## Reassessing the frequency of partial core melt accidents

April 27th, 2011 by Thomas B. Cochran, National Ressources Defense CouncilThere have been enough partial core-melt accidents that we can ask whether the operational nuclear power plants throughout the world are safe enough as a group.

12 nuclear power reactors have experienced fuel-damage or partial core-melt accidents: The Sodium Reactor Experiment (SRE), Stationary Low-Power Reactor No. 1 (SL-1), Enrico Fermi Reactor-1, Chapelcross-2, St. Laurent A-1 and A-2, Three Mile Island-2, Chernobyl-4, Greifswald-5 and Fukushima Daiichi-1, -2 and -3. (see Table 1 in paper). Eleven of these (all except SL-1) produced electricity and were connected to the grid during some period of their operation, and all are now permanently shut down. In assessing the historical core melt frequency among nuclear power reactors, the number counted depends on how the issue is framed. SL-1 is excluded because it was an experimental reactor, and the design was abandoned after the accident. Although it was the first U.S. reactor to supply electricity to the grid, the SRE could be excluded because it was primarily a research reactor. Chapelcross-2 and St. Laurent A1 and A2 were dual use military reactors, producing plutonium for weapons and electricity for civilian use. From the data available to this author it is unclear whether any fuel actually melted in Greifswald-5. In five cases then, i.e., SRE, Chapelcross-2, St. Laurent A1 and A2, and Greifswald-5, the fuel melt or damage did not result in immediate closure of the plant; rather the damage was repaired and the reactor was restarted.

Worldwide, there have been 137 nuclear power plants that have been shut down after becoming operational with a total generating capacity of about 40,000 MWe and 2,835 reactor-years of cumulative operation (1). Thus, one in twelve [137/11 = 12.5] or fourteen [excluding SRE: 136/10 = 13.6] shut down power reactors experienced some form of fuel damage during their operation. Of the power reactors that have been shut down one in 23 [137/6 = 22.8] were shut down as a direct consequence of partial core melt accidents; one for every 500 reactor-years [2,835/6 = 472.5] of operation. Only about seven of eight giga-watts (GW) [40,000-5,250.5)/40,000 = 0.87≈ 7/8] of nuclear power plant capacity have been closed without experiences a fuel damage accident. One out of 13 GW [40,000/3,011 = 13.3] of nuclear power plant capacity have been closed as a direct result of a fuel melting accident.

Worldwide, there have been 582 nuclear power reactors that have operated approximately 14,400 reactor-years (1). Thus, to date, the historical frequency of core-melt accidents is about one in 1,300 reactor-years [14,400/11 = 1,309], or excluding SRE, about one in 1,400 reactor-years.

Worldwide, there have been 115 Boiling Water Reactors (BWRs) that have operated approximately 3,100 reactor-years. Thus, to date, the historical frequency of core-melt accidents in BWRs is about one in 1,000 reactor-years [3,100/3 = 1,033].

Worldwide, there have been 49 BWRs with Mark 1 containments (the type at Fukushima) and 12 with Mark 2 containments. Five with Mark 1 containment (Millstone Unit 1 and Fukushima Daiichi Units 1-4) have been permanently shut down. These 61 BWRs have operated for 1,900 reactor-years to date. Thus, to date, the historical frequency of core- melt accidents in BWRs with Mark 1 and 2 containments is about one in 630 reactor- years [1,900/3 = 633].

In July 1985, the U.S. Nuclear Regulatory Commission’s (NRC) Advisory Committee on Reactor Safeguards (ACRS) stated (2):

We believe that the Commission should state that a mean core melt frequency of not more than 10-4 per reactor year [one in 10,000 reactor- years] is an NRC objective for all but a few, small, existing nuclear power plants, and that, keeping in mind the considerable uncertainties, prudence and judgment will tend to take priority over benefit-cost analysis in working toward this goal.

On August 4, 1986, the NRC published a final policy statement on safety goals, which said (3):

Severe core damage accidents can lead to more serious accidents with the potential for life-threatening offsite release of radiation, for evacuation of members of the public, and for contamination of public property. Apart from their health and safety consequences, severe core damage accidents can erode public confidence in the safety of nuclear power and can lead to further instability and unpredictability for the industry. In order to avoid these adverse consequences, the Commission intends to continue to pursue a regulatory program that has as its objective providing reasonable assurance, while giving appropriate consideration to the uncertainties involved, that a severe core damage accident will not occur at a U.S. nuclear power plant.

