Balanced design of a CANDU 6 NPP: Insights from full and simplified PRA models

Abstract of the technical paper presented at:
Pressurized Heavy Water Reactor (PHWR) Seminar
June 23–26, 2014
Ottawa, Ontario, Canada

Prepared by:
A. Patel, Y. Akl, G. Banaseanu, R. Gheorghe
Canadian Nuclear Safety Commission
Ottawa, Ontario, Canada

Abstract

The balanced design of a nuclear power plant means that no particular feature or postulated initiating event makes a disproportionately large or significantly uncertain contribution to the overall risk. Probabilistic Safety Assessment (PSA), as part of the Safety Analysis, is a comprehensive and integrated assessment of the safety of the plant used, among other aspects, to demonstrate that a balanced design has been achieved.

The purpose of the paper is two-fold: (1) to investigate the balance of design of a Canadian Deuterium Uranium (CANDU) 600 MWe NPP, including the impact of recovery actions, and (2) to illustrate the benefits of using the CNSC simplified PSA model to assess the balanced design.

The paper presents the insights regarding the balanced plant design, obtained from the results of the PSA for a CANDU 6 NPP operating in Canada. The focus is on the contribution of the initiating events (IE), and accidents sequences to the risk metrics Core Damage Frequency (CDF). In addition, the balanced design has been further explored considering the accident sequences as functions of the initiating events and Conditional Core Damage Probability (CCDP). The relationship between the IEs, and CCDP, offered a new perspective on the plant response to each IE in the full spectrum of initiating events, from anticipated operational occurrences to beyond design basis events; the paper also shows the impact of the recovery actions on the risk of the plant.

The insights have been gathered either from ready-available information from the performed CANDU 6 PSA or they have been obtained by running the licensee’s full PSA model and CNSC’s simplified risk model. The CNSC model has been derived from the full PSA model for internal events, retaining the most significant event sequences and IEs. This results in significant computational time saving leading to swift generation of accident sequences. The simplified PSA model provides a reasonable snapshot of the full model. The experience of developing and use of a proof-of-concept simplified PSA model for CNSC is hereby presented.