Tiza Group

Home / Guides / Codes and grid connection: US, UK, CA, AU

G99 Loss of Mains Protection: ROCOF and Vector Shift

G99 Issue 2 loss of mains protection: ROCOF at 1 Hz/s with a 500 ms delay, why vector shift is not accepted, stability tests and a worked island example.

By Shadai García, CEO and protection engineer at Tiza Group.

DNO networknetworkbreakertrapped load 3.2 MWVTinterfacebreaker81R RoCoF1 Hz/s, 0.5 sGsynchronous, 5 MVAH = 2.5 sPowerMeasure / controlisland: 4.0 MW against 3.2 MW trapped load0.8 MW: 0.16 x 50 / 5 = 1.6 Hz/s, above 1 Hz/s0.5 s continuous, then trip: about 0.6 s1 Hz/s needs 0.5 MW = 10 % of 5 MVA2 % mismatch: 0.2 Hz/s, not seenvector shift is not accepted as LoMHz/smismatch (MW)00.51.012threshold 1 Hz/s0.8 MW: 1.6 Hz/s0.5 MW0.1 MW: 0.2 Hz/sillustrative island; constant mismatchLoss of mainsNetwork opensIsland forms81R RoCoF0.5 s delayTrip© tizagrp.comDNO networknetworkbreakertrapped load 3.2 MWVTinterfacebreaker81R RoCoF1 Hz/s, 0.5 sGsynchronous, 5 MVAH = 2.5 sPowerMeasure / controlisland: 4.0 MW against 3.2 MW trapped load0.8 MW: 0.16 x 50 / 5 = 1.6 Hz/s, above 1 Hz/s0.5 s continuous, then trip: about 0.6 s1 Hz/s needs 0.5 MW = 10 % of 5 MVA2 % mismatch: 0.2 Hz/s, not seenvector shift is not accepted as LoMHz/smismatch (MW)00.51.012threshold 1 Hz/s0.8 MW: 1.6 Hz/s0.5 MW0.1 MW: 0.2 Hz/sillustrative island; constant mismatchLoss of mainsNetwork opensIsland forms81R RoCoF0.5 s delayTrip© tizagrp.com
G99 loss of mains: a 5 MVA synchronous generator with a 3.2 MW trapped load, the network breaker opening first, the interface breaker tripped by the 81R relay fed by the VT at 1 Hz per second with a 0.5 second delay, and the rate of change of frequency against power mismatch with 1.6 Hz/s at 0.8 MW, 1 Hz/s at 0.5 MW and 0.2 Hz/s at 0.1 MW not detected. Location in the system and operating principle. Illustrative: real values come from the nameplate and the manufacturer's manual.
Short answer

G99 Issue 2 requires loss of mains protection for Type A, B and C modules and sets the rate of change of frequency (RoCoF) setting at 1 Hz/s with a 500 ms delay that must be held continuously. Voltage vector shift is stated in clause 10.4.11 to be not acceptable as loss of mains protection. The relay must still ride through a vector shift of 50 degrees and a RoCoF ramp of 0.95 Hz/s in the stability tests. Large facilities rely on intertripping instead.

Short answer: the G99 loss of mains settings

Loss of mains (LoM) protection detects that a generator is still feeding a section of network that has lost its connection to the grid. For Type A, B and C modules, G99 sets the rate of change of frequency (RoCoF) trip at 1 Hz/s with a 0.5 s time delay. Voltage vector shift is not an acceptable loss of mains protection (clause 10.4.11). The source is ENA Engineering Recommendation G99 Issue 2, 10 March 2025, Table 10.1 and section 10.4, made openly available by the Energy Networks Association.

For the other interface protection settings see G99 protection requirements. The overall picture is in the UK electrical codes and grid connection guide.

Why loss of mains matters

If the network supply fails and the generator keeps running, two problems appear (clause 10.4.3). When supply returns, the generator may be out of synchronism with the network and suffer damage. And customers on the island can be supplied at voltage and frequency outside the legal limits, with network earthing arrangements no longer valid.

Clause 10.4.4 explains why detection is hard. If the generator can match the island load almost exactly, there is little change in voltage or frequency to see. Large mismatches are detected easily by voltage and frequency protection. The difficult case is a small mismatch, and that is the case LoM protection is designed to catch.

What G99 requires

LoM protection is required for all Type A, B and C modules. For Type D the network operator advises whether it is needed (clause 10.4.2). The accepted techniques are:

  • RoCoF, which measures how quickly frequency changes, based on the mains voltage cycle period (clause 10.4.11);
  • reverse power or reverse reactive power flow, where the site always imports (clauses 10.4.8 and 10.4.9);
  • active methods such as reactive export error detection or frequency shifting (clause 10.4.7);
  • intertripping from the network breaker that opened.

When the site normally exports, reverse power cannot detect the loss of supply, so RoCoF or power factor techniques are used (clause 10.4.10).

How the 500 ms delay works

The delay in Table 10.1 is not a number of cycles. The text under the table sets the rule. The delay starts when the measured RoCoF exceeds the threshold, and it resets if the rate drops below. The relay may not trip unless the rate stays above the threshold continuously for 500 ms. A relay that counts cycles would trip in under 500 ms when the real rate is far above the threshold. A cycle count therefore does not meet the requirement.

