Power Quality & Harmonics
A long interruption is a loss of service, not a short voltage disturbance — supply is effectively lost at one or more points until the network is restored. Performance is judged not by waveform severity but by service continuity: how often supply is lost, how long it stays lost, and how much load is affected. This guide covers the frequency, duration, availability, severity and cost indices, and why the event definition matters more than any single number.
Long interruptions are assessed over a reporting period — usually one year — on service continuity rather than waveform severity. Performance is described by three kinds of index together:
Section 1
A long interruption occurs when supply is unavailable at a customer, delivery point, reception point or defined connection point for longer than a specified duration. The threshold between momentary and sustained interruption depends on the utility or standard — a common practical split is one minute, though some systems use 3 minutes or 10 seconds:
The event runs from loss to restoration, and the duration is simply the time between:
Section 2
Long-interruption indices fall into a few categories — frequency, duration, availability, severity and cost. No single index captures everything: a system may have few interruptions but very long restoration times, or many short interruptions with little energy lost.
| Category | Main Question | Example |
|---|---|---|
| Frequency | How often does supply fail? | Number of interruptions, SAIFI |
| Duration | How long is supply unavailable? | SAIDI, SAIRI, restoration time |
| Availability | How much of the time is service available? | Circuit availability |
| Severity | How large is the impact? | Energy not supplied, system minutes |
| Cost | What is the economic consequence? | Interruption cost index |
Section 3
The simplest index is the number of interruptions, \(N_{\text{int}}\) — but that can mean incidents, delivery points or customers, and these are not the same. One transmission fault may be one incident, two delivery-point interruptions, and many customer interruptions:
| Counting Basis | Meaning |
|---|---|
| Incident count | One physical event or disturbance |
| Delivery-point count | Number of supply points affected |
| Customer count | Number of customers affected |
| Feeder count | Number of outgoing feeders interrupted |
| Phase count | Whether single-phase interruptions are included |
Without this definition, interruption statistics from two utilities may not be comparable at all.
Section 4
SAIFI is the average number of interruptions per delivery point in the reporting period:
So \(\mathrm{SAIFI}=0.5\) means each delivery point saw, on average, half an interruption over the year. Many utilities split it into \(\mathrm{SAIFI}_{MI}\) (momentary) and \(\mathrm{SAIFI}_{SI}\) (sustained), because the two have different causes, impacts and remedies.
Section 5
SAIDI is the average total interruption time per delivery point:
Usually expressed in minutes/year or hours/year, so \(\mathrm{SAIDI}=20\ \text{minutes/year}\) means each delivery point was without supply for about 20 minutes over the year. SAIDI mixes frequency and duration: it can rise because there are many interruptions, or because a few took a long time to restore.
Section 6
SAIRI is the average restoration time per interruption event — “when an interruption occurs, how long until supply returns?”:
SAIDI averages over delivery points; SAIRI averages over events. With a consistent counting basis, the three are linked:
This split is useful: it separates how often interruptions happen from how well the network restores them.
Section 7
Energy not supplied is a severity index — the energy that was not delivered because of interruptions:
With \(P_i\) in MW and \(t_i\) in hours, ENS is in MWh. It reflects both the size and the length of the interruption — a short interruption of a very large load can exceed a long interruption of a small load:
Section 8
System minutes normalise energy not supplied to the size of the power system:
\(\mathrm{SM}=1\) means the severity is equivalent to losing the whole system demand for one minute. Related forms compare unsupplied energy with annual demand — a severity index and an equivalent average interruption time:
These are valuable for comparing large events across years or systems — but only if the system-size reference is defined consistently.
Section 9
Availability is the proportion of time a circuit, supply point or service is available:
and for a set of circuits, conceptually:
Availability is not the same as interruption performance: a circuit can be unavailable through an outage while customers are still supplied by an alternative circuit — availability drops, but no customer interruption occurs. This distinction is central in meshed transmission systems.
Section 10
Some systems estimate the economic cost of interruptions by weighting energy not supplied by a cost per category:
| Factor | Reason |
|---|---|
| Customer type | Industrial, commercial and domestic costs differ greatly |
| Notification | Planned interruptions usually cost less than unplanned |
| Duration | Longer interruptions usually have higher impact |
| Time of day | Peak business hours may be more costly |
| Process sensitivity | Continuous processes lose heavily even from short events |
Cost indices drive regulatory incentives and investment decisions, but they rely on agreed cost assumptions.
Section 11
A planned interruption gives advance notice (maintenance, reinforcement, connection work); an unplanned one does not (faults, equipment failure, weather, third-party damage). The two have very different customer impact, so the report must say what is included:
| Type | Description |
|---|---|
| Planned | Advance notice is provided |
| Unplanned / forced | No notice; caused by fault or failure |
| Momentary | Short-duration loss of supply |
| Sustained | Longer-duration loss of supply |
| Major event | Storm, earthquake or widespread system incident |
Major events can dominate annual indices and mask good day-to-day performance, so many utilities report total performance and, separately, underlying performance excluding major events:
| Reporting Basis | Meaning |
|---|---|
| Including major events | Total customer experience |
| Excluding major events | Underlying network performance |
| Listing major events separately | Transparency on exceptional incidents |
Customers experience the total; planners often need the underlying figure to guide investment.
