The Green MechanicsEnergy, engineering and sustainable mobility

How do you calculate SAIDI? SAIFI, CAIDI and MAIFI explained with a worked example

A step-by-step guide to SAIDI, SAIFI, CAIDI and MAIFI under IEEE 1366: formulas, which interruptions count, major event days and a fully worked example.

Line illustration for Engineering

SAIDI is the single number most often used to describe how reliable an electricity distribution network is: the average number of minutes of supply interruption a customer experiences over a period, usually a year. In May 2012 the index was a live topic in Sabah, where a state-wide blackout in the early hours of 30 April came just as Sabah Electricity Sdn Bhd (SESB) was reporting a sharp improvement in its SAIDI. This guide sets out how SAIDI and its companion indices are defined, how to calculate them from outage records, and what the Malaysian figures of that period showed.

The four indices at a glance

IEEE 1366, Guide for Electric Power Distribution Reliability Indices, is the reference most utilities and regulators use. The current edition is IEEE 1366-2022, which superseded the 2012 edition. Its core indices are:

Index What it measures Formula Unit
SAIDI Average outage duration per customer served SAIDI = Σ(ri × Ni) / NT minutes per customer per period
SAIFI Average number of sustained interruptions per customer served SAIFI = Σ Ni / NT interruptions per customer per period
CAIDI Average restoration time for customers who were interrupted CAIDI = Σ(ri × Ni) / Σ Ni = SAIDI / SAIFI minutes per interruption
MAIFI Average number of momentary interruptions per customer served MAIFI = Σ IMi × Nmi / NT momentary interruptions per customer per period

where:

  • ri = restoration time for each interruption event (minutes)
  • Ni = number of customers interrupted by event i
  • NT = total number of customers served in the area at the end of the period
  • IMi, Nmi = number of momentary interruptions and customers affected

The product ri × Ni is the customer-minutes of interruption (CMI) for an event. SAIDI is simply total CMI divided by customers served. Note that the denominator is all customers served, not just those affected, which is why one large event can dominate a year’s figure.

Which interruptions count?

The indices only mean something if everyone agrees where a “momentary” interruption ends and a “sustained” one begins.

  • IEEE 1366 classifies an interruption of more than five minutes as sustained; anything of five minutes or less (typically a recloser or breaker auto-reclosing) is momentary and is counted in MAIFI, not SAIDI or SAIFI.
  • Malaysian practice uses a shorter threshold. TNB’s Electricity Supply Application Handbook (ESAH) glossary defines an interruption as a loss of service for a sustained duration of more than one minute. An outage of, say, three minutes therefore counts towards SAIDI and SAIFI in Malaysian reporting but would be momentary under IEEE 1366.

When comparing figures between countries or utilities, check the threshold first, and check whether planned (scheduled) interruptions and interruptions originating on the transmission system are included. Regulators differ on both points.

Major event days and the 2.5-beta method

A typhoon or a grid-wide collapse can produce more customer-minutes in one day than in the rest of the year combined. IEEE 1366 therefore separates major event days (MEDs) so that underlying performance can be tracked, while still requiring the MED figures to be reported.

The 2.5-beta method works on daily SAIDI values:

  1. Collect daily SAIDI for the previous five years (or all available history if less).
  2. Discard days with zero SAIDI; take the natural logarithm of each remaining value.
  3. Compute α, the mean of the logarithms, and β, their standard deviation.
  4. Set the threshold TMED = e^(α + 2.5β).
  5. In the following reporting year, any day whose daily SAIDI exceeds TMED is a major event day. An interruption that spans midnight is attributed to the day it started.

Conceptually, daily SAIDI is roughly log-normally distributed, so 2.5 standard deviations in log space isolates the rare, extreme days. As a purely illustrative example, if five years of data gave α = −1.20 and β = 1.10, then TMED = e^(−1.20 + 2.75) = e^1.55 ≈ 4.71 minutes per day. Researchers at Lawrence Berkeley National Laboratory tested alternatives and found none performed uniformly better than the standard’s method.

A worked example

Consider a hypothetical distribution area serving NT = 20,000 customers. Over one year its outage log records six events:

Event Description Customers interrupted (Ni) Duration (min) Customer-minutes
1 11 kV feeder fault 2,400 95 228,000
2 Distribution transformer failure 150 240 36,000
3a Cable fault, first stage restored 2,000 30 60,000
3b Cable fault, remainder restored 1,100 150 165,000
4 LV fuse operation 80 300 24,000
5 Recloser trip, auto-reclosed 4,500 0.5 (momentary)
6 Switching transfer 1,000 3 see below

Event 3 shows how staged restoration is handled: each group of customers is counted with its own restoration time, but the 3,100 customers are counted once as interrupted.

Customers off supply against time for a cable fault with staged restoration; the shaded area equals customer-minutes of interruption 3,100 1,100 0 30 150 minutes 2,000 × 30 1,100 × 150 = 165,000 customer-minutes customers off supply
Figure: Event 3 as an area. Customer-minutes are the area under the "customers off" curve: 60,000 + 165,000 = 225,000.

