IBR Fault Response & Ride-Through

A synchronous generator can't help making fault current: its magnetic flux pushes current into a fault whether you want it to or not. An inverter is different. It is a fast switching power supply with a control computer deciding, every few tens of microseconds, what current to produce. Its semiconductors (IGBTs or SiC MOSFETs) are destroyed within milliseconds if overloaded, so the controls limit output to typically about 1.0–1.5 pu of rating, whatever the fault. The limit is set by the OEM (utility-scale PV inverters are commonly 1.1–1.2 pu, BESS converters are often specified at 1.2–1.5 pu), and it applies after a sub-cycle transient that can be higher, before the current controller acts. Take the actual value, and how it is split between sequences, from the OEM short-circuit model.

The direction of that current is also a design choice. During a fault the controller can produce mostly reactive current to hold voltage up, mostly active current, or (in older designs) stop altogether. Grid codes now specify that choice.

The lab below puts a synchronous machine and an inverter side by side, each on its own rating base. Set the inception angle to 90° for the worst DC offset, then make the fault shallower with the residual-voltage slider. The machine's current falls a lot. The inverter's barely moves, because it is already at its limit.

The machine model uses representative constants (E = 1.0 pu behind X″d from the slider, X′d = max(X″d + 0.05, 0.25) pu, Xd = 1.8 pu, T″d = 30 ms, T′d = 0.8 s) with no pre-fault loading correction, so treat its numbers as illustrative. The peak ratio readout compares instantaneous peaks: the synchronous peak includes the DC offset, so it overstates the ratio of symmetrical RMS currents by the asymmetry factor, up to nearly 2× at a 90° inception angle.

After several large disturbances in which many inverter plants tripped or stopped injecting current during ordinary transmission faults, the industry changed its expectation. IBRs are now expected to stay online and help. The main documents are:

• IEEE Std 2800-2022: interconnection and interoperability requirements for transmission-connected IBRs. Clause 7 covers ride-through and fault response, and Clause 11 covers measurement and monitoring. Requirements apply at the reference point of applicability (RPA), usually the POI/POM.
• NERC PRC-029-1: mandatory frequency and voltage ride-through requirements for IBRs. PRC-028-1: disturbance monitoring equipment at IBRs. PRC-030-1: identifying and analysing unexpected IBR output reductions, with corrective action plans.
• NERC PRC-024-4: frequency and voltage protection settings, now only for synchronous generators, Type 1/2 wind and synchronous condensers. IBRs moved out of PRC-024 and into PRC-029-1.

Points a P&C engineer has to design around:

• Plant protection (27/59/81, inverter self-protection, PPC logic) must not trip inside the ride-through envelope. Settings that used to be "conservative" can now be violations.
• The plant must ride through phase-angle jumps (this course's research cites up to 25° at the RPA under IEEE 2800; verify against the standard) and specified RoCoF without spurious tripping, and frequency for protection should be measured over a window of several cycles, not instantaneously.
• Tripping for repeated voltage dips (auto-reclose, sequential clearing) is allowed only beyond defined counts within defined time windows. Encode those counters exactly as written in the governing text and do not invent your own. PRC-029-1 has its own multiple-excursion provisions, which may differ from IEEE 2800: PRC-029-1 is mandatory, and IEEE 2800 applies as adopted by the TO, so apply whichever governs.
• During low-voltage ride-through the default is reactive-current priority, with dynamic positive- and negative-sequence current injection.
• Momentary cessation (stopping current injection during a disturbance) is restricted inside the mandatory-operation regions of IEEE 2800 and PRC-029-1, and permitted only where the standards allow it. Quote the exact voltage boundary and any legacy-equipment exemptions from the governing standard, not from this summary.

The inverter's sequence analyser splits terminal voltage into $V_1$ and $V_2$ within a cycle or so, and the current controller then regulates both sequences. The lab uses the following example control law. It is a common OEM-style form, not text quoted from IEEE 2800:

Positive-sequence reactive current raises $V_1$. Negative-sequence current leading $V_2$ by about 90° behaves like a reactance in the negative-sequence network, which pulls $V_2$ down and makes the IBR's I₂ look more like a machine's. All of this has to fit inside the current limit. The lab uses the conservative bound $|I_1| + |I_2| \le I_{max}$, which caps the worst phase current.

When the limit binds, priority rules decide what is curtailed. Under reactive priority active current goes first. If reactive current alone exceeds the limit, the two reactive components are reduced together. A rule commonly drawn from IEEE 2800 keeps the positive-sequence increment at least as large as the negative-sequence increment ($\Delta I_{q1} \ge \Delta I_{q2}$). Read the exact clause text before encoding it in a specification.

