Research On Source Inversion For Nuclear Accidents Based On Variational Data Assimilation

In a nuclear accident,release source term is the key issue of nuclear emergency preparing,emergency response,accident classification and accident assessment.Source inversion is a way to estimate the source term based on environmental monitoring data during the accident period.It receives a wide attention after the Fukushima nuclear accident.The modelling of source inversion for nuclear accidents based on variational data assimilation?VAR?is able to use and balance all monitoring data in the whole research domain and period to get the global optimal result,which is promising in many applications.However,the result of VAR is greatly influenced by the error of the atmospheric dispersion model used in source inversion.Its accuracy need to be further enhanced.In this thesis,VAR is investigated and improved with two thoughts:to reduce the dispersion model error,and to reduce the influence of the dispersion model error on the quality of the source inversion.With the first thought,the CALMET/SWIFT-RIMPUFF atmospheric dispersion modelling is proposed to cope with the complex terrain and the building layout of nuclear power plants.In this modelling,CALMET and SWIFT are coupled with RIMPUFF which is the atmospheric dispersion model of Chinese nuclear accident consequence assessment.And it adopts the near-field diffusion coefficient correction scheme of the Atmospheric Relative Concentrations in the Building Wakes Computer Code?ARCON96?.The validation against a wind tunnel experiment of Sanmen nuclear power plant demonstrates that the simulation of this modelling is fine inside the emergency monitoring domain of nuclear power plant.With the second thought,VAR with the dispersion model error?DME-VAR?and VAR coping with truncated total least squares?TTLS-VAR?are proposed.These two modellings are applicable to the continuous monitoring data from stationary monitoring stations near the nuclear power plant and all kinds of monitoring data,respectively.In DME-VAR,the dispersion model error at monitoring stations and the release source term are estimated at the same time.The influence of the dispersion model error on the inversion result is partial removed.Evaluation with the wind tunnel experiment data shows that the accuracy of the result calculated by DME-VAR is obviously improved.Moreover,the analysis of the distribution of the estimated dispersion model error reveals the monitoring stations whose error are inaccuracy.Removing these stations and doing the inversion again is illustrated to be able to further improve the quality of the source inversion.In TTLS-VAR,the truncated total least squares is adopted to mathematically correct the dispersion model operator and monitoring data vector,and smooth the estimated result for the ill-posed problem in source inversion.Evaluation with the wind tunnel experiment data shows that the accuracy of the result calculated by TTLS-VAR may be improved,because its correction is mathematical but not physical.In this thesis,the ¹³⁷Cs source term released from the Fukushima Daiichi nuclear power plant is estimated by VAR and TTLS-VAR based on ¹³⁷Cs activity concentration in the air at mesoscale and large-scale.The total released activity of ¹³⁷Cs is estimated to be close to the result published from IAEA.