IEEE 1366 basis (sustained > 5 min). Events 1 to 4 are sustained; events 5 and 6 are momentary.

  • Total customers interrupted: ΣNi = 2,400 + 150 + 3,100 + 80 = 5,730
  • Total customer-minutes: Σ(ri × Ni) = 228,000 + 36,000 + 225,000 + 24,000 = 513,000
  • SAIDI = 513,000 / 20,000 = 25.65 min per customer per year
  • SAIFI = 5,730 / 20,000 = 0.287 interruptions per customer per year
  • CAIDI = 513,000 / 5,730 = 89.5 min (check: 25.65 / 0.2865 = 89.5)
  • MAIFI = (4,500 + 1,000) / 20,000 = 0.275 momentary interruptions per customer

Malaysian basis (sustained > 1 min). Event 6 now becomes sustained, adding 1,000 customers and 3,000 customer-minutes:

  • SAIDI = 516,000 / 20,000 = 25.80 min
  • SAIFI = 6,730 / 20,000 = 0.337
  • CAIDI = 516,000 / 6,730 = 76.7 min
  • MAIFI = 4,500 / 20,000 = 0.225

The comparison makes a practical point: a shorter threshold barely moves SAIDI here but raises SAIFI and lowers CAIDI, because many short interruptions pull the average restoration time down. CAIDI on its own can therefore look better while customers experience more interruptions.

What the Malaysian figures showed

In the period of the original post, the Energy Commission (Suruhanjaya Tenaga) published SAIDI for both major utilities:

Year Sabah (SESB), min/customer/year Peninsular (TNB distribution), min/customer/year
2009 2,868 not shown here
2010 687 (target 700) not shown here
2011 about 495 69.11
2012 556.82 (target 650) 60.46

The 2009 and 2010 Sabah values are from the Commission’s 2010 performance report, which recorded a 76% reduction and fast-track improvement funding for SESB. The 2011 Sabah figure is given as 495 in the Commission’s report and as 494.64 in a parliamentary reply. The 2012 values are from the Commission’s 2012 annual report and its 2012 industry statistics. Sabah’s 2012 figure rose again after two years of improvement; it covered a year that included the 30 April state-wide blackout, although the published reports do not break the total down by event.

The gap between the two networks reflected very different starting points: a long, lightly meshed grid serving dispersed rural loads in Sabah against a denser, more interconnected peninsular system. For context on supply coverage in the state during the same period, see 100% power supply coverage in Sabah.

Using the indices well

  • Check the definitions (sustained threshold, planned outages, transmission-originated events, MED exclusion) before comparing two numbers.
  • Report SAIDI with and without MEDs. The “all events” figure is what customers lived through; the “normalised” figure shows whether the network itself is improving.
  • Read SAIDI, SAIFI and CAIDI together. Fewer but longer outages and more but shorter outages can give the same SAIDI.
  • Audit the data. Indices depend on accurate customer counts per transformer and feeder and on accurate restoration timestamps, increasingly captured automatically by SCADA and smart meters.

Related reading: calculating power factor from meter readings and lightning and surge protection, since lightning is a leading cause of overhead-line interruptions in Malaysia.

What has changed since

IEEE 1366 was revised in 2012 and again in 2022; the index definitions above are unchanged in principle. TNB reported a distribution SAIDI of 47.88 minutes per customer for 2024, against a target of 48 to 50 minutes. In Sabah, The Star reported in December 2024 that annual interruption durations had ranged between 189 and 332 minutes from 2017 to 2021, well below the 2012 level but still several times the peninsular figure. Since 3 January 2024, electricity supply in Sabah has been regulated by the state’s own Energy Commission of Sabah (ECoS) rather than the federal Energy Commission, so Sabah reliability statistics are now published under that regime.

Sources and further reading

  1. IEEE 1366-2022 Guide for Electric Power Distribution Reliability Indices, IEEE
  2. Evaluating the Performance of the IEEE Standard 1366 Method for Identifying Major Event Days, Lawrence Berkeley National Laboratory
  3. ESAH glossary and definitions (definition of interruption), Tenaga Nasional Berhad
  4. Energy Commission Annual Report 2012, Suruhanjaya Tenaga (Energy Commission)
  5. Electricity Supply Industry in Malaysia 2012: Performance and Statistical Information, Suruhanjaya Tenaga (Energy Commission)
  6. Performance and Statistical Information on Electricity Supply Industry in Malaysia 2010, Suruhanjaya Tenaga (Energy Commission)
  7. Supply reliability (SAIDI and SAIFI, 2024), Tenaga Nasional Berhad
  8. Energy Commission of Sabah (ECoS) as regulator of electricity supply, Energy Commission of Sabah
  9. Sabah's green power ecosystem, The Star