Why Does the Electricity System Allow 0.002% of Energy to Go Unserved?

Geoffrey Chen

When Numbers Start Making Decisions · Season Five, “Things That Have Not Happened Yet” · Article 5

1. Reliable does not mean never short of electricity

The reliability standard for Australia’s National Electricity Market does not promise that every unit of electricity must be supplied under every imaginable condition. Clause 3.9.3C of the National Electricity Rules sets a maximum expected unserved energy, or USE, of 0.002% of total energy demanded in a region in a financial year for generation and inter-regional transmission. National Electricity Rules: clause 3.9.3C

Suppose an AEMO outlook changes a region’s estimated USE from 0.0019% to 0.0021%. Crossing the line does not automatically disconnect customers. It changes the forecast classification from expected to meet the standard to expected not to meet it. That can invite a market response and bring contracts, reserves or intervention into consideration. A movement deep in the decimal places changes present commitments to future capacity.

The current 0.002% standard is often translated into a long-term average of roughly ten minutes of generation-shortfall outages per customer per year. It is not an allocation of ten minutes to every household and not an annual outage budget. It is a regional expected proportion of demand that generation and interconnector capacity cannot supply.

An extremely small percentage can therefore influence billions of dollars in generation, batteries, transmission, demand response and reserve. It measures electricity that has not yet failed to arrive and organises investment today.

2. USE covers only one kind of outage

Power at a home can fail because a tree falls on a distribution line, local equipment breaks, fire or storm damages the network, or maintenance is underway. USE in the reliability standard principally concerns an insufficiency of generation, demand response and inter-regional transmission to meet demand.

The AEMC’s 2026 review material observed that roughly 99% of customer outages arise from network issues rather than a shortage of generation. The boundary of the indicator matters. A region can meet the USE standard while a particular distribution feeder remains unreliable. A storm can leave many customers without electricity without proving that the generation reliability standard failed.

Political debate produces bad remedies when it combines those events. Additional generation cannot repair a fallen pole. Stronger local distribution cannot by itself resolve a market-wide shortfall at the evening peak. A metric has decision value only when its object remains correctly named.

3. The standard is an economic and ethical compromise

Higher reliability generally requires more capacity, fuel, storage, transmission and contracts that may rarely be used. Consumers, taxpayers or investors ultimately pay. The standard seeks a socially valued degree of reliability rather than an imagined condition of absolute safety.

That calculation draws on the Value of Customer Reliability: what different users would pay to avoid unserved electricity. Households, hospitals, data centres, smelters and small businesses suffer different losses. A brief interruption is unlike a multi-day event. Aggregating those differences into an annual regional percentage creates a market tool but flattens the distribution of harm.

The 0.002% value is not a natural constant. The Reliability Panel reviews it within a legal and modelling framework. In April 2026, the final review recommended 0.003% expected USE for the period beginning 1 July 2028, equivalent to a long-term average of about sixteen minutes of generation-driven interruption. At the time of writing, 0.003% is a recommendation for the next regulatory period, not a revision of today’s rule. AEMC: 2026 Reliability Standard and Settings Review

Changing the threshold reallocates two possible errors: building too much backup and imposing excess cost on everyone, or building too little and carrying a larger risk of shortage in rare conditions.

4. How a forecast shortage triggers action now

AEMO forecasts USE using assumptions about demand, retirements, committed projects, fuel, weather, transmission and storage. When a forecast shows a region may exceed the standard, it signals opportunity for market investment and can prompt contracts, reserves or other policy responses.

The rules also contain an interim reliability measure of 0.0006%, a tighter supplementary trigger linked to arrangements such as the Retailer Reliability Obligation and interim reserves during the transition period. A market can therefore possess a long-run efficiency standard and a more cautious temporary line for tail risk at the same time.

These figures do different institutional work. One helps set long-term market settings; another gives policymakers additional protection during a changing system. Without an explanation of use, however, the smaller number simply appears “safer” and conceals both cost and scope.

