基于稀疏系统辨识的空气污染源定位算法  

Sparse system identification algorithm for air pollution source localization

作　　者：余潇潇[1] 张林[1] 李安国[2] 

机构地区：[1]清华大学电子工程系,北京100084 [2]香港大学电机电子工程系

出　　处：《清华大学学报（自然科学版）》2013年第8期1077-1081,共5页Journal of Tsinghua University(Science and Technology)

摘　　要：空气污染已经成为现代人类社会面临的严峻挑战,快速准确地定位城市污染源能够有效提高对污染物排放的监管和治理。传统布设静态监测站的方法检测范围有限且成本高昂。该文提出了一种基于移动车载传感器网络的空气污染源定位算法,将污染源定位问题建模为稀疏系统参数辨识问题,依靠不完全的随机采样数据利用L1-范数正则化的最小均方误差优化方法重构稀疏信号,克服了无法采用少量样本数据对污染源进行及时准确定位监控的问题。仿真分析了在完全观测和部分观测情况下,对于污染场监测的准确性。结果表明:该算法在有限时间稀疏采样的场景下仍可快速准确地恢复出污染源位置及强度,实现对污染源的有效监测。Air pollution has been a severe threat for the contemporary human society. Timely and accurate detection of the locations of the pollution sources is very important for the management of waste disposal and rapid response to environmental disasters. However, due to the high costs of pollution monitoring devices, few and sparse monitoring stations are located in the vast area to be monitored, resulting in pollution information with poor resolution. An air pollution source localization algorithm was developed based on vehicular sensor networks, which models the problem as a theoretical problem, namely, stochastic system identification, with the problem with limited samples solved using L1-norm regularization optimization. Simulations show the validity of the proposed algorithm both with full observations and partial observations.

关 键 词：随机系统辨识 时空关联 环境监测 移动传感网 

分 类 号：TP393.2[自动化与计算机技术—计算机应用技术]