Online transient behavior prediction in nuclear power plants

This article describes an online simulator that works on a fuzzy network mo del to predict transient behavior of a nuclear power plant. The model can b e a causal fuzzy, qualitative fuzzy, or quantitative fuzzy network dependin g on which simulation, fuzzy, qualitative fuzzy, or quantitative fuzzy simu lation, is requested. The networks are derived from underlying mathematical models about the system. This deep design knowledge empowers the simulator to do transient behavior prediction. Specifically, each network contains p roperly grained fuzzy operators to relate system variables together. The op erators are represented as mapping tables so that time-consuming fuzzy comp utations can be reduced to efficient table-lookup during online simulation. In addition, the simulator is equipped with a mechanism to dynamically dec ide a best time increment between two simulation steps, which makes the sim ulation even more efficient. Finally, the simulator includes analysis knowl edge, operational knowledge, and input-output records to support diagnosis, prediction, and learning capabilities, which allow the simulator to make f ast responses, answer anytime status queries, propose reasonable solutions, and perform adaptive modeling. It has been shown that it can successfully predict transient behavior of the temperature and pressure of a steam gener ator with ruptured U-tubes undergoing the operator recovery process in a nu clear power plant. This helps the plant operator immensely in correctly mak ing and successfully carrying out emergency response plans to deal with cri tical accidents, whether or not they are unprecedented.