The angle window comes from IEEE 2800-2022 Clause 7.2.2.3.4 (negative-sequence current injection): I₂ should lead V₂ by about 90°–100° for full-converter IBRs, and the standard allows a wider 90°–150° band for Type III (doubly-fed) wind turbines (both bands are taken from secondary summaries of the clause; verify against the clause text). A likely reason for the wider band is that a doubly-fed machine's rotor-side converter can't fully control the stator's negative-sequence current. That wider band is the lighter wedge on the lab's gauge. IEEE 2800 specifies response times and priority behaviour, not a mandated K-factor: the gains (K₁ = K₂ = 2 here is only an example), deadbands and the high-voltage absorption limit are set by the OEM to meet the TO's interconnection requirements. In this example law ΔIq₁ switches on below 0.9 pu (so it steps to 0.1·K₁, or 0.2 pu at K₁ = 2, at the threshold; the same step occurs at the 1.1 pu absorption threshold), ΔIq₂ has no deadband, and absorption in the high-voltage region is capped at 0.3·Imax.

An inverter has to know the grid angle to synchronise its current. A phase-locked loop (PLL) tracks it by rotating its own reference until the q-axis voltage is zero. When a fault suddenly shifts the voltage angle, the PLL has to slew to catch up, and while it slews its estimated frequency swings hard.

A frequency element measuring over a window $T_w$ treats a phase step $\Delta\varphi$ as an apparent frequency excursion:

The PLL in this lab is a teaching loop: its raw frequency includes the proportional term and isn't amplitude-normalised, so the size of its spike says nothing about any OEM's frequency-trip filtering. The window arithmetic above is the transferable part.

Things to try:

• Raise the PLL bandwidth: tracking is faster, but the raw frequency spike grows. Lower it and the phase error persists longer.
• Lower the dip voltage: the loop gain falls with V, so the PLL slows down during the very event it has to track.
• Add a 5 Hz/s RoCoF. That is a real frequency change, and the 81 element is supposed to see it.

NERC and Texas RE investigated the 2021 and 2022 Odessa disturbances in West Texas. Normally cleared transmission faults were followed by large, unexpected reductions in output across many solar plants. The reports found several causes, not one:

• Inverter PLL loss-of-synchronism tripping on phase jumps.
• Inverter AC overcurrent and DC-side protection (e.g. DC bus voltage) operating during the fault transient.
• Instantaneous or erroneous frequency measurements tripping on transients.
• Momentary cessation and slow active-power recovery, plus PPC interactions.
• Plant-level AC under/overvoltage settings and feeder protection that did not match equipment capability.

Several of NERC's subsequent IBR disturbance reports repeat the last point (verify the specific events and findings against the reports before citing them): plant-level feeder and collector protection set inside the inverters' own capability tripped healthy generation. The collector relays you set are part of the ride-through problem, not just the inverters (see the feeder-island discussion in Modules 1.2 and 2.2).

The envelope in this lab is a representative approximation, not the normative table. It models single-stage definite-time 27/59 on one voltage magnitude and has no repeated-dip counters, so it can't check the consecutive-deviation rules discussed above. Those have to be checked against the governing standard (Module 4.3).

Everything above describes grid-following (GFL) inverters: a PLL locks to the grid voltage and the controller injects a commanded current. Grid-forming (GFM) converters, increasingly specified on new BESS projects, instead regulate an internal voltage and angle and behave like a voltage source behind an impedance, within their current limit.

• In the first cycles of a fault a GFM converter responds immediately, like a machine, rather than after a PLL and current loop act. Once it hits its current limit it falls back to a current-limited mode that can look much like a GFL plant.
• Its negative-sequence behaviour depends on whether the OEM forms a balanced voltage (which presents a low negative-sequence impedance, so it carries I₂ naturally, like a machine) or limits I₂ to protect the converter. Don't assume the GFL 90°–100° behaviour. While a GFM converter is current-limiting, the angle of its fault current can also be OEM-specific, which matters for the directional and phase-selection elements in Module 4.1.
• Requirements for GFM performance are still emerging. Treat the OEM model and the TO's requirements as the source of truth, and re-check any protection that relies on IBR sequence currents (Module 4.1) when a GFM plant is involved.

Protection settings are only as good as the fault model behind them. An IBR can't be represented as a voltage behind X″d: it is a voltage-dependent current source whose output depends on the terminal voltage it sees, which in turn depends on its output.

• Phasor short-circuit programs (ASPEN OneLiner, CAPE, PowerFactory and similar) represent IBRs with OEM-specific or generic models of IEEE 2800-compliant behaviour (IEEE 2800 does not define a model form), as tables or control equations of current versus voltage, and solve them iteratively. Generic models are a starting point, not a substitute for the OEM's.
• EMT models (PSCAD and similar), supplied by the OEM as compiled real-code models, capture the sub-cycle transient, PLL behaviour and controller limits. Use them to validate the phasor model and to study cases the phasor model can't represent, such as phase jumps, weak grids, islanding and TOV.
• What to ask the OEM for: the short-circuit model for your program and firmware version, the current limit and its sequence split, the I₂ angle behaviour, the response and settling times, any momentary-cessation thresholds, the EMT model with its validation report, and notice of any firmware change that alters these.