Reliability indices — SAIDI / SAIFI / CAIDI / MAIFI / LOLE / EUE
Distribution (IEEE 1366): SAIFI = Σ N_i/N_total (sustained ≥ 5 min); SAIDI = Σ (U_i·N_i)/N_total; CAIDI = SAIDI/SAIFI; MAIFI for momentary. Typical US: SAIDI 60-300 min/yr; SAIFI 0.5-2.0. MED β-method excludes storms. Generation adequacy (NERC): LOLE ≤ 0.1 days/yr (1-in-10); EUE in MWh; PRM 15-20%. Capacity accreditation via ELCC. EEI benchmarking + PUC PBR. PSPS, microgrids, DERMS, resilience.
Step 1 — Reliability indices: how to quantify supply continuity
Reference notes
Power-system reliability is measured at two levels: distribution (customer-side outage experience per IEEE Std 1366) and generation (capacity adequacy per NERC LOLE / EUE / PRM). Distribution metrics drive utility capital investment, performance-based regulation, and customer experience. Generation metrics drive resource adequacy planning, capacity markets, and renewable integration.
IEEE 1366 — Distribution Reliability Indices
| Index | Formula | Meaning | US typical |
|---|---|---|---|
| SAIFI | Σ N_i / N_total | Sustained interruptions per customer-year (≥ 5 min events) | 0.5-2.0 |
| SAIDI | Σ (U_i · N_i) / N_total | Sustained minutes per customer-year | 60-300 |
| CAIDI | SAIDI / SAIFI | Average duration per interruption (minutes) | 60-200 |
| MAIFI | Σ momentary_i / N_total | Momentary interruptions (< 5 min) per customer-year | 1-10 |
| CAIFI | Σ N_i / N_affected | Interruptions per AFFECTED customer-year | — |
| ASAI | 1 − SAIDI/(8760·60) | Service availability (99.97% typical) | 0.9997 |
| CEMI_n | % w/ > n interruptions | Fraction of customers with ≥ n events / year | — |
Sustained vs Momentary
- IEEE 1366 defines sustained ≥ 5 minutes; shorter events are momentary.
- SAIFI / SAIDI / CAIDI count only SUSTAINED.
- MAIFI tracks MOMENTARY events separately. Most momentaries are reclosing operations after temporary faults (lightning, vegetation flashover, wildlife) — clear themselves within seconds.
- Momentaries matter for sensitive industrial loads (semiconductor fabs, paper mills, data centers) — trips processes even though supply returns.
Major Event Day (MED) Exclusion
- β-method per IEEE 1366: identifies storm days as outliers in log-space distribution of daily SAIDI.
- Compute ln(daily SAIDI) for each day in 5-year reporting period; mean μ + standard deviation σ.
- Days with ln(daily SAIDI) > μ + 2.5·σ flagged as MEDs.
- Headline SAIDI / SAIFI reported MED-EXCLUDED; MED-only metrics reported separately.
- Critical for fair comparison year-over-year and utility-to-utility — without exclusion, one hurricane dominates annual stats.
- Climate change increasing MED frequency; recent push to report BOTH MED-included and MED-excluded.
Outage causes — typical US distribution
| Cause | Typical % | Mitigation |
|---|---|---|
| Weather (non-MED) | 25-40% | Hardening, undergrounding, fire-resistant poles |
| Equipment failures | 15-25% | Asset management, age-based replacement, condition monitoring |
| Vegetation | 10-20% | Vegetation management (NERC FAC-003 transmission; state-regulated distribution) |
| Animal contact | 5-15% | Animal guards on transformers, bushings, lightning arresters |
| Third-party damage | 5-10% | One-call laws ("call before you dig"), pole-protection bollards |
| Operational errors | 1-5% | Training, switching procedures, smart-grid automation |
| Overloads | 1-3% | Capacity expansion, demand response |
| Unknown | 5-15% | Improved fault location, AMI data, FLISR |
- PSPS (Public Safety Power Shutoffs) for wildfire prevention adds 5-15% to total customer-minutes in California / Oregon / Washington in extreme years.
