Ride-through LVRT / HVRT

What it is (precise)

When a fault strikes the transmission or distribution network, voltage at the BESS terminals can momentarily collapse (a sag) or, once the fault clears or load sheds, briefly rise above nominal (a swell). Ride-through defines a no-trip envelope in per-unit voltage versus time.

The abbreviations travel under aliases: LVRT is also written UVRT (under-voltage ride-through) and, in European codes, FRT (fault ride-through); HVRT likewise appears as OVRT (over-voltage ride-through). A typical LVRT curve requires surviving near-zero voltage for roughly 150 ms (a normally cleared transmission fault), then a staged recovery plateau around 0.5 to 0.9 pu lasting from hundreds of milliseconds to several seconds.

HVRT curves mirror this above nominal, commonly near 1.2 pu for a second or less, with the continuous ceiling near 1.05–1.1 pu depending on whether the plant sits under IEEE 2800 (~1.05) or IEEE 1547 (~1.10). The swells HVRT guards against come from load rejection after a fault clears, single-phase faults raising the healthy phases, and capacitor-bank or line switching.

They stress the hardware differently from a sag: the DC link needs voltage headroom above the elevated AC peak or the inverter loses current control, sustained overvoltage eats into surge-arrester TOV margin, and the inverter absorbs reactive power to pull the voltage down rather than injecting it. Outside the envelope the plant is permitted to disconnect — and under IEEE 1547's shall-trip regions on distribution, required to.

Ride-through is a PCS-and-controls behavior, not a battery-cell property: the DC side just needs voltage headroom and the contactors must stay closed, so the requirement is met by inverter firmware and the plant controller, whether the plant runs Grid-following or Grid-forming control. Modern codes pair LVRT and HVRT with dynamic support: during a sag the inverters prioritize injecting reactive current (a capacitive boost drawing on the machine's P-Q capability) to prop up voltage, and during a swell they absorb it.

Reference point matters: codes define the envelope at the Point of Interconnection, but the inverter measures at its own terminals behind the Transformer impedance, so the two voltages differ throughout the event. And because the support current and the delivered power come out of one current limit, the envelope is only half the requirement — what the current does inside it is the other half.

Why it matters in a real grid-scale project

Ride-through is a mandatory Interconnection requirement, not an optional feature. The interconnection agreement and grid code make it a condition of energization, and the plant must prove compliance through model validation (in the US increasingly via IEEE 2800-2022 and the IEEE 2800.2 conformity-assessment recommended practice) and often a field or hardware-in-the-loop test before the utility grants permission to operate.

Fail it and you block the commercial operation date and the revenue behind it, because retrofitting inverter firmware or protection settings after energization is slow, negotiated work that can reopen the whole study.

Regulators care because the failure mode is systemic: if many inverter-based resources trip on one remote fault, the network loses gigawatts at the worst moment and the disturbance can cascade. The 2016 South Australia blackout and the 2016/2017 California events (the Blue Cut and Canyon 2 fires, where roughly 1,200 MW and 900 MW of solar dropped through inverter tripping and momentary cessation) drove much of the tightening in today's standards.

For the owner the stakes are commercial too: nuisance trips on recoverable sags mean lost dispatch, missed ancillary-service obligations tied to Frequency response, and penalties or curtailment from the system operator for persistent non-compliance.

Typical values and standards

In the United States the governing documents are IEEE 1547-2018 for distribution-connected systems and IEEE 2800-2022 for transmission-connected plants, with NERC PRC-024 barring protection settings that would trip bulk-system generation inside its voltage and frequency boundaries (a settings standard, not a performance one) — though for inverter-based resources PRC-024 is being superseded by PRC-029, the ride-through standard NERC wrote for IBRs under FERC Order 901, so check which one the interconnection agreement actually cites.

IEEE 2800 notably restricts momentary cessation (the inverter briefly ceasing current injection while staying connected), the behavior that made the California solar-loss events so large. In Europe the framework is the ENTSO-E RfG network code (Regulation 2016/631) as implemented nationally, for example Germany's VDE-AR-N 4110 for medium voltage and 4120 for high voltage. Project-specific deviations live in the interconnection facility study, which governs over the generic curve.

Representative envelope values for utility-scale plants: ride through roughly 0 to 0.05 pu retained voltage for about 150 ms, a recovery ramp or plateau back to near 0.9 pu within roughly 1.5 to 3 seconds, HVRT tolerance near 1.2 pu for around 1 second or less, and continuous operation typically between about 0.9 and 1.1 pu. Envelopes are written in per unit precisely so one curve serves every POI voltage.

How high the healthy phases actually climb during a ground fault, though, is a grounding question rather than a code question: where the system meets the classical effective-grounding test (X0/X1 no more than 3 and R0/X1 no more than 1, both positive) the unfaulted phases stay near 80% of line-to-line voltage, against the full √3 an ungrounded system delivers — so the HVRT setting, the arrester rating and the collection-system grounding arrangement are one decision, not three.

Many codes also require riding through multiple successive faults (auto-reclose sequences), which stresses control recovery, not just survival of one dip. Frequency ride-through is specified in the same documents, typically covering excursions of a few Hz around nominal.

Dynamic reactive-current injection usually begins once voltage deviates more than about 5 to 10 percent from nominal, with a proportional gain — the k-factor, commonly set around 2 and adjustable from 0 up to roughly 10 in European practice — and a required response within a few tens of milliseconds. Read the gain's definition as carefully as its value, because codes differ on whether it multiplies the whole deviation or only the part outside the deadband, which moves the commanded current materially on shallow sags.

