Ask a utility executive and a regulator what “reliability” means and you will often get two different answers — because one of them is really describing resilience. The distinction is not semantic. It determines what gets measured, what gets mandated, and where billions of dollars in capital get spent. Get it wrong and you buy the wrong insurance: hardening against hurricanes when the real problem is tree limbs, or trimming trees when the real problem is hurricanes.
Reliability: the everyday standard
Reliability is the grid’s performance under normal and expected-abnormal conditions — the routine failures that happen constantly at small scale. A car hits a pole. Lightning strikes a substation. A transformer fails from age. A squirrel bridges a bus bar. Reliability engineering asks: how often do these events interrupt customers, and for how long?
It is measured with standardized indices, reported annually by every utility under IEEE Standard 1366:
- SAIFI (System Average Interruption Frequency Index): the average number of sustained interruptions per customer per year. For illustration, a typical U.S. utility reports a SAIFI around 1.0 to 1.5 — meaning the average customer sees roughly one sustained outage per year.
- SAIDI (System Average Interruption Duration Index): the average total outage minutes per customer per year — typically 100 to 200 minutes excluding major storms.
- CAIDI (Customer Average Interruption Duration Index): the average duration of each interruption, i.e., SAIDI divided by SAIFI. It measures restoration speed once an outage occurs.
These metrics feed directly into utility regulation. Many states tie rate-case outcomes, penalty and reward mechanisms, and sometimes executive compensation to SAIDI/SAIFI performance. Reliability spending — vegetation management, pole inspections, equipment replacement, feeder automation — is routine capital, reviewed in ordinary rate cases and recovered from all customers.
Crucially, major event days are excluded from the headline numbers. IEEE 1366 defines a statistical threshold (based on each utility’s own history) above which a day’s outages are classified as a Major Event Day and removed from reported SAIDI/SAIFI. The logic is sound: a single hurricane would swamp years of everyday signal, making the metric useless for tracking routine performance. But the exclusion creates a blind spot — the metric utilities are judged on is, by construction, silent about catastrophes. That silence is the first clue that something else is needed.
Resilience: the extreme standard
Resilience is the grid’s ability to withstand, adapt to, and recover from high-impact, low-probability events: hurricanes, ice storms, heat domes, wildfires, floods, cyberattacks, physical attacks. Where reliability is about the frequency and duration of the routine, resilience is about anticipation, absorption, adaptation, and speed of recovery when the routine breaks down.
There is no universal resilience metric comparable to SAIDI. Practitioners work with frameworks instead, organized around four capabilities:
- Harden: make assets harder to damage — steel poles instead of wood, undergrounding critical feeders, elevating or flood-proofing substations, winterizing generation.
- Redundant design: build networked feeders that can be reconfigured around damage, add tie switches, deploy microgrids that can island critical facilities.
- Situational awareness: sensors, advanced distribution management systems, storm and wildfire modeling that tell operators what is failing and where, before customers call.
- Recovery: pre-positioned crews, mutual-assistance agreements between utilities, stockpiles of spare large transformers (which can take a year or more to manufacture).
The closest thing to a headline resilience measure is recovery time — how fast critical load is restored after a major event — sometimes expressed as customer-outage-hours or the time to restore 90 or 95 percent of customers. But recovery time depends heavily on the storm’s severity, making cross-utility comparison difficult. The industry is still searching for standardized resilience metrics, and that search is itself a policy battleground.
The bulk-system layer: NERC standards
On the high-voltage system, reliability is not voluntary — it is federal law. The North American Electric Reliability Corporation (NERC) writes mandatory standards covering operations, planning, cybersecurity, and physical security, enforceable with fines that can reach millions of dollars per violation per day. Key families include transmission planning performance requirements (TPL standards, which mandate the system survive defined contingencies), emergency operations, and critical infrastructure protection (CIP) for cyber assets.
Note the boundary: NERC’s jurisdiction covers the bulk power system — generally facilities at 100 kV and above. The distribution system, where most customer outages occur, is regulated by the states. This split explains a persistent pattern: the bulk system is held to rigorous, audited, continent-wide standards, while distribution reliability varies widely by state regulatory philosophy and utility investment.
Why confusing them is expensive
The policy failure mode looks like this. A regulator, reacting to a catastrophic storm, demands better “reliability” — and gets a utility plan optimized for SAIDI: more tree trimming, more routine maintenance, more feeder automation. All useful, all nearly irrelevant to the next hurricane. Conversely, spending resilience dollars on everyday reliability problems buys expensive over-engineering: burying every suburban feeder to shave a few SAIDI minutes is a poor use of capital compared to hardening the specific feeders that serve hospitals, water plants, and emergency services.
The distinction also shapes who pays and how. Reliability investments are steady, predictable capital recovered through normal rate cases. Resilience investments — storm hardening programs, microgrids for critical facilities, strategic undergrounding — are lumpy, contested, and hard to justify on annual metrics because their benefits accrue in rare, catastrophic years. States have responded with dedicated cost-recovery mechanisms: storm-hardening riders, securitization of storm costs (spreading a hurricane’s repair bill over decades of bonds), and targeted resilience surcharges. Each mechanism is an argument about intergenerational fairness: should today’s customers pay to protect against a storm that may not come for twenty years?
