ICONE19-43403 Inversion Method of Source Term in Nuclear Accident based on Gaussian Puff Model

Abstract

The inverse problem of source terms information estimation in nuclear accident is important for emergency response. In this study a review of data assimilation applied on atmospheric dispersion is given. For the atmospheric dispersion model is nonlinear and with model errors, ensemble Kalman filter is adopted for data assimilation. The dispersion consequences is described by Gaussian puff model, and the source term emission rate and release height is estimated real-time. To determine the best first guess parameters' value and errors, more than 10 twin experiments have been carried on. The results show that the ensemble Kalman filter can be applied successfully to estimate the source term information when there are one or two unknown parameters, the estimated accuracy is related to first guess value, and is impacted by the standard deviation of perturbation. To reduce the estimation error, first guess value setting to the half to two times of true value is recommended.