EDG Load Rejection Peak Voltage: The Technical Specification Limit and How Other Plants Raised It

Expanded from Tip of the Week, 2 March 2019. When a diesel generator rejects a heavy load, the voltage spikes…

That is not a fault. It is what a machine does when you suddenly stop asking it for current: the field is still where it was a moment ago, the governor and the voltage regulator need time to catch up, and in the interval the terminal voltage overshoots before settling back. Every emergency diesel generator does this. The surveillance exists to confirm it does it within bounds.

The bound is in your technical specifications, and it is easier to exceed than most people expect.

What the surveillance is actually checking

The full load rejection surveillance verifies two things: that the EDG will not trip on overspeed when its load disappears, and that it stays inside acceptable voltage and frequency limits while recovering.

The overspeed half is intuitive. The voltage half is the one that catches plants off guard.

In older technical specifications the peak voltage limit is often 4800 V. In newer ones it may be 5000 V. The limit can be customized if adequate technical justification is provided — but absent that justification, the default is probably what you are held to.

The trap: grid voltage creep

Here is the part worth internalizing.

The peak is referenced to where you started. If grid voltage rises by even 100 volts, your load rejection peak is likely to rise with it. It can rise enough to violate the technical specification limit while the diesel itself is behaving perfectly — no overspeed, no equipment problem, nothing wrong with the machine at all. The surveillance fails because the starting point moved.

Grid voltage is not a constant. It drifts with system conditions, and it moves when you move it. Which leads to the practical warning:

Keep this in mind if you are doing any modification that affects site auxiliary/reserve transformers or their tap settings.

A tap change that looks entirely benign on a load flow study can put your next load rejection surveillance over a limit written decades ago. That connection is not obvious from either end: the transformer engineer is not thinking about diesel surveillances, and the diesel surveillance is not obviously a function of transformer taps.

Raising the limit is a license amendment

If the analysis supports a higher peak, the limit can be raised. But it lives in the technical specifications, which means a license amendment request, which means the NRC staff has to agree.

Several plants have already been through this, and their submittals are public. That matters more than it sounds, because ADAMS full-text search is not kind to anyone hunting a problem by description rather than by accession number. If you are in this bind, the precedents below are the shortest path I know to seeing what a successful argument looks like.

The DC Cook precedent

Indiana Michigan Power submitted a license amendment request for Donald C. Cook Nuclear Plant, Units 1 and 2 (Dockets 50-315 and 50-316) on 12 April 2006, seeking to revise SR 3.8.1.11 — the EDG full load rejection surveillance — from a limit of ≤ 5,000 volts to < 5,350 volts.

What makes it useful as a template is the shape of the justification. The submittal does not argue from margin in the abstract. It goes to the equipment that would have to live with the higher voltage and gets each vendor on the record:

• The generator. General Electric confirmed that “the transient overshoot voltage of 5350 volts that may be experienced every 18 months does not adversely impact the generator.”
• The cable. Okonite stated the cables “are suitable for continuous operation at 5 percent above the 5,000 volt nominal rating and for 15 minutes of operation at 10 percent above the 5,000 volt nominal rating.”
• The switchgear. ABB stated there would be “no adverse affect on the ability of the switchgear to operate following a voltage transient of 5350 volts for short durations.”

Running underneath all three is the duration argument: the overshoot is “a transient condition typically lasting for only a few seconds, with the peak voltages lasting for a much shorter period,” and it occurs on a surveillance interval rather than continuously.

So the categories you need to evaluate come into focus: the generator itself, the cable insulation, the switchgear, and the frequency and duration of exposure. Your own submittal will need considerably more depth than an initial cover letter shows, but this tells you where to dig.

Here’s the DC Cook engineering story, for posterity: the Unit 2 AB DG measured 4993 V during the March 26, 2006 surveillance, which became 5048 V once the 55 V instrument uncertainty was added — over the 5000 V limit. Voltage regulator replacement, retuning, transfer switch contacts, a manual regulator card, and repeated retunes all still landed at 5042–5105 V. GE eventually confirmed the regulator was working as designed. The 5000 V figure came from NUREG-1431 and had been adopted in the 2005 improved-STS conversion; the previous TS specified no voltage limit at all.

Here’s the interesting thing: it was the TS change that surprised DC Cook, not a grid voltage shift. They implemented the change, and then got caught flat-footed after running the same surveillance test as always. The only difference was their TS now specified a limit that didn’t previously exist.

Source: DC Cook license amendment request, ML061240382

Source: NRC Safety Evaluation to support amendment approval: ML061010512

Wider industry precedent

Duke assembled a presentation on EDG surveillance testing changes that collects several successful industry precedents, including the DC Cook request. If you are building a case, it is a faster way to find your neighbors than searching ADAMS cold.

Source: ML17044A089

If you are in this right now

Pull the submittals. Find the make and model for the equipment installed in your plant. Get your vendors on the record the way DC Cook did — a letter from the manufacturer saying the machine tolerates the transient is worth more than any calculation you can produce internally.

Better to think about this in the design phase before any mod is installed.