核事故源项反演中多重不确定性的矫正方法研究

Source term describes the radionuclide release rate in a nuclear accident, which is an important information for nuclear accident classification, social emotion stabilization, and emergency response. The Fukushima accident demonstrates that source inversion is one of the very few feasible approaches for source term estimation, in the case that no reactor status data are available. However, the accuracy of source inversion suffers from multiple uncertainties, including various model biases in air dispersion models and incomplete data set. ..This project is focused on Fukushima accident and proposes a joint correction of both biases and source correction methods with sparsity constraints, in order to reduce the multiple uncertainties from these two sources. .For the model biases in air dispersion, the online WRF-Chem program is used to reduce numerical errors. The wind tunnel experiment is used to reduce the uncertainties in the deposition model, which is the main source of model biases. In addition, a robust inverse model is calculated based on the improved WRF-Chem model. Furthermore, the project combines black box strategy and blind deconvolution method to extend the source inversion to a joint correction model of both the total biases and the source term, which corrects the total uncertainties from other less important models in air dispersion models such as turbulence and diffusion. .For the data incompleteness, this project investigates the temporal feature of the source and compensates the information loss by developing reconstruction algorithms based on sparsity constraint. The sparsity reconstruction is coupled with the joint correction model so as to improve the accuracy of source inversion, which promotes the capability of the nuclear emergency response of nuclear power plant, in the case that no reactor status data are available.

源项描述了核事故中放射性核素向大气释放的逐时速率，是事故分级与核事故应急的核心依据。福岛事故表明：缺少反应堆工况数据时，反演是获取源项的少数可行途径，但其准确性受多重不确定性的制约，包括：大气扩散模式中沉积、湍流、扩散等多个模型的偏差和数据的不完备。.本课题围绕福岛核事故，通过模式、实验与算法的研究，建立稀疏性约束的“偏差-源项”联合矫正方法，综合降低多重不确定性。针对大气扩散模式偏差，采用“气象-扩散”耦合模式，降低数值计算误差；开展风洞实验，校正模式中不确定性最大的沉积模型，建立稳健的反演模型；采用黑箱策略和盲反卷积方法，将反演拓展为“偏差-源项”联合矫正，降低模式中湍流、扩散等次要模型偏差造成的总体不确定性。针对数据的不完备，挖掘源项的时序特征，构造先验知识，采用稀疏性约束重建方法，补偿信息缺失。通过联合矫正与稀疏性约束的耦合，提高反演准确性及核电站在无反应堆工况数据时的应急能力。

本项目从大气扩散模式不确定性的矫正和不完备观测数据的补偿两方面入手，研究建立了源项反演中多重不确定性的矫正方法，提高源项反演的准确性及核设施在无工况数据时的应急能力。.针对反演模型自身存在固有偏差的问题，本项目首先提出了一系列的自纠偏的源项反演方法，基于观测数据，同时矫正偏差并反演源项。本项目首先提出了“烟羽-源项”同步矫正方法，该方法可以矫正大气扩散模式所模拟的烟羽范围，并同步改进源项，能将风洞试验中源项反演误差降低到0.1%以下。本项目针对实际核事故中观测数据不完备的问题，提出了全变分约束的源项反演方法，能有效恢复观测数据缺失带来的释放信息缺失的问题。进一步地，本项目将时域稀疏性、全变分和“偏差-源项”联合矫正方法结合起来，建立了多重稀疏性约束的“偏差-源项”联合矫正方法，能准确反演洲际、区域和厂区尺度的释放源项的峰值与释放时间窗口。最后，针对福岛核事故，将反演源项和正演源项结合起来，建立耦合估计模型，获取包含长、短寿命核素的详细释放源项。.针对区域尺度大气扩散模式中沉降模型不确定性大的问题，本项目开展干、湿沉降的风洞试验，测量了典型的沉降系数，并分析了降雨、地形、风向等因素的影响规律。基于WRF-Chem，将25种云内-云下湿沉降模型进行在线耦合，建立了区域尺度的“气象-扩散-湿沉降”耦合模式，并通过3km NHM-LETKF和1km NHM-LETKF的高分辨率气象数据模拟，揭示其不确定性的特点。针对局地尺度大气扩散模拟中建筑和地形效应的难点，提出了湍流参数方案，矫正MSS模式在中尺度烟羽范围低估的问题，分析了MSS模式在小尺度的不确定性特点，提出了适用任意网格的三维辐射场快速计算模型。.本项目的研究成果能有效减低源项反演与事故后果评价的不确定性，满足我国核安全亟需稳健源项反演技术的需求。