Vegetation management: the unglamorous workhorse
If you ask reliability engineers where the most cost-effective SAIDI improvements come from, the answer is usually tree trimming. Vegetation contact is among the leading causes of distribution outages — branches bridging conductors in wind, trees falling across lines in storms. Utilities spend billions annually on cyclical trimming programs, typically on 3-to-5-year cycles per feeder, and regulators scrutinize these budgets closely because the spending is large, recurring, and visible to customers who dislike seeing their trees cut.
The frontier is precision: LiDAR-equipped helicopters and drones mapping every span’s clearance, machine-learning models predicting which spans will grow into conductors before the next cycle, and targeted “hazard tree” removal beyond the right-of-way. Some utilities are moving from calendar-based cycles to risk-based trimming — prioritizing spans by outage history, tree species growth rates, and exposure. The reliability return per dollar is among the highest in the industry, which is why vegetation management survives every budget review.
Undergrounding: the economics of burying wires
Undergrounding — replacing overhead lines with buried cable — is the most requested and most expensive reliability and resilience measure. Buried lines are immune to wind, ice, vegetation, and most vehicle strikes; they fail far less often. But they cost roughly 5 to 10 times more than overhead construction per mile (for illustration; the multiple varies widely with terrain and density), take longer to repair when they do fail (fault location on buried cable is slow, specialized work), and remain vulnerable to flooding and dig-ins.
The economics therefore favor strategic undergrounding: burying the specific feeders or feeder sections with the worst outage history, the highest wildfire risk, or the most critical customers — not the whole system. Several states have approved multi-billion-dollar strategic undergrounding programs, usually funded through dedicated riders with explicit cost-benefit showings. The programs that survive regulatory review are the ones that target the worst-performing circuits first and publish the SAIDI improvement per dollar. Blanket undergrounding, by contrast, is a good way to double rates for modest reliability gains.
Mutual assistance: the restoration machine
When a major storm hits, no utility staffs for the peak — it keeps crews for normal operations and calls in mutual assistance for catastrophes. Through regional mutual assistance groups and national coordination, utilities lend crews, trucks, and materials to the stricken utility, which reimburses their costs. A major hurricane response can mobilize tens of thousands of lineworkers from across the continent, staged at fairgrounds and parking lots just outside the damage zone.
This system is one of the grid’s genuine resilience strengths — a standing, practiced, continent-scale disaster response capability with no government command structure. Its limits are scale and simultaneity: mutual assistance works when one region is hit and others can spare crews. A season with multiple major hurricanes, or a continent-scale winter event, strains the pool. Utilities are supplementing with contractor pre-commitments, materials stockpiles, and increasingly, mobile substations and transformers that can be trucked to a destroyed substation to restore service in days rather than months.
The measurement frontier
The most important unresolved question is how to measure resilience well enough to regulate it. Proposals include probabilistic risk assessment (expected customer-outage-hours per year across simulated storm catalogs), critical-load restoration curves, and resilience “scores” for utility plans. None has achieved SAIDI-like standardization, partly because resilience is inherently scenario-dependent: a system hardened against hurricanes may be no better against ice storms.
What is converging is the planning practice. A growing number of states require utilities to file dedicated resilience, storm-hardening, or climate-vulnerability plans alongside traditional reliability reporting. FERC has focused on correlated bulk-system risks — gas-electric interdependence, common-mode weather failures of generation — that sit at the boundary between reliability standards and resilience planning.
A note on metrics gaming
Any metric that determines money gets gamed, and reliability indices are no exception. Because Major Event Days are excluded from SAIDI/SAIFI, utilities face a subtle incentive at the margin: the statistical threshold for exclusion means a bad storm day can either count fully or vanish entirely, with nothing in between. Regulators audit these classifications, but the line between a major event and a bad normal day involves judgment.
More consequentially, SAIDI/SAIFI measure averages, which can mask deteriorating performance for specific customer groups. A utility can improve system-average SAIDI while the worst-performing feeders — often serving lower-income or rural areas — stagnate or worsen. Some states now require reporting of worst-performing circuits and targeted improvement plans, an acknowledgment that the average can hide inequity. The emerging best practice is layered reporting: system indices for trend tracking, circuit-level data for targeting investment, and customer-level outage data (from smart meters) for ground truth. Averages are the start of reliability analysis, not the end.
The practical rule
Reliability asks “how often and how long” about the events you can count on happening. Resilience asks “how bad and how fast do we recover” about the events you hope won’t. A grid plan that only answers the first question is incomplete — and after every major event, regulators rediscover the second one, usually at great expense and under political pressure. The utilities that navigate this best maintain two distinct investment programs, measured two distinct ways, and never let one substitute for the other.