عدم قطعیت جامع و تجزیه و تحلیل حساسیت برای محاسبات کد توأم از گذر نیروگاه VVER

The development of coupled codes, combining thermal-hydraulic system codes and 3D neutron kinetics codes, is an important step to perform best-estimate calculations for plant transients of nuclear power plants. For applications in safety analysis, these coupled codes should be validated by benchmark calculations and, preferably, by comparison with plant transient data from operating plants. In addition, the results should be supplemented by applying uncertainty and sensitivity analysis methods, which allow to identify relevant parameters of models and solution procedures affecting the results and to quantify their relative importance. Both objectives were part of the VALCO project. The aspect of validation is presented in [S. Mittag, et al., 2004. Neutron-Kinetic Code Validation against Measurements in the Moscow V-1000 Zero-Power Facility, in press; T. Vanttola et al., 2004. Validation of coupled codes using VVER plant measurements, in press], the aspect of a comprehensive uncertainty and sensitivity analysis for coupled code calculations is the topic of this contribution. The results and experiences obtained by the analysis for two plant transients in a VVER-440 and a VVER-1000, respectively, are presented and discussed.

The coupled codes combine thermal-hydraulic system codes and 3D neutron kinetics codes to perform best-estimate calculations of plant transients of nuclear power plants (NPPs). Such codes improve the capability to analyze plant conditions, which are determined by a strong coupling between the coolant flow in the primary circuit and the nuclear power generation in the reactor core affected by the reactivity feedback. Meanwhile, each thermal-hydraulic system code like ATHLET, CATHARE, RELAP or TRAC has been coupled with 3D neutron kinetics codes. The validation of these codes is part of international co-operations. Within the OECD framework, several benchmark problems have been defined and analyzed, e.g. the PWR main steam line break (MSLB) benchmark (Todorova et al., 2003) and the BWR turbine trip (TT) Benchmark (Solis et al., 2001). Corresponding work for VVER NPPs has been performed in projects with participants from MOE countries and Russia funded by the European Commission (EC). In project SRR1/95, data of plant transients from VVER-440 and VVER-1000 have been collected and evaluated for validation, and two plant transients have been chosen to perform detailed calculations by coupled codes. These transients are: a load drop of one turbo-generator in Loviisa-1, a VVER-440 (Hämäläinen et al., 2002) and a switch-off of one out of two working main feed water pumps in Balakovo-4, a VVER-1000 (Mittag et al., 2001). Within the VALCO project (Weiß et al., 2003), funded by the EC in the frame of the FP5 programme, the efforts of code validation have been continued (Mittag et al., 2004 and Vanttola et al., 2004). In addition, a work package was dedicated to the comprehensive uncertainty and sensitivity analysis of coupled code calculations. The GRS uncertainty and sensitivity method based on the SUSA code package (Krzykacz et al., 1994 and Hofer, 1999) was applied to both transients previously analyzed in the SRR1/95 project. Chapter 2 describes the main steps of the uncertainty and sensitivity analysis method developed by GRS, which is based on the statistical code package SUSA. Chapter 3 includes a description of the plant experiment performed in Loviisa-1 NPP, the considerations to determine the list of uncertain parameters for this transient and the results of the uncertainty and sensitivity analysis. The results are completely presented to give an overview on the analysis. Chapter 4 includes corresponding information on a Balakovo-4 plant transient, but the presented results are limited to particular aspects of the results. General conclusions from the uncertainty and sensitivity analysis performed for these VVER plant transients are summarized in Chapter 5.

The main results of the study performed for the two VVER plant transients can be summarized for the following view points. 5.1. Application and performance of the GRS uncertainty and sensitivity analysis method The GRS uncertainty and sensitivity analysis method was successfully applied to the plant transients. This method is based on proven statistical procedures. The application of this method is only based on variations of input parameter values. No internal code adjustments are needed. The amount of calculations needed is in the order of one hundred. This number of calculations was feasible also for coupled codes combining thermal hydraulic system codes with 3D neutron kinetics. 5.2. Results of the analysis of the plant transients The most important step is the systematic evaluation of the transient and the discussion of its dependence on modelling features and plant system actions. The analysis of experimental plant data is the most valuable source for the validation of physical models in large integrated computer codes. The interaction between analysis of experimental data and modelling by computer codes is an important activity to improve knowledge. The result of the uncertainty and sensitivity analysis is a quantification and ranking of effects. The experience was that some expected modelling effects were confirmed as relevant, but, as well, some expected sensitivities were not confirmed, because other effects were more dominant. In this way, the GRS uncertainty and sensitivity analysis is a fully quantitative method. The expert judgement defines the knowledge status by the uncertain parameters and their subjective probability distributions, but the quantitative evaluation is part of the method. The two-sided lower and upper limit values for relevant output parameters can be determined by proven statistical procedures. The results of the sensitivity analysis give indications for which parameters the uncertainty of knowledge should be reduced in order to reduce the uncertainty of results most effectively. The variation of input parameter values as a consequence of uncertain knowledge can activate system actions causing quite different transient evolutions. This gives indications about possible plant conditions that might be reached from the initiating event assuming only small disturbances. In this way, the uncertainty and sensitivity analysis reveals the spectrum of possible transient evolutions. Therefore, the modelling of the plant should be as complete as possible, including the plant specific control and protection system. Generally, it is necessary to get more experience for the uncertainty and sensitivity of results for safety relevant transients. Each transient defines particular requirements in view of the uncertainty of model parameters.