89. Fire Risk Analysis for Nuclear Power Plants

From a public health and risk standpoint, the integrity of the nuclear fuel is the main safety concern at nuclear power plants. Absent sufficient cooling, the heat generated from radionuclide decay can lead to fuel melting and the potential release of radioactivity into the environment. Past studies have shown that accidents involving the loss of cooling to the reactor core (see Figs. 89.1 and 89.2) are the dominant contributors to risk, and these accidents continue to be the focus of current risk studies. Recent analytical studies, as well as the March 11, 2011 accident at the Fukushima nuclear power plant in Japan, have indicated that accidents involving used (“spent”) nuclear fuel outside of the core may be more risk significant than previously thought. Studies to re-assess the potential importance of these accidents are ongoing.

This paper was prepared, in part, by employees of the United States Nuclear Regulatory Commission. It presents information that does not represent an official staff position. The NRC has neither approved nor disapproved its technical content. The views and conclusions in this chapter are those of the authors and should not be interpreted as necessarily representing the views or official policies, either expressly or implied, of the U.S. Nuclear Regulatory Commission.

The U.S. Nuclear Regulatory Commission (NRC) designates its staff-prepared reports using the nomenclature “NUREG.” NUREG/CR contractor reports are reports prepared by NRC contractors.

Among other things, this statement indicates that the NRC intends to increase its use of PRA technology “in all regulatory matters to the extent supported by the state-of-the-art in PRA methods and data…”.

NUREG/CR-6850/EPRI TR-1011989 will be referred to as NUREG-6850 for simplicity.

Technically, the CDF is the expected (in a statistical sense) number of core damage events per unit time.

http://​www.​nrc.​gov/​reactors/​operating/​ops-experience/​fire-protection.​html

Guidance documents on a number of specific fire PRA issues can be found on the NRC public web site. Agency documents are maintained on its Agencywide Documents Access and Management System (ADAMS) at http://​www.​nrc.​gov/​reading-rm/​adams.​html#web-based-adams. A number of fire PRA methodology enhancements and clarifications have been developed as a part of industry and NRC efforts to implement risk-informed performance-based fire protection based on the NFPA-805 standard. These can be found using the search term “NFPA 805 FAQ”.

“Targets” in this context are PRA components and cables near the fire location and “target sets” are simply collections of individual targets that may be threatened by a given fire.

As one minor difference, NUREG/CR-6850 defines two stages of physical fire analysis; namely, a preliminary fire source screening task called “scoping fire modeling” and a follow-up task called “detailed fire modeling”. In the ASME/ANS standard, these two tasks have been incorporated into one technical element called “fire scenario selection and analysis”.

The ASME/ANS Standard includes a requirement to address Seismic-Fire Interactions that is not shown in Fig. 89.6. Current methods for analyzing these interactions are mainly qualitative and have no direct impact on the fire PRA process or results. Furthermore, since a seismic event is the actual initiating cause, the consequential fire scenarios should, following typical PRA conventions, be included in the seismic portion of the PRA rather than the fire portion. (Of course, care should be taken to ensure the issue does not fall through any gaps.) With recent events in Japan, including seismically-induced fires at the Kashiwazaki and Onagawa plants in Japan, the NRC and the nuclear power industry are considering developments to address the risk of such events quantitatively.

The terminology surrounding the naming of fire PRA spatial partitions has a somewhat checkered history. Various terms used in fire protection engineering have definitions that were not developed for fire PRA. For example, in fire protection engineering “fire zone” is often defined in the context of coverage areas for a fire detection or suppression system. Similarly, “fire area” has a very specific meaning in the context of the NRC fire protection regulations (see Regulatory Guide 1.189 for related discussions). NUREG/CR-6850 avoids the use of these and other preexisting terms and instead refers the spatial units as “fire compartments,” although the fire PRA spatial partitions may or may not be enclosed spaces. The more accurate, albeit fairly technical phrase used in the ASME/ANS Standard is “physical analysis units” or PAUs. Fire PRA spatial units may correspond to a fire zone, a fire area, a compartment, a subset of any of these, or indeed a superset of smaller compartments. For the sake of simplicity, this chapter uses the very generic term “plant area.”

Spurious operations are defined as a circuit-fault mode wherein an operational mode of the circuit is initiated (in full or in part) due to failure(s) in one or more components (including the cables) of the circuit; examples are a pump spuriously starting, or the spurious repositioning of a valve.

Consistent with typical PRA modeling practices for many hazards it is generally assumed that fire frequencies are constant over time. One uncertainty regarding fire frequencies is whether or not this is a good assumption. The U.S. nuclear industry has documented fire event data going back to 1965, but the appropriateness of using of older events in a current analysis has been questioned. Prior analyses of fire frequency trends have reached divergent conclusions. One review of fire event data published in 2001 [44] showed that the frequencies of reported fires in key U.S. nuclear power plant compartments had not changed dramatically when comparing the periods 1965–1985 and 1986–1994, and thereby supported the use of the Poisson model for U.S. nuclear power plant fire occurrences. However, a more recent analysis demonstrated an apparent shift towards lower fire occurrence rates around 1990. The fire PRA community has not reached consensus on the subject, and as discussed in the “Current Activities and Future Directions” section below, an ongoing effort to gather additional fire event data is being performed, in part, to resolve the question.

In the context of a PRA study, scenarios are “quantified” by estimating their frequencies of occurrence. Ideally, the estimates are expressed in terms of probability distributions for these frequencies.

These procedures include Emergency Operating Procedures (EOPs), Annunciator/Alarm Response Procedures (ARPs), and Abnormal Operating Procedures (AOPs), as well as specific fire response procedures.

As with nuclear power plant PRA in general, fire PRA addresses two types of uncertainty: aleatory and epistemic uncertainty [10]. Aleatory uncertainty, also called “random uncertainty” or “stochastic uncertainty,” is that associated with inherent, potentially observable variability in the events and behaviors being modeled. Epistemic uncertainty is associated with limitations in the PRA analyst’s state of knowledge, and can involve such things as uncertainties in the true value of an input parameter for a fire model or in the appropriateness of the model itself. Unlike aleatory uncertainty, epistemic uncertainty can be reduced through the collection of additional information (e.g., via experiments).

The fundamental structure of fire PRA is aimed at assessing aleatory uncertainty, as the fire-induced CDF is a measure of aleatory uncertainty (it addresses the likelihood of a random event—the occurrence of a core damage accident due to fire). The uncertainty analysis element discussed in this section deals with the epistemic uncertainty in the fire PRA inputs, models, results, and insights.

In addition to data collection, the development of a component-based fire frequency approach that goes beyond that provided in NUREG/CR-6850 will require further analysis. This analysis is needed to characterize the relationship between the number of ignition sources and fire frequency, since this relationship may not be linear. (For example, a plant with 50 pumps may not have twice as many pump fires as a plant with just 25 pumps.)