Every threshold should travel with its purpose. Otherwise, an audience cannot tell which decimal governs market price settings, which informs retailer obligations and which reflects an interim political concern.

5. An annual expectation can hide the tail

Two futures can have the same expected USE. One contains numerous tiny, brief gaps. The other contains a rare crisis that lasts much longer. The expected total is the same, while the lived distribution is radically different.

As wind, solar, storage and weather-sensitive demand grow, heat, low renewable output, generator failures and fuel constraints can combine in tail events. The AEMC and Reliability Panel have therefore considered whether the form of the standard should communicate duration, frequency and depth more clearly rather than merely moving the decimal.

This is a validity problem. The total may be accurately calculated and still fail to express what policy fears. Spread a long shortage across annual demand and the percentage becomes tiny. For a person dependent on medical equipment or unable to escape extreme heat, the consequence does not.

Distribution across users matters too. Regional compliance does not mean equal protection. Hospitals provide backup power; many vulnerable households cannot. The common market standard must work with critical-infrastructure duties and social protections.

6. Forecasts can change the thing they predict

A published reliability gap may attract batteries, delay retirement or increase demand response. The shortage then does not occur. Calling the forecast exaggerated ignores the action it produced.

The opposite risk also exists. Announced projects can be delayed. Weather correlations may be underestimated. Demand may change quickly. A central scenario that meets the threshold is not evidence of safety in the tail.

Reports should separate committed, under-construction and merely proposed projects, show multiple weather years and retirement scenarios, and identify the assumptions that drive the result. The further the forecast reaches, the more action should be staged: preserve options and shorten review intervals before making irreversible commitments.

Post-event review should not ask only whether load was shed. It should examine the forecast available at the time, the market response, intervention cost and the risks prevented. Successful prevention is hard to see, so institutions must preserve counterfactual evidence.

7. A system standard is not an individual service guarantee

A household cannot usually challenge the 0.002% standard merely because its power failed. The standard is not a retail promise of continuous supply. Consumers, industry and governments can contest its inputs and level through reviews, consultation and rule-making.

Minimum transparency should include the definition and regional denominator of USE, forecast scenarios, project assumptions, the time and size of any gap, the expected cost of intervention and the separation between network outages and generation shortage. If the standard changes, the decision should identify who saves cost and who carries additional tail risk.

Remedies for an individual outage belong to other systems: retailer protections, network service standards, life-support customer rules and disaster response. Macroeconomic compliance cannot erase a specific service failure.

8. Reliability is partly constituted by the failure we permit

Electricity reliability is not a permanent property of one machine. It is maintained across generation, networks, weather, demand, prices, user response and reserves. The 0.002% figure compresses those relationships into a governable interface so a system without a single all-knowing operator can still invest.

A reliable system is not one without failure. It is one that makes constrained, revisable arrangements about which failures can be accepted, who bears them and how recovery works. The Reliability Panel and governments must own the compromise rather than describe it as a purely technical answer. A model can estimate shortage but cannot decide whether a rare, prolonged event is acceptable for vulnerable people.

Conclusion: permit a tiny risk without letting the average conceal a disaster

Expected USE should remain the basis of the NEM generation reliability standard. The 0.002% line supplies a common scale for investment, market settings and intervention. Attempting absolute zero shortage would be unrealistic and could impose disproportionate cost.

The next stage should not debate only 0.002 versus 0.003. It should show duration, frequency, tail scenarios and affected users, and should repeatedly explain that the standard excludes most distribution-network outages. The 0.0006% interim measure also needs a clear purpose and end point.

A mature electricity system does not promise never to fail. It makes the definition of acceptable failure public, prevents the annual average from hiding the shape of catastrophe and remains able to revise investment as evidence changes.

Primary sources

Discover more from Geoffrey Chen

Subscribe to get the latest posts sent to your email.