Abstract: Leakage accidents from storage tanks with dangerous chemicals may cause catastrophic consequences. For the purpose of improving the predicted accuracy of heavy gas dispersion during release incidents, the dispersion coefficients were improved in SLAB to obtain the SLAB-D model. Then SLAB-MD model was developed by integrating the optimization algorithms of dispersion coefficients based on the SLAB-M. The Jack Rabbit II (Trials 1, 7, and 9) liquid chlorine leakage experiments were employed as the dataset to evaluate and validate models' performance. The outcomes revealed that the four models' performances were ranked as SLAB-MD>SLAB-D>SLAB-M>SLAB, and the statistical performance measurements (SPMs) of the SLAB-MD model were close to the ideal values, among which the mean relative bias (MRB), geometric mean (MG), mean relative square error (MRSE), geometric variance (VG), and factor of 2 (FAC2) were 0.27, 1.34, 0.18, 1.24, and 0.93, respectively, which proved that SLAB-MD was excellent in accurately predicting the dispersion of heavy gas leaks. Finally, the model was used to analyze three typical heavy gas leakage and dispersion accidents at domestic and international, which obtained key information such as heavy gas concentration distribution and three-dimensional concentration surface. Additionally, the range of potential casualties was predicted based on the protective action criteria values (PACs). The potential impact zones at the downwind distance, including fatalities, serious injuries, and minor injuries, were 0.15, 1.75, and 7.6 km2, 0.22, 1.98, and 7.03 km2, and 0.12×10-3, 0.29×10-3, and 0.88×10-3 km2 in the context of liquid chlorine continuous release, liquid chlorine instantaneous leakage, and vinyl chloride continuous discharge, respectively. At sensitive points, the longest escape time was 28 min for liquid chlorine continuous leakage accident, and aftereffect time was 40 min for vinyl chloride continuous release accident. In summary, these predictions provide critical information and effective technical guidance that can be used for emergency response planning and management involving hazardous chemical material spills.

Key words: air pollution, heavy gas dispersion, meteorological parameters, model optimization, dynamic simulation

摘要： 危险化学品储罐一旦发生泄漏事故，将极有可能造成灾难性的事故后果。为了提高重气扩散模型模拟结果的精确性，改进SLAB模型中的扩散系数得到SLAB-D模型，然后基于SLAB-M模型改进扩散系数的算法，建立SLAB-MD模型；随后采用Jack Rabbit II液氯泄漏1, 7, 9号试验对模型性能进行评估。结果表明，四种模型的性能为SLAB-MD>SLAB-D>SLAB-M> SLAB，其中SLAB-MD模型表现优异，统计性能评估指标接近最优值，平均相对偏差, 几何平均偏差, 平均相对平方差, 几何平均方差和FAC2的值分别为0.27, 1.34, 0.18, 1.24和0.93。利用SLAB-MD模型对三起典型危化品泄漏事故进行模拟分析，预测重气浓度分布、三维浓度曲面等关键信息，并根据保护行动标准值预测事故伤亡范围。这些预测结果可为危险化学物质泄漏的应急救援提供有效的技术指导。

关键词: 大气污染, 重气扩散, 气象参数, 模型优化, 动态模拟