Section 12
The event definition strongly affects the statistics — including whether a back-fed delivery point counts as an interruption when customer load stays supplied. A robust report defines its terms:
| Definition Item | Why It Matters |
|---|---|
| Incident | Physical disturbance or outage |
| Delivery point | Interface where supply is delivered |
| Customer interruption | Customer actually loses supply |
| Alternative supply | Back-feed may avoid a customer interruption |
| Multiple events | Whether repeated trips are counted separately |
| Single-phase interruption | May or may not be included |
Duration depends on the reporting basis too: for a delivery point it is \(t_{\text{restored}}-t_{\text{interrupted}}\), but an incident affecting several points has one incident duration and several different point durations. The momentary/sustained split (\(t<1\) min vs \(t>1\) min) separates auto-reclose and transient faults from events needing repair or manual switching — though for sensitive industrial customers a momentary interruption can be almost as damaging as a sustained one. Transmission reports must also state whether single-phase trips and auto-reclose operations are counted, as this strongly affects the frequency index.
Section 13
The standard reporting period is one year, \(T_{\text{report}}=1\ \text{year}\), which captures seasonal, weather, loading and maintenance effects. Monthly values reveal seasonal trends, but multi-year trends are usually more meaningful because annual figures vary so much.
| Reporting Period | Purpose |
|---|---|
| Monthly | Identify seasonal or operational trends |
| Annual | Main reliability comparison |
| Multi-year | Long-term trend and investment impact |
Transmission networks also serve generation and interconnection reception points, not only load delivery points, so liberalised-market assessments may consider both. And every severity index needs a clear system-size reference — the same \(\mathrm{SM}=\tfrac{\mathrm{ENS}\times 60}{P_{\text{reference}}}\) gives different answers if \(P_{\text{reference}}\) is the current-year peak, the previous-year peak or the historical maximum demand, so it must be stated.
Section 14
Different indices answer different questions, and a complete assessment uses more than one:
| Purpose | Useful Index |
|---|---|
| Interruption frequency | SAIFI |
| Annual outage duration | SAIDI |
| Restoration performance | SAIRI |
| Large-event severity | ENS or system minutes |
| Supply availability | Availability index |
| Economic regulation | Interruption cost index |
| Investment justification | Multi-year trend of SAIDI, SAIFI and ENS |
SAIFI shows how often, but not how long; SAIDI shows total duration, but not whether it is many short events or one long one; ENS shows load impact, but can under-weight small, sensitive customers. A bare \(\mathrm{SAIDI}=15\ \text{minutes/year}\) is incomplete — it does not say whether planned interruptions, major storms, momentary events or third-party events are included, nor whether counting is delivery-point or customer-based. A defensible statement looks like:
Section 15
Long interruptions are reduced by cutting fault occurrence, shortening restoration or limiting the load affected. The right measure depends on which index is poor:
| Mitigation Method | Main Effect |
|---|---|
| Network reinforcement | Reduces interruption exposure |
| Automation & remote switching | Reduces restoration time |
| Redundancy / alternative supply | Reduces customer interruption |
| Improved protection coordination | Limits the affected area |
| Condition monitoring & maintenance | Reduces equipment failure and forced outages |
| Vegetation & lightning management | Reduces weather-related faults |
| Spares strategy | Reduces repair duration |
| Customer backup generation or UPS | Reduces the consequence of interruption |
If SAIFI is high, target fault frequency and network design; if SAIDI or SAIRI is high, target faster restoration and operational response; if ENS or system minutes is high, protect large loads and prevent widespread loss.
Long interruptions are service-continuity events, assessed over a year by frequency (\(\mathrm{SAIFI}\)), duration (\(\mathrm{SAIDI}\)), restoration (\(\mathrm{SAIRI}\), with \(\mathrm{SAIDI}=\mathrm{SAIFI}\times\mathrm{SAIRI}\)), severity (\(\mathrm{ENS}=\sum P_i t_i\) and system minutes) and cost. No single number tells the whole story, and the decisive issue is definition: incident, delivery-point or customer counting; planned, momentary and major events included or excluded; and the system-size reference. A robust assessment must specify the event definition + duration threshold + counting basis + reporting period + included/excluded events + severity calculation + system-size reference — only then can interruption performance be compared fairly and technically.
Eight-Part Technical Series
An eight-part guide to power-quality indices and assessment objectives — how harmonic voltage, voltage flicker, voltage unbalance, voltage dips and long interruptions are measured and indexed, how the measurement data is reviewed, and the recommended indices and objectives for planning and reporting.
How long interruptions are assessed — frequency (SAIFI), duration (SAIDI), restoration (SAIRI), energy not supplied and system minutes, and why the event definition matters.
