |
A system where four fifths of daytime demand can be met by generation the operator does not directly meter, followed by a ramp of that size, is a system whose forecasting depends entirely on the quality of its distributed energy data. That is what these seven proposals are about.
What is being proposed
Energy Policy WA groups seven proposals under three principles. The issues paper carries the full drafting; the summary below is in plain terms.
Principle one, increasing the quality of standing DER data. Require all inverters up to 30 kVA to be CSIP-AUS enabled to the SA TS 5573:2025 technical specification, extending an obligation that currently sits with Synergy to all retailers. Establish formal Western Power functions to gather, verify and distribute distributed energy data for demand forecasting. Collect electric vehicle charger data indirectly, balancing what the system needs to know against keeping connection simple.
Principle two, making it easier for businesses to connect. Streamline data requirements for larger installations, enabling a single connections process for all systems up to 1 MVA. Require Western Power to publish additional distribution network capacity detail, so projects can be sited against real constraint rather than guesswork.
Principle three, understanding what virtual power plants can do. Examine mechanisms for virtual power plant operators to supply capacity information to AEMO, and link Western Power's network capacity publishing to virtual power plant customer opportunities.
| Principle | Proposals | What it changes |
|---|---|---|
| Standing data quality | 1 to 3 | What is known about a device at connection, and who holds it |
| Easier business connection | 4 and 5 | One process to 1 MVA, and published network capacity |
| VPP grid mitigation | 6 and 7 | What an aggregator tells the market operator |
Read together, the first five are a single argument: the operator cannot forecast what it cannot see, and it currently cannot see enough. The last two extend that from visibility to dispatchable response.
Where the public record breaks today
We build the SWIS distributed energy picture from public sources, and the places where that work is hardest are, we think, the places this consultation should aim at. Three are worth stating plainly, because each is a limitation we have had to engineer around rather than a theoretical concern.
Installation records arrive by postcode, not by network location. Clean Energy Regulator small-scale records are published against postcodes. Postcodes are a postal artefact and align with neither statistical geography nor network topology. To place solar and battery records against 246 local areas we allocate each postcode fractionally across statistical areas using ABS residential mesh-block shares. That is a defensible method and it is still an estimate. Proposal 2 is the one that could remove the need for it.
The most recent year of installation data is always the least complete. Small-scale certificates may be created up to twelve months after installation, so recent counts mature upward for a year after the fact. Any comparison of a recent window against an earlier one understates the recent period. In our own cohort the earlier twelve-month window holds 33,498 records against 31,021 in the most recent window to May 2026, and that gap is not a demand signal. It is a lag artefact, and a forecasting function built on unadjusted recent counts inherits it.
| Cohort window | Records | Caution |
|---|---|---|
| Earlier 12 months | 33,498 | Substantially matured |
| Most recent 12 months, to May 2026 | 31,021 | Still maturing; certificates may be created up to 12 months late |
Capacity is published where constraint is what matters. What a project developer needs, and what Proposal 5 gestures at, is the constraint surface underneath: where hosting capacity binds, at what time of day, and how that moves as adoption grows. A published list of opportunities is a filtered view of that. The underlying series is the more useful disclosure.
What we would say in a submission
Three points follow, and each is narrow enough to be actionable.
Publish the allocation, not only the aggregate. If Western Power takes on a verification and distribution function under Proposal 2, the highest-value output is distributed energy data already resolved to network location, with the method published. That removes an estimation step for every downstream user and makes independent replication possible.
Publish maturity, not only counts. Any installation series intended for forecasting should carry an explicit vintage or completeness flag, so a user can tell a genuine slowdown from certificate lag. This is cheap to add at source and expensive to reconstruct afterwards.
Treat the capacity series as the product. Proposal 5 is stronger if what is shared is the time-varying constraint series rather than a curated list of opportunities, with the same versioning discipline the rest of the framework is getting.
Submissions go to the address published on the consultation page by 5pm AWST on Tuesday 18 August 2026. This report explains the proposals; it is not a submission, and it does not speak for any party that makes one.