基于遗传算法的空调风系统突发污染快速溯源

Genetic Algorithm for Sudden Contaminant Source Identification in Ventilation System

摘要:

为了在空调风系统内突发污染发生的短时间内辨识到污染源，进而采取相应的应急措施保障室内人员安全，建立了不同污染场景下的风系统传感器模拟浓度数据库，基于遗传算法，以传感器模拟计算浓度与实测浓度差的平方为适应度函数，研究了风系统采用不同的传感器数目、不同的传感器监测位置与时间、传感器性能参数变化及节点间距对反演结果总平均相对误差(ξ)的影响.算例结果表明：针对某一特定风系统，经位置优化后的传感器数目越多，ξ越小，同时考虑ξ的降低幅度及传感器成本，该风系统内宜设置3个传感器；在三传感器系统条件下，溯源计算收敛子代数少且收敛时间短(<5 s)，可实现快速溯源；传感器监测位置与首次监测到污染物的时间对溯源误差的作用是耦合的，传感器监测误差对源位置x0反演没有影响，而传回数据时间间隔越大，源位置x0反演相对误差越大；最后风系统内节点总数存在一个适当范围兼顾全局寻优及降低计算负荷的作用.

Abstract:

This paper illustrated a source identification method in air duct system, aiming at detecting contaminant source within a short period of time after a bioterrorist attack. The method was based on genetic algorithm (GA) with minimal difference between calculated concentration and measured concentration as fitness function. We established a database of calculated concentration of sensors considering different releasing scenarios. Then we discussed the impact of the number of sensors, the location and measurement time of sensors, the capability of sensors, and the distance between nodes on the overall average relative error of inversion results (ξ). Results of a case study showed that the ξ decreased as more sensors were set in the ventilation system. The optimized number of sensors in this case was supposed three, considering both the decrement provided by each sensor and the high cost of each sensor. Meanwhile, the convergence generations were few while the convergence time was short. Then, the impact of sensor location and detection time on the ξ was coupled. The inversed source location x0 is sensitive to the time interval of feedback data but nonsensitive to the detecting error of sensors. Finally, there existed an appropriate number of nodes distance in air duct system, which gave consideration to lower calculated load and global optimization.