ML16245A086

Reliability of Emergency AC Power at Nuclear Power Plants

ML16245A086
Person / Time
Issue date:	05/19/1992
From:	Ward D Advisory Committee on Reactor Safeguards
To:	Selin I Advisory Committee on Reactor Safeguards
References
D920519
Download: ML16245A086 (4)

Text

D920519 The Honorable Ivan Selin Chairman U.S. Nuclear Regulatory Commission Washington, D.C. 20555

SUBJECT:

RELIABILITY OF EMERGENCY AC POWER AT NUCLEAR POWER PLANTS During the 385th meeting of the Advisory Committee on Reactor Safeguards, May 6-9, 1992, we discussed the issue of the reliabili-ty of emergency ac power at nuclear power plants. This topic was reviewed in the context of the proposed amendment to 10 CFR 50.63 (Station Blackout [SBO] Rule) and the associated Revision 3 to Regulatory Guide 1.9, that deals with the assurance of adequate reliability for the emergency diesel generators (EDGs) normally used to provide this power. This was also discussed during a joint meeting of the ACRS subcommittees on Probabilistic Risk Assessment and Control and Electrical Power Systems held on April 22, 1992.

The Committee has most recently commented on the proposed amendment to the SBO Rule in a letter to you dated December 20, 1991. We also had the benefit of the referenced documents.

This is a particularly illuminating subject to review, since it carries with it elements of many recent actions of both the Commission and its staff, and poses such a formidable challenge to coherent regulation. These actions include the SBO Rule (itself nonquantitative, but translated into quantitative terms through other regulatory documents), the issues of performance-based maintenance and regulation, the use by the staff of probability and statistics, and finally the complex of proposed actions with respect to the reliability of EDGs the specific subject of this letter. We mention all these other interlocking issues because the underlying theme emergency ac power reliability provides a fine opportunity to proceed in the direction of coherent, performance-based regulation. Both the Commission and we have often supported such a direction. This letter is specifically devoted to EDG reliability, but we think it important to see the subject in its context.

The problem begins in the SBO Rule, where the list of requirements to be met by a licensee includes a demonstration of ability to cope for an appropriate length of time with a loss of all ac electrical power. That capability is in turn dependent (in part) on the "reliability of the onsite emergency ac power sources." For most plants, these are diesel generators. In the implementation of the SBO Rule, licensees have been given the option of declaring the underlying reliability of their EDGs to be either 0.95 or 0.975, depending on other details. While there is no problem in declaring an intent to meet such a reliability standard, there is equally well no realistic possibility of demonstrating that it has or has not been met for any particular plant's EDGs with the current failure-rate data. That is the problem.

It is easy to see. NUMARC has kindly provided us with INPO data on actual industry EDG experience for the years 1988-1990, showing, in round numbers, approximately a hundred failures to start and a hundred failures to load, out of 20,000 and 15,000 attempts respectively. This is for about 200 diesel generators, so the average diesel has a failure of some kind about once every three years. This is simply not the kind of data accumulation that lends itself to individual statistical analysis, or to the determination of individual reliability. There is no evidence (or, to be generous, at best marginal evidence) that any single generator is below its required reliability.

The industry average for those three years is better than 99.5 percent reliability for start and 99 percent reliability for load.

This is in fact far above the requirements. But there is a policy issue here is it the function of this regulatory agency to seek out malefactors (if there are any) and punish them for their own sake? In any population, there will always be a worst performer should one seek to bring the worst up to the norm (thereby creating a new worst), or to seek a prescribed level of protection for the public? These are not the same objectives. A test whose main objective is to find below-par diesels will inevitably trap large numbers of innocents.

For example, consider the last fifty attempts to start, more than a year's experience for the average EDG. For an EDG of reliability 0.95, there is a greater than 10 percent probability that it will have 5 or more failures. That means that if one were to rely on a criterion of an apparent drop to 0.90 reliability (5 failures in 50 tries), one will trap 10 percent of the innocents with reliability 0.95 in return for less than an even chance of catching a culprit with reliability degraded to 0.90. That is a poor trade in any enforcement environment. (Statisticians call these Type I and Type II errors, and artifices like double triggers help very little in the absence of adequate data.) These numbers become more compli-cated for the various double trigger thresholds, but the point remains the same the data will not support defensible enforcement action, because they will not separate the innocent from the guilty.

