Reactivity estimation of nuclear reactor combined with neural network and mechanism model

The reactivity control is an important mean to ensure the safety operation of a nuclear power plant. But the reactivity of the nuclear reactor usually can not be directly measured. It should be computed with certain estimation method. In this paper, artificial neural network and mechanism model are combined to estimate the total reactivity of the nuclear reactor by taking advantage of neural network´s nonlinear mapping ability. Base on analysis of the reactivity affecting factors, a mechanism model is built for fission poison influence calculation and several neural network models are built separately for rod, boron, doppler effect and moderator effect. The influence factor of fuel is considered as a constant. The total reactivity of nuclear reactor is obtained by a summative function. Extensive simulation tests show that satisfied results can be achieved with the proposed approach. It presents a new idea to estimate the nuclear reactor´s reactivity for a nuclear power plant.