The NRC cites core-melt frequency estimates from probabilistic risk assessment (PRA) studies in the ranges from 2 x 10-5 to 1 x 10-4 event/reactor-year,5 i.e., from 1 to 5 per 10,000 reactor-years; and for Peach Bottom Unit 2, a GE BWR with Mark 1 containment, 1.202 x 10-5,6 i.e., 1 in 10,000 reactor-years.

Clearly, the historical frequency of core melt accidents worldwide does not measure up to the safety objectives of the NRC. On the whole the operational reactors worldwide are not sufficiently safe. If nuclear power is to have a long-term future greater attention must be given to the safety of current operational reactors worldwide. Older obsolete designs should be phased out rather than having their licenses extended. We should also revisit whether the newer reactor designs currently under construction worldwide and those on the drawing board are safe enough.

Thomas B. Cochran, Natural Ressources Defense Council

P.S. This post is an excerpt of my Statement on the Fukushima Nuclear Disaster and its Implications for U.S. Nuclear Power Reactors Joint Hearings of the Subcommittee on Clean Air and Nuclear Safety and the Committee on Environment and Public Works United States Senate Washington, D.C (available here)

(1) This sum excludes the US reactors, SL-1, Ml-1, PM-1, PM-2A, PM-3A, SM-1, SM-1A and Sturgis. The German KNK-I and KNK-II reactors are treated a one reactor.

(2) ACRS letter from D. A. Ward to N. J. Palladino, Subject: ACRS comments on proposed NRC safety goal evaluation report (17 July 1985); cited in David Okrent, “The Safety Goals of the Nuclear Regulatory Commission, Science, 236, 296-300 (17 April 1987).

(3) Nuclear Regulatory Commission, Federal Register 51, 28044 (4 August 1986); cited in David Okrent, “The Safety Goals of the Nuclear Regulatory Commission, Science, 236, 296-300 (17 April 1987).

March 31st, 2012 at 2:11 pm

Are probabilities the right criteria to assess the risk of core meltdown in nuclear power plants?

In your post you shows that there is an obvious contradiction between the theoretical probabilistic risk assessment (PRA) carried out by the NRC and the historical frequency of core melt accident. This fact is beyond any controversy, however maybe would it be interesting to see in detail how is this probability calculated, what is it supposed to represent and the use that is done or should be done of such numbers.

How is this probability per reactor calculated?

To evaluate the probability of an accident, a deterministic approach was taken by the NRC. Therefore, one first has to imagine all the possible initiating events of an accident. Then, for each initiating event we have to build a tree of all the potential events that could influence the first one. At the end of each path lie the consequences of such a sequence.

Then come the probabilities. Indeed, a probability has to be attributed to each node of the tree. A major assumption is then made, stating that all events are independent from each other. This is quite a controversial choice to make; yet it allows to easily compute the probability of a given path by simply multiplying the probability of each node it crosses. In the end, the probability of a core meltdown is simply the sum of the probabilities of all paths from all initiating events that lead to such an accident.

How is this number used, and how should it be used?

In the United States, the NRC decided that this absolute probability had to be less than 10-4 per reactor and per year. Yet several issues arise from such a narrow decision on nuclear safety. To be efficient, it would require the NRC to investigate all the possible initiating events, all the imaginable paths leading to a core meltdown. This is not realistic. What about the assumption concerning the event’s independence?

Moreover, this method really narrows the scope of safety inspectors such as the NRC. Regulators ought to assess the safety of a power plant, investigate all aspects of potential risks. This cannot be achieved by checking that the sum of some probabilities does not exceed a certain limit.

In France, the ASN (Autorité de Sûreté Nucléaire, the French NRC counterpart) does not use the absolute value of the probability. This tree-based decomposition of the probability of core meltdown is used to assess the importance of a given equipment, a given function or a given human decision in the probability. This makes sense; if the probability computed is 10-6 per reactor and per year, it is interesting to see what equipment is the biggest contributor. If we realized for instance that 50% of the probability is due to a potential loss of coolant flow, we can improve this number by adding flywheels in the vessel or by adding additional pumps or additional power backups. In the ASN mindset, the tree-based probability computed is just a tool that allows regulators and decision maker to pinpoint weaknesses that call for improvement.

Such probabilities should only be used for the relative values they can yield. The absolute value cannot pretend to be a fair image of the nuclear power plant safety. Its actual use in the US discredits their safety policy. How can they maintain their use of such numbers when they are so completely out of range compared to the frequency of accidents?