It is even worse for the proposed criterion of seven consecutive successful starts as a condition for return to service after a "problem" indication. Here a 0.95 EDG has a better than 70 percent chance of passing the test, higher still if it is allowed some failures before the successful string. It would be far better to inspect the EDG and repair the presumed cause of the failure as would surely be done anyway.

This is a problem the Commission has brought on itself (and ACRS must share the blame) through inattention to the implications of the unfortunate wording of the SBO Rule. If the proposed changes are made, they will only further ossify an untenable position.

We believe that this case provides an excellent opportunity to create a paradigm for performance-based regulation, but it will require coherent consideration of all the elements of emergency ac

power reliability, not just the EDGs, and also require Commission guidance on the emphasis of regulation punitive or protective.

We recommend that the current initiative be scrapped as statisti-cally flawed, that the requirement in the SBO Rule for diesel reliability be interpreted as applying to populations rather than to individuals, that it be monitored on an industry-wide basis (where the statistics are adequate over a period of years), and that any modification in the maintenance and testing requirements for EDGs be pursued on an issue-by-issue basis. Some of the proposals contained in the proposed Revision 3 to Regulatory Guide 1.9 dealing with EDG testing and maintenance appeared reasonable if divorced from the flawed statistical bases for their implementa-tion.

To be very specific, the premise that there are bad diesels out there (made to us verbally by some staff members on several occasions) cannot be supported with any reasonable assurance by the data. It may be true, but neither the staff nor we know it to be.

Finally, there is legitimate regulatory interest in knowing to just what extent it is in fact possible to use the individual and community diesel failure data to learn something about the reliability of individual diesels. Such information would be useful in assessing the more detailed health status of the diesel population. There exist a number of reputable statistical tech-niques which can be brought to bear on this problem, and which could also serve to sharpen some of the arguments we have made above. They have not been used by the staff in formulating the proposed amendment to the Station Blackout Rule.

Additional comments by ACRS Members James C. Carroll, Ivan Catton, Carlyle Michelson, and Paul G. Shewmon are presented below.

Sincerely, David A. Ward Chairman Additional Comments by ACRS Members James C. Carroll, Ivan Catton, Carlyle Michelson, and Paul G. Shewmon While we support much of what is said in the Committee's report, we do not agree with two of the Committee's four recommendations.

We do not believe that "the current initiative should be scrapped."

We note that the staff has revised the package to make it clear that "the trigger concept should not be viewed as a statistical estimate of the EDG reliability, but rather as a method to identify the potential degradation of reliability." We support the effort in progress by the Committee's consultant on statistics and reliability theory and believe that it could lead to statistically based trigger values instead of the present empirically based

values.

Given the current industry EDG reliability experience, we believe that the proposed amendment to the SBO Rule and its associated regulatory guide provide a reasonable, performance-based regulatory basis for ensuring EDG reliability. Promulgation of this Rule will provide valuable experience to the NRC and licensees in perfor-mance-based regulation that will be useful in the implementation of the maintenance rule. Licensees with good EDG maintenance programs and root cause analysis techniques will have little difficulty in staying below any of the proposed trigger values. (We do agree with the Committee's comment that the proposed criterion of seven consecutive successful starts as a condition for return to service after a "problem" indication should be changed.)

Finally, we do not agree with the Committee's recommendation that any modification in the maintenance and testing requirements for EDGs should be pursued on an issue-by-issue basis. We fail to see how this provides a viable enforcement basis for dealing with individual licensees with poorly performing EDGs.

References:

1. 57 Federal Register, 14514, April 21, 1992, Proposed Rule, Loss of All Alternating Current Power, U. S. Nuclear Regulato-ry Commission.

2. Draft Regulatory Guide DG-1021 (Second proposed Revision 3 to Regulatory Guide 1.9), "Selection, Design, Qualification, Testing, and Reliability of Emergency Diesel Generator Units Used as Class 1E Onsite Electric Power Systems at Nuclear Power Plants," dated April 1992.

3. Letter dated March 5, 1992 from A. Marion, NUMARC, to H.

Lewis, ACRS, transmitting data on U.S. nuclear power plant EDG performance for years 1988-1990.

4. Memorandum dated March 30, 1992, from P. Boehnert, ACRS, to W.

Minners, NRC, RES, transmitting H. Lewis' comments dated March 23, 1992 on "Diesel Reliability."

5. ACRS report dated December 20, 1991,

Subject:

Resolution of Generic Safety Issue B-56, "Diesel Generator Reliability."