基于源项估计的污染气体扩散数据驱动仿真

Data Driven Simulation of Polluted Gas Dispersion Using Source Estimation and Particle Filter

化工园区污染气体扩散的实时预测对于大气污染事故的应急管理具有重要意义。由于大气扩散模型中源项参数的未知以及输入参数的误差,传统的大气扩散仿真精度有限。因此提出了一种基于源项估计与粒子滤波的污染气体扩散数据驱动仿真方法。基于源项估计获取的源项参数,利用粒子滤波将无人机观测气体浓度数据实时注入扩散模型中以校正系统状态,获取更为准确的预测结果。实验表明,相比传统仿真方法更准确地预测了气体浓度分布,为应急处置提供有力的数据支撑。

The real-time prediction of the air contaminant dispersion in chemical industry park is important to the emergency management of air pollution accident. Due to the unknown source terms and the error of input parameters in the atmospheric dispersion model, the accuracy of traditional simulation is limited. A data driven atmospheric dispersion simulation based on source estimation and particle filter was proposed. Based on the results of the source estimation, particle filter was applied to assimilate the UAV observation into the dispersion model in real time to calibrate the system state and obtain more accurate prediction results. Experiments show that the proposed data driven atmospheric dispersion simulation method can predict the concentration distribution more precisely and provide strong support for emergency treatment.