Total operating time from detection to breaker opening is typically about 100 ms longer than the delay setting.

Vector shift: no trip, and a test to prove it

Clause 10.4.11 states that vector shift is not an acceptable LoM technique. The history sits in the amendment record at the front of G99: amendment G99/1-1 of 23 July 2018 dealt with the regulator’s decision on vector shift protection and new RoCoF requirements for type-tested generation.

Vector shift has not left the forms, however. Where the interface protection is not type tested, clause 15.4.1(d) requires RoCoF and vector shift stability tests on all relays, whatever LoM method they use. The protection must not trip during them. The commissioning test forms (Form A2-4, B2-2 and C2-2) set the values:

  • a positive vector shift of 50 degrees from 49.5 Hz;
  • a negative vector shift of 50 degrees from 50.5 Hz;
  • RoCoF ramps of +0.95 Hz/s from 49.0 to 51.0 Hz and -0.95 Hz/s from 51.0 to 49.0 Hz, each lasting 2.1 s.

The modules must also withstand RoCoF up to 1 Hz/s measured over 500 ms without disconnecting. The exception is a trip by a RoCoF-type LoM relay (clause 11.2.2 for Type A, clause 12.2.2 for Type B).

Adjusting a relay that trips too often

If spurious tripping becomes a nuisance, the generator and the DNO look for the cause together (clause 10.4.12). Raising a setting reduces the ability to detect an island, so any change needs a full evaluation, and Annex D.2 gives guidance for it. Where the DNO requires LoM protection that cannot be susceptible to nuisance trips, the standard gives intertripping as the example.

Worked example: when does RoCoF see the island?

Illustrative data. An island forms with a 5 MVA synchronous generator, inertia constant H = 2.5 s, on a 50 Hz system. The rate of change of frequency is approximately:

df/dt = (ΔP ÷ S) × f0 ÷ (2 × H)

where ΔP is the power mismatch and S the rated apparent power.

  1. Generation 4.0 MW against a trapped load of 3.2 MW gives ΔP = 0.8 MW. df/dt = (0.8 ÷ 5) × 50 ÷ (2 × 2.5) = 0.16 × 10 = 1.6 Hz/s. That is above 1 Hz/s, so the relay starts its 0.5 s timer and trips after about 0.5 s + 0.1 s = 0.6 s.
  2. Mismatch for exactly 1 Hz/s: ΔP = 1 × 2 × 2.5 × 5 ÷ 50 = 0.5 MW, which is 10% of the generator rating.
  3. A balanced island with a 2% mismatch gives df/dt = 0.02 × 50 ÷ 5 = 0.2 Hz/s. The relay does not see it. This is the non-detection zone, and the reason voltage and frequency protection, and intertripping where required, remain part of the design.
  4. The stability ramp of 0.95 Hz/s for 2.1 s changes frequency by 0.95 × 2.1 = 2.0 Hz. That is why the test runs from 49.0 to 51.0 Hz.

The swing equation here assumes a constant power mismatch and no voltage dependence of the load, so it is only a first estimate. Real island studies use measured load and network data from the DNO.

Checklist for the engineer

  • Confirm the type and whether LoM protection is required, or intertripping is expected.
  • Set RoCoF to 1 Hz/s with a 0.5 s continuous delay, not a cycle count.
  • Confirm the relay does not use vector shift as a trip function.
  • Run the vector shift and RoCoF stability tests on the form for your type, and keep the records.
  • Ask the DNO about intertripping requirements and signalling.
  • Check the relay is type tested, or plan the non-type-tested tests.
  • Record any agreed non-standard setting in the connection agreement.
  • Evaluate any change made to reduce nuisance trips before applying it.

From the settings to a plan your engineer can review

Loss of mains settings, the stability tests and the relay configuration have to line up with the DNO’s requirements. We provide remote relay configuration, with the settings plan behind it. The package is ready for your chartered engineer (CEng) or the responsible designer to review and sign. Tiza does not stamp, sign or certify designs in the UK. Contact us with the relay or inverter model and the DNO’s requirements.

For technical background see ROCOF 81R protection, interconnection relays, 81 frequency trip with on-site generation and cogeneration and the grid.

Frequently asked questions

What are the G99 loss of mains settings?

For Type A, B and C modules, rate of change of frequency at 1 Hz/s with a time delay of 0.5 s (Table 10.1 of G99 Issue 2). The delay starts when RoCoF exceeds the threshold and resets if it falls below. The relay may not trip unless the rate stays above the threshold for 500 ms.

Is vector shift allowed as loss of mains protection under G99?

No. Clause 10.4.11 says the voltage vector shift technique is not an acceptable loss of mains protection. Instead, the interface protection must remain stable, with no trip, during the vector shift stability test of 50 degrees defined in the commissioning forms.

What replaces RoCoF at very large sites?

Intertripping. Table 10.1 expects intertripping for Type D modules and for facilities above 50 MW, and clause 10.6.2 says RoCoF-type protection is not installed at sites of 50 MW or more. The DNO decides if loss of mains protection is needed for Type D.

WhatsApp