Prediction of reactor vessel water level using GMDH in severe accidents due to LOCA

Abstract

In certain circumstances of the severe accident, it is essential to confirm major parameters of a nuclear power plant. The reason of confirmation is to check the status of a nuclear power plant and to respond appropriately to each situation. Particularly, the reactor vessel water level is important data in order to confirm the reactor core condition. Therefore, in preparation for the uncertainty of a sensor in severe accident situations, the reactor vessel water level was predicted using a group method of data handling (GMDH) algorithm. The prediction model of a reactor vessel water level was developed based upon numerical simulation data such as development data and test data. These data were generated by simulating the severe accidents of a total of 810 cases using MAAP4 code about the OPR1000 nuclear power plant. As a result of predictions, the prediction performance of the developed GMDH model was quite satisfactory. Therefore, the developed GMDH model could be successfully applied for providing effective information for operators in severe accident situations.