A Comparison of Scenario Binning Methods for Dynamic Probabilistic Risk Assessment

Dynamic Event Tree (DET) analysis is an effective approach for evaluating plant response during the course of a transient in the presence of modeling or stochastic uncertainties. DET analysis produces very large output datasets, which creates challenges for data management and interpretation. In classical Level 1 PRA, scenarios are grouped mostly according to the states of various active plant systems. However, it is possible that scenarios with similar active component states have quite different physical histories. The Mean-Shift-Methodology (MSM) is proposed as a means to group DET scenarios based on their physical characteristics. A DET analysis was performed for a Station Blackout (SBO) scenario of a pressurized water reactor with possible AC power recovery using the MELCOR code coupled to the ADAPT DET generation tool. The advantages of scenario grouping using MSM are illustrated versus the conventional Level 1 PRA binning methodology. 2 SYSTEM AND INITIATING EVENT CONSIDERED In a SBO accident, all offsite AC power is lost as well as AC power from emergency diesel generators. Under these conditions, heat removal from the primary system is provided by the turbine-driven auxiliary feedwater (AFW) system. However, this system will also fail if AC power is not restored before station batteries deplete (assumed to be six hours in this case). Also considered in this scenario is the possibility that a small loss of coolant accident (LOCA) can develop from one of two possible sources: 1) failure of pressurizer valves to close on demand, and, 2) failure of reactor coolant pump (RCP) seals. RCP seals are designed to prevent the leakage of reactor coolant system (RCS) water from the primary system out of the RCPs during normal operation. When exposed to the conditions expected in a SBO, RCP seals can degrade, potentially leading to large leakages (Sankatar, 2003). The binning characteristics used for this work discussed in Section 3.1 come from the NUREG-1150 report for the Zion Nuclear Power Plant (NPP) (Stattison, 1990). At the time of the writing of NUREG-1150, RCP seal failure was an important event considered in the evolution of a SBO as it could lead to a significant loss of reactor inventory. While the likelihood of a hypothetical RCP seal failure is considered to be smaller today due to enhancements in RCP seal materials (NRC, 1991), RCP seal failure was treated in this paper in a manner consistent with that in NUREG-1150 in order to enable a comparison between a traditional approach and a dynamic approach. To that end, a model which predicts the probability of RCP seal failure as a function of the stress and temperature was developed from data in NUREG-1150, as well as experimental data from (Kittmer, 1985) representative of the RCP seal materials of the time. During the course of this hypothetical SBO event, AC power may be recovered which can allow the actuation of emergency core cooling system (ECCS). Both low-pressure injection systems (residual heat removal pumps and passive accumulators) and high-pressure injection systems (safety injection pumps and charging pumps) may become available if needed. When ECCS is recovered, pump systems initially take suction from the refueling water storage tank (RWST). However, if ECCS runs for a long enough period, the RWST may deplete and the ECCS will need to switch to recirculation mode where pump suction is taken from the containment sump. When AC power is recovered, however, ECC systems will not be immediately recovered as operating procedures must be followed to allow the activation of safety systems. For this work, a time delay was assumed between the recovery of AC power and the possible actuation of safety systems to simulate a delay in the recovery of safety systems due to procedure following. While the modeling of procedures was not rigorous, it was meant to be a more realistic representation of system response as compared to recovering ECCS immediately after power is restored. The duration of the time delays were estimated using Westinghouse emergency operating procedures (Westinghouse, 1996) and conversations with Westinghouse personnel (Lutz, 2010). When plant systems are recovered after the initiating event, the possibility of failure upon demand of the following systems is considered: 1. Turbine-driven AFW (TDAFW) and motor driven AFW systems 2. Safety injection and charging pumps 3. Residual heat removal pumps 4. Recirculation system Only the TDAFW systems is available before AC power is recovered and all others listed above may become available when AC power is recovered. 3 SCENARIO BINNING METHODOLOGIES CONSIDERED 3.1 Classical scenario binning approach For Level 1 analysis, these scenario bins are known as plant damage states (PDSs). For this work, the binning characteristics used were taken from the NUREG-1150 analysis of the Zion Unit 1 NPP since the system under consideration is a Westinghouse-type 4-loop PWR. These bin characteristics are shown in Table 1. For each characteristic a certain number of pre-defined values are considered which cover the range of event sequences on the event tree. For example, the first characteristic listed in Table 1, “Status of RCS at Onset of Core Damage” deals with the presence or lack of a break in the RCS. Table 2 gives a listing of the possible values that this binning characteristic can take on. Table 1. Binning characteristics used for Classical PRA sce-