Generation Adequacy — NERC framework
- LOLE — Loss of Load Expectation: expected number of days per year when generation cannot meet peak load. Industry standard target: ≤ 0.1 days/year = "1 day in 10 years" reserve criterion. Used by PJM, MISO, CAISO, ERCOT, NYISO, ISO-NE.
- LOLP — Loss of Load Probability: same as LOLE expressed as probability per period. LOLP = 0.000274 per day = 0.1 days/year.
- EUE — Expected Unserved Energy: expected MWh/year of unserved load. Captures depth + frequency. Typical target < 50 MWh/year for a 50,000 MW system.
- PRM — Planning Reserve Margin: capacity above peak demand needed to achieve target LOLE. Typical 15-20%. RTOs publish PRM annually as part of capacity-market clearing.
Capacity accreditation
| Resource | Accredited capacity |
|---|---|
| Synchronous generator (coal / gas / nuclear) | Nameplate × (1 − EFOR); EFOR typically 5-10% |
| Wind | 5-25% of nameplate (varies by capacity-factor profile and peak coincidence) |
| Solar PV | 30-60% summer peak; ~0% winter peak |
| Battery — 4-hour duration | 80-95% of nameplate |
| Battery — 2-hour duration | 40-70% of nameplate |
| Battery — 1-hour duration | 20-40% of nameplate |
- ELCC — Effective Load-Carrying Capability: modern accreditation method accounting for resource correlation with load and other resources.
- Adopted by CAISO and PJM (2022-2024 transitions) as IBR penetration grew.
- FERC Order 2222 (2020) requires RTOs to allow aggregated DER market participation with appropriate capacity accreditation.
Benchmarking and performance-based regulation
- EEI — Edison Electric Institute: largest US reliability benchmarking program. Member utilities report SAIDI, SAIFI, CAIDI, MAIFI annually using consistent IEEE 1366 definitions; EEI publishes percentile rankings.
- Typical US distribution-utility benchmarks (EEI 2023): SAIDI median 120 min/customer/year; first-quartile < 80 min; fourth-quartile > 200 min. SAIFI median 1.3.
- Best-in-class underground urban: SAIDI < 30 min, SAIFI < 0.4.
- International: European utilities 30-90 min SAIDI (higher % underground + reliability investment). Japan: < 20 min, world-leading.
- State PUCs — PBR (Performance-Based Regulation): tie utility financial outcomes to measured reliability.
- California CPUC: IOU-specific targets (PG&E, SCE, SDG&E).
- New York PSC: REV proceeding.
- UK Ofgem: RIIO program.
- Australia AER: similar incentive scheme.
- Recent push to disaggregate metrics by area (urban/rural) and demographic factor (income, race) for equity reporting.
Smart-grid investment for reliability
- Distribution automation (DA) — remote-controlled switches, sectionalizers.
- FLISR — Fault Location Isolation and Service Restoration — automatic isolation of faulted section and restoration of healthy segments. Reduces SAIDI 20-50% in deployed areas.
- AMI — Advanced Metering Infrastructure — smart meters provide outage detection and verification.
- SCADA + Outage Management System (OMS) — coordinated dispatching of restoration crews.
- Microgrids — local backup for critical loads (hospitals, military bases, college campuses, community centers).
- DERMS — Distributed Energy Resource Management Systems — coordinate distributed solar, batteries, EVs, demand response.
Modern challenges
- Extreme weather — climate change increasing frequency and intensity of weather events:
- 2021 Texas February ice storm — ERCOT load-shed 20 GW for several days.
- 2018-19 California wildfire-driven PSPS by PG&E.
- 2017 Hurricane Maria Puerto Rico — 11-month full restoration.
- PSPS — Public Safety Power Shutoffs: proactive shutoffs to prevent wildfire ignition during high-wind/dry conditions. Customer-minute impact enormous; CPUC tightening rules to minimize duration.
- IBR (Inverter-Based Resource) penetration: changes dynamic-stability landscape; large IBR-trip events affect downstream reliability. NERC standards (TPL-001, PRC-024, BAL-001) being updated.
- EV charging: changes load profile and peak demand patterns.
- Resilience metrics: beyond reliability — ability to recover quickly from disruptions. Customer-minutes lost during high-impact events, restoration time, percent-back-online curves. IEEE 1366 update (2024-25 cycle) addressing resilience + equity + EV/DER impacts.