The setting beside it is current priority: which component the inverter gives up when it cannot supply both. Transmission-level rules for inverter-based resources generally default to reactive priority through a voltage disturbance, while IEEE 1547 leaves it configurable on US distribution; either way the interconnection agreement names the choice, and that choice decides how much real power survives the dip. Always read the exact curve, gain and priority rules from the applicable code; these ranges only orient the design.

The arithmetic is short and worth doing once. A converter is a current-limited machine, and its active and reactive currents add in quadrature, so the prioritized component takes what it needs and the other gets the remainder: Ip is capped at the square root of (Ilim² − Iq²).

Take a sag holding 0.55 pu at the inverter terminals — a 0.45 pu deviation. A k-factor of 2 on the full deviation commands 0.9 pu of reactive current; against a 1.1 pu ceiling that leaves about 0.63 pu for active current, and real power is the product of what is left of the current and what is left of the voltage: 0.55 × 0.63 ≈ 0.35 pu.

The plant is exporting roughly a third of its rating while doing exactly what the code demands. Deepen the sag and the reactive command claims the entire budget — with k = 2 that happens near 0.45 pu retained voltage — and real power goes to zero for the duration, which is compliance, not a fault.

The ceiling is modest because an inverter is not a synchronous machine: fault contribution runs about 1.1 to 1.2 times rated current, with some units allowing roughly 1.5× for a few cycles, against the ~5 to 7× of a spinning generator. That saturation is also the defining constraint on Grid-forming controls.

Recovery is specified as tightly as the dip. Once voltage returns, the plant must restore active power toward its pre-disturbance value within a stated time and along a stated ramp — IEEE 2800 sets the duty for transmission-connected IBRs in the US, national codes implementing ENTSO-E RfG set it in Europe, and the exact figures become project numbers frozen by the interconnection study.

The ramp is bounded at both ends, which is the part that surprises people: too slow fails the study on post-event delivery, while a fleet slamming full power back into a still-weak network drives voltage overshoot and inter-plant oscillation. One case the generic curves rarely address is direction. A BESS may have been importing when the fault struck, so 'return to pre-fault power' can mean resuming a charge — state that in the interconnection agreement alongside its charging-load limit rather than inheriting a vendor default written for a generator.

How it shows up in specs, studies and contracts

On a PCS datasheet, find the LVRT/HVRT curve itself (voltage versus time), the reactive-current gain and its settable range, the current-priority setting, the fault-current contribution as a multiple of rated current, momentary-cessation behavior, and the certification basis: UL 1741 with the SB supplement demonstrates IEEE 1547-2018 conformance, while European units carry unit certificates to FGW TR3 test procedures under national RfG rules.

Do not read the short-term overload ladder as the fault-time answer — one utility-scale PCS family publishes 166% for 100 ms, 150% for 5 s, 120% for 8 s and 110% for 15 s, which is graded thermal capability for ordinary operation, not the limiter that acts inside a fault. A curve on a brochure is not compliance either. The question to put to a vendor is direct: send the type-test report and the exact parameter set loaded in firmware, and confirm the k-factor, current-priority and recovery-ramp values match what the study will freeze.

In interconnection studies, ride-through is where model quality gets exposed. Transmission providers in inverter-heavy regions increasingly demand EMT (electromagnetic-transient) models, usually in PSCAD, alongside positive-sequence RMS models, because ride-through and weak-grid behavior do not appear correctly in RMS.

The study benchmarks the two models, sweeps fault types and grid strengths, and freezes the resulting settings (k-factor, current priority, trip thresholds) into the interconnection agreement. Those frozen settings are the binding constraint on the plant: changing them later usually reopens the study and can slip the schedule, so treat the study-approved parameter set as contractual, not tunable.

Practical checks for a working engineer: confirm which bus the envelope references (POI or inverter terminals) and translate through the Transformer impedance; verify plant protection relays and the ride-through curve do not overlap, so protection cannot trip inside the no-trip zone; confirm auxiliary power for cooling, controls and the BMS rides through the same sag, since a control system that browns out defeats a fully compliant inverter; and check that commissioning includes the utility's ride-through parameter verification, commonly a hold point before permission to operate.

Each is a routine, avoidable cause of a failed test.

Common pitfalls

The classic trip-wires sit at the edges of the inverter, not inside it. Grid-following controls that lean on a phase-locked loop can lose synchronism during deep sags on weak grids — or on the phase-angle jump the fault itself produces — even when the hardware could ride through, which is why weak-grid screening (short-circuit ratio at the POI) belongs in the same conversation.

Auxiliary systems are the other standard failure: HVAC contactors and 24 V control supplies fed from station service can drop on a 150 ms sag unless buffered, taking the plant offline seconds after a textbook ride-through. Successive-fault requirements catch designs tuned for a single dip.

Two framing errors follow. The first is reading the lost megawatts as a defect: real power collapses during a deep sag by design, because reactive priority spends the current budget, so a plant can be perfectly compliant and delivering almost nothing — the reason ride-through is treated as the precondition for market services rather than as one of them. The second is treating ride-through as a pure voltage topic.

The same standards bundle frequency ride-through, rate-of-change-of-frequency withstand and the active-power recovery duty above, and a BESS that survives the dip but comes back badly — too slowly, or oscillating against nearby plants — still fails the study. Treat the whole disturbance sequence, from fault inception to full power recovery, as the requirement, and budget commissioning time to prove it rather than discovering the gap on site.