A forward-backward coupled source term estimation for nuclear power plant accident: A case study of loss of coolant accident scenario

Source term information of radioactive release in a nuclear accident is important for emergency response. Two major categories of source term estimation techniques are forward method based on the status data of nuclear power plant and backward method based on environmental monitoring data. Both data contain considerable uncertainties which can propagate through estimation and result in a biased estimate. To solve the problem, a coupled estimation method is proposed in this study, which combines forward method (Response Technical Manual, RTM-96) and backward method (ensemble Kalman filter). The coupled method builds an evolution model based on a forward estimate and uses it to constrain the temporal correlation of estimates in the backward part, so that the propagation of uncertainties is reduced. Numerical experiments and sensitivity analysis are performed to verify the proposed method, based on a hypocritical loss-of-coolant (LOCA) accident process and the records of a tracer field experiment for Sanmen nuclear power plant. The results demonstrate that the coupled method provides the most accurate estimate in all tests and is more robust to uncertainties in various parameters than both RTM-96 and ensemble Kalman filter.