Peak power prediction method of heat pipe cooled reactor start-up and power-up processes based on ANN

Abstract

The start-up and power-up processes of the heat pipe cooled reactor are essential parts of the autonomous operations. The rapid power fluctuation in the processes can affect the safety of the heat pipe reactor. The fast and accurate prediction of the peak power is significant for the safe operation of the heat pipe cooled reactor. This paper generates the peak power datasets of heat pipe cooled reactor start-up and power-up processes by coupling Monte Carlo sampling, and system analysis program with heat pipe cooled reactor MegaPoweras the research object. A fast prediction model of peak power was developed based on the artificial neural network and evaluated in terms of cost, accuracy, and interpretability. The results show that the artificial neural network model has high prediction accuracy and is suitable for large datasets with complex non-linear relations. However, the training cost is high, and the interpretability is weak. The above characteristics are explained by theoretical analysis, and the ability of ensemble algorithms to improve the accuracy of the artificial neural networks is discussed.